A Historical Phonology of English

A Historical
Phonology of English
Donka Minkova
EDINBURGH TEXTBOOKS ON THE ENGLISH LANGUAGE – ADVANCED

A Historical Phonology of English
Edinburgh Textbooks on the English Language – Advanced
General Editor
Heinz Giegerich, Professor of English Linguistics, University of Edinburgh
Editorial Board
Laurie Bauer (University of Wellington)
Olga Fischer (University of Amsterdam)
Rochelle Lieber (University of New Hampshire)
Norman Macleod (University of Edinburgh)
Donka Minkova (UCLA)
Edgar W. Schneider (University of Regensburg)
Katie Wales (University of Leeds)
Anthony Warner (University of York)
TITLES IN THE SERIES INCLUDE :
Corpus Linguistics and the Description of English

Hans Lindquist
A Historical Phonology of English

Donka Minkova
A Historical Morphology of English

Dieter Kastovsky
Grammaticalization and the History of English

Manfred Krug and Hubert Cuyckens
A Historical Syntax of English

Bettelou Los
English Historical Sociolinguistics

Robert McColl Millar
A Historical Semantics of English

Christian Kay and Kathryn Allan
Visit the Edinburgh Textbooks in the English Language website at www.

euppublishing.com/series/ETOTELAdvanced
A Historical Phonology of
English
Donka Minkova
© Donka Minkova, 2014
Edinburgh University Press Ltd

22 George Square, Edinburgh EH8 9LF
www.euppublishing.com
Typeset in 10.5/12 Janson by

Servis Filmsetting Ltd, Stockport, Cheshire,
printed and bound in Great Britain by
CPI Group (UK) Ltd, Croydon CR0 4YY
A CIP record for this book is available from the British Library
ISBN 978 0 7486 3467 5 (hardback)

ISBN 978 0 7486 3468 2 (paperback)
ISBN 978 0 7486 3469 9 (webready PDF)
ISBN 978 0 7486 7755 9 (epub)
The right of Donka Minkova

to be identified as author of this work
has been asserted in accordance with
the Copyright, Designs and Patents Act 1988.
Contents
Acknowledgements x
List of abbreviations and symbols xii
A note on the Companion to A Historical Phonology of English xv
1 Periods in the history of English 1
1.1 Periods in the history of English 2
1.2 Old English (450–1066) 2
1.3 Middle English (1066–1476) 9
1.4 Early Modern English (1476–1776) 15
1.5 English after 1776 17
1.6 The evidence for early pronunciation 20
2 The sounds of English 24
2.1 The consonants of PDE 24
2.1.1 Voicing 26
2.1.2 Place of articulation 27
2.1.3 Manner of articulation 29
2.1.4 Short and long consonants 31
2.2 The vowels of PDE 32
2.2.1 Short and long vowels 35
2.2.2 Complexity: monophthongs and diphthongs 37
2.3 The syllable: some basics 39
2.3.1 Syllable structure 39
2.3.2 Syllabification 40
2.3.3 Syllable weight 43
2.4 Notes on vowel representation 45
2.5 Phonological change: some types and causes 46
3 Discovering the earliest links: Indo-European – Germanic –
Old English
FOOT-PODIUM, TOOTH-DENTAL, HILL-CULMINATE, THREE-TRIPLE 54
3.1 Family matters: Indo-European – Germanic – Old
English 54
vi A HISTORICAL PHONOLOGY OF ENGLISH
3.2 The Indo-European family of languages 55

3.3 The Germanic branch of Indo-European 59
3.4 Some pre-Old English segmental and prosodic changes 61
3.4.1 Grimm’s Law, or the First Germanic Consonant
Shift 61
3.4.2 Some IE vowel changes in Germanic 68
3.4.3 Early prosodic changes: stress and syllable
weight in Germanic 69
3.4.4 Lengthening of final vowels in stressed
monosyllables 70
3.4.5 West Germanic (Consonant) Gemination (WGG) 71
4 Consonantal histories: Old English
KIRK-CHURCH, DAY-DAWN, SAY-SAW-SAGA, SKIRT-SHIRT,
SHRIFT-SCRIPT, DISH-DISK-DISCUS, LOAF-LOAVES vs SERF-SERFS,
ELF-ELFS-ELVEN, BATH-BATHS-BATHE, BELIEF-BELIEFS-BELIEVE 74
4.1 The consonants of OE 74
4.1.1 Singletons 74
4.1.2 Geminates 76
4.2 Sound-to-spelling correspondences of the consonants in
OE 81
4.2.1 The <g>’s of OE 82
4.3 Palatalisation and affrication of velars in OE 84
4.4 Morphophonemic alternations: fricative voicing. CLOSE
enough to CLOSE the door? 88
5 Consonantal developments in the second millennium
‘AND WOT ’AVE WE ’ERE, GUV’NOR?’, WHICH-WITCH, THOUGH-
TOUGH, FAR-FA, BRIDGESTOWE-BRISTOL, WRITE-RIGHT, IAMB-
IAMBIC, GIGOLO, MATURE-GOCHA, BETTER-BEDDER, MUS GO 99
5.1 H-related histories: AND WOT ’AVE WE ’ERE, GUV’NOR? 99
5.1.1 Phonetic and phonological properties of /h/ in
PDE 99
5.1.2 The velar and glottal fricatives of OE 102
5.1.3 Initial glottal fricatives in ME and EModE:
ABLE-HABILITATE, WHICH-WITCH, WHINE-WINE 105
5.1.4 Non-initial glottal fricatives in ME:
THOUGH-TOUGH 112
5.2 R-related histories 115
5.2.1 Phonetic and phonological properties of the
rhotics in PDE 115
5.2.2 Reconstructing the phonetics of <r> in OE 116
5.2.3 Pre-consonantal /r/-loss 121
CONTENTS vii
5.2.4 Post-vocalic /r/-loss 125

5.2.5 The other liquid: BRIDGESTOWE-BRISTOL
Historical parallels between /r/ and /l/ 128
5.3 Cluster simplification: KNIGHT-NIGHT, WRITE-RIGHT,
IAMB-IAMBIC 132
5.3.1 Initial <kn-, gn-, wr->: KNIGHT-NIGHT, WRITE-
RIGHT 132
5.3.2 Final <-mb, -mn, -ng>: IAMB-IAMBIC, DAMN-
DAMNATION, SINGER-LINGER 134
5.4 Other inventory changes: the adoption of // 141
5.4.1 More alveolar palatalisations and affrications:
s-, t-, d- + -j. GOTCHA, INJUN 143
5.5 The glottal stop [ʔ] and the alveolar tap [ɾ] 145
5.5.1 The glottal stop 145
5.5.2 Voicing of [t] and tapping of [t] and [d]:
MATTER-MADDER 147
5.6 Recent trends: [ts-, ʃm-, ʃl-, ʃt-]: MASH POTATO,
MANAGE CARE, STAIN GLASS 148
6 The vowels in Old English: spelling, pronunciation. PDE
alternations traced back to OE
FOOT-FEET, FULL-FILL, MAN-MEN, CHILD-CHILDREN,
HOUND-HUNDRED 151
6.1 The vowel inventories of PDE and OE: a comparison 151
6.2 Orthography and the reconstruction of OE vowels 153
6.3 I-Mutation: FOOT-FEET, FULL-FILL, SELL-SALE 156
6.4 OE homorganic-cluster lengthening: CHILD-CHILDREN,
HOUND-HUNDRED 164
6.4.1 Are pre-cluster lengthenings prosodically
incongruous? 169
6.5 The late OE vowel inventory in detail 171
6.5.1 The short vowels in late OE 171
6.5.2 The long vowels in late OE:
STREET-STRATFORD 173
6.5.3 Diphthongs and diphthongoids 175
6.5.4 Unstressed vowels 179
7 The vowels in Middle English. Dialects. Spelling innovations.
Vowel quality and quantity. PDE alternations traced back to
ME
DIZZY-BUSY, FURY-BURY, MOON-MONDAY, STEAL-STEALTH,
GAME-GAMMON, GRASS-GRAZE 184
7.1 ME dialects 184
viii A HISTORICAL PHONOLOGY OF ENGLISH
7.2 Notes on ME spelling (vowels): <TAXI>-<ATAXY>,

<TYRE>-<TIRE>, <COME>-<CUT>, <SEE>-<SEA> 186
7.2.1 Letter-to-vowel correspondences in ME (single
letters): SUN-SON, CONE-COME 187
7.2.2 Letter-to-vowel correspondences in ME
(digraphs): BEAT-BEET, ROAD-ROOD 190
7.3 Qualitative changes in ME 192
7.3.1 Short vowels: DIZZY-BUSY, FURY-BURY, MERRY-
MIRTH 192
7.3.2 Long vowels: GAE-GO, OAKE (Somerset) - AIKE
(East Riding of Yorkshire) 200
7.4 The diphthongal system of ME 204
7.5 Quantitative changes: why do they matter? 209
7.5.1 Shortenings: CREEP-CREPT, MOON-MONDAY,
MOUTH-PORTSMOUTH, LEEK-GARLIC 212
7.5.1.1 Pre-consonantal shortening: FEED-FED,
WIDE-WIDTH, SHEEP-SHEPHERD 212
7.5.1.2 Trisyllabic shortening: HOLY-HOLIDAY,
SOUTH-SOUTHERN 216
7.5.1.3 Vowel shortening in unstressed syllables:
BUT-ABOUT, LEEK-GARLIC 219
7.5.2 Lengthenings: GAME-GAMMON, GRAZE-GRASS 220
7.5.2.1 ME open-syllable lengthening 221
7.5.3 Orthography and quantitative changes 225
7.6 Reduction and loss of unstressed vowels in ME 227
8 Vowel quality and quantity in Early Modern English and later

MOTH-MOTHER, DULL-BULL, LOST-POST, FEAR-BEAR, HERE-
THERE, MOOD-STOOD-BLOOD, DEAD-BEAD 234
8.1 Information glut. New sources of phonological
reconstruction 234
8.2 Qualitative changes 236
8.2.1 Short vowels: QUASH-QUACK, WASH-WAX, GOD-
EGAD, PUTT-PUT 236
8.2.2 The Great/Long/English Vowel Shift 248
8.2.2.1 Chronology and dating 252
8.2.2.2 Mechanism and causation 256
8.2.2.3 Further instability and enrichment:
BREW-NEW, DO-DUE, AUNT-HAUNT 267
8.3 The effect of phonotactics on long vowel shifting 271
8.3.1 Shortening in monosyllabic words: LEAD (Pb)-
LEAD, V., DEAF-LEAF, MOOD-STOOD-BLOOD 271
CONTENTS ix
8.3.2 Vowels in relation to /r/: PERSON-PARSON, TEAR,

v. - TEAR, n., FLOOR-POOR 274
8.4 Continuity or reinvention 281
8.4.1 Canadian or long-vowel ‘Raising’: ICE [iS] -
EYES [aiZ] ~ [aεZ] ~ [aə] ~ [a
] 281
9 The evolution of the English stress system
ALWAYS-CAUSEWAYS, PRÉSENT, n. - PRESÉNT, v.,
HARÁSS ~ HÁRASS 284
9.1 Preliminaries: definition of some terms 284
9.2 Syllable structure and syllable weight 286
9.3 Historical sources of information for prosodic
reconstruction 288
9.3.1 Orthographic evidence for word-stress 289
9.3.2 Verse evidence for stress 290
9.4 OE stress placement 294
9.4.1 OE word-stress 294
9.4.2 OE stress above the word level 298
9.5 ME stress placement: the native component 303
9.6 ME prosodic innovations 306
9.6.1 Grammar, meaning and stress-shifting:
PERFÉCT-PÉRFECT, CANÁL-CHÁNNEL 310
9.7 ME compound and phrasal stress 312
9.8 Post-ME prosodic innovations 314
10 Early English verse forms: from Cædmon to Chaucer 323
10.1 Preliminaries: speech prosody vs poetic meter, stress vs
ictus 323
10.2 Alliterative verse 326
10.2.1 Classical OE alliterative verse 327
10.2.2 Continuity and reinvention of alliterative
versification in ME 339
10.3 Introduction of rhyme, syllable-counting, and binary
foot structure 345
10.4 Chaucer and the invention of the iambic pentameter 356
Bibliography 367
Word Index 396
Name Index 412
Subject Index 417
Acknowledgements
This book has grown out of nearly four decades of studying, teaching,
and researching the history of English, yet its chronological and the-
matic breadth presented unforeseen and seemingly endless challenges.
It is a pleasant duty to recognise the encouragement and support I have
received in dealing with these challenges.
My first acknowledgement goes to the University of Edinburgh’s
Institute for Advanced Studies in the Humanities for an annual
Fellowship, giving an East European beginner a career-changing
opportunity to benefit from its library resources and invaluable col-
legial advice and criticism. Angus McIntosh trusted me with my first
Middle English classes and Roger Lass let me loose on his Phonology
Two students. During that year I started to think of myself as an English
Historical Phonologist. The influence of my first teachers will be
obvious throughout the book.
I wish to thank UCLA and more specifically the Council on
Research and the Center for Medieval and Renaissance Studies
for support throughout the lengthy gestation of this book. The
Department of English at UCLA has provided a friendly and accom-
modating working environment and a steady stream of intelligent and
demanding students, without whom such work would be barren. My
colleagues in the Department of Linguistics at UCLA have always
welcomed me in their seminars and discussions; the analytical rigour
and innovation in their work sets the bar high and I feel fortunate to be
part of that community.
A generous grant from the Arts and Humanities Research Council
of Great Britain on ‘The Verse Forms of Middle English Romance’ has
been providing incentives and opportunities for exciting collaborative
projects with Ad Putter and Judith Jefferson of the University of Bristol.
Some of the results of our ongoing research are cited in this book.
I am grateful to the ETOTEL team at Edinburgh University
Press, Esmé Watson, Sarah Edwards, Vicki Donald, Michelle Houston,
x
ACKNOWLEDGEMENTS xi
Gillian Leslie, James Dale and my copy-editor Eliza Wright for their
sustained support in spite of the many delays. I gratefully acknowledge
the assistance of Gillan Adler, UCLA, in the preparation of the indexes.
Special thanks go to Heinz Giegerich, the series editor. His much-
tested patience, his encouragement and his unfaltering faith in the value
of this project helped me persist stubbornly even when I felt that I had
exceeded the limits of metaphorical chewing.
Henry Ansgar (Andy) Kelly, medievalist, Latinist, Satanist, philolo-
gist, was characteristically observant and generous with his comments
on the overlong first draft. I feel lucky that his state of Emeritude has
only enhanced his inexhaustible curiosity, attention to detail and spir-
ited collegiality.
This book should have been co-authored with Robert Stockwell. For
many years he and I talked about, argued about and wrote about English
historical phonology. After 2006 Parkinson’s gradually deprived him
of his keen eye and the ability to assert his views. Without the produc-
tively critical bouncing back-and-forth of drafts with him, completing
the project has been a lonely trip. Dedicating this book to Robert is the
least I can do for one of the finest contemporary English historical pho-
nologists and the most important person in my life.
UCLA
Los Angeles, 18 October 2012
Abbreviations and symbols
AAVE African American Vernacular English

adj. Adjective
AmE American English
ANAE Atlas of North American English
AN Anglo-Norman
Angl. Anglian
AS Anglo-Saxon
AusE Australian English
BrE British English
C Consonant
CE Canadian English
DOE The Dictionary of Old English
DOEC The Dictionary of Old English Corpus
EModE Early Modern English
Fr. French
GA General American
Ger. German
Gk Greek
Gmc Germanic
Goth. Gothic
GSR The Germanic Stress Rule
GVS Great Vowel Shift
H Heavy syllable
IE Indo-European
IPA International Phonetic Alphabet
L Light syllable
LAEME The Linguistic Atlas of Early Middle English
LALME The Linguistic Atlas of Late Middle English
Lat. Latin
LModE Late Modern English
LOE Late Old English
xii
ABBREVIATIONS AND SYMBOLS xiii
LP Linguistic Profile
LVS Long Vowel Shift
MDu Middle Dutch
ME Middle English
MED The Middle English Dictionary
N nasal
n. Noun
NED New English Dictionary on Historical Principles
Norw. Norwegian
NY New York
NZE New Zealand English
OE Old English
OED The Oxford English Dictionary
OFr Old French
ON Old Norse
PDE Present-Day English
PIE Proto-Indo-European
PrG Proto-Germanic
RP Received Pronunciation
SAE South African English
Skt Sanskrit
Sp. Spanish
SSBE Standard Southern British English
SSE Standard Scottish English
V Vowel
v. Verb
WG West Germanic
WGG West Germanic (Consonant) Gemination
WS West Saxon
YE East Riding of Yorkshire
[] phonetic representation
[C] ambisyllabic consonant
<> orthographic representation
* reconstructed form, also unattested form
† obsolete
. syllable boundary
: rhymes/alliterates with
# word boundary
~ alternates with
艐 approximately the same as
< previous stage/the input of a change
> next diachronic stage/the output of a change
xiv A HISTORICAL PHONOLOGY OF ENGLISH
For the full names of the Old English texts, their abbreviations, for
named authors, and bibliographical information see <http://tapor.
library.utoronto.ca/doe/dict/bibl/index.html> (last accessed 27 May
2013).
For the ME authors and titles see <http://quod.lib.umich.edu/h/
hyperbib/> (last accessed 27 May 2013).
A note on the Companion to
A Historical Phonology of
English
A Companion to this book is available at http://www.euppublishing.

com/page/ETOTELAdvanced/Minkova. The Companion includes
additional readings, exercises, comments and further resources. Online
resources will be indicated in the text by the icon :.
xv
In loving memory of
Robert Stockwell
(12 June 1925–28 October 2012)
1 Periods in the history of
English
This book is about the evolution of the phonological structure of

English: the history of individual sounds and their representation, the
history of syllable structure and word stress. Our very first job is to
clarify what we mean by ‘English’; one hardly needs a reminder that no
two speakers of English sound exactly alike. A phone call from a com-
plete stranger immediately triggers associations based on age, gender,
place of birth, education, social status and race. From exposure to the
language we know the typical characteristics for each group of speakers;
we also know that many or all of these characteristics can be unstable.
In spite of the amazing amount of group or individual variation, there
are enough shared properties that allow us to construct a mental rep-
resentation of how an English word, phrase or sentence sounds; thus
we can ignore voice quality and accent and communicate in English
with people from very different backgrounds and regions. Like all
other living languages, English is subject to constant change. What John
Donne cherished and bequeathed as ‘mine English tongue’ (The Will,
1633), is a very different-sounding English in the twenty-first century.
We start, therefore, by acknowledging that the notion of a single
‘English’ is a convenient abstraction and a cover term for a multitude of
clearly distinct yet mutually intelligible varieties of one language.
The use of the noun English with reference to the language spoken
by the Anglo-Saxons, as distinct from Latin or Celtic, must have been
current from the time of the earliest contacts between the original
inhabitants of the British Isles and the ‘newcomers’, who were speakers
of north-west Germanic. Exactly when the first Germanic-speaking
settlers arrived in Britain is not known; recent genetic mapping sug-
gests that the Germanic colonisation of the Celts may have begun
even before the Romans took administrative control of large sections
of the island in the first century AD. What we do know for sure is that
large permanent settlements of Angles and Saxons between c. 400–600
justified a separate name for the linguistic identity of the continental
1
2 A HISTORICAL PHONOLOGY OF ENGLISH
transplants. In writing, the earliest record of the noun English is found in

a ninth-century translation into what we now call ‘Old’ English of the
Ecclesiastical History of the English Nation, originally composed in Latin by
the historian and scholar Bede (c. 672–735). By the middle of the four-
teenth century there appear references to specific regional varieties of
English, such as the English of Kent, and somewhat later, in the sixteenth
century, we find English characterised further as new, old, northern and
southern. In spite of the always-present and growing diversity of English,
especially after the language was exported to other continents, it was
only early in the twentieth century that English began to be treated also
as a count noun; the OED records the first such use in H. L. Menken’s
article ‘The Two Englishes’, which appeared in the Baltimore Evening
Sun, 15 September 1910. A hundred years later there are more Englishes
than ever before, including blends such as Chinglish, Taglish, Frenglish/
Franglais, Singlish, Janglish and more.
1.1 Periods in the history of English

English as spoken in the twenty-first century is only the latest and the
most readily observable stage in a very long history. Like a living organ-
ism, our language both carries the marks that identify it as a member of
a specific branch of the Indo-European family, Germanic, and has its
own peculiarities, setting it apart from its ‘genetic’ relatives. One of the
goals of this book is to describe the phonological features of the modern
language in terms of its development, seeking to reveal how the present
is indebted to the past. A telescopic view back into the evolution of the
language requires us to set up a frame of temporal references and situate
the linguistic findings in their historical, literary and social setting. We
start with a survey of the main chronological divisions: Old English
(OE), Middle English (ME), Early Modern English (EModE) and
Present-Day English (PDE). Notice how the divisions become longer
as we look back: PDE is the language of roughly the last two centuries,
EModE spans about three centuries, ME – four centuries, while the
most distant period, OE, stretches over more than six centuries. The
further back we go, the less familiar we are with the socio-historical
setting of linguistic change, thus the focus in the brief survey below will
be primarily on the older periods of English.
1.2 Old English (450–1066)

Traditionally, Old English, also referred to as Anglo-Saxon, is assigned
a birth date around AD 450. By the beginning of the fifth century pro-
PERIODS IN THE HISTORY OF ENGLISH 3
longed threats from the north, the west and the east contributed to
internal strife and disorder within the Roman Empire. In 400 troops
were withdrawn from Britain to defend Italy against the invasion of
Alaric the Goth and in 407 a large contingent of Roman troops were
transferred from Britain to the Continent to bolster the armies fighting
against Gaul and Spain. In 410 an appeal for support for the remaining
Roman troops in Britain was rejected; that year marks the end point of
what we call ‘Roman Britain’.
After a hiatus of about forty years, during which time the rest of the
Romans must have either left or become assimilated to the local Celtic-
speaking population, a new, extensive and permanent occupation of
Britain took place. According to an entry for the year 449 in Bede’s
Ecclesiastical History of the English Nation, the first Germanic settlers
migrated to Britain within the next seven years. This is the justification
for associating the earliest stage of English, OE, with the year 449 in
text- and reference books, although, as noted above, there is mounting
evidence that members of Germanic tribes had probably lived in Britain
since the second century.1
The demographic balance after the Germanic invasion was originally
in favour of the indigenous Celts who outnumbered the conquerors by
a considerable measure. The estimated number of settlers participating
in the Germanic diaspora ranges between as little as 10,000 and up to
200,000. In some estimates, most of the three and a half million speakers
of Celtic survived the initial conquest.2 However, the Celts had limited
military experience and lacked the organisation to resist the incursions
for more than half a century and by c. 550 larger and larger groups of
Germanic-speaking peoples moved in, pushing the Celts away – those
whom they did not kill or enslave – from the central part of the country
west and south towards Cornwall and Wales, and north to the Lothian
region. By the end of the sixth century the dominant language spoken
on the British Isles was no longer Celtic. Old English had ‘begun’. The
end of the Celtic territorial and political dominance also determined the
direction and the scope of the linguistic influence of Celtic on English:
as is often noted, the transfer of lexical items from the language of the
conquered into a higher-status language, in this case English, can be
expected to be quite limited. This is not necessarily the case with the
1
See Wakelin (1988: 180); the evidence is addressed more specifically in Oppenheimer
(2006).
2
The distribution of Celtic speakers in the different regions was uneven, with the south
more heavily Germanic than the north, where the ratio of Celts to immigrants may
have been as much as 50:1; see Tristram (2002: 113–14).
transfer of phonological and prosodic features, though the latter are

much harder to isolate. Bilingualism and the physical proximity of
Celtic and English speakers in the British Isles has resulted in shared
phonological features between the modern languages Irish and Welsh,
and Irish English and Welsh English.
In addition to the continuing Celtic–OE bilingualism in the geo-
graphical areas of contact, the reintroduction of Christianity called for
the study of Latin. Christianity had been practised in Roman Britain,
but the Germanic settlers of the fifth and the sixth centuries were
pagan, hence the use of ‘reintroduction’ of Christianity. The conversion
began in 597, when a Roman monk, Augustine, and his missionaries
were allowed to preach and spread the Christian faith in the southern
kingdom of Kent. About thirty years later in the north, a Bishop Aidan
came from Ireland to Northumbria, and, through his efforts, the north-
ern kingdoms were converted. : What resulted was a dangerous rivalry
between the version of Christianity introduced in the south from Rome,
and the modes of worship introduced from Ireland in the north. The dif-
ferences were settled in a convocation known as the Synod of Whitby,
which took place in 663–4. The Synod’s decision unified the religious
practices across England; by 669 the Church acknowledged a single
head, a Roman Archbishop named Theodore of Tarsus (602–90). The
promotion of the Roman Catholic mode of worship brought the English
Church into close contact with the Continent and further enhanced the
need and popularity of Latin among English clerics.
The unification of the Church ushered in a period of intense and organ-
ised scholarship, known also as ‘The Golden Age of Bede’. The study of
Latin and the translation of many liturgical and scholastic texts from
Latin into OE was a central concern of the officers of the Church, making
religion an important channel through which learned Latin borrowings
came into the language. Loanwords having to do with trade, warfare and
household objects were characteristic of the earlier, Continental contacts
between Germanic and Latin, while loanwords from Latin after the
middle of the sixth century are mostly associated with religion, books,
learning and writing. Characteristically, the words that came into the
language through the monks’ scholastic culture were introduced through
the written language, though the oral aspect of the transmission is also
important. Silent reading was not a common practice, and instruction
was done mainly through memorisation and repetition. This allowed for
a quick adaptation of the more frequently used loanwords to the phonol-
ogy of the native language. Latin words recorded in OE are of course not
restricted to religious terms; there are words from the fields of medicine,
biology and the arts, and include plain core vocabulary.
Phonologically, the early loanwords were commonly assimilated to

OE: Latin words such as candēla, comēta, which would have had stress on
the penultimate syllable in the source language, shifted their stress onto
the first syllable, following the Germanic model of root-initial stress. In
OE verse such words alliterate on [k-].
The political and administrative history of multi-kingdom Anglo-
Saxon England was unstable.3 In the first half of the seventh century
the kingdom of Northumbria had considerable powers over large areas
south of the river Humber. Throughout that century Northumbria and
Mercia were in continuous rivalry, with Mercia emerging dominant
after 678. The supremacy of Mercia continued into the first quarter of
the ninth century, challenged only by Wessex. During the ninth century
the rivalry between Mercia and the rising power of Wessex became less
important than the need of unity in the face of an outside enemy.
At the same time as the OE language was developing various local
and regional varieties whose names are associated with the political
and cultural divisions of the Anglo-Saxon kingdoms, large parts of
northern Europe were raided and colonised by the Scandinavian seafar-
ing pagan tribes, jointly referred to as the Vikings. Linguistically, the
Vikings’ language differed from the western branch of Germanic that
the Anglo-Saxons spoke; by c. 600 their language was already a distinct
northern variety of Germanic, known as Old Scandinavian or Old Norse
(ON). The Viking Age in Europe is dated c. 750–1050. During that time
Old Norse was spoken not just in present-day Scandinavia but also in
Iceland, Greenland, the Faroe Islands and in parts of Ireland, Scotland,
England, Northern France (Normandy) and Russia. England was not
spared: the Scandinavian incursions and the permanent settlements of
ON speakers in large parts of eastern England and Yorkshire became
an important source of linguistic diversity in OE, though some of the
effects of this linguistic contact are not recorded until later.
The invasions started at the end of the eighth century. The Anglo-
Saxon Chronicle, the most important historical source recording contem-
poraneous events, describes the first invasion under the year 787 when
‘there first came three ships of the Northmen . . . the first Danish ships
that came into England’. Throughout the ninth century the incursions
and settlements by new ON speakers continued; the Anglo-Saxons
3
Before the middle of the ninth century, the country was divided into kingdoms whose
political importance rose and fell; the seven most important kingdoms of the early
period, often referred to as the heptarchy, were Northumbria, Mercia, East Anglia,
Wessex, Essex, Sussex and Kent. Unlike the kingdoms and states of later history, these
kingdoms did not have firmly fixed boundaries and their internal organisation was
based on both kinship and perceived merit.
Figure 1.1 England after 886 (public domain image taken from
<http://history medren.about.com/library/atlas/natmapengland886.htm>,
accessed 16 May 2013)
usually referred to them as Danes, although Norwegian presence is also

identifiable in the North-West Midlands.4 Unable to resist the attacks,
in 878 King Alfred of Wessex signed a treaty with the Scandinavian
ruler of East Anglia establishing a territory north-east of the river
Thames, which came to be called the Danelaw, also spelled Danelagh or
Danelaga, meaning ‘Danes’ Law’. The treaty suspended the continuous
warfare between speakers of OE and ON in that region, allowed the
newcomers to set up new households, till the land and trade with the
4
See Dance (2003: 51, 151); Bibire (2001: 89).
local population, thus putting down permanent roots in England. Most

important for the evolution of English, it legitimised the presence of
another, northern, strand of Germanic in the country and created con-
ditions for an enduring, and most likely peaceful integration of the two
related languages.
Towards the end of the tenth century a new wave of Scandinavian
invasions hit the country. Between 1013 and 1016 the English crown
passed from Danish to English and back to Danish hands. Between 1017
and 1035 the country was ruled by King Canute, who was also King
of Denmark and King of Norway. After the initial takeover turmoil,
King Canute’s reign appears to have been non-violent and orderly,
promoting further linguistic integration. Two of Canute’s sons reigned
briefly after their father’s death. The Scandinavian rule of England
continued until 1042, when the English royal line was once again
restored. During the decades of Scandinavian reign in England the lin-
guistic contacts between speakers of the two varieties of Germanic were
very close, possibly resulting in a kind of mixed, creolised vernacular
language.
It is estimated that between the end of the eighth and the beginning of
the twelfth century, about one thousand common words were adopted
from ON. The new loans were not typically learned words, so they were
transmitted orally, and the record of their adoption into English lags
behind the actual time when these words must have come into circula-
tion. Words whose presence or form in OE can be traced back to ON
are shown in (1).
(1) Words whose presence or form in OE can be traced back to ON:
ceallian ‘to call’ feolaga ‘fellow’
cnif ‘knife’ ragg(g) ‘rough, shaggy’
hæfen ‘haven’ utlaga ‘outlaw’
husbonda ‘householder, husband’ wrang ‘wrong’
Other common words borrowed through contact with Scandinavian
are bank, bull, cast, gape, guess, hap ‘luck, success’, leg, loan, score, skill, sister,
skin, sky, wing. The meanings of many of these words were already fully
or partially covered by OE words, thus sky, originally in ON meaning
‘cloud’, developed also the meaning ‘firmament’ and overlapped with
OE heofon ‘heaven, atmosphere’; ON vengr ‘wing’ replaced one of the
meanings of OE feþer ‘feather’, which used to mean ‘feather’, ‘wing’ and
‘pen’.5
5
The latter meaning was replaced in the fourteenth century by an Old French loan
penne < Latin penna ‘feather, quill-feather used for writing’.
Scandinavian and late OE/early ME frequently shared the Proto-

Germanic (PrG) root of the same word. However, because of the dif-
ferent sound changes within each branch, the forms of the native
continuation and the borrowing differ in PDE, so that we get pairs such
as shirt-skirt, church-kirk, where the first word reflects late OE changes,
while the second word retains the ON pronunciation. The influence
of Scandinavian is attested also in the pronunciation of words such as
again (OE on-ġean), get (OE ġietan), give (OE ġiefan), in which the bold
letters would be pronounced [j], as in PDE yes, yard, had it not been for
the [g-] pronunciation of the Scandinavian speakers (see 4.2.1, 4.3). ON
also contributed to some vowel alternations in OE and ME (see 6.2,
6.5.3, 7.4).
The everyday contacts of the Anglo-Saxons with speakers of Celtic
and Old Norse and the exposure of clerics to Latin complicated the
linguistic realities of Old English. Also, it is logical to assume that
the populations of the various historical kingdoms must have spoken
distinct dialects, though what we can reconstruct is not strictly area-
determined, but based on clusters of texts reflecting the scribal tradi-
tions in major monastic centres such as Durham, Lichfield, Winchester
and Canterbury (Hogg 2006a: 358). Thus the linguistic situation in
Anglo-Saxon England was not unlike the picture we get in any modern
language: time, place and sociocultural factors work against the notion
of linguistic homogeneity. It is therefore evident that normal dialect
divergence and language contact make the single label, Old English,
inadequate to cover six hundred years of turbulent history. : Only
a small part of the non-homogeneous OE language can be recovered,
however, and in this book we will stay with the traditional umbrella
label Old English with references to specific varieties whenever needed.
As for resources, the entire surviving body of OE materials, 3,047
texts, is now available in digitised form. Excluded are only some vari-
ants of individual texts. Web access was made available by the ongoing
Dictionary of Old English Project, Centre for Medieval Studies, University
of Toronto. The Old English Corpus (OEC) is a research resource
which was originally prepared for internal use at the Dictionary of Old
English (DOE). When completed, this will be the most important and
versatile electronic resource for the study of OE. The progress report
for 2010 informed us that ‘more than one third of the Dictionary – eight
of the 22 letters of the Old English alphabet – has been published, and
more than 60% of the total entries have been written to date’.6 The
6
Available at <http://www.doe.utoronto.ca/index.html> (last updated 22 July 2011,
last accessed 16 May 2013).
DOE report of May 2011 states: ‘The editors are making good progress
towards completing the writing of H (the next letter to be published).
We are also drafting entries for I/Y, L, M, and N, and the lemmatization
of S (the largest letter) is continuing.’
1.3 Middle English (1066–1476)

Deciding when and why it is appropriate to label a reconstructed lan-
guage as no longer ‘old’ but ‘middle’ is not easy. The periodisation of
historical English into Old, Middle and Early Modern, originally pro-
posed by the prominent English philologist Henry Sweet in 1874, was
based on an important and orthographically testable criterion: the neu-
tralisation of vowel distinctions in final unstressed syllables as attested
in prototypical texts. Sweet was also fully aware that strict chronological
divisions are unrealistic, and that ‘transition’ periods during which the
linguistic evidence is not clear-cut have to straddle the core periods.
In Sweet’s schema, transition OE stretches from 1100–1200, followed
by early Middle English from 1200–1300, late Middle English from
1300–1400 and, again, transition Middle English from 1400–1500. The
proposed partitions were later challenged by new evidence for vowel
reduction in manuscripts of the second half of the tenth century: ‘The
transition period from Old to Middle English is not the twelfth century,
as the grammarians used to think, nor even the eleventh, as most of them
think today, but rather the tenth’ (Malone 1930: 117) (see also 6.5.4). At
the other chronological end, Kitson (1997) points out the persistence
of OE linguistic features in non-literary documents into the early thir-
teenth century. The problematic nature of such divisions notwithstand-
ing, we still need a label for the linguistic phase that follows OE, so in
this book we will follow the Cambridge History of the English Language
in dating the ‘beginning’ of ME conveniently, if arbitrarily, from a
linguistic perspective, to the year of the Norman Conquest of 1066.
The events surrounding the Conquest are well known: in 1051
William, Duke of Normandy (1028–87) had almost certainly been
promised to succeed the childless Edward the Confessor, the last
King of the Anglo-Saxon royal line. When Edward died early in 1066,
however, a well-liked and powerful English earl, Harold, was elected
King. In pursuit of his alleged royal rights, William assembled a force
of about 5,000 knights, crossed the English Channel in September, and
defeated Harold’s army at Hastings in October. The Norman troops
then moved on to London and William declared himself King of
England on Christmas Day 1066.
The cultural and linguistic consequences of the eleventh-century
occupation of Britain by speakers of Norman French were far-reaching.

The links to Scandinavia became much less important while contacts
with Western Europe were intensified. : The political reins of the
country passed into the hands of Norman landlords, upper clergy
and administrative officers – a foreign minority who spoke little or no
English and who maintained strong cultural and linguistic ties to the
Continent, especially for the first century and a half after the Conquest.
During that same time, and outside the monastic circles, people who
had access to writing and the leisure to engage in any literary activ-
ity had little or no English. The social layering of feudal England was
reflected in linguistic layering too: at the bottom of the scale serfs and
peasants would be monolingual English speakers, the most tenacious
custodians of the mother tongue. At the top end of the social scale, the
sovereign and his barons could barely understand English. Not sur-
prisingly, English was no longer the primary language of record, nor
was it the language of upper-class literary entertainment and learning,
law or administration; these activities were conducted in Latin and
Anglo-Norman (AN).
AN is the variety of Old French (OFr) spoken in England from the time
of the Conquest to approximately the middle of the fifteenth century. :
The Normans themselves – the word comes from Germanic Northman –
were an ethnic mixture of Scandinavian and Frankish people whose
variety of French was influenced by ON. The ties between the insular
and the continental Normans started to weaken after King Philip of
France seized the Norman estates of the Anglo-Norman barons in 1204.
Being thus cut off, during the thirteenth century and after, more and
more members of the upper classes adapted by learning English.
For a linguist digging into the historical records, the ME period
matches, and even exceeds, OE in diversity and complexity. This is
partly due to the finite nature of the surviving OE material compared
with the much richer ME documentation. Another challenge for the
study of ME comes from the fact that it was, at least early after the
Conquest, the language of unrecorded everyday communication. We
are also in the dark about ‘colloquial’ OE, of course, but at least from
the time of King Alfred (871–99) until the Conquest we have evidence
of Anglo-Saxon vernacular learning and transmission. Sermons were
written and popular Latin texts were translated into Old English to
be shared by the laity. The events of 1066 put an end to that practice.
Ultimately, the main source of increased complexity is the introduction
of multi-tiered and lasting second- and third-language influence on the
already diverse base of late OE.
Statistically, the number of monoglot French speakers in England
was never very high; it is estimated that it ranged, roughly, between

2 per cent and 10 per cent of the population, and even that group had
started shifting to some variety of English by the late twelfth century.7
The social boundaries within which French was used were narrow. As
Ostler puts it:
The spread of Norman French would have been limited by the very
rigidity in the social hierarchy over which the Normans presided. Within
the feudal system, the status of every English man and woman was
largely determined by birth, with the church providing the only paths for
advancement through merit, and that was severely limited through the
constraints of celibacy. As a result the French-speaking nobility remained
almost a closed society . . . and there was little or no scope for people to
better their prospects through aping their masters. In feudal England,
people knew their place . . . (Ostler 2005: 461)
This angle on the interaction between English and French speakers
excludes, however, the well-attested bilingualism and even trilingual-
ism within the ‘closed society’ of the aristocracy and the clerics. Spoken
transmission of AN, though demographically limited, continued stead-
ily, and indeed the late thirteenth and the early fourteenth century
are associated with a temporary expansion of the functions of insular
French (Ingham 2012). While Anglo-Norman did not survive as a
true vernacular language much beyond the middle of the fourteenth
century, the written forms of Anglo-Norman lingered on in legal and
parliamentary use until the seventeenth century (Rothwell 2001).
For upwards of two centuries the country’s important affairs were
conducted in Anglo-Norman or Latin, yet from the middle of the
twelfth century onwards, English had to be shared more and more
widely even among the nobility. Manuals for instruction in French
began to appear in the thirteenth century at the same time as English
was gradually regaining ground as a literary language. The Hundred
Years’ War (1337–1453) is a milestone in the development of national
consciousness and unity in England, and with that, the re-emergence
of English as a new national language. Literary activities in English
were resumed, and by the middle of the fourteenth century English
was again, and firmly, not just the leading language in the fields and the
marketplace, which it had never ceased to be, but also the medium of
prose and poetry of interest and lasting value.
The re-establishment of English did not signal the elimination of
French as an important second language in medieval England; it held
7
The demographic details are insightfully summarised in Lass (1987: 56–7).
its position as a learnèd tongue, like Latin. Active political and cul-
tural contacts with the Continent in the fourteenth century meant that
French continued to be a language of prestige and culture, only by
this time there were two types of French in England: the home-grown
Anglo-Norman vernacular, and Old French as it had evolved on the
Continent. Both were developments of vernacular forms of Vulgar
Latin. The two descendant languages were not identical; their parallel
use in England paved the way for the creation of doublet forms of the
same etymological form: from AN we get Karl, cattle, warranty, warden,
matching Charles, chattel, guarantee, guardian from OFr (see 4.3).
The most widely recognised effect of the Norman Conquest on
English was the rapid absorption of words from all spheres of interaction
characteristic of the higher social status of the French-speaking nobil-
ity. The new rulers brought with them legal, administrative, military,
political and cultural terms which often paralleled existing English
words. In some instances the French loans would be indistinguishable
from their ultimately Latin prototypes, which makes it difficult to state
with precision what the source of the borrowing is. Convenient cover-
terms for all of these are ‘Latinate’ or ‘Romance’ loanwords. The most
common practice of recording the etymological source in dictionaries is
to assign ‘origin’ to a word according to the immediate source of borrow-
ing. This means that at times derivatives of the same root will appear
under different etymological labels. In the 1150–1450 time-bracket of
the Chronological English Dictionary (Finkenstaedt et al. 1970) we find:
excuse, n., v. (OFr), but excusable (Lat.); exemplar (OFr), but exemplary
(Lat.); lineage (OFr), but lineation (Lat.); violence (AN), but violent (OFr);
visage (AN), but vision (OFr).
What matters to us in the context of this book is that the word-stock
was undergoing rapid innovation and growth; words with unfamiliar
phonological and prosodic shape were replacing or duplicating words
with prototypical Germanic structure. Some examples of such replace-
ments are shown in (2); the dates in parentheses in the second column are
the earliest OED records for the new entries, cited in their modern form.
(2) Replacements of Old English words with loanwords from AN/

OFr:
OE AN/OFr loan
(ge)mot ‘court, council’, PDE moot assembly (1325)
deman ‘to judge’, PDE deem judge, v. (1290)
deor ‘animal’, PDE deer animal (1398)
treowþ ‘truth’ verity (1375)
wundor ‘wonder’ miracle (1230)
As we will see in Chapters 5 and 9, the loanwords from AN/OFr often

exhibited features ‘alien’ to the structure of OE: seemingly underived
and uninflected words of more than two syllables (assembly, animal,
miracle), word-initial /d-/ (judge) or word-initial /v-/ (verity). Such
features had to be learned by monolingual speakers of English and
eventually became integrated into the basic language structure.
Another window into loanword adaptation is provided by their mor-
phological behaviour. More specifically, if a foreign lexical element,
either a stem or an affix, combines with native stems and affixes, we can
posit a certain degree of ‘nativisation’ or ‘naturalisation’ of the borrowed
element. Such hybrid word-formation is an independent parameter on
which one can determine the ‘nativeness’ of a new loanword. Some
examples of hybrid compounds with their first attestations cited in the
OED are shown in (3).
(3) Hybrid compounds:
English + AN/French AN/French + English
town-clerk (1386) safe-keeping (1432)
breastplate (1386) gravel-stone (1440)
freemason (1376) riverside (1366)
bedchamber (1362) dinner-time (1371)
Newly derived words could combine borrowed roots with native affixes
and native roots with affixes taken from other languages, as shown in (4).
(4) Combinations of roots and affixes:
English + AN/French AN/French + English
husbandry (1290) joyful (1290)
loveable (1340) mannerly (1375)
talkative (1432) colourless (1380)
wizard (1440) foretaste (1435)
When we turn to the history of stress-placement in Chapter 9, we will
see that one of the channels through which new patterns of accentua-
tion enter English is the adoption of new affixes. While suffixation in OE
never changed the stress of the stem, suffixation in PDE can shift the
main stress away from the root: hónour-honourée, állergy-allérgic, ségment-
segmentátion, sólid-solídify).
Another way in which the presence of a French-speaking social
elite in England after the Conquest was felt was the shift in modes of
versification. As we will see in Chapter 10, OE poetic composition
lies firmly within the Germanic alliterative tradition; the verse line
was held together by alliteration, which is consistently associated with
the first sound of the stressed syllable of a word. In Middle English
syllable-counting and end-rhyme, properties characteristic of contem-

porary Continental versification, became increasingly popular, culmi-
nating in the invention of the English iambic pentameter by Chaucer (c.
1343–1400). This culturally based innovation cannot be separated from
the prosodic changes in progress towards the end of the ME period and
thus offers yet another, albeit indirect, way of calibrating the linguistic
effect of the Conquest.
As already noted, the shift of power after 1066 placed English in the
position of a demographically dominant vernacular with a relatively
low social status. This results in the famous fragmentisation of Middle
English. The centralised written ‘standard’ that had emerged in late
OE was abandoned. English obviously continued as a robust vernacular
because valuable literary compositions in English did begin to re-
emerge around the end of the twelfth century, but the literary language
existed and developed without codification. We do not know how local
scriptoria taught their scribes to write English, but they must have been
allowed the freedom to ‘translate’ the texts they copied into their own
regional pronunciation and grammar. The surviving ME literary docu-
ments do not represent a monolithic language with systematic spell-
ing and grammatical rules. Up to at least the middle of the fourteenth
century, it is fair to describe English as a conglomeration of dialects,
each representing distinct, yet often also mixed, accents, grammar and
vocabulary.
Starting roughly in the second half of the fourteenth century, the re-
emergence of English as the leading language in all spheres of life led
to the formation of more systematic scribal practices, especially in the
crafting of official documents in which dialectal differences were lev-
elled out in favour of a unified written standard. The process of stand-
ardising English thus has its roots and primary motivation in the events
that usher in the re-establishment of English as the language of written
communication. The single most important event which accelerated the
formation and spread of a written standard and in many ways shaped
the future of the English language was the introduction of the print-
ing press in England, in 1476, by William Caxton. Although we know
that many linguistic changes associated primarily with Early Modern
English started in the fourteenth century, and that some typically ME
grammatical features persisted beyond 1500, we can take 1476 as the
emblematic ‘end’ of Middle English; the linguistically arbitrary histori-
cal signpost of 1066 that we chose to mark the beginning of the period is
thus matched by another datable historical event at its end.
There are rich digital resources for the study of ME. The Electronic
Middle English Dictionary, in the public domain since 2007, crowns
seventy-five years of lexicographical work. It is an indispensable

research tool that gives searchable access to more than 15,000 pages of
printed material. Two other extremely valuable records of the linguistic
landscape of ME are the Linguistic Atlas of Late Middle English (LALME),
and the Linguistic Atlas of Early Middle English (LAEME). Six of the top
twenty-five most frequently cited authors and works in the OED are
Middle English, including Chaucer in sixth and Caxton in thirteenth
place.
1.4 Early Modern English (1476–1776)

The official use of Latin and French decreased rapidly during the fif-
teenth century. The last French entry from the statutes of Parliament
dates to 1489, during the reign of the first Tudor king, Henry VII
(1485–1509). Clearly, English was firmly and permanently established
as the language of art and literature, the language of trade, administra-
tion and the court. In religious and academic discourse, however, Latin
and Classical Greek continued to be central to the education of scholars
and priests. Intense scholarly engagement with the humanistic herit-
age of the Classics is one of the most salient characteristics of the two
centuries following the introduction of printing.
Academic commitment to Greek and Latin and the development of
new areas of scientific inquiry led to an unprecedented rate of adoption
of Classical vocabulary in EModE. According to one estimate based
on counting entries in the OED, 4,500 new words were recorded in
English during each decade between 1500 and 1700.8 About half of the
new words were derivatives of pre-existing native words, but the rest
were loanwords: over 20,000 words borrowed from the Classical lan-
guages between 1500 and 1700 have survived to this day. The spread
of Classical learning and the density of new items make it much more
likely that words borrowed from Latin and from Greek via Latin during
that period would resist assimilation. For the prosodic history of English
a rate of borrowing from Latin ranging from 35–40 per cent of the new
lexis (Nevalainen 2006: 53) means that such words would tend to pre-
serve the stress of the original, as in alumnus, antenna, cerebellum, curricu-
lum, lacuna, radius. Affixes were also borrowed from Latin, for example
the suffixes -ence, -ancy, -ency, -y and the prefixes ante-, post-, sub-, super-.
In Chapter 9 we will see how the influx of Latin vocabulary in EModE
led to the establishment of a new, parallel system of stress placement.
8
The figures are a recalculation of counts presented in Barber (1997: 220), which cover
2 per cent of the entries in the first edition of the OED.
Among the social and demographic factors influencing the develop-

ment of the language during the EModE period were rapid population
growth and the shift from rural to urban living resulting in the levelling
of dialect differences. : An early seventeenth-century commentary
on English by Alexander Gil, written in Latin, described a variety of
English pronunciation, Communis dialectus, a ‘General’ dialect, which is
socially and not regionally identified and which exists side by side with
the regional Northern, Western, Southern and Eastern varieties. Other
contemporary writers were also recommending the pronunciation of
the aristocracy and the educated upper class as a model to be imitated.
The city of London was the geographic focus of the patterns of pronun-
ciation which became the core of Standard Southern British English
(SSBE) in the course of the eighteenth century.
Early Modern English is marked also by the rapid expansion of
English across the British Isles and overseas. Although English was
spoken in parts of Wales before, and Wales had been annexed to the
English crown since 1284, a new measure in 1536 brought the two coun-
tries under the same legal and administrative system; the common law
of England, rather than Welsh law, was to be used in the Welsh courts,
which guaranteed the spread and prestige of English. After the middle
of the sixteenth century, economic changes resulted in rapid growth of
the use of English in Ireland too, affecting areas beyond Dublin and the
original English Pale. The English of medieval Scotland, Scots English,
was also under sociopolitical pressure: the Union of the Crowns in 1603,
whereby King James VI of Scotland became the first Stuart King of
England as James I, and declared himself ‘king of Great Britain’, brought
about wider use of (southern) English in Scotland and accelerated the
process of ‘Anglicisation’ of Scots English. The blending of Welsh, Irish
and Scots English with English English has not been completed; it is
most advanced in morphology and syntax, while the phonology of the
four varieties is still distinct.
The 1607 Jamestown Colony in Virginia and the Pilgrims’ landing
at Plymouth Rock in 1620 marked the beginning of a new phase in
the expansion of English. In the course of the seventeenth century,
the New World became the second geographic area where English
was the dominant language. ‘Colonial’ English was far from homogene-
ous. The demographic strands of the settlers do not correspond exactly
to the dialect differences in later American English, but they provide
one basis for identifying regional varieties in North America and tracing
them back to the speech of the first colonists. :
For the first century and a half after the colonisation, the linguistic
ties to Britain were very strong, especially along the eastern seaboard.
At the same time, English speakers were exposed to other languages, not
just a variety of American Indian languages but also German, Dutch,
French and Spanish. In spite of many bi- or multilingual groups and
areas, colonial American English is considered more homogeneous than
contemporary British English.
1.5 English after 1776

By 1776, when the colonists declared their independence from England,
the political, cultural and linguistic conditions for identifying American
English as a separate variety were at hand. In 1783 the lexicographer
whose name would subsequently become synonymous with specifically
American dictionary-making, Noah Webster, published his American
Spelling Book; he started work on his American Dictionary of the English
Language in 1807 and completed the first edition in 1828. He was an
influential advocate of American spelling and usage and introduced
many original American entries in his dictionary, which initially sold
more copies in England than it did in the US, contributing much to the
perceived legitimisation of American English. The idea of a separate
American language was widely accepted in the nineteenth century
although British English continued to be admired and imitated well
into the twentieth century. The noun Americanism was first recorded
in 1871 (OED). Predictably, American English itself became more and
more diversified. Political contacts between Canada and the US made
American English a rival of British English and, of course, French,
in Canada, bringing about the formation of a new variety of North
American English, Canadian English. In addition to the establishment
of identifiable regional varieties of North American English, political
and demographic changes in the US have given rise to new ethnic vari-
eties of American English. The best-studied among them are African
American Vernacular English and Latino/Chicano English.
The late eighteenth century also marked the expansion of English
to other continents. Early immigration to Australia in the 1780s and
the settlement of English speakers in New Zealand from about 1840
led to the creation of two other national varieties: Australian and New
Zealand English. These varieties have close historical links to southern
British English, but they also developed features which distinguish
them both from the Cockney dialect of many of the original settlers and
from SSBE.
It is impossible to estimate the number of people who spoke Proto-
Germanic (see 3.2), but we know how the various branches of Germanic
have fared more recently. English has emerged as one of the most
widely used native languages: only Mandarin Chinese, Hindi and

Spanish have more native speakers than English. Stated very broadly,
post-colonial English outside Britain and North America is character-
ised by the diversity of new features introduced through the contact
between colonisers and colonised. The British Raj (1858–1947) fostered
the development of Indian English; today South Asia is one of the three
largest English-using regions in the world. The establishment of the
Cape Colony in 1806 in what is now South Africa and the arrival of
American-Liberians in West Africa in 1822 resulted in new English-
language communities in Africa. Today, in addition to Liberia, English
is the official language of administration in many otherwise multilin-
gual African countries: Botswana, Ghana, Kenya, Lesotho, Malawi,
Nigeria, Sierra Leone, Swaziland, The Gambia, Uganda, Zambia and
Zimbabwe. Other geographic regions have adopted English too, gen-
erating new branches of English such as Caribbean English. English
is now both a world language geographically and a global language
functionally: it is the language of the airlines, an official language of
the United Nations, the dominant language in computing, science and
international communication.
This book focuses on the historical underpinnings of the Present-Day
English phonological system, but as we have seen from the preceding
survey, there is no single, or static, entity that can be labelled ‘Present-
Day English’. Covering the phonological and prosodic details in the
huge variety of present-day Englishes is beyond this book’s remit;
instead, the focus will be on some reference standard accents which
will form the starting point for our journey back in time. These pro-
totypical accents are widely recognised and can be easily related to
varieties spoken outside the narrow confines of the ‘standards’. The two
varieties from which we will project backwards are Southern Standard
British English (SSBE), also known as Received Pronunciation (RP),
and General American (GA), with cross-references to other varieties
whenever appropriate.
If the goal of this book is to discover how the past shaped the selected
present-day standards, we need some understanding of the genesis of
the notion ‘Standard English’, a topic which has received considerable
scholarly attention. First, we have to keep the written and the spoken
standards separate. The orthographic ‘standard’ is the conformity with a
unified written norm that keeps words looking the same on the page or
on the screen. Apart from some well-known national differences, such
as American English versus British English spelling of -or/-our or -er/-re
words, the spelling norms are shared by all varieties of English. It is the
variation in the spoken standards that is our main target of interest, and
since we are limiting our scope to phonology, we can talk of ‘standard

accents’. Here we need to remind ourselves that ‘standard’ is a socially
constructed notion: any accent of English could have become stand-
ard under the right cultural and socio-economic circumstances. It is a
truism in linguistics that no language is inherently better or worse than
any other language; there is no intrinsic value in standards, although
standards do carry considerable social and cultural prestige, or baggage,
depending on the viewpoint.
Historically, none of the standard varieties we cover can be recon-
structed as a linear descendent of one particular historical dialect. The
OE standard represented in textbooks is nothing more than a ‘high level
of agreement in a language community as to what does and what does
not constitute “the language” at a given time’ – this is what sociolin-
guists understand by the term ‘focussing’ (Trudgill 1986: 86). Applied
to Old English, the standard is the relatively stable set of forms found
in the works of Ælfric (c. 955–c. 1025) and Wulfstan (d. 1023) (Hogg
2006b: 401–3). Yet the ancient origins of Standard English are unknown
to us because the particular variety of late West Saxon represented by
Ælfric’s and Wulfstan’s writing is not the basis of any one of the emerg-
ing late Middle English ‘standards’. Those too are most aptly charac-
terised as ‘focussed’ varieties (Smith 1996: 65–77), preserved only in
writing. By the middle of the seventeenth century ‘a high degree of uni-
formity in spelling’ was in evidence in contemporary printed materials,
though the codification did not extend to private correspondence until
much later (Nevalainen and Tieken-Boon van Ostade 2006: 289–91).
Pronunciation standards develop much more slowly than the written
standard; moreover, a standard accent often coexists with various -lects
in the repertoire of a single individual. Pronunciation norms emerge
under the heavy scrutiny and direction of prominent and respected
intellectuals and educational institutions. In England, the educated
accent of London and the court became codified during the nineteenth
century when the term Received Pronunciation (RP) was applied to that
accent. This accent, although not native to the large majority of English
speakers, is recognised widely and its features are constantly updated
to reflect ongoing change. In the US, the corresponding ‘standard’ is
commonly referred to as ‘General American’. GA is much less region-
ally or socially focused and it can be defined negatively: it is identified
as the accent that is neither eastern nor southern. It covers a broad
geographical area comprising parts of the midlands and the west.
This chapter started with observations on the enormous diversity
of Englishes today, making it clear that a single starting point for our
telescopic journey back in time is an artificial construct. Yet we have
to start from somewhere, and the choice of two accents, GA and SSBE,
is motivated by their widespread use, easy identification by speakers of
other varieties and by the author’s own limitations – these are the two
‘standard’ varieties that I have been taught and that I have had most
exposure to as a non-native speaker of English.
1.6 The evidence for early pronunciation

The inconsistencies of letter-sound correspondences in PDE – pairs
like know-no, scene-seen, reel-real, bass-bass, rough-though – are frequently
remarked on. If PDE spelling is so unreliable, how do we know what
sounds correspond to the letters used in the historical documents?
Unlike morphology, syntax and word-formation, where the properties
of the forms are more tangibly attested in the written texts, the pho-
nological properties of a written form are much less physically ‘real’.
Nevertheless, rigorous methodologies for phonological reconstruction
have been developed and tested repeatedly in the last two centuries.
Although scholars will disagree about the precise realisation of specific
form, or forms, there is consensus about the validity of many of the
histories that have become canonical in describing the evolution of
English.
The obvious first step in reconstructing sound systems from written
documents is to look at what the scribes recorded and how consist-
ent they were. The main body of surviving Anglo-Saxon manuscripts
is based on the Roman alphabet. : Along with their conversion to
Christianity, starting in 432, the neighbouring Irish had gradually
adopted and adapted the Roman script, creating their own version of the
letters, known as ‘insular’ script. That ‘insular’ script-form was imported
into England by missionaries after the adoption of Christianity by the
Anglo-Saxons in 597. Monastic culture, not just in Britain, but through-
out Western Europe, was dominated by writings in the Roman alphabet.
Not surprisingly, the OE orthographic system developed on the basis of
Latin writing during the seventh and the eighth century. The Latin con-
nection is helpful in figuring out how the scribes, who were also literate
in Latin, matched sound to letter. Not all OE letters are found in Latin:
two OE letters, thorn <þ> and wynn , were borrowed from the runic
alphabet; another native addition was the letter eth <ð>, also spelled
edh.9 The letters <q, k, x, z> were used very rarely in OE; they became
part of the regular inventory of consonantal letters after the adoption of
9
Angled brackets are used to mark letters, as distinct from sounds, which will be
enclosed in slashes and square brackets; see Chapter 2.
the Carolingian script in ME. Throughout the history of the language

the actual shape of the letters has changed many times, but whether
we find <d, d, d, D, D, D>, or <d>, we can be fairly confident that that
letter matches the first and the last sound in David.10 The interpretation
of scribal evidence is rarely that straightforward; nevertheless, match-
ing the familiar letter values in the Roman alphabet to the values of the
letters used by the scribes is a primary source of information about the
older pronunciation.
Consistency or inconsistency of spelling is another fairly good
measure of the stability of a particular pronunciation. This presupposes
that prior to the introduction of printing and the codification of spelling
there was a high degree of letter-to-sound matching – a safe general
assumption, though the details, especially the ME details, can falsify
such expectations. The variability of spelling and the abandonment
of some forms can also be harnessed for phonological reconstruction.
Thus the dropping of orthographic <h-> in words like OE <hwil>
‘while’, OE <hwit> ‘white’ in some ME texts, where they appear as
<wil>, <wit>, as well as the addition of <h-> to words such as OE
<wit(t)> ‘wit’, <wiht> ‘wight’, ME spellings <whyt(t)>, <whihte,
whith, whyht>, suggests that the scribes could no longer perceive the
difference between a ‘pure’ voiced /w/ and its voiceless counterpart
// (<hw>). If they heard the same initial sound in white and wit, they
would be likely to confuse its representation with and without an <h->.
Comparing forms in genetically related languages can be very helpful
too. If we find OE <hungor, hunger> ‘hunger’, and the form in Old
Saxon is <hungor, hungar>, Old Frisian <hunger, honger>, Old Norse
<hungr>, Old High German <hungar>, German <Hunger>, Middle
Dutch <hongher>, Dutch <honger>, Swedish and Danish <hunger>,
we can be fairly confident that the OE word also began with /h-/, that
the vowel following it was not a long vowel or a low front vowel, that
the third sound is /n/ and that the second syllable starts with /g/.
Language universals are another important basis for hypothesis-
ing about earlier pronunciation. With appropriate caution, commonly
observed cross-linguistic patterns of inclusion or exclusion of particular
sounds and sound combinations can be applied to historical reconstruc-
tion. Some statistically testable universal properties of language, such
as ‘languages have fewer vowels than consonants’, are too general to be
of practical interest. More viable in our plotting of phonological change
10
The general term for a single alphabet item, irrespective of its shape, is grapheme,
while the various shapes of a letter are referred to as allographs. The terms corre-
spond to the widely used terms phoneme and allophone (see 2.1).
are the physically testable principles of production and perception, both

related ultimately to the way language is learned. This is particularly
helpful when we try to reconstruct the direction and the intermediate
stages of sound change: we can posit a historical shift of [a] to [æ]; these
sounds are phonetically close to each other, but we cannot jump from
[a] to [u] without multiple steps in-between. Using the voiced /w/ in
wine and the voiceless // in whine maintains the perceptual difference
between two distinct lexical items, but it also requires some extra effort.
If we merge the pronunciation of the two words, we create homophones,
a perceptual complication arising concurrently with the elimination of
the extra effort. The tension between these forces is always present in
language; we will see how these general linguistic principles can be
applied to the accounts of phonological change in English.
Verse structure is another source of data for phonological reconstruc-
tion. Old English verse is based on alliteration, or identity of the initial
sound or sounds of the first stressed syllable in the word. The matching
of OE <pharaones> ‘of the pharaoh’ with <folce> ‘to the folk’ is evi-
dence that <ph-> in the borrowed word is a spelling for /f/. Similarly,
the alliterative pairing of <Firgilies> ‘Virgil’s’ and <freond> ‘friend’
suggests that the poet or scribe replaced the Latin initial consonant
in Virgil with /f/. Alliterative versification flourished in late ME too;
one of the ways we can ascertain that the digraphs <wr-, kn-, gn->
represented real consonant clusters until about 1500 is by their use in
fourteenth-century alliterative verse where the first consonants alliter-
ate, for example writ: wonder, gnaw: God, knight: kiss. Rhymes also provide
valuable tests for phonological change. A ME rhyme such as honour: flour
‘flower’ suggests that the word honour could still be pronounced with
stress on the second syllable (see 9.6). Drawing further on the creative
use of language we can also go to puns to test the similarity between
sounds. Consider the word-play in Romeo and Juliet (I, iv):11
Mercutio: ‘Nay, gentle Romeo, we must have you dance.’
Romeo: ‘Not I, believe me. You have dancing shoes
With nimble soles; I have a soul of lead . . .
Such use is good evidence that Old French sole ‘sole’ and Old English
sawol ‘soul’, which had different vowels originally, had already acquired
the same value for Shakespeare’s audience.
The development of a fixed system of spelling after the introduction
of printing in the late fifteenth century did not proceed in step with the
11
Compare the pun on reason-raisin in ‘If reasons were as plentiful as blackberries, I
would give no man a reason upon compulsion, I’ (Henry IV, Part One, II, iv).
contemporary phonological changes. The discrepancies between spell-

ing and pronunciation became a major scholarly concern. The people
who, from the sixteenth century onwards, were writing manuals for
native speakers and for foreigners and who were proposing spelling
reforms, are known as orthoepists. Although their testimony may be
biased and inconclusive at times, since they were influenced by Latin,
the orthoepists’ works form the backbone of our reconstructions of pho-
nological change in EModE. Their detailed comparisons with Latin,
French and Italian and the attempts at phonetic description, albeit
imperfect and open to different interpretations, provide a better record
of contemporary variation than any of the previous sources mentioned
above. The first extensive dictionaries of pronunciation appeared in the
second half of the eighteenth century, culminating with John Walker’s
1791 Critical Pronouncing Dictionary.12 The scholarly value of Walker’s
contribution to the history of eighteenth-century pronunciation cannot
be overstated; it is a worthy predecessor of the great pronouncing dic-
tionaries of the twentieth century, Daniel Jones’s in Britain and Kenyon
and Knott’s in the US. Beginning with the invention of the phonograph
by Thomas Edison in 1877, the technology of speech recording has
become more and more sophisticated; the current and future genera-
tions of scholars have incomparably superior instruments and methods
of recording, quantifying, and analysing phonological change.
: Suggested further reading on Companion website.
12
A critical pronouncing dictionary and expositor of the English language . . .: to which are prefixed,
principles of English pronunciation . . . with observations etymological, critical, and grammatical
. . . with directions to foreigners, for acquiring a knowledge of the use of this dictionary, with edi-
tions from 1st (1791) to 28th (1826).
2 The sounds of English
2.1 The consonants of PDE

English is not a ‘new’ language for the reader of this book, but Old and
Middle English are different enough to be considered genuinely ‘new’.
Here is the opening line of Caedmon’s Hymn, the earliest piece of OE
poetry whose composer is known:
Nu sculon herigean heofonrices weard
‘now (we) shall praise heaven-kingdom’s guardian’
meotodes meahte and his modgeþanc,
‘the creator’s might and his thought’
Even with word glosses in place, the OE text is a challenge to read
aloud. : If now was spelled <nu> in OE, how was it pronounced:
like new, like noo, like now or some other way? If shall was <sculon>,
was it pronounced like skull-on, like shoe-l-on or some other way? The
word <meaht> means ‘might’, but no word in our modern dictionaries
contains the sequence of letters <-eaht>. The pronunciation of earlier
stages of English is the topic of much of this book, but in order to estab-
lish a common denominator between the past and the present, we need
to start with the modern language.
A note on terminology and representation is in order. Throughout
the book the terms ‘sound’ or ‘segment’ are used to refer more gener-
ally to entities of pronunciation, without necessarily assigning them
phonemic status, that is, without reference to their ability to distinguish
meaning. A phoneme is a mental image of all the various realisations of
one and the same sound. Thus [v] has always been a sound in English,
but in Present-Day English (PDE) it is also a phoneme /v/. Its realisa-
tion is not determined by word-structure or by adjacent segments; it can
signal meaning contrasts: fan-van, safe-save, leafy-Levy. The glottal stop
[ʔ], articulated with a constriction of the glottis, occurs between vowels
or word-finally in some accents of English, as in letter, bitter, sap, cat, sack.
24
THE SOUNDS OF ENGLISH 25
Unlike /v/, the glottal stop is not a phoneme in English; it is an allo-

phonic realisation of the phonemes /t/, /p/, /k/ in our examples. In
GA the alveolar stops /d, t/ are realised allophonically as voiced alveo-
lar taps [-ɾ-] – ladder [læɾ], capitalise [kæpiɾəlaiz], at ease [əɾi
z] –
but the tap is not part of the inventory of contrastive English consonants
(see 5.5.2). Similarly, the various types of [r], or the types of [l] used in
different regions of the English-speaking world, are allophonic. :
This book follows the accepted practice of enclosing phonemes in
slashes / /, while square brackets [ ] are used for specific realisations.
Previous exposure to the International Phonetic Alphabet (IPA) is
assumed. :
Some of the familiar observations about phonological inventories and
classifications are that:
1. languages have more consonants than vowels;

2. the sounds of any language are either vowels or consonants;
3. consonants are more stable historically than vowels; and
4. the properties of consonants are more easily described and perceived.
All four statements hold for English, although only (1) is truly, and trivi-
ally, beyond challenge. As for (2), the difference between consonants,
produced with some stricture in the vocal tract and not pronounceable as a
syllable, and vowels, produced with little or no stricture and forming a syl-
lable, is mostly clear-cut, yet there are special cases such as the <r> in a syl-
lable such as nerd where the consonant [ɹ] can be syllabic; another special
case is presented by the sounds known as ‘glides’ or ‘semi-vowels’, such as
/w/ and /j/.1 The statement in (3) is also an overgeneralisation; the next
three chapters will convince you that consonant (sub-)systems of English
offer a rich gamut of variation and change, so overall ‘stability’ of the con-
sonants can hardly be claimed for the entire inventory. The impression
that consonants are easily described is also misguided, although one has
to admit that the consonantal features are more accessible to the speaker
for self-examination than the corresponding vowel features.
The production of consonants and vowels is commonly described
with reference to the vocal tract, starting from the lungs, through the
larynx, the pharynx and the upper part of the tract: the oral cavity and
the nasal cavity. Most of the action happens in the oral cavity – the chart
in Figure 2.1 shows the more detailed anatomy of the upper vocal tract.
1
‘For many speakers of American English, the approximant ɹ at the beginning of the
word “red” bears the same relationship to the vowel in “bird” as the approximant j in
“yes” does to the vowel i in “heed”’ (Ladefoged and Maddieson 1996: 323).
nasal cavities
alveolar
ridge
upper hard palate

lip
velum
(soft palate)
blade front
teeth tip
uvula
back
tongue
body
lower pharynx
lip tongue
root
epiglottis
hyoid bone
thyroid
cartilage
larynx
cricoid
cartilage
Figure 2.1 Cross-section of the vocal tract (from Ogden (2009) Introduction
to English Phonetics, Edinburgh University Press, p. 10)
The characterisation of consonants is traditionally done with reference

to three physical criteria: voicing, place and manner of articulation.
2.1.1 Voicing
The sounds that one can sing or hum are voiced. Voicing is dependent
on the movement of the glottis; it is caused by vibration of the vocal
cords and narrowing of the glottis; the glottis itself is the space between
the vocal cords. Since the contrast depends on action in the larynx and
the state of the glottis, questions of voicing can also be discussed with
reference to laryngeal or glottal specifications. Another related prop-

erty of note is that voiced consonants require less muscular effort; they
are also described as lenis consonants, as opposed to fortis consonants,
which are articulated with stronger effort. The lenis–fortis distinction
correlates with voicing, but it is, properly speaking, a distinction based
on the manner of articulation (see 2.1.3).
The vibration of the muscular folds triggers voicing. When the
vocal cords are spread and not vibrating, as in breathing, the result is
a voiceless sound. In English only one subset of the consonants can be
voiceless – /p, t, k, f, θ, s, ʃ, tʃ, h/ – so that the voiced sounds outnum-
ber the voiceless sounds. All English nasals and approximants (/m, n,
ŋ, l, r, w, j/, see below) are voiced. The exact phonetic manifestation
of consonant voicing in English varies considerably. The ‘voiced’ /b,
d, g/ can be realised phonetically very close to [p, t, k] word-initially,
especially after a voiceless consonant as in this boy, this day, this guy.
The voiceless /p, t, k/ appear as aspirated allophones word-initially
and before a vowel in a stressed syllable, [ph, th, kh], as in pin, attain,
cuddle. :
Our knowledge of the degree of glottal opening historically is
inferential. It is based on feature compatibility (no co-occurrence of
/h/ and aspirated stops, no *hp-, *ht-, and so on), and on the fact that
aspirated stops and /h/ occur in the same positions in PDE, as in pat-
hat, betoken-behave. In this book our historical account will follow the
tradition and refer to the oppositions /b/-/p/, /z/-/s/, /g/-/k/, and
so on as ‘voiced’ vs ‘voiceless’, recognising that this shortcut is not
always sufficient to account for all the details that one encounters in
the records.
2.1.2 Place of articulation

The organs of speech in the vocal tract shown in Figure 2.1 are known
as articulators. This cover term conflates two separate parameters:
the active articulators located at the lower area of the vocal tract, for
example lower lip and tongue; and the passive articulators, that is, the
target area towards which the active articulators move and the area in
which the articulation occurs, for example upper lip, alveolar ridge and
hard or soft palate. Self-monitoring the place of articulation for most
consonants is pretty easy if one concentrates on the movement and the
place of contact of the lips and the tongue. The places of articulation
most relevant for PDE are indicated by arrows on Figure 2.1. Here they
are described starting from the front and moving to the back of the oral
cavity:
• BILABIAL: articulated with the lower lip touching the upper lip.
The bilabial consonants in English are the voiceless stop /p/, the
voiced stop /b/, and the nasal /m/. :
• LABIOVELAR: the lips are rounded closer together, the tongue is
bunched up, and the sound is produced at the velum. The labiovelars
are: the voiced labiovelar approximant /w/, as in west, sway, glowing,
which appears in all standard varieties of English, and, in some varie-
ties, a phonemically contrastive voiceless labiovelar fricative // as
in whine, whale, whistle, which is a reflex of the OE consonant cluster
/hw-/ (see 5.1.3).
• LABIODENTAL: articulated with the lower lip touching the upper
teeth. The English labiodentals are the voiceless fricative /f/ and the
voiced fricative /v/.
• DENTAL: articulated most commonly with the tip of the tongue
touching the lower edge of the upper teeth, though the articulation
can also be interdental. The (inter)dental consonants in English are
the voiceless fricative /θ/ as in thick, bath and its voiced counterpart,
the fricative /ð/ as in this, bathe.
• ALVEOLAR: articulated with the tongue contacting or approach-
ing the bony alveolar ridge behind the upper teeth. English has two
alveolar stops, /t/ and /d/, two alveolar fricatives, /s/ and /z/; the
nasal /n/, the lateral approximant /l/, and the central approximant
/r/ are also alveolar.2:
• PALATO-ALVEOLAR: articulated with the middle of the tongue
contacting or approaching the hard palate. The palato-alveolars of
English are the voiceless fricative /ʃ/ as in ship, dash, the voiced
fricative // as in measure, the voiceless affricate /tʃ/ as in chair,
match, and its mate – the voiced affricate /d/ as in jazz, huge. These
consonants are also called post-alveolar.
• PALATAL: articulated with the front part of the tongue body
moving towards the hard palate. In English the symbol for the
palatal central approximant in yellow, beyond is either /j/, fol-
lowing the IPA, used in this book, or /y/, which matches the
orthographic representation of the sound (see also 2.2.2 for [j] ~
[i/i]). :
• VELAR: articulated with the back of the tongue contacting or
approaching the soft palate. The English velars are the stops /k/ and
2
This book follows the practice of previous phonological descriptions of English
(Giegerich 1992; McMahon 2002; Kreidler 2004) in using the symbol /r/ for the
phoneme whose most common realisation in GA and SSBE is the central approximant
[ɹ] (see further 5.2).
/g/, the velar nasal /ŋ/, as in sing, and the voiceless fricative /x/ as
in loch, phonemic only in Scottish English.
• GLOTTAL: articulated with the vocal cords moving closer together.
The only glottal consonant in English is the voiceless fricative /h/,
as in hill, behave. Its articulation foreshadows the following vowel. A
voiceless glottal stop, [ʔ], is used in some varieties of PDE as an allo-
phone of the voiceless stops /p, t, k/. [ʔ] is also inserted optionally
before stressed vowel-initial syllables (see further 5.5.1).
Another useful descriptive parameter refers to the active articulators in

producing consonants: the lips, the tip of the tongue or the body of the
tongue. Involvement of the lips produces labials. When the tongue tip
is involved in the articulation, the consonants have the feature coronal.
The coronal consonants of English are the dentals, the alveolars and
the palato-alveolars, while the tongue-body consonants are dorsal. All
velars are also dorsal.
2.1.3 Manner of articulation

The way in which the airstream travels through the vocal tract deter-
mines the manner of articulation, that is, the degree and timing of
blockage as the air is pushed out of the lungs. During the production
of /p, b, t, d, k, g/, there is a brief instance when the air is completely
stopped; these consonants are stops, also called plosives. During the
production of the consonants /f, v, θ, ð s, z, ʃ, /, the airstream encoun-
ters a constriction, but the air flows through it generating some noisy
friction; these are the fricatives, also called spirants. English combines
stops and fricatives in the production of the consonants /tʃ/ as in chess
and /d/ as in joke, in which the air is stopped for a brief period, and
is then released with a certain degree of friction. The consonants that
combine the manner of articulation of stops (plosion) and fricatives
(friction) are called affricates. The only English stops that can be
affricated are /t/ and /d/.
A different type of contrast depends on the path through which the
air comes out of the vocal tract. Typically, consonants are produced
within the oral cavity; they are called oral consonants, not marked
in Figure 2.2. In the production of /m, n, ŋ/, however, the airflow is
blocked from the oral tract and is instead channelled through the nasal
cavity, so they are called nasals.
The production of the approximants /r, l, j, w/ is associated with the
lowest degree of constriction. Within that set, the air can exit through
the central part of the oral cavity, or it can be diverted: the approximant
/l/ is called lateral because during its production the airflow is deflected
around the sides of the tongue. Also within this group, /r, l/ are called
liquids. As noted above, /j, w/ are also identified as glides or semivowels.
The major manners of articulation relevant to the description of
English consonants are:
• STOPS /p, b, t, d, k, g/: the air is completely stopped for a brief
period heard at the beginning of words (town, down), in the middle of
words (upon, cigar, sucker), or word-finally as in rag, rib, sock.
• AFFRICATES /tʃ, d/: the air is stopped, then released with some
friction, as in chimney, jam, pitcher, Cajun, such, ledge. The affricates are
more ‘complex’ and are commonly represented with a tie-bar over
the two IPA symbols: /tʃ/ and /d/.
• FRICATIVES /(), f, v, θ, ð, s, z, ʃ, , h/: the air passes uninter-
rupted, with a degree of friction, as in wharf, fan, van, thin, this,
sip, zip, sure, Zhivago, hat. The palatal fricative [] is a latecomer to
English (see 5.4).
• NASALS /m, n, ŋ/: the air is released through the nose, rather than
the mouth: man, numb, hung.
• APPROXIMANTS /w, r, l, j/: the air flows considerably more
freely than for the other types of consonants: win, ray, low, yes.
Additionally, the degree of closure of the vocal tract can be important: if
the closure is incomplete, the sound is a continuant. This feature strad-
dles the dividing line between obstruents and sonorants: stops and affri-
cates are non-continuant, and so are the nasals [m, n, ŋ]. All fricatives
and the approximants are continuants, and so are all vowels.
Figure 2.2 presents an inventory of the consonantal phonemes of
Labial Labio- Dental Alveolar Palatal Velar Glottal

dental
STOPS p b t d k g
Obstruent
AFFRICATES tʃ d
FRICATIVES () f v θ ð s z ʃ (x) h
NASALS m n ŋ
Sonorant
Approximants Lateral l
Central w r j
Figure 2.2 The consonantal phonemes of PDE

PDE. In cells where there are pairs of phonemes the one to the left
is voiceless and the one to the right is voiced. Parentheses enclose
consonants which are phonemic only in some varieties of English.
The leftmost column in Figure 2.2 shows another important division
between consonants based on the configuration of the vocal tract. The
upper part of the chart accommodates the class of consonants collectively
known as obstruents. They are produced with some degree of obstruction
of the airstream and they can be voiceless or voiced. One special case in
this set is /h-/, which is practically frictionless. In PDE it is a kind of voice-
less precursor to the following vowel; it is produced with ‘spread glottis’,
a feature which distinguishes between aspiration and lack of aspiration.
The consonants below the heavy line in Figure 2.2 are sonorants.
Sonority is a property associated, loosely, with the acoustic loudness of
sounds. All English sonorants are voiced. They are high on a sonority
scale that applies to the entire inventory of English phonemes, not just
the consonants. Vowels have the highest level of sonority, followed by
glides, followed by nasals and approximants. The least sonorous conso-
nants are the voiceless stops /p, t, k/. In English only consonants at the
upper end of the scale of sonority, that is, the sonorants, can function as
syllable peaks, as in kitten, bottom, muscle (see 2.3).
The description of the distinctive sounds of any language requires
reference to the bundle of features that characterise each phoneme.
The phonetic properties of the consonants described in 2.1.1–2.1.3 serve
as the basis for their phonological representation, so properties such as
nasal, labial, voiced, stop, fricative, obstruent, and so on provide the physi-
cal substance of the more abstract set of phonological features. It is by
reference to the particular features that we can define the classes of
phonemes, their interaction and their historical change.3
2.1.4 Short and long consonants

Consonants can appear as singletons, or they can be ‘long’, or geminate.
In PDE the use of ‘long’ consonants in pronunciation signals morpho-
logical complexity; there are no geminates within the boundaries of a
single morpheme. The orthography can be misleading: doubling of con-
sonants in spelling does not automatically signal consonant gemination.
Thus the pairs in (1a) do not differ in the length of their consonants,
3
For a more comprehensive discussion of the set of phonologically relevant features
see Giegerich (1992: chs 1, 2, 5); McMahon (2002: ch. 4). Additional features for the
description of specific historical changes in English will be introduced in later chapters
as needed.
while in (1b) the extra consonantal length is needed to mark off the
morphological boundaries.
(1) Short and long consonants in PDE:
(a) allow alone (b) full-length
furry fury barrier-reef
pass pace class suit
summon lemon beam me (up)
banner saner unnamed
The salience of consonantal length depends on the type of boundary:
geminates are realised as shorter at affix-stem boundaries, as in unnamed,
dissatisfied, rivalless; they may be longer in compounds, as in full-length,
class suit; and predictably, they are quite audibly long within phrases,
as in big garden, grim measure, fifteen nights, lead down, stop pouting. The
sequence of identical consonants in these cases is pronounced with only
one release and one onset, yet the actual duration of these long conso-
nants may be as long as that of a cluster of two separate consonants. The
realisation of geminates in loanwords is a separate issue. Words of clas-
sical origin whose components are not recognisable may be pronounced
as though they are native words: subbie, aggravate, commute, suffer, attrib-
ute, transcend. When the affix is recognised, however, the consonants can
be realised as geminates as in (1b), for example subbrachial, disservice,
connateness.
The status of geminate consonants has changed from Old to Modern
English; we return to it in 4.1.2.
2.2 The vowels of PDE

The movement of the articulators and the places of articulation for
vowels are not so readily open to self-examination. The most impor-
tant active articulator in vowel production is the tongue body and the
actual phonation occurs in the vocal tract. The airstream is allowed free
passage through the mouth.
The quality of a vowel sound changes with the movement of the lips
and the tongue. Minor shifts in the tongue position, difficult for the
speaker to feel, produce quite distinct auditory and acoustic results,
described along the dimensions of height and backness. Height refers
to the distance of the tongue from the roof of the mouth: the smaller the
distance, the higher (or closer) the vowel. Backness refers to the tongue’s
distance from the front teeth: the further away the highest point of the
tongue is from the teeth, the more pronounced the backness feature is.
Some vowels also require reference to the participation of the lips in the
articulation: if the lips are rounded, the resulting vowel is also round; if
not, the vowel is unround. Rounding is a feature with limited function
in PDE – non-back vowels cannot be rounded, only back vowels can be
rounded, that is, a reference to roundness is redundant for the non-back
vowels. There are, however, two back vowels that are unrounded: the
vowel /ɑ/ in COT in GA, which forms a minimal pair with the rounded
vowel /ɔ
/ in CAUGHT in those varieties of English that preserve this
historical contrast, and the vowel // as in STRUT, which contrasts in
roundness with the vowels in THOUGHT, LOT in varieties that have
/ɔ(
)/ in these words.4
The IPA vowel chart includes twenty-eight different vowel types,
and those can be further modified by diacritics, marking additional
properties such as length (macron, breve, colon), nasalisation (tilde),
centralisation (umlaut), and so on. Figure 2.3 shows the full version of
the IPA vowel chart. :
Fitting the IPA chart into a feature-based phonological chart is not a
straightforward transfer. Nevertheless, the dimensions of height, back-
ness and rounding allow us to present the contrasts schematically in
rows and columns, though with the vowels the positioning is quite dif-
ferent from the categorical placement of the consonants in Figure 2.2
VOWELS
Front Central Back
Close i y ɯ u
iy υ
Close-mid e ø ə ɵ o
ə
Open-mid ε œ ɔ
ɐ
Open a œ ɑ ɒ
Where symbols appear in pairs, the one
to the right represents a rounded vowel.
Figure 2.3 IPA Vowel chart, revised to 2005 (from <http://www.langsci.ucl.

ac.uk/ipa/vowels.html>, available under a Creative Commons Attribution-
Sharealike 3.0 Unported License. Copyright © 2005 International Phonetic
Association)
4
The use of the colon diacritic is addressed in 2.2.1.
FRONT CENTRAL BACK
Unrounded Rounded
Upper high i (FLEECE) u (GOOSE)
––––––– –––––– ––––––––
Lower high i (KIT) υ (FOOT)
Upper mid ei (FACE) / (NURSE) oυ (GOAT)
––––––– ––––––– ə (COMMA) –––––––– ––––––––
Lower mid ε (DRESS) (STRUT) ɔ (THOUGHT)
(TRAP)
Upper low
ɑ (LOT,PALM) ɒ (LOT) (SSBE)
Low
(GA)
Figure 2.4 English vowels
in well-defined cells. Only rounding is treated as strictly binary for the

vowels. Realisations with respect to height and backness are scalar and in
order to define their unique contrastive properties one has to work with
different degrees of the same main features. For height we will use the
three levels, high, mid and low, with additional specifications for each of
the three levels, lower or upper. Note that this corresponds to the IPA
labels close, mid and open. For backness the main dimensions are front,
central and back. Again, positioning of the vowels within the cells is not
absolutely fixed; it is a descriptive convenience to highlight contrasts, but
the realisations will vary with individual speakers or communities.
Figure 2.4 presents an overview of the vowels of English.
Figure 2.4 provides only a basic frame of reference. Except for the
mid central unrounded vowel schwa, [ə], it shows an inventory of the
vowels in stressed syllables; that inventory is considerably impoverished
in the absence of stress, where [ə] may contrast with [i], as in Rosa’s [-ə-]
roses [i] or tilde, salad, balanced, cherub, welcome, hydrangea, nostalgia with
[ə] vs spinach, sandwich, ceiling, language with [i]. : In fast speech [ə]
can alternate with the syllabic form of the sonorants [r, l , m , n]: mutter,
kettle, bottom, weapon. Syllabic sonorants can be functionally equivalent
to a vowel only in unstressed syllables. The ‘neutral’ vowel schwa is the
most frequent vowel in the language; its frequency (10.74 per cent) is
matched only by the frequency of /i/ at 8.33 per cent.5
5
See Cruttenden (2008: 156), whose counts were based on colloquial SSBE. Reports
The upper mid central symbols // (GA) or /

/ (SSBE), as in
NURSE, represent the development of the vowel preceding histori-
cal /r/ in such words as BIRD, WORD, HEARD, CURD, SIR. The phonetic
realisation of the NURSE vowel in GA is [], that is, right-hook schwa
for the r-colouring, which is present throughout the vowel; in faster
speech it could be simply a syllabic [r ]. In r-less/non-rhotic varieties of
English, like SSBE, the realisation of the NURSE vowel is [
] (see 5.2). In
eastern New England, parts of the American South and much of African
American English, the pronunciation varies and can be [
], [ə] and
even [i ~ əi] (Thomas 2001: 48–9).
In most of the American West and New England, American speak-
ers have identical vowels in LOT, THOUGHT and possibly also in PALM,
CLOTH.6 The low back unrounded vowel /ɑ/ for LOT, PALM is a gen-
eralisation for values ranging from the front upper low [æ] to the low
back [ɒ]. The fronted variant of [æ] is part of an ongoing chain shift of
the vowels /ɔ/, /ɑ/, /æ/, /ε/ and // known as the Northern Cities Shift,
whereby /ε/ → // → /ɔ
/ → /ɑ/ → /æ/ → /æ /.7 Salient regional or
ethnic differences will be highlighted in Chapters 6–8.
2.2.1 Short and long vowels

The colon symbol (
) in Figure 2.4 is an IPA diacritic indicating that the
vowel is long; all other vowels are short. : Taken literally, ‘length’, or
‘quantity’, refers to the physical duration of vowels. A reference to vowel
length as a contrastive feature works well for languages like Classical
Latin, where the differences between the vowels are assumed to have
been based on duration. However, as happened in Vulgar Latin, and as
is the case in PDE, the historical long and short vowels are distinct not
only by virtue of duration: they have different articulatory and acous-
tic properties. In many accounts of PDE sound systems, therefore, the
terms ‘long’ and ‘short’ have been replaced by tense and lax, which refer
to the manner of articulation of vowels, more specifically, the overall
muscular effort involved in the production of a sound. : Tenseness
involves increased muscular effort, strong spread of acoustic energy
and movement away from the centre to the periphery of the vowel space,
of measurements in GA produce the same top-frequency vowels. The least frequent

vowel shared by both varieties is the diphthong /ɔi/ (0.14 per cent).
6
The Atlas of North American English (ANAE) (available at <http://www.ling.upenn.edu/
phono_atlas/home.html>, last accessed 16 May 2013) shows a map of the merger:
invariant responses in production and perception are shown on Map 1 (see also 8.2.1).
7
See Labov (1991, 2008: 37–46); Wolfram and Schilling-Estes (2006: 147–9); Labov et
al. (2006: ch. 14).
justifying a reference to another property, peripherality, associated

with tenseness. PDE high and mid long vowels are peripheral. Length or
tenseness are irrelevant for the English low vowels; neither one is used as
a phonemically contrastive feature in PDE. Laxness refers to decreased
muscular effort and movement; lax vowels are relatively short and they
are articulated more towards the central or non-peripheral vowel area;
if a vowel is short, it is also lax and non-peripheral.
The short and long vowels in Figure 2.4 are represented by different
IPA vowel symbols, since in the basic reference varieties, GA and SSBE,
no pair of vowels contrast only in duration; in every instance length
distinctions are accompanied by height, backness and peripherality dis-
tinctions. Actual duration differences are attested in identical environ-
ments. Measurements of RP cited in Cruttenden (2008: 96), for example,
show /i
/ + voiced stop, as in lead, with a duration of 28.5 centiseconds,
while for /i/ + voiced stop, as in lid, the duration is 14.7 centiseconds.
The durational difference can be reversed, however, depending on the
consonant following the vowel: the duration of /i
/ + voiceless stop, as
in meat, is 12.3 centiseconds against the 14.7 centiseconds of the ‘short’
lid vowel. Therefore, many descriptions of the PDE vowel system rely
solely on qualitative differences, using different symbols for the short
and long vowels and dispensing with the length mark. :
The choice to refer both to the quality and, for the non-low vowels,
explicitly to the quantity, of the PDE vowel phonemes in Figure 2.4, is
justified mainly historically. If we recognise the relevance of quantity
in PDE, with all the caveats above, we can project this quantitative
dichotomy backwards as far as we can go. The length dimension thus
serves as a consistent common denominator between reconstructions of
earlier states of English and PDE (see 6.1, 7.5, 10.2).
Using only qualitative specifications, on the other hand, makes it
more difficult to trace the continuity with confidence. First, the pho-
nological function of tenseness is obscured in some PDE varieties.8
Second, the availability of only five vowel letters in the Roman alphabet
is extremely limiting for a language such as English with at least fifteen
contrastive vowels at any time in its history, so that the use of identical
letters is uninformative with respect to qualitative differences. Very
importantly, the agreed-upon reconstructed vowel system of Proto-
Germanic is based on quantity: from Proto-Germanic Old English
inherited short vowels and long vowels. The short–long dichotomy
8
The Scottish Vowel Length Rule (McMahon 2002: 867) lengthens [i], [o] to [i
], [o
]
before [r, v, ð, z, ], before another vowel and before a morpheme boundary, so the
vowels in grief and grieve differ only in duration.
was maintained in the subsequent history of English vowels, which go

through a series of changes described traditionally as ‘lengthenings’
and ‘shortenings’. It is for these reasons that the terms long and short are
adopted in this book, bearing in mind that in most cases quantity alone
is not sufficient to distinguish the FLEECE and KIT, FACE and DRESS,
GOOSE and FOOT vowels of our reference varieties, GA and SSBE. :
2.2.2 Complexity: monophthongs and diphthongs

The relatively ‘pure’, steady-state vowels of English are monoph-
thongs: there is usually no perceptible difference in quality between
the beginning and the end of the vowel in the syllable peak (see 2.3.1).
For the long vowels, on the other hand, maintaining a steady state is less
typical; the additional duration allows for a transition from one vowel
quality to another. :
The representations of the upper mid front and back vowels [ei]
and [oυ] in Figure 2.4 indicate that the vowels of FACE, GOAT are diph-
thongs – the quality of the vowel changes during the pronunciation of
the vowel. The components of a diphthong are not equal: the nucleus
is more easily identified with a specific vowel quality, while the glide
portion, usually the second vowel position, may or may not reach its end.
Diphthongisation involves the addition of a glide: with long vowels the
gliding is usually towards a higher vowel, as in [ei] and [oυ], while with
short vowels the most common gliding is into a schwa-like central vowel,
as in KID pronounced as [kiəd], or New York English OFF pronounced
[ɔəf ~ υəf]. The degree and the direction of vowel diphthongisation is
an important historical index and a salient dialect criterion for PDE.9
The dimension of complexity thus refers to the absence or presence
of gliding from one position to a second position within a single syllable
peak. Only one of the diphthongal components is syllabic; glides are the
parts of diphthongs that are non-syllabic. The direction of the gliding
and the length of the trajectory vary. In addition to the front and the
back glides [i] and [υ], diphthongs formed from a vowel + a following
/r/ insert a central glide [ə], as in SSBE SQUARE, CURE.
9
This book follows the common practice – for example, McMahon (2002); Ladefoged
(2005: 28–9) – of transcribing the PDE diphthongal glides as [i], [υ], [ə] for the front,
back (outgliding) and the central (ingliding) elements. This transcription avoids a
potential confusion with the palatal and labial approximants [j] and [w]. However, the
rationale for choosing [i] over [i], [y], [j] or [υ] instead of [u], [w] could be debated,
and there is no universally accepted solution. Historically, more ‘consonantal’ [j] and
[w] are precursors of diphthongal glides, and we will use them in transcribing earlier
diphthongs in 6.5.3, 7.4, 8.2.2.
All accounts of PDE recognise the existence of three diphthongal

vowels, shown in (2).
(2) PDE diphthongs:

/ai/ as in PRICE
/aυ/ as in MOUTH
/ɔi/ as in CHOICE
Some analyses of English classify the vowel in CUTE as diphthongal,

consisting of a glide, usually represented by [y], followed by [u]/[υ].
(Recall that [y] is another way of representing the IPA palatal approxi-
mant [j].) The analysis of this sequence is a well-known crux in English
phonology. There is no agreement on the tenseness of the vowel when it
is not followed by [ə] due to loss of [r] in non-rhotic dialects, as in PURE,
nor is there agreement on whether /j/ is (part of) the syllable onset or
the syllable peak. Bearing in mind some distributional peculiarities of
the palatal approximant – it appears only syllable-initially when pre-
ceded by one or more consonants, as in beauty, cute, fury, skewer, and the
only vowel allowed after it is /u
/ – we analyse /j/ like /w/, that is, as
part of the syllable onset, and not as part of a diphthong which occupies
the peak/nucleus of the syllable. :
For many GA speakers the sequence [-ju] after coronal consonants –
duty, dues, news, sue, tune – is realised as [u
]; the glide-deletion is part of a
long historical process that started in the sequence [rj-] and now occurs
variably whenever the vowel follows a consonant in the dento-alveolar
series [l, n, t, d, s, z], as in lubricate, news, Tuesday, due, suit, presume (see
further 8.2.2.3).
Complexity is a scalar property: all long vowels are complex to
some degree. The high long vowels in FLEECE, GOOSE are more clearly
monophthongal and for them we can use the more abstract representa-
tion VV, or [V
], where V stands for ‘any vowel’. The vowels of FACE,
GOAT are perceived as changing from beginning to end, but the length
of the trajectory between the two end-points varies, and there are varie-
ties in which the vowels are monophthongs, such as SSE in Britain and
Louisiana Cajun English, South Carolina, and Georgia in the US. The
diphthongal nature of the vowels of PRICE, MOUTH, CHOICE in (2) is gen-
erally strong in most varieties of English, except in the American South.
The clearly perceived transition from one quality to another in them
justifies the more abstract representation VG, where G represents the
glide. VV and VG vowels function in very similar ways (see 2.3.3); this
similarity is crucial in accounting for the famous long vowel shifting in
the history of English (see 8.2.2).
2.3 The syllable: some basics
2.3.1 Syllable structure

Phonological segments are not pronounced in isolation – they are
organised into syllables. Native speakers have strong intuitions about
the division of words into syllables. The way we define a syllable
depends on whether we are interested in the production, the perception
or the function of the syllable. The most elementary definition refers
to production: a syllable is the smallest independently pronounceable
unit of speech. In auditory terms, the syllable is a string of sounds of
different prominence, where ‘prominence’ corresponds to the acousti-
cally measurable sonority of the sounds. Functionally, the syllable is a
domain which hosts and governs the combinations and distribution of
sounds, a domain of prosodic phenomena such as stress and rhythm, and
a domain of interaction between phonology and morphology. A brief
survey of the basic principles of syllable structure and syllable division
will give us the initial reference points for the discussion of historical
phonological processes that occur in specific syllabic environments.
The organisation of sounds into syllables depends on the sonority of
the adjacent sounds. Vowels are always independently pronounceable,
they have higher sonority than consonants and they form the core of a
syllable. Structurally, vowels are positioned in the syllable peak, also
known as syllable nucleus. The segment(s) preceding the peak form the
onset of the syllable, and the segment(s) following the peak form the
coda. The peak and the coda together form the syllabic rhyme. These
constituents are illustrated in (3), where C = consonant, V = vowel,
G = glide and Ø indicates an empty constituent. The sigma (σ) is a
conventional symbol for the whole entity of the syllable.
(3) Syllable structure:
(a) σ (b) σ (c) σ
R(hyme) R R
O(nset) P(eak) Co(da) O P Co O P Co
C V C Ø VG C CC V V C C
s ŋ a t cr i : p s
sing out creeps
The syllable constituents are not equally important – the only obliga-
torily filled constituent is the peak. The peak can be filled by any vowel
, n ], as in the second syllables of mutter,
or by a syllabic sonorant [r, l , m
kettle, bottom, weapon. The onset in PDE (but not in OE stressed syllables
at least in formal style; see 10.2) can be empty, as in out in (3b); it can be
filled by a singleton, as in sing in (3a); or it can be filled by a cluster, as in
creeps in (3c). Similar variability is allowed for the coda: it can be empty,
as in my, see, bow; it can be filled by a singleton, as in (3a) and (3b); or it
can be filled by a cluster, as in creeps in (3c).
The realisation of one and the same consonantal phoneme typically
differs in the onset and the coda, especially in syllables bearing stress.
The voiceless stops in the onset of tip [th-], pit [ph-] are strongly aspi-
rated, while in the coda position the same consonants are unreleased,
that is, there is low vocal-fold activity, no audible burst of air. A well-
known case of allophony based on syllable position in SSBE and SAE
is the different realisation of the lateral liquid: ‘clear’ [l] in the onset, as
in lick [lik], but velarised ‘dark’ [l̃] in the coda, as in [kil̃] (see further
5.2.5). :
Functionally, the onset and the coda are asymmetrical: the onset is
more perceptually salient. Onsets can be articulated more forcefully,
which contributes to the maintenance of phonological contrasts. The
coda, on the other hand, is the location of consonant neutralisation
and loss, illustrated by the voicing neutralisation of the /-d/ past tense
morpheme in English – snapped [-pt], passed [-st], loss of coda <-r> in
non-rhotic varieties, or the historical simplification of coda clusters such
as <-mb, -ng>. Another asymmetry between the onset and the coda
position, observed across the world’s languages, has to do with the pref-
erence for a single consonant between two vowels to associate with the
onset rather than the coda of two adjacent syllables (see 2.3.2).
Yet another difference between onsets and codas has to do with the
arrangement of consonants within these constituents. The ‘ideal’ sonor-
ity slope from the onset to the peak is steep, that is, voiceless stops are
preferred as onsets, while the reverse holds for the peak–coda slope;
codas are preferentially sonorous. When there are consonant clusters,
onsets accommodate sequences with rising sonority – typically an
obstruent followed by a sonorant, for example [pl-, dr-, kr-, gr-], and so
on. Coda clusters are a mirror image of the onset clusters in sonority: in
monomorphemic words the first coda consonant has to be a sonorant or
[s], and the second consonant is typically a voiceless stop, that is, coda
clusters show falling sonority.
2.3.2 Syllabification
In English a syllable is often a whole word: indeed, English is sometimes
referred to as a ‘monosyllabic’ language, since so many of its core vocab-
ulary words have only a single syllable: bread, child, sleep, fight, green, short.
We will see how and why this happened in 7.6. Derived vocabulary and
the borrowed word-stock, however, are polysyllabic, and polysyllables,
as well as connected speech, present the analytic problem of deciding
where to place the syllable boundaries.10
Some syllabification principles are universal and some are language-
specific. Vowels can be syllable peaks in all languages and must be syl-
lable peaks in English. Another widely shared syllabification rule is that
only possible word-initial consonants and consonant clusters can be in
the onset of a word-medial syllable. Which consonants are allowed as
singletons or as clusters word-initially is a language-specific matter.
Thus the fact that English /ŋ/ originates from coda [-ŋg] blocks it from
appearing in word-initial position, therefore a word such as singing
is syllabified sing.ing [siŋ]σ [iŋ]σ, but this would not be the case in a
language like Vietnamese, where the velar nasal is ‘legal’ word-initially.
Except for /ŋ/, and possibly // (see 5.4), all English consonant
phonemes can appear word-initially. Syllabification of intervocalic
singletons places them in the onset of the syllable to the right, as in (4).
(4) Syllabification of intervocalic singletons:
VCV –> V.CV
ra.ven pho.na.tion pa.ra.me.dic
re.ly me.cha.nic pa.li.sade
(4) is in accord with a widely attested syllable structure preference for
a filled onset: all languages have CV- syllables but not all languages
have -VC syllables. This is another instance of the functional asym-
metry between onsets and codas. The principle of filling the onset in
preference to the coda is known as the Maximal Onset Principle or as Onset
Maximalism.
An alternative way of syllabification in PDE is to assume that at
least some singletons preceding an unstressed syllable are ambisyllabic.
Ambisyllabicity is one way of accounting for the realisation of the dental
stops /t/ and /d/ as alveolar approximant taps [ɾ] before an unstressed
vowel in AmE, as in ladder [læɾ], waiter [ weiɾ] (see further 5.5.2).
Our historical account in the following chapters assumes onset-maximal
syllabification, which is crucial in the special recitation style of OE verse
(see 10.2). One should, however, be aware of the possibility of ambisyl-
labic analysis of some consonants as early as our earliest OE records.11
10
Following the IPA, we mark syllable boundaries with a period in orthographic forms,
or with a subscript sigma in phonetic transcriptions.
11
For the history of the research on ambisyllabicity in PDE, the phonetic diagnostics and
Onset maximisation may be overridden by morphological considera-

tions. In dis.able, dis.obey, in.applicable, mis.analyse the prefixes are likely to
retain their final consonant in syllabification.
In addition to placing a singleton to the right of the syllable bound-
ary, onset maximality regulates the syllabic affiliation of the consonants
in clusters. As noted, only clusters that can be word-initial can be placed
in the onset. Other sequences are split between the onset and the coda.
The most common patterns are summarised in (5).
(5) Syllabification of intervocalic clusters:
V.C1C2V: Sea.gram, pa.tron, ho.ne.sty, ver.te.bra
(a) VC1C2V
VC1.C2V: fic.tion, sub.merge, com.pul.sion, pro.pa.gan.da
VC1.C2C3V: em.blem, pan.try,ham.ster, ban.quet, pil.grim

(b) VC1C2C3V VC1C2.C3V: glimp.ses, ant.sy, munch.kin, junc.tion
V.C1C2C3V: in.du.stry, re.gi.strar, or.che.stra
In the upper branch of (5a) the division is based on the availability

of the cluster word-initially: grant, tree, sty, bra. The lower branch
of (5a) splits the sequences since *[kʃ-, bm-, mp-, ls-] do not appear
word-initially.
(5b) follows the same principle. In the top branch [bl-, tr-, st-, kw-]
are kept together in the onset when tri-consonantal medial clusters are
syllabified. The middle branch of (5b) shows how the division of three
consonants in a row avoids [ps-, ts-, tʃk-, kʃ-] in the onset. In the bottom
row C1C2C3 stick together and are syllabified to the right – in.du.stry,
re.gi.strar, or.che.stra – again in accord with the existence of [str-] words
in English. This carries over to VC1C2C3C4V, where the cluster C2C3C4
is cohesive: mon.stro.si.ty, seam.stress.
The schema in (5) does not cover all possible cases and theoretical
interpretations. For example, the very common word-initial clusters
sp-, st-, sk- do not behave consistently in syllabification when they
follow a stressed syllable, so that speakers are likely to produce A-spen
the rules of ambisyllabicity see the overview in Hayes (2009a). The choice of assuming
onset-maximal singleton syllabification in this book is based on the formal style that
the our records represent and the lack of well-worked out arguments in the literature
in favour of ambisyllabicity in Old and Middle English (see Fulk 1997). Some research
into the evidence for syllabification of OE and ME reported in Minkova and Zuraw
(forthcoming) reveals similarities between PDE and the earlier stages of English with
respect to ambisyllabicity.
or As-pen, hi.sto.ry or his.to.ry, ma.scot or mas.cot.12 Further complications

arise from the interaction of phonology and morphology: the syllabi-
fication of compounds and transparently derived words respects the
morphological boundaries rather than the Onset Maximalism, thus
juris.prudence not *juri.sprudence, dis.prove not *di.sprove, close.ness not
*clo.seness (in spite of snow, snide, snap), sick.ly not *si.ckly (in spite of
close, cluster, climb, but compare lu.di.crous), top.less not *to.pless (in spite
of plus, play, pledge).
2.3.3 Syllable weight

In 2.3.1 we noted the correlation between allophonic realisation and the
position of a segment with respect to syllable structure. Syllable struc-
ture is also crucial in determining syllable weight: syllables ending in
short vowels are light, all other syllables are heavy. This statement has
two entailments:
• All syllables with filled codas are heavy.
• All syllables with long vowels or diphthongs in the peak are heavy.
Long vowels and diphthongs are functionally equivalent because they
form the peak of a heavy syllable, as in (3b) out and (3c) creeps, where
the peak is represented as branching. If we take the mora (µ) as the
basic unit of weight, the observation that the syllable peak branches
amounts to saying that it is bimoraic. Non-branching syllable peaks are
monomoraic, as the peak in sing in (3a). Note that there is no neces-
sary correlation between orthography and moraic content. No matter
how bimoraic vowels are spelled, with a single vowel letter or with a
digraph: <hide, rode, raid, coin, noun>, they are single phonemic units,
either monophthongal or diphthongal, and belong to the same syllable.
Thus coi-nage, guile-less, buy-er contrast with go-ing, ki-osk, Mi-a-ta; in
the latter the bold vowel letters correspond to separate syllable peaks.
Short vowels followed by a consonant in the same syllable also form a
heavy syllable; such heavy syllables branch at the level of the rhyme, as
sing in (3a), where the peak is filled by a single mora, and the consonant
in the coda counts as a second mora. We can generalise: heavy syllables
are represented by branching structures, or: heavy syllables are mini-
mally bimoraic. Put differently, syllables that branch at the level of the
rhyme or below are heavy. For the moment, we will stay with the clas-
sification of syllables into light and heavy, but we will see in Chapter 9
12
For a much fuller treatment of syllabification in PDE polysyllabic words see Giegerich
(1992: 167–78).
that syllable weight is a gradient property depending on the nature of

the segments, especially in the rhyme.
It is important to distinguish between length, which is a durational
property of segments, and weight, which is a prosodic property of the
entire syllable. The mora bridges the two notions: both short vowels and
light syllables have a single mora, and long vowels and heavy syllables
have more than a single mora. :
Syllables can also be open or closed. Any syllable with an empty coda
is open: all syllables in ho.ne.sty, pay.ee, Le.vi, mi.ca are open, that is, any
syllable in which the rhyme does not branch is open. Any syllable with
a consonant in the coda is closed. Closed syllables have branching
rhymes; they are heavy by definition. Open syllables can be either heavy
or light – we return to this in 7.5 when we discuss the history of pairs
such as shade-shadow, stave-staff, nose-nostril.
Most of the stressed syllables in the PDE lexicon are heavy; this is
in line with a cross-linguistic preference for heavy syllables to attract
stress. The tendency to place stress on a heavy syllable in English dates
back to Old English, when stress was placed on a heavy syllable in
approximately 80 per cent of the words.
The syllable typology sketched out above correlates also with the
distribution of long, diphthongal and short vowels. In PDE all vowels
can occur in closed syllables. However, unlike OE (see 6.1), if a syllable
in PDE is stressed, open and word-final, it can accommodate only bimo-
raic vowels: say, decree are OK, but forms like *se with [-ε], *decrí with
[-i] would be recognised as un-English. In other words, the short, lax,
non-peripheral vowels [i, ε, æ, υ, ] have a more restricted distribution.
They are also sometimes referred to as checked vowels because they must
be followed by a consonant when they are stressed, whether in the same
syllable or not: sinful, pension, monkey have checked vowels. The periph-
eral long vowels and diphthongs are known as free vowels. The low back
unrounded vowel, namely [ɑ], is neutral with respect to length in GA. It
can equally be treated as a member of the long vowel set and though it
is not phonetically ‘long’, it is behaviourally ‘long’ because it can occur
in both free and checked positions: pa, paw, God, lot, saw, nought. In SSBE
the vowel [ɒ] in LOT is a checked vowel.13
The two types of vowels show some special distributional properties:
13
The low front vowel /æ/ is also difficult to classify: historically it is the reflex of a pho-
nologically short vowel, but it is phonetically longer than the other checked vowels,
especially before voiced consonants. It behaves like a free vowel in at least one very
common adverb in GA, nah ‘not so’, which the Merriam Webster transcribes with the
vowel of ash.
monomoraic vowels combine freely with [-ŋ] and [-ʃ]: ring, flung, mesh,
pang, posh, lush; but bimoraic vowels are generally avoided before these
consonants: *feeng, *poung, *coish, *raish would not be considered ‘normal’
English words in the standard varieties described here. Another unify-
ing property in the history of the long vowels and diphthongs is the
recurrence of changes whereby long vowels develop into diphthongs
and vice versa (see 8.2.2); [ɔi] is the only English diphthong which has
no historical relation to a long monophthong. There is also a distribu-
tional difference based on the prominence of the syllable in which the
vowels appear: it is only in stressed syllables that we find the full inven-
tory of the vowels, while only monomoraic [ə, i] appear regularly in
unstressed syllables in all varieties of English (see 2.2).
2.4 Notes on vowel representation

The English spelling system has not kept apace with the radical and
complex historical vowel changes. The resources were inadequate from
the start: with only seven vowel letters at their disposal – <i, y, e, æ, u,
o, a> – the OE and ME scribes could not represent finer qualitative or
quantitative distinctions and one and the same letter could have multi-
ple values (see 6.2, 6.3). Similarly, in PDE there are different values for
the same letter: in hide, hid, machine, <e> in pet and Pete, <a> in grade
and gradual, <o> in sole and solitude, come, common. Such spellings justify
the statement that English spelling is ‘etymological’. Vowel digraphs,
two vowel letters representing a single phoneme, such as <ea>, <ou>,
<ai> <oo> are useful for marking length, but they can also mask the
difference between short, long and diphthongal vowels
<bread>,
<cousin>, <plaid>, <foot> have short vowels, while long vowels are
commonly spelled with a single vowel letter
<bite>, <bone> <same>,
<fume>. This looks like a chaotic situation, but in fact the histori-
cal information in the following chapters will reveal that English has
a fairly reliable system of orthographic coding of the long vowels. A
useful guide when figuring out the phonological history of a PDE vowel
from the orthography is that vowel digraphs that represent a single
phoneme regularly stand for historically long or diphthongal vowels.
Even without looking up their individual histories, we can expect to
find that the words spelled <stood>, <said>, <head> had long vowels
at some earlier period (see further 8.3.1).
Note on diacritics: The macron symbol <–> indicates vowel length
in the printed early/etymological form of a word.14 The macron is not
14
The use of the macron in the pronunciation guides in many American dictionaries
used in the original texts; it is an orthographic convention adopted by

modern editors of dictionaries and historical grammars. It is also stand-
ardly used in modern editions of OE and some ME texts, a practice
which helps the reader figure out the pronunciation. As noted in 2.2.1,
the phonetic symbol for length is the colon [
], thus the dictionary entry
for Lat. māter corresponds to /ma
ter/; an etymological entry for we, OE
<wē>, corresponds to /we
/.
2.5 Phonological change: some types and causes

Speech sounds are studied in two interrelated ways. Phonologists
explore the sounds’ function in speech: how do we learn them, how do
we store them, how do we keep them apart, how do we combine them,
how do they change? The actual physical aspects of sound: how speech
sounds are produced, how they are transmitted as the sound waves
travel through the air, and how the acoustic signal is perceived are ques-
tions from the domain of phonetics. Since phonetics is an experimental
science, historical phonetics in reference to the earlier periods is an
oxymoron. Only the relatively recent period following the invention
of sound recording in the second half of the nineteenth century offers
opportunities for observing and analysing phonetic details that lead to
change. The absence of historical phonetic records is a discouraging
reality, yet no meaningful hypotheses about phonological developments
in the past can be formulated without knowledge of the basic phonetic
properties of production, acoustics and perception. Consequently, his-
torical phonological analyses have to rely on the phonetic facts estab-
lished for living languages. The phonetic information that we project
back to the older stages of English draws on instrumentally testable pat-
terns of production and perception which provide valuable explanatory
angles on phonological reconstruction.
The inventory of contrastive sounds in any language is finite, while
the allophonic realisation of phonemes allows quite lavish variation.
Variation may or may not produce a restructuring of the inventory, but
it is a reasonable assumption that variable pronunciations underlie all
categorical innovations. Our access to diachronic variation is limited
by the nature of the surviving material, and reconstruction of what
happened in the past must be informed by other general principles of
linguistic change. One such principle is that change can be triggered
both by internal factors such as production, perception, acquisition, the
commonly refers to both quantitative and qualitative differences. This is not in accord
with IPA usage, nor with the usage in this text.
overall shape of the phonological system and language typology, and

by external factors such as language contact, sociolinguistic pressures
arising from group identity, prestige and education. Separating internal
from external factors is analytically expedient. Nevertheless, it has to
be acknowledged that in practice the various sources of innovative
pronunciations are hard to isolate and identify. The focus of this book is
diachronic; the further back we go, the less we know about the external
factors of change, and therefore for Middle and especially Old English
we will rely primarily on argumentation based on internal factors.
The phonetic properties of adjacent segments and their interactions
are important internal triggers of change. When a segmental change is
transparently related to the presence of another segment in the word, the
change is environmentally conditioned. The most frequent environ-
mentally triggered change that we will encounter is assimilation. The
target of assimilation can be one or more of the features associated with
place, manner or voicing. The scope of assimilation can be complete or
partial, and the direction of assimilation can be from left to right, also
known as regressive, and from right to left, also known as progressive.
In PDE assimilations occur commonly in relaxed speech styles; notice
the naturalness of pronouncing miss Sheila with [ʃ - ʃ], this year as [ʃ - j],
to take just one example of regressive assimilation of /s/ to a following
palatal. Some historical examples of assimilation are shown in (6).
(6) Assimilation:
OE hūs ‘house’ [-s] + bonda ‘freeholder’ > husband [-z]
(Voicing, partial, regressive)
OE wīf-monn ‘woman’ [f - m] > late West Saxon wimman [-mm-]
(Manner, voicing, full, regressive)
ME (c. 1225) questiun [-stjə-] > question [-ʃtʃə-]
(Place, manner, partial, regressive)
OE mēt-an ‘to meet’ + /d/ (past tense) > mētte [-tt-]
(Voicing, full, progressive)
The opposite process is dissimilation: the avoidance of adjacent or
close identical or similar segments. The adjectival suffix -al < Lat. -ālem,
as in choral, mortal, natal, is dissimilated to -ar after [-l-], as in cellular,
scalar, vascular. An example of dissimilation is the frequent pronun-
ciation of diphthong in PDE with [-pθ] (stop-fricative) instead of [-fθ-]
(two fricatives). A historical parallel is the change of OE þēofð, þīefð [-fθ]
(<þēof ‘thief’ + *iþa ‘noun’) PDE > theft [-ft], where the inherited frica-
tive of the suffix becomes a stop; compare also height, sleight, drought, all
of them with original [-hθ]. Avoidance of fricative clusters is important
in the account of the earliest consonantal changes in Germanic (see 3.4).
Another, relatively minor, pattern of change rearranges the linear

order of segments – a process known as metathesis. A well-known
instance of historical metathesis in English is the verb ‘to ask’, which
in OE had two forms, one starting with <asc-> and one starting with
<acs->, so that among the more common ME forms we find <ax, ex,
ask, esk>.15 Metathesis is found also in OE cærse (< *cræsse), ‘watercress’;
Horsa ‘a personal name’ < *hross ‘horse’ + a; the dialectal (Southwestern)
purty for pretty.
Assimilation, dissimilation and methatesis preserve the number of
consonant slots in a word. Other linear phonological processes can add
or delete segments. Epenthesis is the general term for the insertion of a
segment. The epenthesis of a vowel inside a consonant cluster is known
as anaptyxis. A familiar PDE example of anaptyxis is the insertion of an
unstressed vowel between a verbal stem ending in dental stops and the
past tense morpheme /-d/
/-t, -d/ + /d/ > [-əd]/[-d], for example
rusted, minded.
When the segment is added word-initially, the process is called
pro(s)thesis. Prosthetic initial e- characterises French borrowings in
English whose roots in Latin had initial sp-, st-, sk-. Thus the Latin root
sta- ‘to stand’ appears in two forms in English: as estate (1225), from Old
French, and as state, taken directly from Latin. Prosthesis accounts also
for the e- in loans such as espouse (1475), especial (1386), escalade (1598) vs
spouse (1200), special (1225), scale (1330).
The insertion of a consonant within a word is simply a case of epen-
thesis. Epenthetic consonants may fill a structural position and improve
the overall shape of the syllable, for example bra.mel, spi.nel, þu.nor, with
empty first-syllable codas and sonorants in the onset of the unstressed
syllable alternate in OE and ME with bram.ble, spin.dle, thun.der, where
the insertion of the voiced stops causes resyllabification whereby the
syllables bram-, spin-, thun- have filled codas, and the stops, the least
sonorous of all consonants, form the onset of the unstressed syllable.
Segments can be added word-finally, though the examples of this
type of change in English are not very numerous. In ME we find forms
such as <inoht> for enough, <boght> for bough, <borcht> for borough,
and after final nasals margent ~ margin (sixteenth century and after),
vermin ~ varmint (dialect and US). Among the ones that have become
codified are: pound, v. < OE punian; sound, n. < AN soun, OFr son; hound
15
The form ax was the regular literary form until c. 1600. It is still used in Midlands
and Southern dialects, though supplanted in standard English by ask, originally the
Northern form (OED). In the US the metathesised form axe is characteristic of the
South and African American English.
‘a clamp of wood’ < early ME *houn. The final consonant in such unety-
mological additions is called excrescent. Analogy with the superlative
affix -est is the OED explanation of final [-t] added to the OE adverbial
ending -es in against, amidst, amongst, betwixt, whilst. :
The deletion of segments at the beginning of words is called aphe(re)
sis. A widely recognised historical example of apheresis is the loss of
the initial consonant in the clusters <kn-, wr-, gn->, as in knight, wrong,
gnaw in English, which stand for singletons today, but were real clusters
in earlier English (see 5.3.1). Some early aphetic forms are ME ches, chess
‘chess’ < AN and OFr eschès (c. 1180); fray < AN and OF affray, effray
(1300); tend < attend (1370), ply < apply (1393), gypsy < Egyptian, n. (1514).
Word-medially, the consonants of English have been fairly stable
except in isolated instances of cluster simplification. Some early simpli-
fications are reflected in the spelling: OE blosma < blostma ‘blossom’, OE
elboga < elnboga ‘elbow’, OE Sæterdæg < Sæterndæg ‘Saturday’, OE endleofan
‘eleven’. The orthography does not reflect consonantal losses that were
not completed at the time when spelling began to be codified, towards
the second half of the fifteenth century: fasten, thistle, castle, fight, sought,
talk, walk.
The loss of a vowel word-medially is known as syncope [siŋkəpi].
Syncope affects unstressed vowels: consider the rapid speech pronun-
ciation of silvery, family as disyllables, and of course every is no longer
trisyllabic, pace OED’s recorded pronunciation [εvəri]. Syncope can
therefore be a diagnostic for the position of the stress at the time of the
vowel loss, as in the case of ME corúne ‘crown’ < OFr corone, corune, Lat.
corōna), for which we find the spelling <cruness> ‘crowns’ (c. 1200).
When two vowels are adjacent, the vowel which bears a lower degree
of stress may be syncopated, as in marriage, carriage. The most common
case of syncope in English occurs in inflectional morphemes, as in the
formation of the <-est> and <-es> of the second and third person
singular present tense, and the noun plural <-es>.
The dropping of word-final segment(s) is known as apocope
[əpɒkəpi
]. The term can apply to consonants, vowels and whole
syllables. Some cases of consonantal apocope are the loss of [-n] after
unstressed vowels – OE ān > a (indef. article), OE mīn > my (adj.
pronoun), the loss of the voiceless affricate [-tʃ] in the pronoun I <
OE iċ, and cluster simplification as in [-ŋg] > [-ŋ-] (sing, thing) (see
further 5.3.2). Syllabic apocope is illustrated by adverbial -ly < OE liċe,
[-ən]-loss in real or pseudo-suffixal -en, -on, -an, resulting in pairs such
as even(ing)-eve, maiden-maid, gammon-game.
The effect of such segmental changes on the overall phonological
structure varies. Original allophones can become separate phonemes –
this would be a case of phonemic split. For example, [v], the voiced
allophone of the OE fricative /f/, appearing originally only word-
medially between voiced segments, became contrastive everywhere in
ME, as in vast-fast, coffin-coven, leaf-leave (see 4.4). The loss of [-g] in the
[-ŋg] cluster, where [ŋ] is an allophone of /n/, as in sing, thing, resulted
in the addition of a new phoneme /ŋ/ to the consonantal inventory (see
5.3.2). For Shakespeare’s contemporaries the vowels of lush and bush
would have been allophones of /υ/, as they still are in some varieties
of English, yet today the // of lush and the /υ/ of bush are phonemes,
as is clear from the contrast of putt-put (see 8.2.1). Conversely, separate
phonemes can lose their distinctiveness and merge into one distinctive
unit – this would be a phonemic merger. In many North American
varieties the vowels of COT and CAUGHT, which are distinct phonemes
historically, have merged into a single vowel. AmE is in the process
of losing the voiceless labial fricative //, as in wharf, whine – it is
currently merging with /w/ (see 5.1.3).
Inventories can be augmented by borrowing. Famous historical addi-
tions to the English phonemic inventory are the voiced palatal fricative
// as in beige, leisure (see 5.4), and the diphthong /ɔi/ as in choice, toy
(see 7.4).
Another aspect of the diachronic behaviour of segments refers to
the notion of segmental strength, that is, historical phonological pro-
cesses can result in weakening/lenition or strengthening/fortition.
The strength of a segment can be described in reference to its sonority.
Recall from 2.1.3 that segments differ in sonority. Lower sonority, which
corresponds to a higher degree of stricture in the vocal tract, means that
the sound is ‘more consonantal’ or ‘stronger’. If there is less obstruc-
tion and less effort in the production of the segment, the sonority rises.
Consonants are lenited when they involve less stricture and when they
are voiced. Lenited consonants are more vowel-like. The ‘strongest’
consonants are stops, and the weakest are the sonorants, and especially
the oral sonorants [r, l, w, y]. In the vowel system, the most open vowels
are also most sonorous, while the high vowels are least sonorous – in
that sense, they are more ‘consonantal’.
The historical probability of weakening may be measured in terms
of intrinsic articulatory and acoustic properties such as degree of con-
striction for the consonants. Fricatives, for example, are weaker than
stops. Another parameter is constriction duration for the vowels: high
vowels take longer to articulate, they are stronger than the mid vowels.
Complexity is another correlate of weakness: schwa [ə] is the weakest
vowel because it lacks the featural complexity of peripheral vowels, thus
the change of any vowel in English to schwa is also weakening or vowel
reduction. In diachronic terms, weakening can be defined as propensity

towards deletion, the resulting zero being the ultimate ‘weak’ segment.
Weakening and strengthening may be associated with the proper-
ties of an individual segment, with the context in which the segment
appears, or, with both inherent properties and context. Segments are
organised into hierarchical prosodic units – we already looked at one
such unit, the syllable, in 2.3. The syllables themselves are organised
into higher-level domains: prosodic feet, prosodic words, clitic groups
(see 9.4.2). Typically, domain-initial positions resist weakening pro-
cesses, while domain-final positions are more vulnerable to weakening.
A very important interaction between prosodic structure and weakness
relates to the presence or absence of stress. The syllable is a stress-
bearing unit and there is a well-established correlation between weak-
ening and lack of stress; one such weakening, the reduction and loss of
final unstressed vowels in English, has had pervasive consequences for
the entire phonological, prosodic, morphological and even syntactic
history of English (see 7.6). :
Another factor that can influence the implementation of sound
change is frequency. The role of frequency in language change is a
rapidly developing area of research. When the phonetic motivations are
relatively transparent, the prediction is that the most frequent words
will change first, but as Phillips (2006) reports for English, if there
is further involvement of other components of the grammar, sound
change can affect the least frequent words first. The availability of elec-
tronic historic corpora has made such hypotheses testable on a broad
set of changes, but one should still bear in mind that the reliability of
quantitative data diminishes rapidly the further back we go in time.
Changes affecting the contrastive and relational status of sounds:
replacements, additions, deletions of segments, phonemic merger and
split, and the positioning of a segment on a strength scale, are closely
linked to the physical correlates of sounds: their articulatory, acoustic
and auditory properties. The phonetic grounding of sound change is
universal, though the individual paths of optimisation in the produc-
tion and perception of sounds are language-specific. One of the main
mechanisms driving phonological change, therefore, is the diachronic
selection of optimal realisations, where the demands on production and
perception are in constant conflict. :
A description of language in terms of its sounds and prosodic units
is an analytical convenience, but it is also highly idealised and artifi-
cial. In practice, phonemes and syllables matter only in so far as they
are part of the communication process. Phonological segments are
the building blocks of morphemes, the smallest meaningful units
in language, and morphemes build words. Although morphological

structure and change is a separate area of inquiry, the interaction
between phonology and morphology is a major factor in the initiation
and propagation of phonological processes. Phonological alternations
occurring in conjunction with morphological processes are known as
morphophonemic alternations, whereby one and the same morpheme
assumes different phonetic shapes, or allomorphs. Taking a very
simple example: /-d/, /-t/ and /-əd/ are allomorphs of a single past
tense morpheme in English. Allomorphy always involves a phonologi-
cal difference. If the difference is systematic, we seek to establish an
underlying form and the rules or constraints that produce the derived
form. Ideally, these refer to general principles of production and per-
ception. Using the same example, we do not store in our memory all
verbs that take /-d/, /-t/ or /-əd/; instead, we store a single affix /-d/
and change it in those verbs in which the voiced /d/ is incompat-
ible with the final voiceless consonant of the verb stem, as in stacked,
paced, limped, with [-t]. For stems ending in dental stops, waited, bonded,
raided, the affix is realised as [-əd] to avoid a sequence [-dd] which
English disallows word-finally (see further 4.4 for this and other cases
of allomorphic variation).
The sources of phonological change discussed so far – phonetic char-
acteristics, structural relations within the sound system, and prosodic
and morphological alternations – are all ‘internal’ to an idealised theo-
retical construct, the ‘linguistic system’. Obviously, linguistic systems
do not exist in isolation; language is a social act and the initiation and
propagation of new linguistic forms is accomplished by speakers. The
relationship between language and society is studied by sociolinguis-
tics. Modern sociolinguistic methodologies involve collecting and
analysing live data from speakers. Crucially, these data are testable and
replicable. Historical sociolinguistic research, on the other hand, has to
rely on written sources. The further back in time we go, the scantier the
sources. For Old English there is an unfillable gap between the language
of the texts and the range of possible forms and variations used by the
speakers. The availability of early sources is predicated on accidents of
social and military history. The records of linguistic variation are limited
also by the purposes for which texts were written and copied: highly
stylised verse compositions, chronicles, religious, archival, administra-
tive texts survived, while we have no casual correspondence, or the
equivalents of short stories or drama for OE and much of ME. Literacy
is another problem because while ‘illiterate’ spellings are a boon for the
historical phonologist, the bulk of the written records for OE are filtered
through standardisation associated with the Schriftsprache in late West
Saxon, where considerable levelling of the written forms may obscure

spoken variants in contemporary OE.
With these caveats in mind we can still take account of the his-
torical and social context of the changes under discussion. Loanword
phonology is essential in plotting the phonological histories of English
sounds – from pre-fifth-century exposure to Continental Latin, to
the continuing Celtic, Latin, Old Norse, Anglo-Norman, Old French
bi- and trilingualism in Britain. Although commonly measured by the
number of borrowed words, the effect of medieval multilingualism
stretches beyond the adoption of lexical items; the introduction of new
words means new phones in new places, and ultimately a reanalysis of
the entire phonological and prosodic system. This is evident from the
recent history of the palatal fricative // (see 5.4), the acceptability of
previously unattested [ts-, ʃm-, ʃl-] (see 5.6) and the prosodic differ-
ences between OE and PDE (see 9.6, 9.8). Dialect mixture in ME is
another area where system-internal factors interact with social factors,
and in Early Modern English and after, the social forces of stand-
ardisation can be as instrumental for the selection of variants as is their
intrinsic phonological structure.

3 Discovering the earliest links:
Indo-European – Germanic – Old
English
FOOT-PODIUM, TOOTH-DENTAL, HILL-CULMINATE, THREE-TRIPLE
The core meaning in the cognate word-pairs in our title is obviously

shared; most speakers will also recognise foot, tooth, hill, three as native
words, that is, they have been in the language since OE. As is evident
from the dates of the first attestations of their semantic mates in the
OED, podium (1743), dental (1594), culminate (1647), triple, v. (1375) are
newer words; they were borrowed from Latin. What is not immediately
evident to a non-specialist is that these pairs exhibit a pattern of regular
consonantal correspondences: /p/ and /d/ in pod(ium) correspond to /f/
and /t/ in foot; /t/ in dent(al) and tri(ple) corresponds to /θ/ (<th>) in
tooth and three; and /k/ in cul(minate) corresponds to /h/ in hill. There
are also vowel correspondences in cognates that are striking, too: mother-
maternal (1481), wind-vent(ilate) (c. 1440), originally ‘to blow away’. The
roots of these correspondences go deeper than the recorded history of
the language; they are pre-Old English.
This chapter will describe some of the most salient phonological
changes accounting for the difference between PDE words attested in
Germanic and OE, and their cognates borrowed in English from Greek
and Latin and their descendants. Before we address the specific changes,
we need to position OE in a more extended historical timeline.
3.1 Family matters: Indo-European – Germanic – Old English

Recall from 1.2 that the beginning of (Old) English is dated roughly to
the middle of the fifth century AD. The Germanic dialect of the Anglo-
Saxon settlers that became identifiably ‘English’ fifteen hundred years
ago was related to other Germanic dialects, and Germanic itself was
related to other branches jointly forming a set of languages known as
Indo-European (IE); IE is the metaphorical ‘family’ to which English
belongs. As will become clear in 3.4.1, family matters when we seek to
explain the puzzle presented by pairs of cognates such as three-triple,
tooth-indent and foot-podium.
54
DISCOVERING THE E ARLIEST LINKS 55
Cognate, literally ‘born together’, means that we need to identify the

common origin of words such as (in)dent-tooth. It is convenient to refer
to them as representing languages that belong to the same family. The
representation of genetically related language groups as a ‘family tree’ is
just a helpful metaphor and not an exact parallel of biological models of
identity and transmission. Languages are clustered together depending
on their shared features. We focus here on the similarities in phonologi-
cal and prosodic structure of cognate words in PDE, but there are many
other features that are very important in establishing common origin;
they include vocabulary and word-formation, grammatical structure,
territorial proximity and a long list of ethnic and cultural markers:
shared religion, rituals and customs, mythology, fables, chanted verse,
architecture, food, clothing, jewellery, weaving. Archaeological arte-
facts such as tools and weapons, wagons and coins can also point to
ancient contacts among the speakers of now separate languages. The
identification of common linguistic ancestry has been pursued also in
terms of human genetics by tracing the Y chromosome in male popula-
tions as the material basis of shared origin. Such research efforts add
to the testable criteria for relatedness and for the classification of the
world’s languages into families. Among the better-studied families
are Native American, Hamito-Semitic, Niger-Congo, Austronesian,
Finno-Ugric, Dravidian, Sino-Tibetan, Korean, Japanese, Vietnamese
and, closest to us, Indo-European, of which Germanic is one of the
surviving branches.
3.2 The Indo-European family of languages

The study of Indo-European has a long history. The ancient Greeks
and Romans were aware of the similarities between their languages.
Throughout the Middle Ages and the Renaissance, the European ver-
naculars had been compared with the Classical languages, but not in
any systematic way, and the usual goal of the comparisons was to point
out the disadvantages of the vernaculars. The first scholarly reference
to ‘familial’ resemblances among European and south Asian languages
(Persian and Sanskrit), was made by a British judge stationed in India,
Sir William Jones (1746–94). A Welshman by birth, he was intensely
interested in language and is reported to have mastered over twenty-
five languages before he died, including French, Italian, Spanish,
Portuguese, Latin, Greek, Persian, Hindi, Sanskrit, Hebrew, Arabic and
Chinese. Addressing the Bengal Asiatic Society in 1786, he announced
that the affinity between Sanskrit, Greek and Latin could not ‘pos-
sibly have been produced by accident’; the languages had to ‘to have
sprung from some common source, which, perhaps, no longer exists’.

Throughout the nineteenth and into the twentieth century, scholars
discovered new similarities among these and other languages, and
more branches were added to the Indo-European family tree. Today
the descendants of the ‘common source’, Proto-Indo-European (PIE),
have been identified and agreed upon, though the representations of
the family vary from branches that resemble a genealogical diagram to
spokes on a wheel, to images of trees with bigger and smaller branches.
Some form of Indo-European, the presumed original, or proto-
language, was spoken in the Black Sea–Caspian Sea area until, roughly,
4500–3500 BC. The latter part of the fifth millennium BC is therefore
considered ‘the latest possible date for the community of Proto-Indo-
European proper’ (Watkins 1992: 2,088). Thereafter, geographical sepa-
ration of the speakers following tribal migrations resulted in divergence
of the proto-language into many mutually unintelligible languages.
This is also the approximate date archaeologists have established for the
spread of the wheel through Europe, thus the dating of the IE diaspora
coincides with the initiation of transport in wheeled carts. The Anatolian,
Indo-Iranian and Greek languages are the earliest attested individual
branches of Indo-European. It is believed that the split between them
and other branches occurred between c. 3500 BC and 3000 BC. By c. 2500
BC most of the other branches were beginning to take shape. The diver-
sification of Indo-European is such that today as many as 439 separate
language codes are assigned to Indo-European (Lewis et al. 2013).
There are no written records of the reconstructed parent language,
Proto-Indo-European. Except for the tablets in Hittite, an extinct
Anatolian language, which date back to the fourteenth to the twelfth
century BC, the earliest texts preserved from any language in the IE
family are in Sanskrit, the religious language of the Hindus, in India.
The oldest literary work in Sanskrit is the Rigveda, or Veda of the Stanzas
/ Wisdom of the Verses, a collection of religious hymns composed c.
1400–1000 BC, and preserved only in documents of a later date. Greek,
whose speakers had moved away to the south-west, is recorded from
c. 850 BC in the largely historical account of the Trojan Wars fought
between the city-state of Sparta and the City of Troy in Asia Minor.
The Iliad and the Odyssey, ascribed to Homer (c. 850–750 BC) include
myths and legends that must have been part of the Greek tradition of
oral transmission long before they were written down in the form with
which we are familiar.
Figure 3.1 shows the division of Proto-Indo-European into some of its
main branches; the dates are necessarily approximate. The vertical con-
necting lines are not intended to suggest that the upper branches develop
Retreat of glaciers in northern Europe
(c. 8000 bc)
*Proto-Indo-European (c. 4500–3500 bc)
*Indo-Iranian -----------Greek
Celtic---------------Later European – I ------------ Italic/Romance
Balto-Slavic --------------- Later European – II ------------- Northwest European
(*Proto-Germanic c. 750–250 bc)
Figure 3.1 Some branches of the IE family of languages
into the lower ones; the branches evolve independently from each other,
though the cessation of human linguistic contact cannot have been abrupt
or complete, and therefore some linguistic continuity must be assumed.
The chart is simplified.1 : An asterisk (*) is used to mark languages
of which we have no physical record, but which are reconstructed
on the basis of the forms found in the daughter languages. Bold are
the branches whose daughter languages have played some role in the
development of Germanic or English.
Starting from the geographically most distant IE relative, we can
say that the influence of Greek on the phonological shape of English is
minimal. Most etymologically Greek words came into English via Latin,
or via Latin and Old French. Some orthography, but not the pronuncia-
tion, carries through, so only the spellings <ch-, ph-, pn-> in chorus,
physics, pneumonia signal their Greek origin. Further, words spelled
with <chth, phth, hy-, ps-, rrh, chr-, pt-, ct-, rh-, x-, sth-, mn-, bd->,
are almost exclusively of Greek origin, but the non-native clusters have
been simplified to fit the system of English consonants. In some rare
cases when a direct Greek loanword survives with its original pronun-
ciation – for example, chthonic or the combining form ichthi(o)- with /
kθ-/, phthisis with /(f)θ-/, sthenic with /sθ-/ – recognisably foreign con-
sonant clusters can be added to the periphery of the inventory of allow-
able consonant groupings in English (see 5.6). Some words have variable
realisations: asthma can be [æzmə], [æsmə], [æsθmə], earlier [æstmə],
and isthmus is [isθməs] [istməs], [isməs] (OED). The initial cluster in
1
For a complete chart of the Indo-European language family see the inside of the back
cover of the American Heritage Dictionary of the English Language, any edition.
the combining form pneo- ‘related to breathing’ is simplified today, but

The New English Dictionary (1907), as the OED was then known, gives the
pronunciation as /pn-/. The Greek stem phthi ‘coughing’ as in phthisic,
was borrowed first via Latin as /t-/, but pronunciations with /θ-/ or
/fθ-/ are recorded in PDE (OED). In clusters such as <ps-, pt-, mn->
the foreign sequence is nativised by dropping the first consonant: psy-
chology, pterodactyl, mnemonic. Greek initial <x-> is pronounced /z/, as in
xylophone. Consecutive vowels often form the peaks of adjacent syllables,
for example theatre, idea, idiom.
The Celtic branch of Indo-European, and especially the Brythonic
languages Cornish and Welsh, were territorially closest to OE. As noted
in 1.2, the presence of Celtic speakers in the British Isles prior to the
fifth-century Germanic incursions and settlements does not appear
to have influenced the Germanic word-stock significantly, at least as
attested in the surviving records. The transfer of phonological and
prosodic features from Celtic into Old English is still a matter of debate
(see also 6.5.3, 9.6).
The Italic branch, with its illustrious principal language Latin, has a
very long recorded history; there are Latin inscriptions dating back to
the sixth century BC, and literature in Latin flourished after the third
century BC. As the primary language of religion and learning in England
for at least a thousand years after the adoption of Christianity in the sixth
century AD, Latin has been an important contact language throughout
the history of English. The influence of Latin on OE phonology, if any,
is hard to identify. Latin loans whose phonological shape did not fit the
OE system were apparently assimilated. Thus, the native system disal-
lowed word-initial [v-] (see 4.4), so that we find spellings such as <fers>
for Lat. uersus ‘verse’, Firgilius for Virgilius (see Campbell 1959: 212).
Prosodically, OE Latin loans such as candēla, OE candel ‘candle’, comēta,
OE cometa ‘comet’ were also adapted to the native model of root-initial
stress. The situation changed in late ME and the Renaissance (see 9.6,
9.8) so that PDE has a layered stress-system.
Within Italic, Anglo-Norman and Old French provide an important
source of phonological innovation in Middle English. Both the conso-
nantal and the vowel inventories were augmented: /v/ and /z/ became
independent phonemes, the diphthong /ɔi/ as in CHOICE was added to
the vowel system (see 7.4). English–French bilingualism is one of the
main early conduits of a new system of stress-placement, allowing main
stress on the final syllable of nouns and adjectives: degree, abysm, entire.
Another Italic language, Spanish, has become increasingly important
for the development of PDE. In the last century, Spanish–English
bilingualism in the American Southwest has resulted in the formation
*Proto-Germanic
(c. 750–250 BC)
North-West Germanic East Germanic (Gothic)
West Germanic North Germanic
Low High Mainland Insular
Afrikaans Austrian German Danish Faroese
Dutch (High) German Swedish Icelandic
English Swiss German Norwegian
Flemish Pennsylvania Dutch
Frisian Yiddish
Figure 3.2 Proto-Germanic and its descendants
of new varieties of Latino Englishes, which in their turn influence GA.

One such influence is the acceptability of initial and stem-internal [h],
as in jojoba, rioja, fajita.
3.3 The Germanic branch of Indo-European

Using the family metaphor again, English is a descendant of the North-
West European branch of Indo-European. Figure 3.2 shows the modern
Germanic languages that share that genealogical line.
Proto-Germanic (PrG) remained a relatively homogeneous group
of languages spoken in the north-western parts of Europe. After c. 250
BC, speakers of Proto-Germanic settled large areas of Central, Western
and Eastern Europe, from southern Scandinavia, the North Sea and the
Baltic coasts, present-day Holland, east to the Vistula River in what is
now Poland and in the lands along the rivers Rhine, Weser and Elbe to
the south.
The most important East Germanic branch of PrG is Gothic, which
became extinct during the sixteenth century. Gothic is important for the
study of the history of the Germanic language family because, except
for a few scattered runic inscriptions, Gothic records antedate those of
the other Germanic languages by three to four centuries. The earliest
Gothic writings date back to the fourth century AD, when a Christian
bishop named Wulfila (c. 311–c. 382) is credited with putting together a
new, Gothic, alphabet, based on Greek and Latin letters for the purpose
of translating the Bible into a Germanic language. The best-preserved
Gothic manuscript copy of Wulfila’s translation, the Silver Bible, or
Codex Argenteus, with 188 leaves surviving, dates from the sixth century.
This translation of the Bible is by far the oldest lengthy text we have in
any Germanic language. Although it gives us little information about
syntax, since it is a rather close glossing of the Greek, and most of
the time follows Greek word order, it is very informative about early
Germanic morphology and is indeed the main source of our reconstruc-
tion of the earliest details of the Germanic inflectional systems. Gothic
provides the phonological anchoring for the reconstruction of the pre-
history of all Germanic languages, both because of the earliness of the
written record, and because of apparent similarities of sound-to-spelling
in Gothic to the better-documented contemporary Greek system.
The living daughter-languages of Proto-Germanic are grouped
further into North Germanic and West Germanic. West Germanic
is further subdivided into ‘High’ and ‘Low’, where ‘low’ refers to the
lowlands of northern Germany. The Low West Germanic group does
not include any of the standard languages referred to as ‘German’ today.
Historically, the languages in this group, especially Frisian, Dutch and
Flemish, are most closely related to Old English. The ‘youngest’ off-
shoot of Low West Germanic is Afrikaans, also known as ‘Cape Dutch’.
It is based on the language of the Dutch and other European settlers of
South Africa after the middle of the seventeenth century.
The original High West Germanic dialects of the mountainous
central and southern parts of Germany have evolved into High German
(Ger. Hochdeutsch), used as a standard language throughout Germany.
The national varieties of German spoken in Austria and Switzerland also
belong to this group. German and Swiss refugees in the US in the nine-
teenth century developed another variety of High German, known as
Pennsylvania Dutch < Deitsch ‘German’, an important Germanic ‘her-
itage’ language in North America. Another High West Germanic lan-
guage is Yiddish, which arose as a fusion of Germanic and Slavic in the
last millennium. Yiddish loan phonology has led to the recognition and
integration of new consonant clusters in English: [ʃm-, ʃl-, ʃt-] (see 5.6).
In the North Germanic branch, Icelandic, Faroese and Norwegian
are descended from Old Norse. There were only minor differences
between Old Norse and the precursor of modern Danish and Swedish.
As noted in 1.2, Old Norse is of particular interest because it was spoken
by the Scandinavian invaders and settlers of England between the
eighth and the eleventh centuries. ON influence is evident both in the
consonantal system (see 1.2, 4.2.1, 4.3) and in the vocalic system (see
6.2, 6.5.3, 7.4) of OE and ME. The most recent addition to the North
Germanic branch is Nynorsk, or New Norwegian, a language created
in the mid-nineteenth century, which combines many rural dialects
and which functions as the second official Norwegian language, the first
being Bokmål, a language heavily influenced by Danish.
We have now looked at the genealogical history of the language up
to the point when it became English nearly sixteen hundred years ago.
The next section introduces some phonological processes that charac-
terise the Germanic branch of Indo-European as distinct from the other
contact languages.
3.4 Some pre-Old English segmental and prosodic changes

The separation of Proto-Germanic from Indo-European is posited on
the basis of archaeological findings, shared lexicon, grammatical struc-
tures and verse tradition. Most relevant in our context, however, is the
phonological identity of the Germanic group – the features shared by
the Germanic languages, but not found in the other IE branches. The
topics included here – the First Germanic Consonant Shift, some vowel
changes from PIE to Germanic, early prosodic changes, final vowel
lengthening in monosyllables in Germanic and consonant gemination
in West Germanic – are chosen because these processes can be use-
fully invoked in the recognition and account of cognates, allomorphic
relations and prosodic properties in the modern language.
3.4.1 Grimm’s Law, or the First Germanic Consonant Shift2

The most salient phonological difference between Germanic and its
ancestors was discovered in the nineteenth century. It refers to a set
of changes affecting three sets of IE obstruents, known as Grimm’s
Law, though the term ‘law’ was not used by its discoverer. : The
changes covered by Grimm’s Law are also more neutrally and more
descriptively referred to as the First Germanic Consonant Shift,
where ‘first’ distinguishes it from a second set of changes, the ‘Old
High German’ or the ‘Second Consonant Shift’, which affects only
Continental West Germanic, more specifically Old High German.
The correspondences included in the First Germanic Consonant
Shift are the key to the puzzle referred to at the beginning of this
2
Material presented in this section is treated also in Minkova and Stockwell (2009:
142–5).
chapter: recall the ‘teaser’ pairs pod(ium)-foot, tri(ple)-three, cul(minate)-

hill. A simplified overview of the main consonantal correspondences
which identify the Germanic languages as a separate Indo-European
branch is shown in (1).3
(1) The First Germanic Consonant Shift:
*Proto- Proto-
Indo- Germanic
European
*p t k → fθh (voiceless stops→voiceless fricatives)
*b d g → ptk (voiced stops→voiceless stops)
*bh dh gh → bdg (voiced aspirated stops→voiced stops)
The time-span during which the First Germanic Consonant Shift
occurred is reconstructed as extending roughly between 750–250 BC,
and possibly as late as the turn of the first millennium AD. In the course
of at least five centuries the Germanic branch was moving further
away from the proto-language. At the same time, various dialects must
have evolved, they must have acquired new distinct properties, but all
Germanic dialects were still probably mutually intelligible. Since all
modern Germanic languages show its effects, it is logical to assume that
the Shift was under way prior to the split of Proto-Germanic into sepa-
rate Germanic sub-branches.
As shown in (1), the Shift involves three sets of stops in the parent
language, reconstructed PIE. Within all three sets the original distinc-
tions in place of articulation are well preserved: before and after the
shift labials remain labials, dentals remain dentals and velars remain
velars. In addition, the change of voiceless stops to fricatives involves
also a change from a bilabial /p/ to a labiodental /f/, a dental /t/ to an
inter-dental /θ/, and a velar /k/ to a glottal /h/.
Looking at the top row in (1), the original voiceless stops */p, t, k/
changed their manner of articulation and emerged as the voiceless frica-
tives /f, θ, h/, most likely going through an intermediate stage of aspi-
rated voiceless stops: /ph, th, kh/. In the second row, the voiced stops */b,
d, g/ became voiceless: /p, t, k/. Unlike Germanic, in Latin and the old
Romance languages, as well as in Greek, the consonant sets in the first
3
The asterisk (*) is used for forms established by comparative reconstruction. In the
PIE reconstructed consonant system the velar stops */k, g, gh/ can also appear in
h
a labialised version: */kw, gw, gw /; in Germanic the secondary labial articulation is
interpreted as the lip-rounded vowel /u/, while the primary velars develop into
/h, k, g/.
Consonant *PIE Latin Greek English Examples

class
*p p p f pedal, podiatry, foot

VOICELESS
*t t t θ triple, triad, three
STOPS
*k k k h cordial,5 cardiac, heart
*b b b p Very rare
VOICED
*d d d t endure, dryad, tree
STOPS
*g g g k cognition, prognosis, know
VOICED *bh f ph b refer, euphoria, bear
ASPIRATED *dh f th d forum, thyroid, door
STOPS *gh h kh g horticulture, chorus, gird
Figure 3.3 PIE cognates in Latin, Greek and English
two rows, /p, t, k, b, d, g/, kept their quality, or a quality very similar to
the original PIE input.
The third set involving the PIE aspirated voiced stops */bh, dh, gh/
represents a more complex development. : In Germanic the aspi-
rated stops are reconstructed as going through an intermediate stage of
voiced fricatives /β, ð, γ/; they result ultimately in voiced stops /b, d,
g/ in West Germanic/Old English. PIE */bh, dh, gh/ first get devoiced
to /ph, th, kh/ in word-initial position, which is the result of the shift
we find attested in Greek. In the Italic branch, as attested in Latin, the
aspirated voiceless stops /ph, th, kh/ become voiceless fricatives /f, f,
h/. Of special note for the shift in Latin is the merger of the labial and
the dental, where the place difference is neutralised in favour of the
labial.4
Figure 3.3 illustrates the First Consonant Shift with native English
words and cognates borrowed from Latin and Greek. 5
4
The confusability of the fricatives /f/ and /θ/ and their merger is perceptually moti-
vated. A parallel merger of /θ/ and /f/ is attested in some English dialects. Though
‘generally associated with Cockney and London pronunciation . . . it is in fact charac-
teristic not only of the Home Counties as a whole but of areas further afield. It can,
for example, be heard in Leeds’ (Wakelin 1972: 98). Wells (1982, II: 329) illustrates
the variability of [f] ~ [θ] in Cockney with a pun: ‘Advertisements for beer award the
brand in question “thirst prize”.’ :
5
The spelling for /k/ in Latin is <c>: as in cor ‘heart’, centum ‘hundred’, celer ‘swift’,
culmen ‘summit’, and so on.
The correspondences shown in Figure 3.3 represent the results of

a chain-shift, an important type of change in historical phonology. A
chain-shift differs from an ordinary ‘shift’ of a single sound, for example
// ‘shifting’ to /w/ in whale, white, where ‘shifting’ is simply a synonym
for ‘merging with’; no other consonants are involved. A chain-shift, on
the other hand, entails the maintenance of some of the original contrasts
within a related set of sounds, known as natural classes. The three sets
in the PIE column – voiceless stops, voiced stops and voiced aspirated
stops – are natural classes because they share manner of articulation
and voicing; the contrast among the members of each set is based on
different places of articulation. Similarly, the output sets of consonants
in the English column are natural classes; they share manner of articula-
tion and voicing and their members maintain the place of articulation
contrasts of the input sets.
The order in which the sets shifted is not certain, but as in other
cases of chain-shifting, for example the English Vowel Shift (see 8.2.2),
one can posit more than a single starting point; the end results can be
reached via more than one possible route. If the voiceless stops changed
their articulation first, creating a full set of fricatives – PIE is recon-
structed as being rather impoverished in that area, having only [s]- and
[h]-like sounds – then the /p, t, k/ slots in the system would have been
vacated. This would trigger the devoicing of /b, d, g/; it is well known
that consonantal systems typically disallow voiced obstruents if the cor-
responding voiceless obstruents are missing. Vacating the voiced /b, d,
g/ slots in its turn makes possible the shifting, or loss of breathy-voice,
of the original voiced aspirated stops. This scenario can be described as
a drag chain-shift.
The avoidance of perceptually undesirable mergers of previously
distinctive entities allows an alternative direction for the shift. In that
reconstruction, the voiced aspirated stops */bh, dh, gh/ are the initial
focus of change since they are universally ‘marked’, that is, they are
rare across the systems of the world’s languages.6 If the voiced aspirated
stops */bh, dh, gh/, shifted to /β, ð, γ/ and then to /b, d, g/ first, the
addition of new items with /b, d, g/ would motivate the devoicing of
the pre-existing PIE /b, d, g/ to maintain earlier contrasts. One way in
which merger can be avoided is by shifting the original /b, d, g/ to /p,
t, k/. Further, one way of avoiding homophony would be for the ‘new’
6
Informally, markedness correlates with the strong tendency across languages for some
elements to be avoided or to become ‘unmarked’. Markedness is part of the human lin-
guistic competence. In the consonantal system, markedness is associated with voicing,
coronality, aspiration and nasalisation.
/p, t, k/ set to be differentiated by reanalysis of aspiration as distinc-

tive followed by a change of the manner of articulation, where one
enabling factor would be the fact that /f/ and /θ/ were unfilled slots in
the PIE system of fricatives. This alternative sequence of events would
constitute a push chain-shift. The documentation for either scenario is
lacking, however, and even in cases where we know much more about
the intermediate stages, as with the English Vowel Shift, push chains
and drag chains remain controversial. :
In the nineteenth century the Shift was taken as a paradigm case of
the regularity of sound change, yet it does not apply in all environments.
The most common and phonetically well-defined exception to the shift
is when */p, t, k/ follow another voiceless obstruent in reconstructed IE.
Thus, /gh/ in PIE ghosti- ‘guest, stranger’ is shifted as expected, while
/t/ remains a stop after /s/. (2a) shows more examples of the regular
operation of the shift from IE to OE, while (2b) illustrates the ‘blocking’
of the shift by the preceding sibilant fricative /s-/.
(2) IE voiceless stops in Germanic:
(a) Regular shifting (b) No shifting
*pleu-, OE flowan ‘to flow’ *spe-, OE spēdan ‘to prosper, speed’
*tā-, OE ðāwian ‘to thaw’ *sta-, OE standan ‘to stand’
*kerd-, OE heorte ‘heart’ *skei-, ON skífa, PDE skive ‘to cut’
The rationale behind the preservation of /p, t, k/ in (2b) is easy to
understand: the implementation of the shift would result in two adjacent
voiceless fricatives, /sf/, /sθ/, /hθ/; such clusters require additional
articulatory effort and their acoustic closeness would make it hard
to perceive the difference between them. : Since /p, t, k/ remain
unchanged in Greek and Latin, loans which contain /sp, st, sk/ are
the same in English and the borrowed item; compare IE *hesti ‘(s)he
is’, Lat. est, Goth. ist; IE *sta- ‘stand’, English stool, stead, and the loans
stance, static, status, statistics; IE skabh ‘to scratch’, OE (ON) sceabb ‘scab’,
and the loans scabies (c. 1400) ‘scabies, itch’, scabrous (1657); IE *spe- ‘to
prosper’, OE sped ‘success, (God)speed’, later loans prosper (c. 1350),
despair (1325).
After the operation of the First Consonant Shift, a prosodically
motivated set of changes resulted in correspondences that look like
apparent exceptions to the shift. The key to the changes in question is
the position of the stress in the original word. It is generally held that
in PIE the accent could fall on any syllable in the word (see 3.4.3). In
Proto-Germanic the voiceless fricatives /f, θ, h/ resulting from the First
Consonant Shift were voiced to /β, ð, γ/ intervocalically when the IE
stress did not fall on the immediately preceding syllable; /β, ð/ merged
with the corresponding voiced stops (/b, d/) in OE. This regularity in
the phonology of early Germanic was explicated by Karl Verner in
1876 and is known in the literature as Verner’s Law. :
(3) Prosodically triggered post-Grimm’s Law fricative voicing in PrG
(Verner’s Law):
(a) Skt bhrátar, OE brōðor ‘brother’ Lat. frater
(b) Skt pitár-, OE fæder ‘father’7 Lat. pater
In (3a) the Shift operates following the model in (1): /-t-/ becomes
/-θ-/, which has a voiced allophone /ð/ between vowels (see 4.4). In
(3b) the intervocalic /-t-/ follows an unstressed syllable, where the
reconstructed chain of events is /t/ > /θ/ (Grimm’s Law) > /ð/ > /d/
(Verner’s Law).
The changes covered by Verner’s Law are shared by all dialects of
Old Germanic except Gothic. Chronologically, the voicing must have
preceded the shift of stress to the first root syllable in Germanic (see
3.4.3). The effects of fricative voicing after an unstressed syllable are
recognisable in OE in the strong verb paradigm of class 2 weak verbs,
where the past plural and past participle show the voicing; compare
OE flēon ‘to flee’ < *flēohan ‘flee’ with intervocalic loss of /-h-/ in the
initially stressed infinitive, vs flogen (past part.) ‘flown’; OE sēoðan ‘to
seethe’ – past participle sodden.
Under the same prosodic circumstances, original /z/, an allophone
of /s/ between vowels, became /r/, possibly passing through a palatal
fricative stage [ř], like Czech /ř/, as in Dvořák. The transition from /s/
to /r/ is known as rhotacism, familiar outside Germanic from paradig-
matic alternations in Latin, for example os ‘mouth’, gen. sg. oris ‘of the
mouth’, opus ‘work’, pl. opera. In Germanic some pairs of cognates also
show the effect of rhotacism. Disregarding vowel changes, the PrG root
*rīs- corresponds both to OE rīsan ‘to rise’ and to OE rœ̄ran ‘to rear’; the
OE form lār ‘lore’ corresponds to PrG *laiz- ‘to teach’; the /z/ ~ /r/
alternation in the past tense forms of the verb to be – sg. was - pl. were –
and the present tense and the adjectival participle of the verb lose-(for)
lorn from earlier (for)loren, are pairs which illustrate historical rhotacism.
An interesting parallel stress-dependent voicing is the so-called
‘Verner’s Law in English’, a sixteenth-century tendency to voice frica-
tives after unstressed syllables, manifested in the voicing of /s/ to /z/
in loanwords, mostly, but not exclusively after prefixes – resign, trans-
act, example, Alexander – and even against the spelling in possess, dessert,
dissolve.
7
For the change of [-d-] to [-ð-] in father see 5.2.3, 5.5.2.
Within Germanic, the regular results of the First Consonant Shift

are still in evidence in English, Dutch, other Low German languages
and the Scandinavian languages. The shift took place within a restricted
span of time prior or up to the beginning of the Christian era. This
entails that already in OE speakers and writers familiar with Latin
would encounter pairs of cognates where the borrowed item and the
Anglo-Saxon word exhibit the relevant correspondences, as the follow-
ing pairs show; they did not, of course, know about Indo-European or
Grimm’s Law.
(4) Cognates showing Grimm’s Law in OE:8
tripartita istoria, þæt is, þryfeald ‘threefold’ (gereccednyss.)
pater noster qui es in celis. þæt is ure fæder ‘father’ (þe eart on
heofenum.)
in corde suo; þæt is on Englisc . . . on his agenre heortan ‘heart’
For a PDE speaker, the way in which the obstruents in words of
shared Indo-European origin appear today is a good chronological
indicator: if a word existed in Germanic prior to the development of
its separate dialects, it shows the effects of the First Consonant Shift,
but if the word was borrowed after the split, it preserves the conso-
nants of the donor language, most commonly Latin or Greek. What
we know about the Shift makes it evident that in a pair such as know-
ignorant, the latter was borrowed after OE, and so were agnostic, prog-
nosis, diagnosis, ignore, while the related can, con, cunning, ken, kenning,
are Germanic.
We have now traced the two possible paths of entry of cognates into
PDE: one is the direct continuation of the PIE form into Germanic, Old,
Middle and Modern English. Then, independently, and after Grimm’s
Law had stopped operating, the language could borrow a word from
Greek, Latin and other IE languages. Such later borrowings are obvi-
ously not subject to any Germanic changes, though they may undergo
their own shifts specific to their branch, as we saw in Figure 3.3 with the
voiced aspirated stops in Greek and Latin. This dual route that the same
root has taken historically in entering English is responsible for the
correspondences in so many interesting pairs of the type FOOT-PODIUM,
TOOTH-DENTAL, HILL-CULMINATE, THREE-TRIPLE. The list in (5) includes
more examples of PIE cognates coexisting in PDE; in the second
column, the root consonants illustrate the results of the First Germanic
Consonant Shift. The dates in parentheses are the first attestations of the
words in English cited in the OED.
8
The examples are from the homiletic writings of Ælfric (c. 955–1020).
(5) Grimm’s Law and the chronology of cognates in English: 9, 10

PIE Germanic/ Later loan
English
*pod-/ped- foot, fetter podium (1743), pedal (1611)
*trei- three triple, v. (1375), triplex (1601)
*kl-/kel-/kol- hill culminate (1647), column (1440)
*dent-/dont- tooth dental (1594), (perio)dontal (1899)
*gal- call, clatter glasnost (1972)9
*bher- bear, bairn fertile (1460), euphoria (1684)
*dhwer- door forum (1460), thyroid (1693)10
*ghos-ti- guest host (1290), hospitality (1375)
Consonantal changes in cognates represent only one aspect in which the
Germanic languages differ from the Indo-European parent language.
The next section turns to some vowel correspondences between IE and
early Germanic.
3.4.2 Some IE vowel changes in Germanic

In the vowel system, the line of transmission from PIE to Germanic to
OE and later is interrupted or diverted much more often than with the
consonants. Therefore, it is only rarely that we can recognise continuity
in the vowels from PIE to PDE without the additional knowledge of the
processes covered in Chapters 6–8. For now, we will just summarise some
relevant changes from IE to Germanic, skipping over some complex IE
interactions between the class of laryngeal sounds and the vowels.
(6) Some vowel mergers from PIE to Germanic:
PIE Germanic Examples
a PIE *al- ‘grow’, Lat. aliment, Gmc *alda ‘old’
PrG a
o PIE *ghos-ti- ‘guest’, Lat. hostis,Goth. gasts
ā PIE *māter-,Lat. māter,OE mōdor ‘mother’
PrG ō
ō Lat. plōrare ‘weep’, OE flōd ‘flood’
9
From Slavic glas ‘voice’; glasnost is ‘giving something public voicing and transparency’.
10
Lat. foras, foris ‘out of doors’; Gk thura ‘door’, thureos ‘shield’; hence the seventeenth-
century formation thyroid ‘door/shield-shaped’.
i PIE *tit-/kit- ‘tickle’, Lat. titillate, ME kittle
PrG i
e Lat. ventus, OE wind ‘wind’, Lat. sedeo,Gmc*sitjan ‘sit’11
11
This selective list of changes may be quite mystifying at this early

stage. It is intended primarily as a bridge between the unfamiliar PIE
forms and the cognate forms that one will see cited in an etymological
dictionary.
A final set of segmental changes that may help one recognise some
etymological connections refers to the set of syllabic sonorants in PIE.
(7) The syllabic sonorants of Indo-European:12
r → <ur> PIE mr -ti-, Lat. mortal, Gmc mur-þra ‘murder’
˚ ˚
l → <ul> PIE wl kwo-, Lat. lupus, OE wulf ‘wolf’
˚ ˚
m → <um> PIE gwm-, Lat. venire, OE cuman ‘to come’
˚ ˚
n → <un> PIE *nter, Lat. inter, OE under ‘under’
˚ ˚
3.4.3 Early prosodic changes: stress and syllable weight in

Germanic
Another important phonological difference between Indo-European
and Germanic is observed in the stress system of the Germanic group.
PIE is reconstructed as having a free accentual system, where accent was
marked by high pitch. In principle, PIE accent could be placed on any
syllable of the word. An IE noun like *pətér- ‘father’ had accent on the
second syllable, while the noun *bhrāter- ‘brother’ was accented on the
first syllable.
The prosodic properties of PIE are still debated. What is agreed
upon with regard to Germanic is that all early dialects of that branch
developed a stress-based system of marking prominence. Moreover,
11
The raising of IE /e/ to /i/ occurs before a covered, and occasionally, before a single
nasal, and before /j, i/ in the following syllable, which was commonly lost in OE.
12
The syllabic sonorants of PIE are represented here with an under-ring following the
common practice by Indo-Europeanists, for example Watkins (1985). Elsewhere in
this book we follow the IPA practice marking syllabicity by a subscript vertical stroke
under the corresponding letters: r, l, m
, n . Sonorants functioning as syllabic peaks can
be treated as equivalent to the more familiar schwa /-ə-/ followed by a sonorant –
/-ər, -əl, -əm, -ən/ – as in PDE mutter, kettle, bottom, weapon.
Germanic stress became fixed on the first syllable of the root. : The
establishment of the innovative system has to be dated later than the
operation of Verner’s Law, because the latter requires IE-type accen-
tuation. One likely approximate date for the fixing of Germanic stress
on the first syllable is about 500 BC (Salmons 1992: 162–4).
Initial prominence in native roots continues to be a defining prosodic
feature of English to this day. This prosodic characteristic present
throughout the history of English is known as the Germanic Stress
Rule (GSR). The GSR refers to the default root-initial stress in non-
monosyllabic words inherited from Old English, the pattern of áfter,
hállow, blóssom, becóme. The preference for marking the left edge of nouns
in English persists and it can affect loanwords whose first syllable was
originally a prefix; such are the synchronically monomorphemic nouns:
cóntrast, conflict, éscort, óbject, présent, próduce, súrvey. :
Primary prominence on the first syllable of the root entails that syl-
lables following the root in non-derived words were unstressed. Already
by c. 100 BC fully unstressed syllables were undergoing reduction, that is,
the vowels and consonants in these syllables were losing their distinc-
tive properties. In English the process of vowel and consonant reduction
in unstressed syllables developed more fully than in other Germanic
languages. Reduction in unstressed syllables eventually resulted in
merger and loss of unstressed vowels from the final syllables of poly-
syllabic words, so that all short unstressed final vowels were dropped
(see 6.5.4, 7.6). To this day English maintains a full inventory of vowels
only in syllables bearing primary and secondary stress, while the set of
vowels allowed in fully unstressed syllables is limited (see 2.2).
3.4.4 Lengthening of final vowels in stressed monosyllables

In 2.3.3 we noted the absence of word-final stressed short, lax, non-
peripheral vowels [i ε æ υ ] in PDE; words such as *se with [-ε], *decrí
with [-i], *bru with [-υ] are ill-formed. This constraint on the shape
of the final stressed syllable in English can be traced back to OE and
even earlier. In all North-West Germanic languages the final vowels
of lexical monosyllables became uniformly long, that is, they were
lengthened if the original vowel was short. In some instances the long
vowel appears to be a compensation for the loss of PrG final /-z/: PrG
*/hwaz/ > OE hwā ‘who’, PrG /wiz/, OE wē ‘we’. Lengthening without
loss of a syllable coda is attested in OE nū ‘now’, Goth. nu, OE swā ‘so’,
Goth. swa; in this second set the process is evidently driven by the pref-
erence for co-occurrence of stress and syllable weight noted in 2.3.3.
Lengthening under stress is the inverse of the reduction of final vowels
in unstressed syllables: having a long vowel in a stressed monosyllabic

word increases the word’s perceptual salience.
The length of the vowel in some of the pronominal monosyllables in
OE – he ‘he’, me ‘me’, þu ‘thou’ – fluctuates depending on the position of
the pronoun in the phrase or clause: the long vowel goes with stressed
position, but when the pronoun was unstressed, the vowel was short.
The subsequent history of the vowels in these particular pronouns is
based on their pronunciation in stressed position; the vowels are long
and they go through the English Vowel Shift. The opposite process,
shortening, can occur if a monosyllabic pronoun ends in a consonant, as
in OE ūs ‘us’ (see 7.5.1.3). What is important for us here is, again, that the
pre-OE constraint on the distribution of word-final stressed vowels is
still active in PDE and renders monosyllables other than function words,
like bi *[bı], lo *[lɔ], se *[sε], ru *[rυ] unacceptable. Similarly, underived
words of the shape *ponrí with [-i], *sebrú with [-υ], could not occur in
OE and are still blocked from the PDE lexicon. The lengthening of
short final vowels in stressed monosyllables is thus the earliest stage in
a long process of regularisation of the weight of stressed syllables in the
language history. By eliminating lexical monosyllables ending in a short
vowel, it is arguably also the process that makes English subject to the
constraint on word-minimality, whereby a minimal lexical word should
be prosodically heavy – either -VV or -VC.
3.4.5 West Germanic (Consonant) Gemination (WGG)

Recall from 2.1.4 that PDE disallows the use of geminate consonants
within the boundaries of a single morpheme, and therefore long conso-
nants in PDE serve as a juncture signal in compounds and transparent
derivatives; compare pen-knife [-nn-], with Pennines [-n-], midday [-dd-]
with middle [-d], misspell [-ss-] with missive [-s]. In this respect PDE
differs from OE, where geminate consonants appeared freely in word-
medial position (see 4.1.2).
The geminate consonants of OE can be traced back to various sources.
Some consonantal geminates arose already in PrG from assimilation.
(8) Proto-Germanic assimilations resulting in geminates:
IE PrG OE Gloss
*wl ə-nā- *wullō wull ‘wool’
˚
*ster-la *sterron- steorra ‘star’
*kl -ni *hul-ni hyll ‘hill’
˚
By far the most important source of consonant gemination in Old English
is a process known as West Germanic (Consonant) Gemination
(WGG), which was under way, but not necessarily completed by the
early fifth century. The gemination is in evidence in the earliest OE
written records, around the first quarter of the eighth century; compare
Goth. sibja ‘amity’ with OE megsibbi ‘affection among relatives’ in the
Épinal Glossary (c. 700?).
The process can be described as doubling of single consonants other
than /r/ in the environment V̆ – /j, w, r, l/: a short vowel followed by
a single consonant (= light syllable), followed by an approximant or a
liquid. The process is most regular before /-j/. Since there was no gemi-
nation in Gothic and in Old Norse it was restricted to the velar stops /k,
g/, we can use actual recorded forms as evidence for the pre-gemination
status of the consonants. The examples in (9) show how gemination
works in the large group of OE verbs formed with the -jan suffix. :
(9) West Germanic Consonant Gemination:
PrG OE Gloss
ON sitja sittan ‘to sit’
Goth. saljan sellan ‘to give, to sell’
Goth. hlahjan hlæhhan (Angl.) ‘to laugh’
Goth. bidjan biddan ‘to bid’
The examples in (9) may appear straightforward, but they also over-
simplify a very complex set of phonological events. While /j/ triggered
gemination in all consonants except /r/, gemination occurred in other
environments, but in a more limited way. The other triggers of WGG,
/w, r, l/, operate only on voiceless stops, for example OE æppel ‘apple’ <
PrG *apl, OE bit(t)or ‘bitter’, ON bitr. Further complications arise from
the gemination of the velar stops, where the end results are not /-kk-/
and /-gg-/, but /tʃ/ written <-cc->, as in PrG *wak-jan, OE weccan ‘to
wake, watch’; WG *strakkjan, OE streccan ‘to stretch’; and /d/ written
<cg> as in Goth. bugjan, OE bycgan ‘to buy’. On the interpretation of
/tʃ/ and /d/ as singletons or geminates see 4.1.2.
As the examples in (9) show, one source of opacity in the distribu-
tion of geminates is that the attested results in OE show no trace of
the trigger /-j-/. Clearly, the loss of this approximant in word-medial
position also belongs to the pre-OE period, most likely after, possibly
even during, but not before the consonant gemination was well under
way. The absorption of /j/ cannot have happened much before the
seventh century. The loss of the approximant is directly related to syl-
lable structure. In 2.3 we discussed the principle of syllabification which
disallows word-medial syllable onset clusters that are not also attested
word-initially. Projecting this principle to the stages of West Germanic
Gemination suggests that the lack of word-initial /Cj-/ in Germanic
can be the structural reason for the loss of /j/ following WGG, thus bid.
jan > *bid.djan > *bid.dan. The intermediate stage *bid.djan could have
developed into *bid.di.an by vocalisation of the /j/, but instead, /j/ was
deleted and the form retained its original disyllabic structure. :
The preference for the disyllabic solution is related to a process
known as High Vowel Deletion (HVD). HVD is sensitive to the pro-
sodic shape of the word: /i/ deletion occurs only after a heavy syllable.
The foot structure of the stem of the reconstructed *bid.di.an would be
H(eavy)L(ight), triggering deletion of the high vowel /i/. Compare this
with e-ri-an ‘to plough’ where the first foot is LL and it exhibits no HVD.
Since /r/ resisted gemination, it would be syllable-initial; /rj-/ is an
impossible syllable onset, however, which results in the vocalisation of
/j/ to /i/. Thus in PrG, /V̆ r-/ stems + /j/ > /V̆ -ri-/. The option real-
ised in OE is /i/ or /e/ after /r/: PrG *arjan > OE erian ‘to plough’, PrG
harjaz > OE here ‘army’. As we will see in 6.3.1, Proto-Germanic medial
/j/ is one of the triggers of I-Mutation, another major co-articulatory
change which affected the vowels of the preceding syllables.
Another related question has to do with the orthographic representa-
tion of the geminates that OE inherited: what makes us confident that
the double consonant spellings in OE words such as æppel ‘apple’, sellan
‘to give, sell’, biddan ‘to bid’ reflect genuine ‘long’ consonants, compared
with PDE apple, sell, bidding, where the doubling is only orthographic? A
simplistic, but still credible response would be that the literate people
recording Old Saxon, Old English, Old Frisian, Old Low Franconian
and Old High German could not have had shared training, and therefore
the scribal practices of doubling consonants we find in all of these Old
Germanic dialects must reflect actual geminate pronunciations. Within
English, the argument for positing geminates in OE and subsequent loss
of gemination are quite subtle. As we will see in 7.5.2.1, the short vowels
before medial geminates, for example OE æppel ‘apple’, behave differ-
ently from the short vowels before singletons, for example OE apa/ape
‘ape’. Another important indicator of the reality of geminates in medial
position in OE relates to the patterns of heavy and light syllables found
in OE verse (see 10.2.1). Finally, the ‘solid’ arguments for the existence
of geminates in OE refer to consonantal length in word-medial position.
There were no word-initial geminates in Germanic. Whether word-final
geminates were actually realised in OE, or whether spellings such as pytt
‘pit’, hyll ‘hill’ were, or became, conventional, remains an open question.

4 Consonantal histories: Old
English
KIRK-CHURCH, DAY-DAWN, SAY-SAW-SAGA, SKIRT-SHIRT, SHRIFT-
SCRIPT, DISH-DISK-DISCUS, LOAF-LOAVES vs SERF-SERFS, ELF-ELFS-
ELVEN, BATH-BATHS-BATHE, BELIEF-BELIEFS-BELIEVE
Some consonantal correspondences in Indo-European non-Germanic

and Germanic cognates: plenary-full, cordial-hearty, and so on, were
covered in 3.4.1. This chapter turns to a different set of consonantal cor-
respondences in cognates, as in kirk-church, the inherited alternations
in day-dawn, say-saw-saga, and the fricative voicing in the paradigm of
a single stem, as in loaf-loaves, bath-bathe, grass-graze. Understanding the
origin of these correspondences takes us back to Old English, so we start
with an overview of the OE consonantal system (4.1). Some background
information on OE consonant spelling is provided in 4.2. Section 4.3
explores specific sound changes: palatalisation and affrication in (late)
OE involving the velars [k, g], relevant for the differences in cognates
such as kirk-church, shrift-script, dish-disk-discus, skirt-shirt. The voicing
alternations in three OE fricatives – [f ~ v], [s ~ z], [θ ~ ð] – and their
continuity and marginal survival in PDE in pairs such as loaf-loaves,
bath-bathe, grass-graze, are covered in 4.4.
4.1 The consonants of OE
4.1.1 Singletons
Figure 4.1 shows the set of consonant phonemes reconstructed for OE;
the classification matches the PDE chart in Figure 2.2.
Whenever there are two consonants in a single cell, as is the
case with all stops, the affricates and the velar fricatives, the one to
the right is voiced. Shaded cells indicate that the respective con-
sonants are reconstructed as functionally different from their PDE
counterparts, including the unfilled shaded cells corresponding to
the PDE voiceless labial fricative // and the velar nasal /ŋ/. The
parentheses around the affricates [tʃ] and [d] and the glottal fricative
[h] also signal differences from PDE (see 4.3 for the affricates and 5.1
for [h]).
74
CONSONANTAL HISTORIES : OLD ENGLISH 75

dental
STOPS p b t d k g
Obstruent
AFFRICATES (tʃ d)
FRICATIVES f θ s ʃ x γ (h)
NASALS m n
Sonorant
Central w r j
Figure 4.1 The consonantal phonemes of Late Old English (singletons)
The inventory in Figure 4.1 makes it obvious that the bulk of the dif-
ferences between OE and PDE affects the obstruent set and within that
set, the subset of fricatives. In terms of place of articulation, it is the
consonants in the palatal and the velar area that have been most prone
to change. The only difference in the inventory of the sonorant set is the
empty cell for the velar nasal /ŋ/ in OE. Phonetically, the nasal /n/ is
predictably velarised before a following velar in the same syllable; it is a
straightforward case of regressive assimilation. Although [ŋ] must have
existed as an allophonic realisation of /n/ before tautosyllabic velar
stops in OE in words such as cyning ‘king’ [-ŋg], lang ‘long’ [-ŋg], þank
‘thought’ [-ŋk], it was the loss of final /-g/ in Early Modern English (see
5.3.2) that resulted in the phonemicisation of /ŋ/, evidenced by PDE
contrasts such as fang-fan, thing-thin.
Within the obstruent set, the stops have been fairly stable. The voiced
velar /g/ is shaded; its addition as a contrastive element was an innova-
tion within the OE system. In Germanic and early OE the corresponding
consonant was a voiced velar fricative /γ/. By the middle of the tenth
century, as evidenced by the alliterative practice, initial /γ/ merged
with pre-existing /j/ before front vowels, but before back vowels and the
sonorants /r, l, n/ it was subject to fortition, that is, it became the voiced
velar stop /g/. : A singleton voiced velar fricative [γ] was preserved
until late OE only between back vowels (see further 4.2.1).
The palatal fricative /ʃ/ developed in late OE from the etymological
sequence /s + k/, usually spelled <sc> in OE, as in sceadu ‘shade’, scearp
‘sharp’, flesc ‘flesh’ (see 4.2). The coarticulation of vowels with velars in OE
and the origin and status of the affricates /tʃ/ and /d/ are discussed in 4.3.
The shaded empty cell for the voiceless labiovelar fricative // in

Figure 4.1 highlights the absence of contrastive // in OE, while some
PDE varieties have different onsets in whale-wail, whet-wet (see 2.1.2,
5.1.3). Historically, // goes back to an OE bi-segmental cluster /hw/,
spelled <hw>, for example <hwæl> ‘whale’, <hwīt> ‘white’.1 The evi-
dence that the onset in such words was identified as a velar/glottal, and
not a labial, comes from verse, where <hw> alliterates regularly on [h-].
(1) OE <hw> alliteration:
Ac se hwita helm / hafelan werede Beowulf 1448
‘but the white helm / head guarded’
hea hornscipe, / ofer hwæles eðel Andreas 274
‘high horned (beaked) ship, / over the whale’s dominion’
Hwalas ðec herigað, / and heofonfugolas Daniel 386
‘whales hear you, / and heavenly birds’2
The [h-] in this position must, however, have been weak, close to
marking just aspiration, which allows an allophonic interpretation of
<hw-> as []. Already by late OE, we find the first spellings of <hw>
with just <w> () and alliterations of <hw> with <w>-initial words.
(2) Late Old English alliteration of <hw->:
þa hwile þe hi wæpna / wealdan moston3 Maldon 83
‘as long as they their weapons / could wield’
and he þar wunode / ða hwile þe he lyfode Death of Alfred4 21
‘and he lived there / for as long as he was alive’
The variation between // ~ /w/ has been in the language for the
whole of the second millennium, and the situation continues to be
unresolved (for more on this see 5.1.3; for the history of the fricatives
/f, θ, s/ see 4.4).
4.1.2 Geminates
Figure 4.1 includes only the single consonants of OE. Most of these
consonants could also appear as geminates arising both from WGG
1
The letter <w-> is a modern convention; the Anglo-Saxon scribes used a runic letter
known as wynn. Its manuscript form in Old and early Middle English is þ, p.
2
Here and throughout, for the full names of the Old English texts, their abbreviations,
for named authors, and bibliographical information see <http://tapor.library.uto-
ronto.ca/doe/dict/bibl/index.html> (last accessed 27 May 2013). For the ME authors
and titles see <http://quod.lib.umich.edu/h/hyperbib/> (last accessed 27 May 2013).
3
Also Maldon 272.
4
The Death of Alfred is dated 1036.
(see 3.4.5) and from various assimilations. The only consonants that did
not appear as geminates in OE were the approximants /w/ and /j/. :
In addition, in words in which the velars /k/ and /γ/ were subject to
WGG, the result was not *[kk] and *[γγ] but [tʃ], spelled <cc>, and
[d], spelled <cg>, for example WG *jukkjan > OE gyccan ‘itch’, PrG
*brugjō- > OE brycg ‘bridge’.
Phonetically, the palatal affricates [tʃ, d] are combinations of stops
and fricatives, not unlike PDE /tʃ/, /d/ (see 2.1.3). In view of the
origin of these affricates, however, it is of relevance to establish when the
original bi-segmental clusters started to function phonemically as sin-
gletons, as they clearly do today.5 There is no direct evidence on which
such a reconstruction can be firmly based, but some indirect evidence is
supplied by later changes in open syllables (see 7.5.2.1): the stressed syl-
lables of words with medial geminates developing into affricates resisted
lengthening, for example OE (ge)mæcca ‘spouse, match’; compare with OE
wacian ‘to wake’. Also, in OE verse words with medial affricates appear to
resist resolution, which requires that the stressed syllable should be light
(see 10.2.1). Such phonological behaviour can be interpreted in two ways:
the affricates were either still perceived as ‘composite’ and rendered the
syllable to the left heavy, or they were treated as ambisyllabic singletons,
again rendering the stressed syllable heavy. The latter interpretation
is of considerable interest since it provides historical evidence for the
possibility of variable contribution of consonants to syllable weight – a
question that has not been explored for the history of English. :
As noted in 2.1.4, 3.4.5, today long consonants occur only when two
identical consonants are positioned back to back across a morphological
boundary, as in pace-setter, set terms, grim monster, unneeded, dissatisfied,
rivalless. Since the morphological compositionality of such words or
phrases is transparent, the two consonants are independently identifi-
able, while ‘real’ geminates are fully cohesive, meaning that they cannot
be separated by an epenthetic vowel or otherwise interrupted. In the
morphologically created long consonants, only one of the consonants
can go through a phonological process, while real geminates behave in
a unitary way, in spite of the fact that the closure duration for geminate
stops in languages that have consonantal length contrast is up to three
times longer than for the singleton stops.6
The reconstruction of consonantal length in OE relies on com-
parative evidence, orthography, minimal pairs and the behaviour of
geminates in various phonological processes.
5
See Cruttenden (2008: 180–4). :
6
See Ladefoged and Maddieson (1996: 92).
Recall from 3.4.5 that some Proto-Germanic assimilations resulted

in long consonants, for example IE *ster-la > PrG *sterron- > OE steorra
‘star’. West Germanic (Consonant) Gemination (WGG) is another
source of inherited geminates, thus Goth. saljan, OE sellan ‘to give,
to sell’. These well-studied processes are shared by other dialects of
Germanic, making the reconstruction of geminates in OE more plau-
sible. As for orthography, we can appeal again to the logic of widely
shared assumptions: OE spelling was closer to being phonemic than the
etymologically based spelling of PDE. This increases the likelihood that
the pronunciation of OE æppel ‘apple’ was with /-pp-/, sellan ‘give’ with
/-ll-/, cyssan ‘kiss’ with /-ss-/, and so on, while in PDE apple, sell, kiss,
and so on, have the same consonants as ape, seal, rice. In isolation, spell-
ing might not be a strong indication of consonantal length even in OE,
yet consistency of double-consonant spellings in a much freer orthog-
raphy also works in favour of positing consonantal length: of well over
a thousand <bb> spellings for abbod ‘abbot’, there is only one attestation
of <abud> in the DOE.
Turning to the test of minimal pairs, we find a small number of items
where the presence of a stem-internal geminate after a short vowel is
contrastive.
(3) Singletons and geminates in OE (minimal pairs):
bitela ‘beetle’ bitter ‘bitter’
bite ‘bit, morsel, cut’ bitte ‘bucket’
hopian ‘to hope’ hoppian ‘to hop’
manu ‘mane’ manna ‘man’
sæp(e) ‘sap’ sæppe ‘spruce, fir’
tela ‘well, thoroughly’ tellan ‘(re)count’
Since WGG occurred only after short vowels, the most common co-
occurrence was of a short vowel followed by a geminate (-V̆ C1C1-),
though geminates due to assimilation, usually across a morpheme
boundary, could arise also after long vowels, for example cyˉþan ‘to make
known’ > past sg. cyˉþ-de ~ cyˉdde; lœ̄dan ‘to lead > 3rd sg. lœ̄d-þ > lœ̄dt >
lœ̄ t(t) ~ lœ̄d; mētan ‘to meet’ > 3rd sg. mēt-þ > met(t). Such morphologi-
cally conditioned assimilations are the source of the most notable set of
‘new’ geminates in OE, one that has had a lasting effect on the structure
of some weak verbs in English. The long consonant in the past tense of
weak dental-final verbs with a long vowel in the stem – lœ̄dan ‘to lead’ >
past sg. lœ̄dde ‘led’; mētan ‘to meet’ > mētte ‘met’; also blēdde ‘bled’, brēdde
‘bred’; fēdde ‘fed’, spēdde ‘sped’ – triggers shortening of the stem vowel,
creating a new pattern of ‘irregular’ past-tense formation for originally
‘regular’ verbs (see 7.5.1).
A distributional restriction on OE geminates noted above was the

absence of */ww/ or */jj/, a typologically predictable situation, because
of the closeness of these approximants to vowels. WGG of the voiced
velar fricative /γ/ resulted in OE [d]. A medial /-gg-/ did occur, albeit
rarely, as in OE frogga ‘frog’, though its source is unclear.7 The rarity of
/-gg-/ correlates well with the aerodynamics of voicing, combined with
the additional factor of difficulties associated with the back articulation
of [g]. This results in a scale of difficulty, where [gg] is harder than [dd]
and [dd] is harder than [bb].8
The voiced fricatives [v, ð, z] were allophones of /f, θ, s/ between
voiced sounds in OE, from which it follows that voiced fricative gemi-
nates are also precluded. Phonetically, geminate fricatives are more
‘effortful’ than geminate stops.9 Hogg (1992a: 32) remarks on the rarity
of /-ff-/, describing it as a ‘marginal member of the normal phonologi-
cal inventory of OE’. The phonological significance of this fact is not
obvious; since other fricative geminates are common, it could simply be
an accidental gap in the lexicon.
There were no word-initial geminates in Germanic, or Indo-
European. This is not surprising: word-initial geminates in particular
are typologically rare, though initial long stops are attested in some
non-Indo-European languages.10
Another sequential distributional restriction on the occurrence of
geminates in OE was that they could occur only post-vocalically within
the confines of a single word, where ‘single word’ is defined to include
inflectional suffixes, that is the environment (-)VC1C1-. Thus we find
nǣddre ‘adder’, winneð ‘strives’, libban ‘to live’, þridda ‘third’. Again, the
restriction is typologically supported: geminates are most common
intervocalically and least common when adjacent to a consonant. :
This restriction does not apply in compounds, where gemination
occurs freely at the juncture of two roots, as in beorhhliþ ‘city-slope’,
gærsstapa ‘grasshopper’, handdæda ‘perpetrator’, feorhhus ‘body’, woruld-
dream ‘earthly joy’. Similarly, geminates are sequentially unrestricted
at the prefix–root junction, as in ymbboren ‘surrounded’, þurhhefig ‘very
heavy’. As in PDE, the long [-bb-] and [-hh-] in these words serve as a
morphological boundary signal.
7
The OED tells us that in OE ‘the ending -gga occurs in several other names of animals:
cf. stagga, docga, wicga. It is possible that frogga may owe its form to the analogy of other
animal names with this termination.’ Other items in this set of words of uncertain
etymology are pig and hog.
8
See Hayes and Steriade (2004: 7–12).
9
Kirchner (2004).
10
Ladefoged and Maddieson (1996: 91–5).
The strongest argument for positing phonemic word-internal gemi-

nates in OE comes from the different treatment of singletons and
geminates in the verse, and their different effect on the preceding vowel:
singletons provide the proper context for lengthening, while geminates
cause vowel shortening or block lengthening (see 7.5.1.1, 7.5.2.1).
In spite of the minimal pairs in (3) and the occasional appearance of
long consonants after a long vowel, it is evident that already in OE the
contrastive function of the geminates was limited both by the nature
of the consonant and by position. : By the tenth century, geminates
became unstable in unstressed medial syllables, for example gyldenne
‘golden’ with earlier [-nn-] > gyldene with [-n]. The functionality of
geminates in the present tense verbal inflectional system, the difference
between, for example, hlyhhe ‘I laugh’ and hlihe ‘laugh!’, was no longer
maintained, as the orthographic records show. Consonantal length is
unstable at domain-edges; there were no word-initial geminates, and by
the eleventh century the spelling of word-final geminates indicates that
they were also simplified, so words like OE bed(d) ‘bed’, sib(b) ‘relation-
ship’, were pronounced as in PDE. Nevertheless, the full simplification
of stem-internal geminates is an ME development.
Degemination in ME was a function of multiple factors. The set of
words in which medial geminates were phonologically contrastive with
the corresponding singleton was quite limited already in OE. There
were no geminates following long vowels in ME, because original gemi-
nates triggered shortening of the preceding long vowel, as in OE mētte >
ME met(t)(e) ‘met’, blēdde ‘bled’, brēdde ‘bred’, and so on. Another factor,
not yet explored in the literature, is the possibility of language contact
in this process. Of the languages that speakers of ME could have been
exposed to, AN and Latin had no long consonants, and in British Celtic
voiceless geminate stops had merged with the singletons. :
Most importantly, the progressive loss of final unstressed vowels in all
types of words, and the syncope and loss of inflectional syllables, jeopard-
ised the status of the original geminates. To illustrate: when steorra ‘star’
lost its <-a>/[-ə], the original geminate would be in final position, and
therefore it would not be differentiated from a singleton (no geminates at
domain-edges). Similarly, when an inflection, for example <-eð>/[-əθ]
in winneð ‘strives’, is syncopated to [-θ], or <-est>/[-əst] in dippest ‘dipst’
becomes [-st], the resulting *[-nnθ] or *[-ppst] would be phonotactically
ill formed. Since the loss of final unstressed vowels in ME was apparently
more advanced in Northern ME than in the south, the full demise of
word-internal geminates is equally regionally bound. In the north, Orm’s
(late twelfth-century) orthographic innovation to use double consonants
to indicate that the preceding vowel was short is the first reliable test of
Letter Value Examples
[θ] ðrēo ‘three’, þīn ‘thine’, bæþ ‘bath’ (at word-edges and before C)
<ð>/<þ>
[ð] hwæþer ‘whether’, baþian ‘bathe’, hræðe ‘rather’ (between voiced sounds)
[k] bōc ‘book’, cræft ‘craft’, cweðan ‘say’ (adjacent to back vowels and C)
<c>
[kj/c]/[tʃ] ċild ‘child’, ċīese ‘cheese’, rīċe ‘kingdom’ (adjacent to front vowels)11
[j] ġiet ‘yet’, mæġden ‘maiden, dæġ ‘day’, reġn ‘rain’ (next to /front vowels)
<g> [g] grund ‘ground’, gast ‘ghost’, gnætt ‘gnat’ (initially before back V and sonorant)
[γ] dragan ‘draw’, lagu ‘law’ (between back vowels)
<-g> ~ <-h> [x] ~ [h]: mearg ~ mearh ‘marrow’, dāg ~ dāh ‘dough’ (finally after back V and
sonorant)
Digraphs
<cg> [d] brycġ ‘bridge’, secġ ‘sedge’
<sc> [ʃ] sċip, ‘ship’, sċulan ‘must’, sċieppan ‘create’
Figure 4.2 Some letter–sound correspondences for OE consonants
the loss of phonemic gemination in his dialect. The south lags behind,
but degemination there must have been very advanced or even fully
accomplished by the end of the fourteenth century. 11
4.2 Sound-to-spelling correspondences of the consonants in OE

Projecting the present into the past inevitably requires information on
the written shape of the words in the older language. Linking spelling
to pronunciation provides the philological foundation for figuring out
the intended pronunciation of OE. With some extra effort and time, one
can develop the skill to read edited OE texts aloud and appreciate the
‘sound’ of the older language. To complement the classification of the
OE consonants in Figure 4.1, Figure 4.2 lists some common correspond-
ences between consonant letters and letter combinations used in OE
11
The overdot on the letters <ċ> and <ġ> is not in the manuscripts. It is a common
editorial practice indicating palatalisation in OE scholarly and pedagogical materials.
This text will use the overdot only when the palatality of the consonant is of some
consequence.
and the sounds that they represent. Figure 4.2 does not include letters
that had the same value in OE as they do today: <b, p, d, t, f, m, n, r, l,
h-, w>. In the interest of simplicity, we eschew references to the actual
handwritten shapes of the letters – the reference here is to the letters
commonly used in modern editions of the surviving texts.
The first set of letters, <ð> (‘edh’) and <þ> (‘thorn’) were used by the
Anglo-Saxon scribes indiscriminately for both the voiced and the voice-
less dental fricative allophones, [ð] and [θ]. Both letters were gradually
replaced by the digraph <th> in ME, though thorn continued in general
use up to the fifteenth century, and it can be found in private papers and
in some printed books up until the seventeenth century.12
The letter <c> was the only commonly used letter for the voiceless
velar stop /k/ in Old English. The letter <k>, though used occasion-
ally in OE, was not fully adopted until after the Conquest when it was
popularised by Anglo-Norman scribes. The only possible pronuncia-
tions of <c> in classical OE were either [k], the realisation of <c> when
adjacent to back vowels or consonants, as in carian ‘to care’, cuman ‘to
come’, cnēo ‘knee’, clerc ‘clerk’, meoluc ‘milk’, or the palatalised variant
[kj] (= IPA [c]) when adjacent to etymologically front vowels (see 4.3),
where it can be the precursor of the late OE/early ME [tʃ], as in ċēosan
‘to choose’, dı̄ċ ‘ditch’, lı̄ċ ‘body’. The value [s] for the letter <c>, as
in PDE city, cell, race, mice, decision, is another innovation in ME due
to Anglo-Norman scribal practices. The spelling <cc> represents the
West Germanic gemination of /k/ where the end result was not /-kk-/
but [tʃ], as in PrG *wak-jan, OE weċċan ‘†wecche, watch’, WG *strakkjan,
OE streċċan ‘to stretch’ (see 3.4.5, 4.3).
4.2.1 The <g>’s of OE

The letter <Y,7>, known as yogh and used in some editions of OE texts,
descends from <g>, the insular form of <g>. For OE, the distinction
between <7> and <g> is, however, unnecessary, and <g> will be used
instead of yogh in citing OE forms in this book. Yogh began to alter-
nate with the Continental <g> in early ME. In the ME manuscripts
yogh was commonly replaced by <z> or <y>, but some forms of yogh
survived throughout ME. In fifteenth-century vocabularies the words
beginning with yogh are at the end of the alphabet. Caxton uses the
12
See Scragg (1974: 2). After 1400 <þ> ‘fell more and more out of use, and in some
scripts was represented only by the y-form in the compendia ye, yt or yat, yei, ym, yu = the,
that, they, them, thou, and the like, many of which continued to be extensively employed
in manuscript in the 17th and 18th centuries’ (OED).
symbol sparingly, chiefly before final <-t>. The use of <z> for the
palatal realisation [j] continued in Scotland as in the spelling of names
like Menzies [miŋiz] and Dalziel [di
εl].
The letter <g> stands for three etymologically distinct categories
in OE. First, it is used for the palatal approximant /j/ from PrG, as in
ġēar ‘year’, ġeoc ‘yoke’, ġeōmor ‘sad’, ġeong ‘young’, ġiest ‘yeast’. All of the
numerous inherited /j/-initial pronouns and function words in OE – ġe
‘and, either, or’, ġē ‘you’, ġea ‘yea’, ġeon ‘yon’, ġief ‘if’, ġiese ‘yes’, ġit ‘you
two’ – belong in that etymological group. Second, it was used for the
palatalised allophone [j] of the PrG voiced velar /γ/ < IE /gh/, as in
*ġeonian ‘yawn’, ġiefan ‘to give’, ġiernan ‘to yearn’. Third, word-finally
after back vowels and sonorants /γ/ was subject to devoicing, most
likely to the voiceless velar fricative [x].
As shown in Figure 4.2, pre-OE /γ/ developed variable realisations
depending on the adjacent segments; the forms are contextually deter-
mined. As with fricative voicing (see 4.4), such contextual constraints
are responsible for the different shape of morphemes within the same
paradigm. As a basic guide to reading aloud, here is how the realisation
of <g> in OE was conditioned by the phonological environment:
• Adjacency to and other front vowels results in palatal assimila-

tion, as in ġiernan ‘yearn’, hāliġdæġ ‘holiday’, ġearn ‘yarn’, mæġden
‘maiden’, maniġ ‘many’. The ‘new’ /j/ merged with the pre-existing
/j/. The reconstruction of the palatalisation to /j/ is based on the
subsequent history of the sound and on variant spellings ~ <ig>:
dæi, meiden, wæi. :
• Adjacency to a sonorant and a following back vowel (/a/, /o/, /u/)
results in /g/, thus Engle, growan ‘grow’, gadrian ‘gather’, Gota ‘Goth’.
As noted in 4.1, the addition of the voiced velar stop phoneme /g/
was an OE innovation after the middle of the tenth century. The rare
OE <gg> stands for geminate /gg/ (see 4.1.2): frogga ‘frog’, stagga
‘stag’. :
• The only unfamiliar value for the letter <g> in OE was the voiced
velar fricative [γ]. It occurs initially before c. 950, as in gāt ‘goat’, gōd
‘good’, grāfa ‘grove’, and medially between back vowels /a/, /o/ or
/u/: dragan ‘to draw’, ōga ‘terror, awe’, lagu ‘law’. In initial position
the fricative became a stop /g/ after c. 950. The normal development
of medial [γ] in late OE and ME is further lenition to the approxi-
mant [w], ME, for example OE [laγu] > ME [law(ə)] ‘law’, which
formed a diphthong with the preceding vowel (see 7.4). The pres-
ervation of medial [-g-] in later borrowings from Old Norse could
result in etymological doublets (see (4) below).
• In word-final position after back vowels or the sonorants /r-, l-/,

Germanic /γ/ was devoiced and realised as [x], thus WS spell-
ings <burg ~ burh ~ burgh> ‘-burgh, city’, <beorg ~ beorh> ‘hill,
barrow’, <plog ~ ploh> ‘plough’. This devoicing was part of a more
general avoidance of voiced fricatives in the coda of stressed mono-
syllabic words in OE, which in itself is a subset of a strong trend for
final consonant devoicing, including stops, characteristic of West
Germanic. :
4.3 Palatalisation and affrication of velars in OE

2.4.1 makes it evident that the OE inherited velars were particularly
sensitive to phonological context. Drawing a parallel with PDE, the
conditions for palatalisation were similar to the conditions for the
process known as ‘velar softening’, characteristic of paradigmatic alter-
nations between [k] ~ [s] and [g] ~ [d] in the Romance vocabulary
of PDE, as in critic-criticise, medic-medicine, fragment-fragile. The specific
palatalisation results were different in OE, however.
The contextual variability of PrG /γ/ in OE is shown in (4); the third
column lists later borrowings of the same Germanic etymons, mostly in
Old Norse loans.
(4) OE <g> in OE, ME, and ON:
Old English Middle English Loanword
dæġ ‘day’ dai, dei
dagian ‘dawn’13 dauen, dawen
dragan ‘to draw’ draw(en) drag (1440) < ON draga
dræġe ‘drag-net’ dray
haga ‘enclosure, hedge’ haw (now dial.) hag(g) (1470) < ON hagi
heġe ‘enclosure, fence’ hay (now dial.) (The) Hague (< Dutch)
sagu ‘saying, saw’ saw(e) saga (1709) < ON saga
sæġe ‘say’, v. imp. say
The different realisations of PrG /γ/ in OE allow some three-way
correspondences in modern English, as in draw-dray-drag, saw-say-saga.
The first item shows the development of OE [γ] between back vowels
(draw, saw), and the second item has palatal [j] in the context of front
vowels which may arise independently in the same stem (dray, say).
13
The form dawn may be from the past tense dagode ‘dawned’ with [γ]; compare OE
infinitive herġian ‘to harry’ - p.t. hergode, the latter base producing ‘harrow’. The failure
of <-i> in the verbal derivational affix -ian to trigger palatalisation is addressed in 6.3.
The palatalisation of the original Germanic velar stop [k] to the

voiceless palatal stop [c] (= [kj]) when adjacent to the high front vowel
[i] or to [j], was already under way in OE. Phonetically, it is a case of
assimilation, whereby a non-coronal segment becomes coronal. The
step following palatalisation was affrication to [tʃ]. The exact dating of
the affrication of [c] to [tʃ] is not reliably reflected in the orthography
and is hard to establish. The process of palatal assimilation must have
started with the allophonic realisation of [k] as [c] in typically weak
positions: in unstressed syllables and in coda position. The allophone
[c] was a complex segment simultaneously preserving the fronted
dorsal place of articulation of [k] (involving the back of the tongue), and
having also the feature coronal (involving the blade of the tongue). In
the onset of stressed syllables, affrication appears to have been delayed.
At least until the late tenth century, and even later, the alliterative
practice allows matching of [k-] to the palatalised allophone [c-], for
example ðe clene Cudberte / on cildhade (‘The clean Cuthbert / in child-
hood’, Durham 16).14 This treatment of <c-> by poets and their scribes
prompts the assumption that [tʃ] was not phonemic until after c. 1000.
The history of the adjacent front vowel is also important: palatalisa-
tion and affrication are carried through in the environment of adjacent
original front vowels, as in OE ċild ‘child’ < PrG *kilþo-; OE ċin ‘chin’ <
PrG kinnjo-; OE dīċ ‘ditch’ < PrG *dīk-; OE benċ ‘bench’ < PrG *banki-z.
If initial [k-] was positioned before ‘secondary’ front vowels derived
from PrG back vowels through I-Umlaut (see 6.3), the fronting of the
consonant remained allophonic, as in OE cyning ‘king’ < PrG *kuningaz;
cymeþ ‘arrives’ < PrG *kum-iþi. The development of [k] to [c] to [tʃ] was
also blocked by a following obstruent, thus spricþ ‘speaks’, ricsode ‘ruled’,
uictor ‘victor’. Affrication does not occur in the Old Norse cognates. This
produces paradigmatic alternations which have come down to PDE: OE
sp(r)æċ ‘speech’ vs sp(r)ecan ‘to speak’, where the words are clearly ety-
mologically related but their consonants have different histories with
respect to palatalisation. Two such pairs are shown in (5).
(5) Pronunciation of OE <c>:
(a) PrG *kara- (b) WG *kirika < Gk kūriakon
[k] [c]/[tʃ] [c]/[tʃ] [k]
OE caru OE ċeariġ OE ċiriċe ON kirkja
‘care’ ‘chary, careful’ ‘church’ ‘kirk’
14
Durham is dated c. 1100. A full account of the velar palatalisations in OE is found in
Minkova (2003: ch. 3).
In (5a) the vowel following PrG /k/ remains a back vowel in the OE noun
caru ‘care’, while in ċeariġ ‘chary’ the vowel undergoes early fronting to
*cæriġ, hence the resulting palatalisation of the initial velar. (5b) illustrates
the OE palatalised reflex of WG /k/ in ċiriċe ‘church’, while the ON form
of the word preserves the original velar. The PDE vocabulary shows
many such pairs. The presence or absence of palatalisation of an etymo-
logical [k] to [tʃ] is responsible for the different initial consonants in cold-
chill, kettle-chettel (dial.). Alternative pronunciations in a derivational set
or in OE and ON are also behind the [k] ~ [tʃ] alternation in the histories
of bench-bank, birch-birk, chest-kist (Scots and Northern), milk-milch, muckle
(dial.)-much, drink-drench, stink-stench, seek-beseech, wake-watch, wreak-
wretch, place-names in -wick vs -wich (Berrywick vs Greenwich), -caster
vs -chester, and surnames whose second element ends in OE -rīċ < rīċ(e)
‘power, rule’: Goodrick-Goodrich, Aldrick-Aldrich, Rickman-Richman.15
In the orthography, the affricate [tʃ] was typically represented as
<c>, or <cc> in the words with historical geminates, for example OE
gyċċan ‘itch’ < WG *jukkjan. Some <ch> spellings appear for the pala-
talised consonant already in late OE, for example <ælche> ‘each’. The
most common ME spellings for [tʃ] were <c, ch, cch>.
Yet another source of early [k-] ~ [tʃ-] alternations are found in
some lexical items independently borrowed from Anglo-Norman and
Old French, where AN preserved /k/ and OFr had /tʃ/: AN catch, Karl,
cattle vs OFr chase, Charles, chattel.16
As shown in Figure 4.2, the digraph <cg> stands for [d], as in OE
bryċg ‘bridge’ < PrG *brugjō-, OE eċg ‘edge’ < PrG *agjā-. The origin
of the affricate is gemination, palatalisation and affrication of WG /γ/
(see 3.4.5). Not just the early but also the later OE verse treats [d] as
bi-segmental, as in on þa briċge stōp ‘on the bridge stepped’ (Maldon 78b),
where the first syllable of briċge ‘bridge’ must be heavy, otherwise the
verse is defective. :
The digraph <sc> represents the West Germanic cluster [sk] in Old
English. Once again, dating the palatalisation is difficult. Some items
with etymological [sk] show metathesis, thus <ascian> ~ <acsian> ‘to
ask’ < PrG *aiskōjan; <fiscas> ~ <fiscas> ‘fishes’, Goth. fisks. In the verse
<sc> is treated as compositional, blocking resolution after short vowels,
thus asca þryþe ‘glory of ashes’ (Wanderer 99b); ofer fisces bæð ‘over the fish’s
15
Ælfric, the prominent Anglo-Saxon writer (c. 955–c. 1010), is now most commonly
pronounced with [-k], but the reversal to [k] from an earlier [tʃ] must have happened
in Middle or even Early Modern English, possibly under the influence of the Latin
form of the name, Ælfricus.
16
Differences in the consonantal systems of AN and OFr are also reflected in the shape
of PDE warranty, warden from AN vs guarantee, guardian from OFr.
bath’ (Andreas 293b). Post-vocalic [ʃ] continues to be treated as compo-

sitional in ME, where it appears only after short vowels in native words,
a restriction that does not apply to any singletons in the system. :
The palatalisation must have spread gradually, but by the end of OE
<sc> is assumed to stand for (conservative and Northern) [sc], alter-
nating with a fully palatalised [ʃ] in all environments. Since this was a
specifically OE development, the absence of palatalisation of [-k-] in
the cluster [sk-] in Scandinavian is the reason for etymological doublets
exemplified in (6).
(6) [*sk-] in Old English and Old Norse:
PrG LOE/EModE ON PDE
*scaf-/scab- sċeafian skabb- shave, scab
*skaljō- sċell, sċiel skāl ‘bowl’ shell, scale, skoal
*skipa sċip skipari ‘shipman’ ship, skipper (1390)
*skot sċ(e)ot skot shot, scot (-free)
*skraw- sċrēade (MDu) schrode shred, scrod (1841)
*skrub sċrybb (Norw.) skrubba shrub, scrub
*skuf- sċufan skuff shove, shuffle,
scuffle
*skurt- sċyrte skyrta shirt, skirt
*– sċīr *skir- shire, Skirlaugh
(YE)
A similar pairing of [ʃ] and [sk] is found in etymological cognates, one
of which existed in OE, and the other of which was a post-OE borrow-
ing from outside Germanic, thus fish-piscatorial, mesh-mask. Later loans
from Greek, and Latin loans where the <sc> is followed by a back
vowel or a consonant, preserve [sk], while palatalisation to [ʃ] occurs if
the cognate form was inherited from OE.
(7) [*sk-] in Old English and later loans:17
OE Later loan Source
sċēadan ‘to separate, schism (1382) Gk skhizein ‘to split’
shed’
schizo- ‘to split’ (1870)
sċrift ‘prescribed script (1375) Lat. scrīpt- ‘to write’
penalty, penance’ scripture (1300)
17
The history of [sk] in words borrowed via French varies since the cluster was simpli-
fied in Old French as in disciple, discipline, visceral, rescind, and so on. Both in French and
in English, direct reference to the original etymological [sk] can restore the cluster or
produce doublets; compare sceptre with [s-] and sceptic, skeptic with [sk-].
fisċ ‘fish’ piscary (1475) ‘fishing Lat. piscā-ria

right’
piscator (1674) ‘fisherman’
disċ ‘flat discus (1656) Lat. disc-us < Gk
round plate’
disc/disk (1715)
The digraphs <cg> and <sc> were abandoned after the end of OE. In
ME, the post-vocalic affricate [-d] was represented most commonly
by <gg>, <g> and <dg(e)> from French after the fifteenth century.
Word-initial [d-], a ME innovation based on the adoption of French
vocabulary, was represented by <g>, or <j-> ~ <i->. The letter <j>
was a variant of and was treated as such until the seventeenth
century, when the shape of was preferred for the vowel and <j> for
the consonant, not unlike the similar differentiation of the value of 
for the vowel and <v> for the consonant, from earlier <v> in initial
position, in medial and final position. :
Palatal [ʃ] had multiple representations in ME: <sc, ss, sh, sch> are
found in the earlier texts. Later ME spellings for [ʃ] include also <ch,
ssh, ssch, schch, schs, sshs>.
4.4 Morphophonemic alternations: fricative voicing. CLOSE enough

to CLOSE the door?
The consonantal alternations producing pairs such as draw-drag, church-
kirk, shirt-skirt, shrift-script, are based on differences in cognate forms
in OE and non-OE/borrowed words. A reference to divergent pho-
nological histories in different languages is not the only way in which
morphemic alternations arise. As is clear from the examples in (4) and
(5a), the difference between, say, OE caru ‘care’ with [k]- and OE ċearig
‘chary’ with [c] > [tʃ] is not a matter of different pedigree; in that pair the
palatalisation and affrication in chary occurs within the same language
and it is conditioned by the different vowels in the second syllable of the
words. In ċearig the adjectival ending -ig causes the fronting of the vowel
following the original [k-] (see 6.3); the secondary front vowel then trig-
gers palatalisation and affrication of [k-] > [c] > [tʃ]. Put differently, the
front vowel /i/ in the derivational morpheme -ig creates a phonological
context responsible for the different pronunciations of the same original
morpheme. OE car- in ‘care’ [k-] and OE ċear- in ‘char-y’ [c/tʃ] are
allomorphs of the same morpheme; the alternation itself, as noted in 2.5, is
known as allomorphy.
It is a common assumption in linguistic theory that we store the basic

forms of morphemes in the lexicon; they are our lexical entries. When
we speak, we use the lexical entries to produce words by assembling
roots, derivational and inflectional morphemes. Building words from
the base forms of these morphemes is the morphological component of
our grammar. Once the morphology of a word is in place, the phono-
logical rules and constraints of the language are activated and we come
up with a surface phonetic realisation of the segments in the word. The
final phonological shape of the word therefore reflects changes reflect-
ing the different morphological composition of words derived from the
same lexical entry.
A familiar example of allomorphy at work in PDE is provided by the
morpheme /-d/, spelled <-ed>, forming the past tense of regular, or
weak verbs; the final shape of that morpheme depends on the properties
of the final segment of the verbal stem. :
(8) PDE past tense allomorphy:
Stem ends in Past tense Realisation
Voiceless [t] iced, lacked, laughed
Voiced /-d/ [d] framed, tried, nagged
[-t, -d] [-əd]/[-d] added, rusted, lauded
A more restricted instance of paradigmatically induced phonological

change in PDE is the voicing of the final fricative [-f] of some singular
nouns: shelf, loaf, wife, to [-v] in their plural forms shelves, loaves, wives.
A similar voicing may accompany the derivation of verbs, as in belief-
believe, grief-grieve, proof-prove. Voiceless-voiced fricative pairs are found
also in house, n.-house, v.; grass-graze; glass-glaze, peace-appease; cloth-
clothe; bath-bathe. These are marginally productive morphophonemic
alternations, which may appear unmotivated to a modern speaker. The
alternation of the labial fricative [f] ~ [v] is somewhat familiar; the
alternation [θ-] ~ [ð-] less so; and [-s] ~ [-z] appears to be fossilised in
pairs such as close, adj.-close, v. In cases like this reference to the history
of the forms is quite illuminating: the current alternations are relics of a
robust Old English morphophonemic pattern of voicing of fricatives in
well-defined environments.
As noted in 4.1.1, one of the striking differences between PDE and
OE is the absence of contrastive voiced fricatives in the older language.
There were no minimal pairs in OE in which only the presence of a
voiced or a voiceless fricative was sufficient to signal a new meaning,
similar to fine-vine; sit-zit; mouth, n.-mouth, v. in PDE. This does not mean
that the Anglo-Saxons could not produce or hear the voicing contrast
in the pairs [f-v], [s-z], [θ-ð]; it means only that the voiced allophones
were not functional independently of the environment in which they
appeared. In linguistic terms, there were three fricative phonemes, /f,
θ, s/, in OE. The realisations [f-v], [θ-ð] and [s-z] were allophonic – they
appeared in mutually exclusive environments. Voiceless and voiceless
fricatives were in complementary distribution (see 2.1).
A legitimate question regarding allophones in complementary dis-
tribution is why the voiceless allophones in OE are assumed to be
more ‘basic’ than their voiced counterparts. There are two reasons for
selecting /f, θ, s/ as the unmarked state of the fricatives in OE. First,
the voiceless variants appear in most environments: word-initially,
word-finally and when adjacent to other obstruents (stops, affricates and
fricatives), while the voiced allophones have a much more restricted
distribution. Second, taking /f, θ, s/ as the unmarked case is in line with
the observation that voiced fricatives are more effortful from an articu-
latory point of view; they are also cross-linguistically less frequent.18
Other than that, the choice is not analytically significant; the crucial
point is that in Old English the voiced allophones appeared only when
they were flanked by voiced segments; otherwise the fricatives were
realised as voiceless.
The distribution of voiced and voiceless fricatives in OE is illus-
trated in (9), using angled brackets for spelling and square brackets for
pronunciation.
(9) The realisation <f, s, ð, þ> in OE:
<f> <s> <ð>/<þ>
[f] [v] [s] [z] [θ] [ð]
OE: wulf wulfas græs grasian bað baþian
‘wolf wolves’ ‘grass to graze’ ‘bath to bathe’
Orthographically, the representations for the labial and the sibilant

were <f> and <s>. The angular allograph of , the letter <v>, was
introduced only in ME. The letter <z> was used only rarely in OE to
represent /ts/. Somewhat surprisingly from a modern point of view, in
spite of the availability of two letters for the dental fricative, <ð> ‘edh’
and <þ> ‘thorn’ (see Figure 4.2), the OE scribes used them indiscrimi-
nately, so there is no systematic orthographic correspondence between
18
See Flemming (2004); Simon (2008) and references therein.
the letters and their phonetic values; <ð> does not have to represent
[ð]. :
The phonetic nature of the process is straightforwardly accounted
for by coarticulation: when the fricative is flanked by two voiced
sounds (vowels or sonorants), the configuration of the vocal cords
remains stable throughout and the input fricative consonant is partially
assimilated to its neighbours. Voiced allophones occur in the onset of
an unstressed syllable either stem-internally – OE ofer ‘over’, broðor
‘brother’, nosu ‘nose’, or across a stem and an inflectional boundary – OE
wif-a ‘of wives’, að-as ‘oaths’, hus-es ‘of the house’.
The voicing patterns show clearly the asymmetrical nature of affixes:
in OE only inflectional affixes can supply the appropriate voicing envi-
ronment. No voicing occurs if the fricative is the onset of the stressed
root syllable, so in OE befǽstan ‘to fasten’, biþéncan ‘bethink, recall’,
gesǽlig ‘happy, silly’, the fricatives are not voiced because be-, bi-, ge-
are prefixes. Derivational suffixes, both when they are fully unstressed
and when they carry non-primary stress, behave like roots in this
respect; thus the fricatives in the onset of OE -sum ‘-some’, -ful ‘-ful’,
-fæst ‘-fast’ -feald ‘-fold’ are not affected in OE wilsume ‘desirable’, synful
‘sinful’, twifeald ‘twofold’.19 There is no voicing in compound words at
the boundary of the two roots: OE tóðæ̀ ċe [-θ-] ‘toothache’, OE hláf-æ̀ ta
‘loaf-eater’, OE gærs-ierþ ‘grass-earth, land’. The lack of voicing in such
forms is a confirmation that OE treated the elements of compounds as
prosodically independent entities, a property also testable in the verse
(see 9.4, 10.2.1).
The phonemic split (see 2.5) of [f-v], [θ-ð] and [s-z] occurred in
Middle English. It resulted in six fricative phonemes in PDE, where
voicing or lack of it differentiates between ferry-very, fast-vast, life-
live, seal-zeal, sink-zinc, race-raise, thistle-this. The development of the
fricatives in ME is of considerable interest because it highlights the
importance and interaction of multiple factors: (a) the influence of
loanword phonology, (b) the effect of system-internal phonological
changes obscuring the evidence for previously existing complementary
distribution, (c) the role of prosody in the categorisation of the contrast,
and (d) the spread of independently occurring initial fricative voicing
from limited dialect areas (Kentish, possibly late OE West Saxon) to the
emerging supra-regional variety. It is instructive to look at these factors
in some more detail.
(a) Loanword phonology. The impact of loanword phonology is not
19
Voicing can occur in nouns derived from the verb base, where the fricative was voiced,
as in PDE teeth-ing, teeth-er, hous-ing, hous-er, shelv-ing, shelv-er, glaz-ing, glazier.
equally important for the three pairs of fricatives. Latin and OF/AN
loans appear to have been the main driving force behind the phone-
micisation of the [f]-[v] contrast to /f/ and /v/. The influence of such
loans is most clearly observed in word-initial position. Old English
had already borrowed some [v-] words, but these isolated items were
assimilated early to the native template of initial voicelessness, thus
OE fann ‘fan’ < Lat. vannu, OE fiddle < Lat. vidula, OE fers ‘verse’ < Lat.
versus. In ME, the influx of [v-] words was much more vigorous and
pervasive. It was the adoption of over 800 items with word-initial [v-]
after the eleventh century that disabled the earlier constraint on voiced
labial fricatives in that position. Already in LAEME, prior to c. 1325, we
find many contrasting minimal pairs.
(10) Early ME [v-]-[f-] minimal pairs:
Old French Middle English
vēle ‘veal’ fēle ‘many’ (OE fēla)
verrien ‘validate’ ferien ‘to transport’ (OE ferian)
veyn ‘vain’ fain ‘glad’ (OE fægen)
vile ‘vile’ file ‘file’ (OE fil)
The nativisation of /v/- initial words is further confirmed by the ability
of the loans to combine with native affixes, thus varīe ‘different’ < OFr
vairié, produces vārī-en (v.), variant-lī (adv.), vārī-inge (ger.), various-
nesse (n.), variāble-nesse (n.).
Word-medially, it is again the labial fricatives for which the bor-
rowed vocabulary contributed most significantly to the demise of the
OE pattern of voicing. If the ME loanword had a voiceless labial fricative
between voiced sounds, this would obscure the inherited coarticulation
pattern.
(11) Word-medial inter-voiced [-f-] in ME loanwords: :
Middle English Compare with
cofin (1330) < Lat. cophinus < Gk ‘coffin’ OE cofa ‘cove’
ofice (c. 1250) < Lat. officium < OFr ofice OE ofer ‘over’
sacrifice (1250) < Fr sacrifice OE drifan ‘to drive’
sulphur (12th c.) < AN sulf(e)re OE culfer ‘culver, dove’
wāfer (1212) < AN wafer ‘wafer’ OE wœ̄fre ‘wavering’
In word-final position the conditions for voicing in OE and OFr were
the same; if the labial fricative appeared in the coda, it was realised as
[-f], as in ME bref ‘brief’, chef ‘chief’, motif ‘motive’, serf ‘serf’.
Compared with the labial fricatives, the share of loanword phonol-
ogy on the history of the sibilants [s]-[z] is less critical. There are only
about thirty <z>-initial words in ME, many of them infrequent items,
for example zephyr, zeal, zone, zeugma. Latin did not have initial [z-], so
most of the borrowings in this set are from Greek. Since OFr intervo-
calic [-s-] was voiced to [-z-], and final [-z] was devoiced to [-s], there
was no difference in the native and loanword phonology with respect to
the sibilants.
The development of the dentals [θ]-[ð] into independent phonemes
in ME was not influenced by loanword phonology.
(b) System-internal phonological changes also obscure the evidence for the
earlier complementary distribution of voiced and voiceless fricatives in
Middle English. The changes most closely associated with the phone-
micisation of the voicing contrast for the fricatives were degemination
and the loss of final unstressed vowels.
Recall from 4.2.1 that OE geminate fricatives could only be voiceless,
[-ff-, -ss-, -θθ-]. Degemination started in OE in word-final position and
in unstressed medial syllables, but geminates persisted in the sequence
VC1C1V, for example ME blissen ‘to bless’, syþþen ‘since’, graffe ‘graft, twig’
< OFr. By the early thirteenth century, however, the vowels in final
unstressed syllables were subject to loss, especially in hiatus and in inflec-
tional endings, destroying the inter-voiced environment for the original
geminates which became word-final, where gemination was not sustaina-
ble (see 7.6). Thus <blisse> would have variants in [-issə] ~ [-isə] ~ [-is].
The gradual degemination of the sequence -VC1C1ə to -VC1 did not
occur in isolation. As schwa loss gets under way and spreads to more
environments, single fricatives in historically intervocalic position
(VCV), where the vowel to the right was [-ə], would have the voiced
variant appear word-finally, creating minimal pairs such as <blis(se)>
with [-s] vs <wise> with [-z]; <graffe> ‘graft’ with [-f]- vs <grave>
‘grave’ with [-v]. More abstractly, the voicing contrast between VC1C1ə
(voiceless) and VC1ə (voiced), where C1 = [f, s, θ], used to be signalled
by the geminate versus singleton realisation of the consonant, while
after the loss of [-ə] voicing became a property of the word’s lexical
form, not dependent on the phonetic context.20
In principle, the system-internal phonological factors should be con-
sidered equally important for all three pairs of consonants in final posi-
tion, though as noted in 4.1.2, word-medial /-ff-/ was rare in OE and
therefore [-ff] to [-f] degemination would not be a major driving force
behind the establishment of a word-final /f/-/v/ contrast. Geminate
[-θθ-] was also rare in OE. However, since [-ð] did not appear in
loanwords, all of the word-final [-ð] instances in PDE are directly
attributable to historical schwa loss: bathe, breathe, seethe, swathe, smooth,
20
For the stress-related voicing in dessert, dissolve, possess see 3.4.1.
loathe, clothe, mouth, soothe, teethe, wreathe, and so on. Indeed, the only
[-ð] items without the possibility of analogy from a verbal form are the
preposition with and the adjective lithe, both of which are variable.21
The /s/-/z/ contrast in final position was also reinforced by the loss
of final [-ə], for example lease ‘glean’ (now dialectal), with [-z] < OE
lesan, v. vs lease, v. ‘grant’ (1292) < AN lesser; phase, n. ‘Passover’ [-z] (OE)
vs face, n. [-s] (1300); nose [-z] < OE nosu vs gross [-s] < Fr. grosse. New
verbs with the suffix -ize (< Lat. -izāre, Fr. -iser) borrowed especially
freely after the sixteenth century consolidated the contrast in word-final
position. Interestingly, the sibilants continue to show some variability
in word-medial position; compare PDE usage, unison, gosling, greasy both
with [s] and with [z].22
(c) Prosody. The third factor in establishing a voicing contrast is pro-
sodic: fricatives were voiced in prosodically weak positions in ME. The
role of prosody is most prominent with the dentals. Initial voicing of the
dental fricatives – outside the dialect voicing discussed in (d) below –
is limited to function and pronominal words: the, this, these, those, that,
there, thus, thine, then, thy, thou, thence. All major-class <th-> words in
English are [θ-]-initial: thumb, thin, thrust, three, thatch, thunder (OE),
theme (Lat. < Gk), thyroid (Gk), thermos (trade name based on Gk), and
so on. We know that the voicing in these words was an innovation and
cannot be projected back to OE because of the alliterative practice of
the OE versifiers. The featural identity of voicing for the consonants in
OE alliterative verse was stringently observed. There are no instances
of alliteration of [p-]:[b-], [t-]:[d-] or [k-]:[g-]. Alliteration of the type
shown in (12) is therefore a fairly reliable indication that the poets were
alliterating on [θ-].
(12) OE alliteration of initial dental fricatives:
ðys dogor þu / geþyld hafa Beowulf 1395
‘This day thou / patience have’
þær ic, þeoden min, / þine leode Beowulf 209523
‘There I, chieftain mine, / thy people’
The voicing of [θ-] to [ð-] in the function and pronominal words must
have been gradual, with the definite article leading the way, since it
would be the least stressable item in the set. The scribal practice is not
21
The OED gives only [-ð] for the now infrequent adjective lithe, but Merriam Webster
Online Dictionary gives both. The letter-name edh, first recorded in 1846, is clearly an
echo of the sound it represents.
22
Also exclusive vs excluzive; the full form of Mrs with medial [-s] or [z] (see Ching 1996);
note also electrizity for electricity for some young speakers of AmE.
23
Similarly Beowulf 2131.
revealing, but we know that voicing is an option in the late fourteenth

century because it is attested by rhymes in Chaucer’s southern variety
of ME;24 the dating of the change in the north is not clear.
The voicing of [-s] to [-z] in the inflection <-es> and in prosodically
weak is, has, was, his must also have been under way during the fourteenth
century. The evidence for that is the use of <> to represent the French
letter <z> by the scribe of the Pearl manuscript; it is usually transcribed
as <z> in print: <cloudez> ‘clouds’, <hillez> ‘hills’ <rennez> ‘runs’,
<watz> ‘was’. Note, however, that if the inflectional vowel in <-es> was
lost earlier, <-s> would be word-final and voiceless in a monosyllabic
word, as in hence < hennes, twice < twies, similarly truce, thrice, pence.
Prosodically induced voicing of [-f] and [-θ] is also attested in func-
tion words, as in of [əv], with [wið], where the increase of sonority is
associated with the lower degree of effort and absence of stretching in
the lower laryngeal area in unstressed syllable codas. More specifically,
phonetic studies show that stretching the cartilages at the bottom of the
vocal cords is linked to voicelessness, while lack of stress correlates with
no stretching, as in of pronounced with [-v], or was, is with [-z]. :
(d) Dialectal developments account for the occurrence of initial fricative
voicing in Kentish, possibly late West Saxon in OE, and the south and
south-west Midlands in ME, where we find spellings <valle> for fall,
<vox> for fox, <zwyn> for swine, <zwo> for so. In these areas it was initial
[f-] and [s-] that had to be imported from the emerging supra-regional
variety. Only a few dialectal forms were eventually adopted by the stand-
ard: vane, vat, vixen. The area of initial fricative voicing has been shrink-
ing; it is no longer a feature of the dialects of Kent and Sussex, although
it is still attested in Somerset, also Devon, Dorset, parts of Cornwall,
Gloucestershire, Wiltshire and Hampshire (Trudgill 1999a: 30–1). The
only trace of this feature in the emigrant communities of North America
is recorded in Newfoundland and Labrador English: vish, vin, varket, vir
for fish, fin, forked, fir (see Kirwin 2001: 446–7). The long-term effect of the
historical initial fricative voicing in the southern dialects is thus mainly
local; it changed the distribution of voiced and voiceless fricatives by
limiting the latter to word-final and obstruent-adjacent positions, but it
did not contribute independently to the establishment of phonemically
contrastive fricatives in the standard language.25
24
The statement is based on infrequent rhymes such as soothe: to the ‘the truth: to thee’
(The Canon Yeoman’s Prologue 662–3), swithe: hy thee ‘swiftly: hasten thee’ (The Canon
Yeoman’s Tale 1294–5).
25
Another historical dialectal change, well attested in south-eastern dialects from the
middle of the fifteenth century, and considered characteristic of Cockney until the
end of the nineteenth century, is the interchange of /v-/ and /w-/ (see further Wyld
The extent to which fricative voicing survives as a morphologically

triggered change today: shelf, loaf, wife - shelves, loaves, wives, is hard to
determine. Arguably, for the labial alternation [f] ~ [v], morphological
conditioning is a recognisable pattern, both in noun-verb and especially
in singular-plural alternations, as in the examples in (13); as usual, the
dates in parentheses are first dates cited in the OED.
(13) Continuity of [f]:[v] alternations:
Noun Verb Singular Plural
belief (1175) believe (OE) hoof hoofs/hooves (OE)
grief (1225) grieve (1225) dwarf dwarfs/dwarves (OE)
proof (1225) prove (1175) elf elven, fem. and pl. (OE)
relief (1330) relieve (1375) turf turfs/turves (OE)26
serf (1483) serve (1303) scarf (1555) scarfs/scarves
In the noun-verb set in (13) the voicing context was present and pre-
served longer in the verbs: verbs had more vowel-initial inflections, and
the OF/AN loans were modelled after recognisable correspondences
that replicated the reif, n. - reave v., grass, n. - graze, v. pattern inherited
from Old English. The second set in (13) is the more common pattern
of voicing in the plural. Here belong also historical doublets such as sg.
cloth, pl. clothes (OE), cloths (after c. 1660); kerchief (< AN courchief c. 1300),
pl. kerchiefs ~ kerchieves; roof (OE hrōf), pl. roofs ~ rooves. The OED lists
the forms dwarves and elven as being brought back and popularised by
Tolkien; the mere fact that this can happen testifies to the continuing
‘reality’ of the voicing alternation in the modern language. :
The synchronic picture is further complicated by a slightly different
pattern: the pairing of members of the same derivational or inflectional
set, where the fricative in the first form is historically flanked by vowels,
the one on the right now ‘silent’, while in the other form it is followed
by an obstruent.
(14) Continuity of [-v]: [f + obstruent] alternations:
(be)reave (be)reft leave left
cleave cleft shrive shrift
five fifth sieve sift
give gift thrive thrift
heave heft weave weft, woof
1949: 210; Wakelin 1972: 95–6). This interchange is most famously recorded in Sam
Weller’s/Sam Veller’s speech in The Pickwick Papers; the ‘vulgarism’, as it is described
in Wyld, died out in the first part of the twentieth century.
26
The verb turve ‘to cover with turf’ is recorded for the fifteenth to seventeenth century
(OED).
Further evidence for the historical continuity of complementary dis-

tribution of [f] ~ [v] is provided by double borrowings from French
such as naif, n. adj. (1531, 1598); naive, fem. (1614); digestive, adj. (1532);
digestif, n. (1908). Such examples of apparent complementarity in PDE
reinforce the familiar sg.-pl. model of wife-wives, leaf-leaves.
For the sibilants, the old pattern of glass-glaze, grass-graze, brass-brazen
originally ‘made of brass’, is mirrored in the variability in word-medial
position in the ME noun-verb pattern peace n. (1160)-appease (1330);
grease, n. (c. 1340 < OFr graisse)-(variable) grease, v. (1440); or goose (OE
gōs)-(variable) gosling (1425).
Paradigmatic voicing of [θ] is found not just in inherited mouth,
n. - mouth, v., but, as noted above, regularly in verb forms derived
from native [-θ]-final words, as in bathe, breathe, seethe, wreathing, and so
on.
While such alternations testify to the partial survival of the OE
model of complementary distribution, there is also good evidence that
the development of a context-free contrast between voiced and voice-
less fricatives in ME, especially in word-initial position, undermined
significantly the recognition of historical voicing correspondences
based on the phonetic environment in verbs or the plural of nouns.
The plural inflection of Romance loans, originally just [-s] (unlike the
native [-əs]), and the derived verb forms of the nouns, the latter all of
relatively recent vintage, as in the last column of (15), generally block
morphologically driven voicing correspondences.
(15) Paradigmatic preservation of [f]:
Noun sg. Noun pl. Verb
brief, n. (1225) briefs brief, v. (1837)
beef, n. (1300) beefs/beeves beef, v. (1870)
carafe, n. (1786) carafes
coif, n. (1325) coifs coif, v. (1530)
golf, n. (1457) – golf, v. (1800)
oaf, n. (1638) oafs/oaves oaf, v. (1876)
surf, n. (1685) surfs surf, v. (1831)
serif, n. (1785) serifs serifed, adj. (1889)
The mixed evidence above suggests that the correspondences discussed
in this section changed from fully productive to only marginally pro-
ductive in PDE. It is difficult to determine a cut-off point for the shift
because of the strong intuition that native speakers of PDE continue
to show for the plural voicing of [f] to [v]. A productivity experiment
(wug test), reported in Hayes (2009b: 198) confirms this: the subjects gave
the plural of the nonce word [hif] as [hifs] 58 per cent vs 42 per cent of
[hivz].27 Nevertheless, new words that have entered the language in the
last two centuries tend to preserve the same form of the fricative [-f] in
plural nouns, for example pouff (1817), chef (1842), spoof n. and v. (1889),
digestif n. (1908), gaffe (1909), smurf (1958), boyf (1990).
In conclusion, the history of the three fricatives [f, s, θ] illustrates
the complexity of the interaction between phonology, morphology,
loanword phonology and prosody. To those interested in etymology,
knowing that OE disallowed voiced fricatives word-initially offers a
window into the origin of all [v-, z-] words in PDE: such words are
loanwords or recent formations; for example, vagrant, vahana, valet, vase,
vodka, zephyr, zloty, zone, zombie.
The next chapter turns to consonantal changes associated primarily
with the history of the language in the last millennium.
27
See also Becker et al. (2012), who found some level of voicing in, for example, myth,
vermouth, plaintiff, pontiff, also giraffes ~ gira[v]es, photographs ~ photogra[v]es, psychopaths
~ psychopa[ð]s.
5 Consonantal developments in
the second millennium
AND WOT ’AVE WE ’ERE, GUV’NOR?’, WHICH-WITCH, THOUGH-TOUGH,
FAR-FA, BRIDGESTOWE-BRISTOL, WRITE-RIGHT, IAMB-IAMBIC, GIGOLO,
MATURE-GOCHA, BETTER-BEDDER, MUS GO
This chapter starts with a discussion of the fate of the glottal fricative
/h/, a notoriously unstable segment in many varieties of PDE. Then
we move on to the history and present state of the English rhotics,
another major regional and social marker in PDE. The next topic is the
simplification of consonantal clusters. The rest of the chapter covers
other inventory changes: the addition of // and /ŋ/ to the phonemic
system, innovative patterns of palatalisation and affrication, alveo-
lar stop tapping and contact-induced influences on the consonantal
phonotactics of PDE.
5.1 H-related histories: And wot ’ave we ’ere, guv’nor?1
5.1.1 Phonetic and phonological properties of /h/ in PDE

The consonant [h] in PDE presents an interesting case of instability
which can cause insecurity and social anxiety in some speakers. Unlike
other letter-sound correspondences in the consonantal system, typi-
cally transparent as <f>-[f], <m>-[m], <h> can be ‘silent’. We drop
the [h-] in heir but not in heritage, in honour but not in honey, and herb,
hostler, humour can be pronounced either way. The words maharaja,
prohibition, vehicle, vehement have two fully acceptable pronunciations:
with or without the medial [-h-]. As recorded in John Walker’s famous
Pronouncing Dictionary (1791), the first such dictionary of English, the
words host, humble, hospital were pronounced without the initial [h-] two
centuries ago. Though somewhat old-fashioned, it is still acceptable to
treat the initial sounds in historic, humorous as vowels and write an historic
event, An Humorous (Day’s Mirth). :
How one sounds one’s (h)aitches can be one of the more subtle
signals of a foreign accent. French speakers show a predictable
1
Cartoon by by Ariel Molvig from The New Yorker, 8 February 2010.
99
Figure 5.1 ‘And wot ’ave we ’ere, guv’nor?’ (© Ariel Molvig (2010) The New
Yorker)
tendency to replace English [h] with strong aspiration, while speakers

of Chinese, Greek, and many Semitic and Slavic languages, whose
native phonology includes the velar fricative /x/, tend to add more
friction to [h], producing [x]. The [x] is, of course, a notable feature
of SSE, as in loch, dreich ‘dreary’; it is found in place names in Scotland,
South Africa (influenced by Afrikaans), and in the unassimilated pro-
nunciation of loanwords, as in Bach with [-x] rather than the common
[-k], chutzpah with [x-], rather than the usual [h-]. Then there are
areas in the English-speaking world where all <h->’s are com-
monly dropped. In England h-dropping occurs in all dialects outside
Northumberland and the Eastern Counties. Australian English, New
Zealand English and occasionally even South African English also
show h-dropping, especially in varieties considered ‘non-standard’, so
that the retention of [h] has become a socially significant linguistic
shibboleth.
The fugitive nature of /h/ has given rise to different analyses. As
the changing realisation of <h> is an excellent example of how history
CONSONANTAL DEVELOPMENTS IN THE SECOND MILLENNIUM 101
rationalises a seemingly unmotivated discrepancy between sound and

spelling, we will take some time to explain how this situation came
about.
The phonetic roots of variable h-dropping and its different interpreta-
tions lie in the segment’s nature. Ladefoged describes it as follows:
In English [h] acts like a consonant, but from an articulatory point of
view it is simply the voiceless counterpart of the following vowel. It does
not have a specific place of articulation, and its manner of articulation
is the same as that of a vowel, only the state of the glottis is different.
(Ladefoged 1982: 33–4)2
The lack of place and manner features motivates the exclusion of [h]
from the inventory of PDE consonants (Ladefoged 1982: 33). Further,
the articulatory basis for the instability of /h-/ in English lies in the
shape of the vocal tract, which for /h/ is the same as that of the follow-
ing vowel. Acoustically, spectral information on English [h] shows that
it has the least energy of all the fricatives, which provides further confir-
mation that [h] is effectively a voiceless cognate of the following vowel
(Tabain 1998). Alternatively, and when [h] is included in the system,
which is the more common practice, it can be given a slot with either the
obstruents or the sonorants, which highlights the difficulties inherent in
fitting this sound within the current taxonomic grids. :
Turning to function: within the English consonantal system, /h/
aligns in many ways with the other consonants, so it ‘acts’ like a con-
sonant, as in the selection of the indefinite article allomorph a hedge - a
ledge - an edge. A possible argument for treating /h/ as [+ consonantal]
comes from its retention in borrowings in the onset of unstressed sylla-
bles: alóha, màharája, fáham; in this context note twentieth-century AmE
colloquial doohickey [du
hiki] (1914), yeehaw [ji
hɑ] (1929), formed
natively.
In terms of distributional restrictions, /h/ is not isolated in the
system; compare the restrictions on the sonorants /m, n, l, r/, which
cannot be followed by another consonant in the syllable onset: hat, mat,
rat are fine, but *hlat, *msat, *rdat are not. Another symmetry argument
which prompts classifying /h/ as a consonant on a par with the other
English obstruents is voicing: all English sonorants are voiced, while
/h/ is voiceless. It has to be acknowledged, nevertheless, that within
the obstruent system as a whole, PDE /h/ is still asymmetrical: it is the
only obstruent that does not have a voiced counterpart in any variety
2
See also Ladefoged and Maddieson (1996: 325–6), who describe [h-] as a voiceless or
breathy voiced counterpart of the vowel[s] that follow it.
of PDE, in contrast to the glottal state pairing of all other obstruents

/p/:/b/, /t/:/d/, /f/:/v/, and so on.3 : Since the origin and the earlier
history of /h/ in English warrant its analysis as a consonantal entity,
throughout this book /h/ will be treated as [+ consonantal] [-sonorant],
that is, as a voiceless fricative.
5.1.2 The velar and glottal fricatives of OE

Except for synchronically still marginal items (maharaja, pahit, vehicle),
the distribution of /h-/ in PDE is almost exclusively found in (a) the
onset of stressed syllables (happy, heather, behave, inhibit) and (b) word-
initially, both in stressed and in unstressed syllables (Ha.vána, ha.bítual,
ha.llúcinate, he.llénic, he.lló, Hun.gárian). This restricted distribution is the
result of a series of historical changes whereby /h/ in all other positions
was gradually lost. Let us look again in Figure 5.2 at the inventory of
consonants in late OE, here repeated from Figure 4.1.
Old English had two velar fricatives: voiceless /x/ and voiced /γ/.
The glottal fricative [h] is in parentheses because it was possibly an
allophonic realisation of /x/ in the onset of stressed syllables. Another
allophone, not shown in Figure 5.2, but see (1) below, is the voiceless
palatal fricative [ç]. The two velar fricatives, /x/ and /γ/, were fully
contrastive only in early OE, as in hramma [xr-] ‘cramp’ vs grama [γr-]
‘rage’; hāt [xa
t] ‘hot’ vs gāt [γa
t] ‘goat’; hōd [xo
d] ‘hood’ vs gōd [γo
d]

dental
STOPS p b t d k g
Obstruent
AFFRICATES (tʃ d)
FRICATIVES f θ s ʃ x γ (h)
NASALS m n
Sonorant
Central w r j
Figure 5.2 The OE consonantal inventory
3
The only other obstruent without a voiced counterpart is the voiceless labiovelar
fricative // in whine, whale, whistle, in some varieties of PDE, a continuation of the
OE consonant cluster /hw-/ (see 5.1.3).
‘good’. After the middle of the tenth century, the voiced velar /γ/
in initial position, and when adjacent to consonants, was no longer a
fricative – it was subject to fortition or ‘strengthening’ to /g/ (see 4.2.1).
In medial position between back vowels [γ] remained a fricative, as in
dragan ‘to draw’, ōga ‘terror. In late OE and ME this [γ] underwent leni-
tion, or ‘weakening’ to [w]; the approximant [w] formed a diphthong
with the preceding vowel, so OE dragan ‘to draw’ > ME drawen, OE/
ON ōga ‘awe’ > ME awe. In final position [γ] merged with the voiceless
velar fricative /x/, as in burg ‘-burgh’, beorg ‘hill, barrow’ (see 4.2.1). (1)
summarises the status of the OE velar and glottal fricatives.
(1) The velar and glottal fricatives in OE
Pre-c. 950 Allophones After c. 950 Allophones Examples
[h] [h] heard ‘hard’
/x/ [ç] /x/ or /h/ [ç] niht ‘night’
[x] [x] sohte ‘sought’
[j] /j/ giellan ‘to yell’4
/γ/ [γ] /g/ grund ‘ground’
[ ] 5
[] or /w/ lagu ‘law’
[x] /x/ or /h/ dāg ~ dāh ‘dough’64,5 ,6
The somewhat simplified summary of a more complex situation in the

velar and glottal area in OE in (1) will serve as the starting point for the
description of the ME and EModE changes which shape the outcomes
in the PDE system. As usual, the decision to assign specific allophonic
realisations of the contrastive entities is based on spelling, subsequent
history and typological considerations. The logic of positing an optional
4
The stem yell, also yelp, -gale (as in nightingale), corresponds to IE *ghel ‘to call’, so the
OE /j/ here is a case of merger of the allophone [j] < /γ/ < IE */gh/ with a pre-
existing Germanic /j/, as in gēar ‘year’, geong ‘young’ corresponding to IE *yēr- ‘season,
year’, yeu- ‘vital force’.
5
The IPA symbol [] stands for a voiced velar approximant. It is posited here as the
intermediate stage between /γ/ and /w/ when /γ/ was flanked by back vowels, as in
dragan ‘to draw’, lagu ‘law.
6
The lenition to [-x] ~ [-h] is indicated by the spelling of etymological /γ/ word-
finally after back vowels and sonorants, as in mearg ~ mearh ‘marrow’, dāg ~ dāh
‘dough’.
/x/ for early OE is based on two arguments: the fact that it corresponds
to IE /k/ (see 3.4.1) and the fact that [x] is the most frequent realisa-
tion of the velar fricative cross-linguistically. The reconstruction of the
three allophones [h, x, ç], as in Modern German, is defensible on the
basis of observable phonetic effects of coarticulation: initially, <h->
is likely to have been the same as PDE [h-], and post-vocalic <h> is
likely to have been velar [x] after back vowels and palatal [ç] after front
vowels. :
Using only /x/ for early Old English but allowing the indeterminacy
of either /x/ or /h/ by the end of the tenth century is justified by some
early evidence of initial <h-> dropping, which suggests that in the
syllable onset the consonant was undergoing lenition to a glottal [h-].
The assumption that [h-] was pronounced with reduced friction is sup-
ported by early evidence of loss in unstressed syllable onsets between
voiced segments, which presupposes a shift from [x] to [h], as in PrG
*sehw- > *seohan > *seo-an > OE sēon ‘to see’. Word-initially too, there is
scribal omission in pronominal and other unstressed forms: OE <æfð>
for <hæfð> ‘hath’; <is> for <his> ‘his’; <hefre> for <æfre> ‘ever’.
Such spellings suggest that the situation was similar to the PDE omis-
sion of /h-/ in ’im for him, ’er for her, ’ve for have, all of which are normal
casual-speech variants for all varieties and registers of English today,
where pass him and passim, Calder and called her are homophones. In
stressed-syllable onsets orthographic h-dropping is rare in OE, though
admittedly a limited number of manuscripts do show occasional forms
such as <ondweorc> for <hondweorc> ‘handiwork’, <yngrade> for
<hyngrade> ‘hungered’, <happel> for <appel> ‘apple’. :
At first sight such evidence prompts the question whether the letter
<h> in late OE stood for an entity which is as difficult to classify as that
of PDE /h/. The answer for OE appears to be easier, and that is not
a corollary of the limited nature of the evidence. The status of earlier
/x/, later /h/, as an obstruent phoneme in OE is supported indepen-
dently by orthographical and by phonological arguments. In spite of
the possible interference from Latin, the OE scribes did preserve initial
<h-> with considerable regularity. Forms such as <ondweorc> for
<hondweorc> ‘handiwork’, <yngrade> for <hyngrade> ‘hungered’ are
infrequent: there are two attestations of <yngr-/ungr-> for the stem
‘hunger’ against 364 attestations of <hyng(e)r-/hung(e)r->; the ratio
is characteristic for other such orthographic doublets in the corpus of
OE texts.
Phonologically, the consonant spelled <h> was fully integrated into
the system. It could appear in onsets as a singleton, as in hū ‘how’, hēah
‘high’, or as the first member of onset consonant clusters such as /hr-,
hl-, hn-, hw-/, as in hring ‘ring’, hlot ‘lot’, hnappian ‘to nap’, hwīt ‘white’.
It was found word-medially followed by another voiceless obstruent,
as in mihtig ‘mighty’, wrohte ‘wrought’, and it could appear in codas by
itself or in the clusters /-rh, -lh/, as in þeah ‘though’, þurh ‘through’,
holh ‘hole’. Like all other consonants except the approximants /w/ and
/j/ (see 4.1.2), the voiceless velar fricative could be geminated, as in
hlæhhan ‘laugh’, cohhetan ‘cough’. It paralleled the stops /p, t, k/ in that in
early OE it was in phonological contrast with the voiced velar fricative
/γ/, as in hāt [xa
t] ‘hot’ vs gāt [γa
t] ‘goat’. Not least, the alliterative
practice confirms further the strong obstruent nature of the segment:
in the poetry <h-> behaves like any other consonant, namely it alliter-
ates with itself, irrespective of the nature of the following consonant or
vowel. Thus <hr-, hl-, hn-, hw-> words alliterate freely among them-
selves and with <h->; the practice is the same in the earlier and the later
OE verse (Minkova 2003: 339–45).
5.1.3 Initial glottal fricatives in ME and EModE: ABLE-HABILITATE,

WHICH-WITCH, WHINE-WINE
Whether the OE inventory of fricatives is analysed as including either

the velar /x/ or the glottal /h/, it is evident that after the middle of the
eleventh century, the segment written with <h> must have been /h-/ in
initial position. Orthographic records and the alliterative practice of the
fourteenth century show that starting in early ME to PDE, the history
of /h-/ is one of loss and instability. The phonetic grounds for this were
discussed in 5.1.1: when followed by a vowel, the vocal tract configura-
tion for /h/ is the same as that of the vowel itself. Phonologically, after
the split of the voiced velar fricative /γ/ into /g/, /j/ and /w/ (see
5.1.2), and the development of voicing contrasts for /f/:/v/, /s/:/z/
and /θ/:/ð/ (see 4.4), /h/ remained the only fricative with no voiced
counterpart in the system.
Another dimension of the change involves the exposure of English
speakers to phonological systems lacking the corresponding segment.
The presence of /h-/ in onset position in Latin, the main source of
borrowing in OE, was inconsistent. Latin had already started abandon-
ing the fricative in the third century, and the loss was complete by
the end of the seventh century, but <h-> was often preserved in the
orthography. In words whose transmission was not exclusively literary,
Latin h is lost in OE: iācin(c)tus (< hyacinthus), istoriam. It was also lost
in popular proper names, for example Ercol (Hercules), Elene. In purely
literary words it is retained in the spelling and pronounced like English
h-: hōlocaustum, Hōlofernus (see Pyles 1943: 909).
Though socially stigmatised, /h/-dropping was spreading in the

Gallo-Roman regions.7 This meant that for post-Conquest Anglo-
Norman speakers /h/ would be a foreign element and would pose
acquisition problems. Second-language learners and speakers of Anglo-
Norman would also be aware that the segment was not common in the
French vocabulary. We say ‘not common’ because the language contact
situation is further complicated by the fact that OFr had reintroduced
[h-] in initial position in some words borrowed from Germanic, such
as haste, hardy, heron; these items were possibly [h-]-ful in AN too. The
ME options, based on etymological source and prosodic prominence,
are shown in (2).
(2) The pronunciation of <h-> in Middle English:
Etymology Anglo-Saxon Anglo-Saxon French words Romance

stressed unstressed words of Germanic words
words origin
Examples hold, heaven, (h)ave, (h)im, haste, hardy, horrible,

high, hundred (h)e, (h)er heron, herald humour,
hermit, humble
Pronunciation [h-] Ø [h-] Ø
In the first three columns, the situation as a whole does not differ from
what we encounter today in SSBE and GA. The parentheses around
(h) for the unstressed words indicate that in addition to etymological
source, prosodic prominence is an important factor in the retention
of [h-]. If <h-> is in the onset of a stressed syllable, it is much more
likely to preserve its consonantal properties, in line with the universal
syllable-structure preference for a filled onset and a sonority increase
from the onset to the peak (see 2.3.1, 2.3.2). In unstressed syllables,
however, lenition of [h] to Ø is much more likely; unstressed syllables
are the typical ‘weakening’ domain; compare ‘Yes, we have’ with ‘We
(h)ave shown’.
The distributional details are much more complex than the over-
view in (2) suggests. ‘Middle English’ is a convenient chronological
label which covers a range of quite distinct regional, and obviously,
7
St Augustine (AD 354–430) wrote that ‘it was deemed a greater offence to drop the h of
hominem than to disregard the law of Christian charity’ (Confessions, I, Section 181, cited
in Pope 1961: 91).
social, varieties. Information about socio-linguistic variability in ME

is sketchy, but it is well established that the Northern dialects resisted
/h-/ loss, while the rate of /h-/-lessness in the Midlands and the south
was much higher.
The last column in (2) refers to h-dropping in Romance loanwords.
Throughout ME, [h-]-dropping in such words was a feature of the
spoken language, though it could be resisted by the scribes and the
educators. Variability between [h-] and Ø was widespread: words like
able (1325), ability (1380), ostler (1325), arbour (1300) ‘a bower of trees
and shrubs’ (Lat. herba ‘grass’) came into English without [h-], directly
from Anglo-Norman/Old French. Many of their cognates, however,
show the orthographic influence of the Latin etymon and preserve
h- both orthographically and in pronunciation: able, ability (1380) vs
habil, adj. (1425), habiliments ‘appropriate garments’ (1491), habilitation
(1612); ostler, but hostel (1250); arbour, but herbage (1390), herbal (1516). In
this group the effect of stress is also noticeable: (h)istóric is much more
likely to be h-less than stressed [h]-initial hístory; hármony, but the <h>
in philharmónic is realised variably, with silent <h> dominant in AmE.
Thus both in native and in borrowed h-initial words the instability of
/h-/ is well attested.
Orthographic standardisation, especially through printing after the
end of the 1470s, was an important factor shaping the later fate of ME
/h-/. Until the beginning of the sixteenth century there was no evi-
dence of association between h-dropping and social and educational
status, but the attitudes began to shift in the seventeenth century, and
by the eighteenth century [h-]-lessness was stigmatised in both native
and borrowed words (Mugglestone 2003: 98ff.). In spelling, most of
the borrowed words kept initial <h->; the expanding community of
literate speakers must have considered spelling authoritative enough
for the reinstatement of an initial [h-] in words with an etymological
and orthographic <h->. New Greek loanwords in <h->, unassimi-
lated when passing through Renaissance Latin, flooded the language;
learned words in hept(a)-, hemato-, hemi-, hex(a)-, hagio-, hypo-, hydro-,
hyper-, hetero-, hysto- and words like helix, harmony, halo kept the initial
aspirate in pronunciation, increasing the pool of lexical items for which
h-dropping would be associated with lack of education. The combined
and mutually reinforcing pressure from orthography and negative social
attitudes towards h-dropping worked against the codification of h-less
forms. By the end of the eighteenth century, only a set of frequently
used Romance loans in which the <h-> spelling was preserved were
considered legitimate without initial [h-]. Here is Walker’s instruction
on how to pronounce the letter <h>:
At the beginning of words, it is always sounded, except in heir, heiress,

honest, honestly, honour, honourable, herb, herbage, hospital, hostler, hour, humble,
humour, humorous, humoursome. Ben Johnson leaves out the h in host, and
classes it in this respect with honest. (Walker 1791–1826: 53)
Today heir, honest and hour (and herb in AmE) and some of their cog-
nates are the only surviving instances of a once widespread phonetic
attrition. :Along with [h]-less heir we find heritage, inherit with [h-];
along with honest we get honorarium with [h-] ~ Ø; and along with hour
we find horary, horæ ‘a book of hours’, horology with [h-]. All OE <h>-
initial words that are still used in PDE, except, possibly, the pronoun
it, OE <hit>, preserve /h-/ in careful speech in the ‘standard’ varieties
to this day : Irrespective of etymology, initial <h> in all words has
become an important shibboleth, and a ‘symbol of the social divide’.8
In Victorian England, h-dropping and its counterpart, unetymologi-
cal insertion of [h-] in onsetless syllables (as in Harab for Arab, Hirene
for Irene), was stigmatised as ‘vulgar’. The great nineteenth-century
phonetician Henry Sweet characterises the ‘proper’ pronunciation of
[h-] as ‘an almost infallible test of education and refinement’. In his
authoritative treatise on The Letter H, Leach states:
it is no exaggeration to say that, socially, H is of English letters the most
important, and that a systematic trifling with half the vowels and conso-
nants of the alphabet would not be visited with such severe social rep-
robation as is the omission or misplacement of an H. (Leach 1880: 9) :
Our focus so far has been the pre-vocalic loss of [h-], because its varia-
ble survival is most easily recognised by speakers of PDE. However, loss
in pre-vocalic position did not occur in isolation from the developments
of /h/ in other contexts. Indeed, the earliest stage of h-loss is attested
in onset clusters of /h-/ + sonorant, for example hlot ‘lot’, hring ‘ring’,
hnappian ‘to nap’, hwīt ‘white’ (see 5.1.2). The loss of [h-] in word-initial
consonant clusters was already under way in the eleventh century and
it progressed rapidly in the next century. LAEME and the MED show
no <hr-, hl-, hn-> spellings after c. 1250.
(3) Early ME loss of /h-/ + sonorant:
OE <hlot> ME <lot> ‘lot’
OE <hring> ME <ring> ‘ring’
OE <hnappian> ME <napp(en)> ‘to nap’
The bi-segmental realisation of the clusters in (3) is testable in OE
8
The phrase is the title of chapter 4 in Mugglestone (2003).
alliterative verse.9 Since there is no comparable method of testing the

transitional period between c. 1000–1150, and since the orthographic
traces of such clusters after 1150 are found in documents copied from
OE, it is safe to assume that h-loss in these clusters had been completed
by the beginning of the thirteenth century.
The simplification of /hw-/ to // to /w-/ started around the
same time as in the other clusters. The history of this cluster, however,
is more complex and more interesting from the point of view of the
modern speaker. As noted in 4.1.1, OE <hw-> represented a fricative-
initial bi-segmental onset, as attested by the alliterative choices. The
most common practice found in the main body of OE verse was for the
poets to match <hw-> to <h->, also to <hr-, hl-, hn->. Assuming that
/x-/ was lenited to [h-] in late OE, it is not very surprising that there is
some evidence in late OE verse that the sequence [hw-] started losing
its initial fricative, and coalescing with etymological /w-/, allowing
alliterations such as hwile ‘while’, wæpna ‘weapons’: wealdan ‘wield’ (see
4.1.1, (2)). Such practice can only be ascertained for compositions from
the southern parts of the country.
The tendency for simplification became more pronounced in early
ME. Loss of [h-] was especially common in weakly stressed inter-
rogative words: what, which, where. The map in Figure 5.3 shows the
distribution of <h->-less forms of which in LAEME.
In the Midlands and in the South, cluster simplification affected also
major class words whose initial sound, historically, had been a [hw-]: wheat,
white, whale, wharf, and so on. The continuing rivalry between [(h)w-] and
[w-] is confirmed by some sixteenth- and seventeenth-century respellings
of original <w-> words: OE wōs, n., wēsan ‘ooze’ appears as <wheeze>;
whiff < early ME weffe; whisk < ON visk.10 In one limited subset of original
<hw-> words – who < OE hwā, whose < OE hwœs, whom <OE hwām and
whoop ~ hoop (onomatopoeic) – the fricative is preserved and the [w] is lost
before the following high back rounded vowel.
The merger of the two previously distinct entities, /hw-/ and /w-/,
persists in PDE varieties in which there is no contrast in the onsets of,
for example, whale-wail, which-witch, whine-wine. However, in ME the
merger in fully stressed words was never complete, and it did not occur
in the Northern varieties of Middle English, where it was represented in
spelling by <qu(h), qw(h)>. This particular spelling is one of the ways in
9
The evidence is discussed in Minkova (2003: 342–9).
10
The OED comments that in whip < the base wip- and in whisk ‘The spelling with
wh was adopted as being symbolic’. The [hw] ~ [-] in these words is a spelling
pronunciation.
Figure 5.3 WHICH spellings with initial <w> and in EME (from A
Linguistic Atlas of Early Middle English © 2007–, The University Court of
the University of Edinburgh, Old College, Edinburgh. Reproduced with kind
permission of Margaret Laing)
which we can ascertain the stable nature of <hw-> in the north, where
its realisation was probably [xw-]. Another confirmation of the differ-
ent status of the cluster in the north and in the south comes from the
practice in fourteenth-century alliterative compositions: behold: quareon
‘whereon’, quilke ‘which’: hert ‘heart’ are unmistakably Northern allitera-
tions. Indeed, the matching of etymological <wh-> onsets with <w->
onsets, as in await: white, word: where is an indication of non-Northern
origin of a composition.11
11
The areas of preservation of [xw] in Middle English are Scotland, Northumberland,
A note on spelling: recall that by the seventh century /h/ was lost
in Latin, whereby <h> was ‘freed’ for use as a diacritic in combination
with other consonant letters. The digraphs <ch> and <th>, both famil-
iar to the OE scribes, are derived from Latin orthography.12 The <ph>
digraph from Greek was also used in OE loanwords and names (pharaoh,
philosoph(e), Phillipp, Joseph, Stephanus). In ME the practice of spellings
using the letter <h> in second position was extended to include <gh,
rh, sh>. These spellings were not uniquely or necessarily associated
with the fricative [h], so at a time when [xw-] > [hw-] > [w], the scribes
reversed the order of the letters in the old spelling <hw-> to <wh->,
thereby aligning the digraph with all other <consonant + h> digraphs.
The <wh> spelling helps reinforce the impression for the literate
speaker that there could, or should be an [h-] in the pronunciation of
<wh-> words.
The sociolinguistic reaction to the which-witch merger in early
Modern English was parallel to the attitude to initial h-dropping.
Throughout the eighteenth and nineteenth centuries the [wh-] ~
[w-] alternation was ‘a socially sensitive variable’;13 as late as 1880,
the popular treatise on pronunciation mentioned above, The Letter H,
identifies one of the book’s goals as ‘to seek redress for the digraph WH’
(Leach 1880: 5). It was only during the last century that the histori-
cally based identity of which-witch became standard for most varieties
of British English, except for Scotland, Northumberland, partly Welsh
and Irish English, but even in those historical strongholds of which-
witch contrast, recent studies indicate a tendency for weakening of the
contrast in large urban areas such as Edinburgh and Glasgow. For New
Zealand English, Bauer (1994a: 395–6) reports that the contrast is reced-
ing, adding that ‘This feature is kept alive by overt teaching: /hw/ is
perceived as being a prestige pronunciation.’
The merger of the originally distinct sounds is also increasingly
popular in North America. In Canadian English the which-witch
homophony is fully established (Brinton and Fee 2001: 430). The
merger has also been spreading in the US. According to reports on
early/colonial American pronunciation, [hw-] was widespread, espe-
cially in the South. This feature of early AE is attributed to the influence
of settlers coming from areas other than southern England, especially
during the eighteenth century (Montgomery 2001: 143). By the middle
Cumberland, Durham, Westmorland, large sections of Lancashire and Yorkshire (see

Minkova 2004).
12
Scragg (1974: 46–7).
13
See further Mugglestone (2003: 186–9, 196).
of the twentieth century, it was mostly the American South that pre-
served the contrast, but by the end of the century the preservation was
no longer cleanly localised.
In summary, the phonological history of the bi-segmental /hw/ cluster

from OE to the present is part of the overall history of /h-/ in the syl-
lable onset. The alternation [h-] ~ Ø, first in weak prosodic positions,
and then in any onset position, undermined the stability of [hw-]. The
next stage in southern ME was dominance of [w-], while in the north
the original [hw-] continued to be realised as a complex segment. /w/
and // are both complex labiovelar segments and the place identity
between them increases the chance of merger in spite of the different
glottal states: constricted vs spread glottis. Phonologically, // is iso-
lated in the sense that its voiced counterpart, /w/, is the only approxi-
mant that has a contrastive voiceless realisation; this could be another
reason for the historical instability of //. The tension between the
phonetic and structural factors promoting loss of //, and the sociolin-
guistic and possibly semiotic factors promoting its retention and resto-
ration, continues. The timeline of the changes of OE /hw-/ is presented
in (4); the current variables are enclosed in square brackets and bold
indicates dominant realisations. :
(4) Variability of /hw-/ ~ /w/ in the history of English:
LOE ME 16th & 17th c. 18th c. – PDE c. 2010
North w
/-/ /-/ /-/ /-/ [ -] ~ [w-]
/hw-/ ~ /w/ /-/ (careful, literate)
South /w-/ /w-/ [w-] ~ [-]
/w-/ (everyday casual)
5.1.4 Non-initial glottal fricatives in ME: THOUGH-TOUGH

In intervocalic position [-h-] had been lost in early OE, as in PrG *sehw-
> *seohan > *seo-an > OE sēon ‘to see’; PrG *teuh-an > OE tēo(ha)n ‘to
tug’. The glottal fricatives were preserved in word-internal geminates:
hlæhhan> ‘laugh’, cohhetan ‘cough’, geneahhe ‘enough’, but such geminates
were rare. When geminate consonants were simplified in ME (see 4.1.2),
the resulting single fricative developed in the same way as <-h> in any
coda position. :
The parallelism of what happened to [h] in word-medial and in final
position is based on two factors. First, when word-medial [h] was fol-
lowed by another consonant, it would automatically be in the coda, as
in soh.te ‘sought’. Second, the attrition of final unstressed vowels in ME

(see 7.6) led to the placement of intervocalic [-h] in the syllable coda. In
words like ME laughen ‘to laugh’, the fricative [-h-] was originally either
ambisyllabic or in the unstressed syllable onset, but in the paradigm of
the stem laugh-, past tense lough-, the inflection was increasingly missing,
so that [-h] was also increasingly frequently just the coda of the mono-
syllable. The number of items with coda /-h/ was augmented by words
in which the Germanic voiced velar fricative /-γ/ was devoiced and
realised as [-x] in word-final position after back vowels or the sonorants
/r, l/: <burg/burh> ‘-burgh’; <beorg/beorh> ‘hill, barrow’; <plog/
ploh> ‘plough’ (see 4.2.1).
In coda position, both the single fricative and the fricative in <-ht>
clusters were subject to changes based on coarticulation with the pre-
ceding vowel. : The realisation of <-h> after back vowels was prob-
ably a velar [-x], while after front vowels it would have been a palatal
[ç], not unlike the Modern German ach-Laut [x] and ich-Laut [ç]. This
coarticulation is likely to give rise to transitional glides: a back glide [w]
before [-x], and a front glide [j] before [-ç]. These glides are interpreted
as part of the syllable peak, forming diphthongs with the original vowel
(see 6.5.3, 7.4).
The developments of non-initial [-x] and [-ç] are behind one of the
most striking discrepancies between spelling and pronunciation in the
consonantal system of PDE, illustrated by the realisation of <-gh> in
the pair THOUGH-TOUGH. As noted already, <-gh> is a ME re-spelling
of <-h>; this only tells us that the presence of a fricative was stable
long enough to be encoded in the orthography. There is more to say,
however, about the difference in the PDE reflexes of the ME <gh>,
as in high, bough, sought, with a complete loss of the fricative, vs enough,
laughter, with the labial fricative [f].
Full lenition of /-h/ in the syllable coda is a typologically predictable
change, though it is obviously not a necessary one. The realisation must
have been variable [-h] ~ Ø, starting as early as the thirteenth and four-
teenth centuries, as evidenced by incidental orthographic loss of coda
<-h>: <hei(e)> ‘high’, <brit> ‘bright’, <bow(e)> ‘bough’, <brout(e)> for
‘brought’, with full-scale evidence found during the fifteenth century. :
The variability of [-x(t)] ~ [-f(t)], as in enough, rough, draught, is also
attested orthographically in the fourteenth century, both in items in
which PDE has [-f], as in <ruff> ‘rough’; <inowf> ‘enough’, and in
items in which the [-f] variant was abandoned in favour of the vocal-
ised pronunciation: <þof> ~ <thouh> ‘though’, <þurf> ~ <trugh>
‘through’; <dofter> ~ <douter> ‘daughter’. The timeline of these
changes is shown in (5).
(5) Development of [ç, x] in Middle English:

Early ME 14th c. 15th–16th c. PDE Examples
[ç]
Ø/[h]/ Ø Ø thigh, sight, bough, caught
[x]
[f]/[x]/Ø [f]/[x] [f] enough, tough, dwarf, trough
The vocalisation of the lenited [-h] occurs after both front and back
vowels, while the change of [-x] to [-f] can only occur after back vowels.
For medial [x] following a back vowel, there was widespread variation
between Ø and [-f]. The phonetic motivation for the shift to [f] involves
an acoustically based reanalysis which ‘strengthens’ the consonant.
Specifically, the trigger of the [-x] > [-f] change is plausibly attributed
to perceptual confusion; the two consonants shared acoustic feature [+
grave], which correlates with the fact that the development is attested
exclusively after back vowels. :
Both the lenition of [x] > [h] > Ø, and the shift of [x] > [f], shown in
(5) are phonetically grounded. The triggers refer to different aspects of
sound production: ease of articulation in the case of vocalisation, and
similarity of the acoustic signal in the case of [f]. The latter ‘mishearing’
was initially socially stigmatised. It has been characterised as originating
in ‘dialectal and vulgar speech’; in words like laugh, cough, rough [x] was
‘the normal pronunciation of good speech’ until the seventeenth century
(Dobson 1968: §371). Dialectal forms such as <barf> < OE beorh ‘hill,
mound’, <dofter/dafter> < OE dohtor ‘daughter’ were abandoned, but the
fully lenited pronunciation and the [-f] pronunciation survive in the pair
dough-duff ‘a dumpling’, the latter a continuation of an earlier Northern
form, OE dah.
The various options in the realisation of the voiceless velar fricative
in the history of English are summarised in (6).
(6) Realisations of OE <h> in PDE:
Position of /h/ PDE OE ME PDE gloss

outcome
Onset: -V [h-] hus hous house

Coda: V- Ø sihð sight(e) sight
þeah þough though

:
[-f] ruh rugh rough
Ø hlot lot lot

Onset: -C
[w-]/[] hwit whit white
Ø þurh þrough through

Coda C-
[-f] dweorh dwargh dwarf
The overall picture is clear: in native words /h-/ remains stable only
if it forms the single onset of a stressed syllable, which is tantamount
to saying that in the native vocabulary this consonant is historically
restricted to stem-initial position.14 The distribution of /h-/ is broaden-
ing, however. In the last century, Spanish–English bilingualism in the
American Southwest has resulted in the formation of new varieties of
Latino Englishes, which in their turn influence AmE. One such influ-
ence results in the acceptability of stem-internal /h-/, as in jojoba, mojo,
fajita. Also, as noted in 5.1.1, /h-/ is retained in borrowings in the onset
of unstressed syllables: rioha (1611), alóha (1825), mája (1832), fáham
‘an orchid native to Mauritius’ (1850), and it is formed natively as in
Soho (1818), colloquial AmE doohickey ['du
hiki], yeehaw ['ji
hɑ], most
recently the blend WeHo ['wi
hoυ] ‘West Hollywood’ (2006). :
5.2 R-related histories

Doe, a deer, a female deer
Ray, a drop of golden sun
Me, a name I call myself
Far, a long, long way to run . . .
5.2.1 Phonetic and phonological properties of the rhotics in PDE

The term ‘rhotics’ comes from the name of the Greek letter rho <ρ>,
Roman <r>. Using /r/ as a generalised phonemic representation for
orthographic <r>, we start the historical account of the English rhotics
by recognising that the realisation of /r/ in PDE is not uniform; the
various dialects of English exhibit at least six of the eight different
14
The suffix -hood goes back to a noun – OE hād, ME had(e), hōd ‘person, condition,
quality, rank’; it carried secondary stress, as is clear from the shift of the vowel [o
] to
[u
] to [υ].
forms of rhotics found in the world’s languages.15 Perhaps the most

widespread articulation of <r> is the alveolar approximant [ɹ], used in
RP and in AmE, but we also find the alveolar coronal trill [r] (in some
accents of Scots), alveolar tap [ɾ] (intervocalically in ‘Refined RP’, also
the most common rhotic in Scots), the retroflex [] (South-West British
English, North American, similar to [ɹ]), the uvular continuant [r],
or the uvular fricative [ʁ] (North-East England, Lowland Scots), and
even the realisation of /r/ as a labiodental continuant [υ] or [], the
velar approximant similar to an unrounded [w] in London/Estuary
English.16 Moreover, in varieties of English that undergo historical /r/-
loss, the non-rhotic varieties, the effect of the loss goes well beyond the
consonantal system and leads to a restructuring of the vowel system
(see 8.3.2). This justifies elevating /r/ to the ‘single most salient factor
differentiating today’s varieties of English’ (McMahon 2009: 113).
5.2.2 Reconstructing the phonetics of <r> in OE

Although the most broadly used articulation of /r/ in PDE is as an
approximant [ɹ] or retroflex [], this may not have been the case histori-
cally, at least for all dialects. For a start, in the North-West Germanic
dialects, /r/ came from two distinct sources: original IE */r/, and the /r/
that resulted from the voicing of IE */s/ by Verner’s Law (see 3.4.1). For
the latter some runic inscriptions have a special symbol, distinct from
the letters for /r/ and /s/, usually interpreted as /z/ (Wakelin 1988:
41). There is no evidence, however, suggesting that the etymological
distinction between the two types of /r/ was maintained in OE. Two
conflicting interpretations of the phonetic nature of OE /r/ are found
in the literature: OE <r> reconstructed as velar/velarised/uvular,
or, more loosely, some back variety of /r/, or as an apical/anterior/
coronal/alveolar approximant or tap. Such contradictory reconstruc-
tions of the phonetic properties of OE /r/ are due to the intrinsic vari-
ability of the rhotics: their stability, realisation and coarticulatory effects
in the modern Germanic languages and dialects vary significantly. :
An important argument in favour of reconstructing a velarised
allophone of /r/ in OE comes from the process known as OE Breaking,
orthographically recognisable by the insertion of a back-vowel letter,
usually <o, a>, after the short front vowels /i, e, æ/ if <-r> was in the
15
Ladefoged and Maddieson (1996: 235–6). Magnuson (2007) provides a comprehensive
phonetic account of the rhotics and their connection with the laryngeal and oral vocal
tract.
16
Cruttenden (2008: 221–2); Davenport and Hannahs (2010: 32).
syllable coda and adjacent to another consonant to the right. Thus, in

West Saxon, we find the forms in (7).
(7) OE Breaking before <-rC>:
Source West Saxon Gloss
*arma- earm17 ‘arm’
*hertan heorte ‘heart’
*lirnojan liornian18 ‘learn’
Whether the vowel spellings in the second column represent phonemic
contrasts in this or any variety of OE is not of interest here – we return
to the question of Breaking in 6.5.3. What matters for now is that such
digraph spellings are fairly consistent and allow the assumption of a
uvular/velar version of the rhotic. This is not the only possible interpre-
tation, however. Typologically, apical rhotics, for example [r], and the
uvular rhotics, for example [r], are acoustically close and they can have
the same effect on the preceding vowels leading to the development of a
glide (Howell 1987: 325). : In articulatory terms, arguably a clear way
of characterising the PDE rhotics, it has been shown that rhotics are
produced with both dorsal and coronal constrictions in English, which
also allows for the realisation of a transitional glide, especially since the
front/coronal gesture is delayed with respect to the dorsal gesture.19
Another observation relating to the historical nature of /r/ has to do
with the geographical area where PDE has the uvular continuant [r], or
the uvular fricative [ʁ], namely North-East England and Lowland Scots.
This dialectal area is presumably a continuation of Northumbrian OE,
that is, part of Anglian, where Breaking before <rC> is unstable, espe-
cially after <a>: Anglian warp ‘threw’, barn ‘bairn’ for WS wearp, bearn.20
Of interest in the context of reconstructing a single velar rhotic for OE
is also the fact that clusters that should have exerted the strongest velar
effect, namely <r> + velar consonant in Anglian, in fact repair the clus-
ters not by weakening of the rhotic, but by inserting an epenthetic vowel
between /r/ and the velar, thus <burug> ~ <byrig> for <burg> ‘city’.
Such observations support the assumption that OE had different
rhotic allophones in different dialects and in different structural posi-
tions. The higher scribal stability indicating Breaking in the ‘focused’
17
The orthographic <a> of the source would have been fronted to [ɑ] > [æ] prior to
Breaking.
18
The <io> spelling is early West Saxon.
19
See Delattre and Freeman (1968). For full details on the gestural overlap of liquids in
English and elsewhere see Proctor (2010).
20
These objections are recorded in Lass and Anderson (1975: 89–90, fn. 1); see also
Howell (1987).
classical and late West Saxon records would be consistent with the
reconstruction of a central approximant, a type of [ɹ], in the syllable
coda. The reconstriction of a coronal trill [r] is also appealing because
it is the most common type of rhotic across the world’s languages
(Maddieson and Ladefoged 1996: 217). Comparison with the other
Germanic languages is also suggestive: in Afrikaans, Faroese, Frisian,
Icelandic, Norwegian and Swedish the dominant realisation of <r> is
also a coronal trill. This would be in line with the documented behav-
iour of both RP English (Heselwood 2009) and American English. The
realisation of /r-/ in stressed-syllable onset in OE probably involved
more friction. Dialect differences in that position are beyond recon-
struction; all we can say is that there is a high likelihood of a strong
apical component and greater constriction, matching the historical
stability of the rhotic in onsets.
The effect of /r/ on the preceding short vowel shown in (7) is
usually interpreted as ‘diphthongisation’ due to the nature of the OE
rhotic. OE Breaking is constrained by the properties of the preceding
vowel – it occurs only after short front vowels. This constraint is com-
patible with interpretation of the transitional glide either as creating a
positional diphthongised allophone of the short vowel, or as marking
the post-vocalic rhotic as a velar allophone. The higher probability of
a transitional element after front vowels is associated with a relatively
high tongue position for the rhotic and the rounded quality of the glide
(Stockwell 2002a; Denton 2003: 29). The absence of orthographically
recorded /r/-adjustments after long vowels is unsurprising, since the
extra phonation time for the long vowel will mask the special nature of
the coda.
In the majority of instances of Breaking, either the stem, or high-
frequency derived or inflectional forms, are non-monosyllabic, as in
heorte ‘heart’, liornian ‘learn’. Since the sequence (C)VRCV, where R
stands for a rhotic, will be syllabified VR.CV, the environment for OE
Breaking can be defined as short vowel + tautosyllabic /r/. There
are also the instances of stem-final geminate /r/, for example feorr
‘far’, where the -/r/ is most likely a singleton coda. This modifies
the common assumption that Breaking requires a ‘covered’ /r/, but
further considerations regarding OE and phonetic considerations from
current studies of rhotic effects warrant this slight amendment of the
canon. :
The position that coda /-r/ is a sufficient condition for OE Breaking
does not negate the importance of the following consonant in the
long-term history of English /r/. There are clear cases of Breaking
where the /r/ is followed by a consonant in the same syllable: wearp
‘threw’, sg., heord ‘custody’. The cluster -RC# alternated paradigmati-

cally with -R.C- as in the past sg., 1st and 3rd pers. of Class 3 strong
verbs: ċeorfan ‘cut’, steorfan ‘die’, weorpan ‘throw’, weorðan ‘become’. The
significance of such paradigmatic alternations has not been explored.
The progressive weakening of the final vowels would also bolster the
-RC# group, another source of the inclusion of the consonant in the
traditional description of Breaking. As we will see below, however, the
real significance of -RC# can only be established in post-OE records.
Two other processes in OE relate directly to the properties of /r/:
metathesis and epenthesis. Some examples of metathesis are shown
in (8).
(8) Metathesis involving /r/ + short vowel in OE:
brinnan ~ birnan ‘to burn’ timbrede ~ timberde ‘built’
cerse ~ cresse ‘cress’ (to)þræsċ ~ þærsċ ‘threshed’
(Sige)-ferð ~ (Reð)-frið21 ðirdda ~ ðridda ‘third’
ræn ~ ærn ‘ran’, sg. worhte ~ wrohte ‘worked, wrought’
Hogg (1992b: 110–11) describes this process as ‘perhaps more structur-
ally organized’ than metathesis not involving /r/. The alternation of (C)
VRC ~ (C)RVC is indicative of the strong sonority of /r/, not surpris-
ingly so, as /r/ is the consonant of highest sonority in PDE, except for
the glides /j/ and /w/. If the short vowel in the sequence -VRC- was
realised as a rhotacised vowel [], as in PDE bird, the perceptual simi-
larity of the transition of [] to the following consonant and [ɹ] + short
vowel + C would make metathesis likely. Note that the process contin-
ues into PDE: bright, burn, burst, cress, crud, dirt, and so on are survivals of
metathesised forms, and throughout ME and in many PDE dialects one
often finds /r/-based metathesis.
(9) Metathesis results in ME and metathesis in PDE:
ME spellings PDE regional variants
frist ~ first (OE fyrst) burches ~ breeches
gridel ~ girdle (OE gyrdel) cross ~ corse
grost ~ gorse (OE gorst) garston22 ~ grass
-thorp ~ -throp (OE ðrop) grin ~ girn
thrist(e) ~ thirst (OE þyrst) purty ~ pretty
The cases of metathesis in (8) and (9) are phonetically ‘natural’ in the
sense that the lower F3 of the rhotic overlaps with an adjacent short
21
The shared Germanic root in the names is frið ‘peace’, as in Frederick, Friedman,
Friedrich, and so on.
22
Dialectal, ‘a grassy enclosure, a paddock’ (OED).
vowel, thus obscuring the sequencing of the rhotic and the vowel.23
Therefore, the persistence of metathesis from OE to PDE points to
salience of the rhotic, making the proposed coda-weakening of /r/
in OE (Howell 1991; Lutz 1994) an unlikely scenario. Metathesis is a
structure-preserving process; focusing on the ‘inherent weakness’ of the
rhotic as its determining property in OE is not enlightening in terms
of its effect and it contradicts the OE orthographic records, which are
remarkably uniform in preserving scribal <r> in all positions. Another
suggestion relating to /r/ metathesis, that it improves the syllable
structure by reducing the weight of the coda by shifting the cluster to
the onset (Windross 1994), fails to account for the bidirectionality of
metathesis we find in English – the coda would be ‘lightened’ in the case
of OE fyrst > frist, but that motivation does not apply to, for example,
OE cros > ME cors ~ cros, or the survival of items such as gorse, first, girdle,
thirst.
Epenthesis or anaptyxis (see 2.5) is another process associated with
the perceptual similarity between -CRV- and -CVR-. In this case, a
short, schwa-like vowel is inserted into a -CR- cluster, in effect increas-
ing the number of syllables in the original word; for example, inflected
forms of the WG root *wɑtr- ‘water’ when followed by a vowel alter-
nate between <wætr-> (x406) ~ <wæter-> (x892). Other examples of
anaptyxis involving /r/ in different environments in OE are shown
in (10).
(10) OE anaptyxis involving /r/:
(a) Onset CR- > -CVR
æˉ fre ~ æfere ‘ever’
bremel ~ beremel(e) ‘bramble’
broþor ~ beroþor ‘brother’
wrohte ~ worohte ‘wrought’
(b) Coda RC(C) > RVC(C) :

beorht ~ beoroht ‘bright’
beorc ~ berec ‘birch’
burg ~ buruh ‘-burg’
fyrn ~ firen ‘transgression’
(c) Coda -CR > -CVR:

apuldr ~ apuldur ‘apple-tree’
bebr ~ beber ‘beaver’
frofr ~ frofor ‘consolation’
23
Blevins and Garrett (1998); Denton (2003).
Ignoring details regarding the etymology, chronology and dialectal

distribution of the epenthetic forms in OE, we can say that the vari-
ation illustrated in (10) was ‘persistent and continuing’ (Hogg 1992a:
239). The motivation for the process is perceptual: the transition from
high-sonority /r/ to the low-sonority adjacent obstruent. The sonority
is falling in the case of initial clusters and in stressed syllable codas (10a,
b), and rising in unstressed syllable codas (10c). The unifying principle
in all instances is that the /r/ in the cluster mimics the transition from
an obstruent to a vowel, rendering the epenthesised CVR and RVC
strings perceptually confusable with CR or RC strings.24
The processes discussed in this section are associated with /r/, but
the other liquid, /l/, and the sonorants /n, m/ can produce similar
effects – we return to these below. Staying within the topic of the past
and the present of the rhotics in English for now, we turn to the most
salient process for speakers of PDE: the historical loss of coda /r/.
5.2.3 Pre-consonantal /r/-loss

Although there are no systematic indications of <r>-loss in the OE
corpus, some early forms foreshadowing later changes should be rec-
ognised. As shown in (11), /r/ due to rhotacism (see 3.4.1) could be
dropped in OE; the loss could result in lengthening of the preceding
vowel.
(11) OE loss of /r/ < PrG /z/:
Gothic mizdō > OE meord ~ mēd ‘meed, reward’
PrG *twizna > OE twīn ‘twine’, compare German Zwirn
PrG *hwaz > OE hwa ~ hwā ‘who’, compare German wer
Gothic weis > OE we ~ wē ‘we’, compare German wir
Sporadic loss of <r> occurs in some onset clusters: sp(r)ecan ‘to speak’,
sp(r)æċ ‘speech’, p(r)æˉtiġ ‘clever, pretty’; and in unstressed syllables:
cwearte(r)n ‘prison’, bere(r)n ‘barn’ < bereærn ‘barley-house’. Such exam-
ples are isolated, however; more significant numbers of forms in which
one can detect a definite pattern of loss appear only in post-Conquest
records. Some of the earliest instances of /r/-loss are shown in (12).
(12) Early instances of /r/-loss before dentals:25
OE cerse - ME cesena ‘of cress. pl.’ (c. 1050)
OE gorst - ME gost ‘gorse’ (11th c.)
24
Bermúdez-Otero and Hogg (2003) provide a comprehensive account of the data on
anaptyxis in OE in relation to the theoretical issues of sound change.
25
Cited in Hill (1940). :
ON karskr - ME kaske (c. 1300) ‘active, vigorous’

OI fors - fosse (1440) ‘waterfall’
ME harsk - haske (1440) ‘rough’
ME mornings - monyngys (c. 1452) ‘mornings’
ME horse - hosse (1473) ‘horse’
Place names (1086 entries are from The Domesday Book):
OE Deorling-tun > Dallingtune ‘Darlington’ (1086)
OE mersc > Messe (1086)
ON Þorst- > Tostenland (1086)
ME Wurðsteda > Wosted (1330)
Pre-consonantal loss of /r/ had occurred in some OFr dialects and
in AN, attested in thirteenth-century rhymes such as sage: large, cors:
enclos (Pope 1961: §1,172). Variant forms such as AN morsel ~ ME mossel
(1290), ordinance ~ ME odinance (1389), parcel ~ passel (c. 1468), tarsel ~
tassel (1459) reinforce the native tendency towards /r/ + dental simpli-
fication illustrated in (12). To this one can add inverse spellings such
as OE wæter, ME water spelled warter (MED, 1156–7, 1463). A strong
indication of the potential for perceptual confusion of the sequence
/-rC/ comes also from some ME rhymes (Jordan 1974: §166; Ikegami
1984: 341).
(13) ME rhymes showing /r/-loss:
bass ‘perch’ (< OE bærs): wers: mess (1330)26
fors: clos ‘prison’ (1400–25)27
worst: adust ‘treated with heat’ (1450)28
ars ‘arse’: was: passe; neke-verse: casse ‘box’: fers ‘fierce’29
The common denominator in the early instances of /r/-loss is the adja-
cency of a dental consonant to the right of the rhotic (Hill 1940). The
path of loss Hill proposes is (a) assimilation, and (b) simplification of the
long consonant. Assimilation is indeed the most plausible account of
the process. Its likelihood is supported by the exclusion of heterorganic
consonants from the environment for early /r/-loss. Hill defined the set
triggering /r/-loss ‘loosely’ as [t], [d], [n], [l], [θ], [ð], [s], [ʃ], [d] and
[tʃ]: this is the full set of the coronal consonants of ME. This suggests
that the rhotic was also coronal, which does not tell us much about its
26
Guy of Warwick (Language of London/Middlesex border. LALME Middlesex
Linguistic Profile (LP) 6510).
27
Castle of Perseverance 1400–25, Norfolk, LALME: LP 58.
28
Cited in Wyld (1953: 298).
29
From the morality play Mankind, fifteenth century, Norfolk.
manner of articulation – all we can say is that there is a strong probabil-

ity of /r/ being an alveolar trill. We should also note that while bun-
dling together all coronals as triggering early /r/-loss is a convenient
generalisation, the data point to /-rs/- cluster-being simplification as
‘ahead of the curve’. In the words of a sixteenth-century commentator,
/r/-loss before /s-/ is a ‘widespread vulgarism’ found in East Anglian
texts.30
Labelling the process ‘assimilation’ does not address the question
of whether it was triggered by articulatory factors, perceptual factors,
the more abstract notion of syllable structure optimisation, or maybe
all three. The loosely shared place of articulation involving the blade
of the tongue for the rhotic and the dentals could be an assimilatory
factor. Both [ɹ] and [s] involve some lip-rounding, also conducive to
coalescence, yet articulatory similarity is hardly sufficient to account for
the entire development. The noise spectrum of sibilants compared with
other coronals after /r/ would contribute to stronger perceptual cues
for [s]. Perceptual place assimilation involving apicals is ‘invariably
progressive’ (Steriade 2001); following that approach, an assimilation
account must involve an intermediate stage of retroflex [s, !, "], and so
on, so [baɾs] > [bas], where the [-s ] is unstable and merges with the pre-
existing [s]. The articulation of retroflex consonants is more complex
than plain coronal/alveolar articulation (Hamann 2005), therefore one
phonetically likely change is loss of retroflexion: [bas] > [bas]. : In
terms of sonority, clusters of /r/ + obstruent are well-formed codas, so
that parameter offers no compelling phonological reason for simplifica-
tion. In terms of syllable structure, one can treat the end-result of the
assimilation as preference for simple codas over complex codas, and
/r/-loss rather than /s/-loss as preference for a sharper sonority drop
in the syllable coda – the transition from the peak to /s/ is preferable
to the transition from a vowel to /r/, in line with the postulates of ‘The
Coda Law’ (Vennemann 1988: 21). Since none of the potential triggers
excludes reinforcement from the others, a full account of the process
must evaluate each of these factors.
Surprisingly, the instances of pre-consonantal loss shown in (12) are
not traditionally considered precursors to a general /r/-loss in coda
clusters (Hill 1940; Wyld 1953: 298–300; Dobson 1968: §427; Lass
1992a: 66–7; Lass 1999: 114–15). One reason for that is that loss before
non-coronals – labials and velars – is incipiently attested only in the
late seventeenth century. Another reason for separating the two types of
/r/-loss is that early loss may leave a preceding short vowel unaffected,
30
Dobson (1957: 112); the commentator is William Bullokar (c. 1531–1609).
while later loss appears to trigger lengthening, thus ass < arse, cuss <
curse with short vowels vs arm, turf with long vowels in the non-rhotic
varieties. :
The occasional survival of an unlengthened vowel is only a tenuous
argument against continuity, however: early instances of /-rC/ simplifi-
cation often show PDE forms with a lengthened vowel, as in gorse, horse,
morning, Darlington. Of relevance also are the small number of pre-/-rC/
lengthenings in early ME, in which the vowel digraph spellings indicate
lengthening before -rC clusters, as in board (OE bord), hoard (OE hord),
earl (OE e(o)rl), hearken (OE he(o)rcnian), heart (OE he(o)rte), mourn (OE
murnen).31 Further, some arguably ‘late’ cases of /-rC/ simplification
fail to result in a long vowel in the peak: gal < girl, [kldi] for clergy,
[wdin] for virgin, and so on (Hill 1940). Predictably, in unstressed
syllables simplification of /-rC/ clusters does not trigger lengthening,
as in -wards [-wədz], or in scissors, colours, letters, lectern, yogurt in non-
rhotic varieties. It is likely, therefore, that the earlier and the later cases
of /-rC/ simplification represent a single historical process stretching
over more than six centuries and affecting different dialects and differ-
ent lexical items unevenly. Vowel lengthening is one possible outcome
rather than an essential stage in the process of /r/-loss. Delayed codi-
fication in the non-rhotic varieties, the occurrence of hyper-rhoticity,
as in warsh for wash, larst for last, incipient derhoticisation in essentially
rhotic varieties of English, and reversal to rhoticity in previously
categorically non-rhotic accents, are clearly points on one historical
continuum. :
While we can subsume early /r/-loss in coda clusters under the
general umbrella of assimilation, many questions remain: the nature of
the rhotic in various dialects, the exact distribution of /-rC/ forms, the
position of the cluster – within the same syllable, stressed or unstressed,
or straddling two adjacent syllables – the relevance of lexical frequency,
are some of the areas that await further research.
Another coarticulatory change involving /r/, and occasionally /l/,
resulted in variation between the voiced intervocalic dental stop /-d-/
and the fricative /-ð-/ in the onset of /-ər, -r/ and /-əl, -l/ syllables.
In front of syllabic /r/ the variation was first recorded in the fifteenth
century, when core-vocabulary items in etymological <-der> began to
be spelled <-ther>, the form in which they have survived to this day.
The list of original [-d(ə)r] words includes father, mother, gather, weather,
31
These items are usually ‘filed’ under ‘homorganic-cluster lengthening’ (see 6.4), but
they can equally well be pre-/r/ schwa-insertion, resulting in a lowered and length-
ened /r/-coloured vowel which resists the Long Vowel Shift (see 8.3.2).
hither, thither, together, whither. They were subject to extensive variation

in the sixteenth to eighteenth century. Britton (2007a: 261) cites the
forms <altherman> ‘alderman’, <elther> ‘elder’, and <chyltheryn>
‘children’ in Machyn’s written language (1550–63). In some items the
fricative is preserved only regionally today: adder, bladder, ladder, fodder
are attested with [ð] in Yorkshire, Leicestershire, Scottish and Irish
English (Jespersen 1909: 208–11). The variability between the stop
and the fricative can be motivated on both articulatory and perceptual
grounds. Fricative articulation and loss of occlusion of [-d-] before
an apical trill is assimilatory. It would be facilitated by the acoustic
similarity of [dr] ~ [ðr].
The reverse development, namely /-rðər, -rðr/ > /rdr/ as in burden,
murder, afford (OE byrðen, morðor, geforðian), is also attested. The other
liquid, /l/, can have a similar effect. One item in this group goes back
to OE, næðl ~ nedl ‘needle’; here belong also spider < OE spiðre, first <d>
spelling c. 1440, and fiddle < OE *fiðele, first <d> spelling c. 1450. The
change in these instances is dissimilatory, also facilitated by the confus-
ability of [d] and [ð] in the context of syllabic liquids. In murder, spider,
fiddle, the second syllable acquires a phonologically preferred onset,
since /d/ is more strongly consonantal than /ð/.
The discussion of the history of rhotics in English so far has revolved
around strategies of avoiding suboptimal transitions from and to an
/r/ and an adjacent consonant. Now we turn to general post-vocalic
loss, the process which justifies the consensus that rhoticity is the most
salient dialect criterion in PDE.
5.2.4 Post-vocalic /r/-loss

Early spellings, as in (14), indicating that coda /r/ in unstressed syl-
lables was unstable, have been taken as evidence that historically coda
/r/-loss was most advanced in prosodically weak syllables. :
(14) Loss of /-r/ in unstressed codas:
<mero> (1434), <merowe> (1475) ‘mirror’
<fathe> ‘father’, <mothe> ‘mother’ (The Cely Letters, 1472–88)
An account which associates consonantal weakening with a weak pro-
sodic position is appealing on typological grounds, yet the evidence
adduced in support of this hypothesis in the case of post-vocalic /r/-loss
in English is sparse. The examples in (14) are isolated; word-final <-r>
appears to be quite stable in the fifteenth and sixteenth century. The
hypothesis of early /r/-loss in unstressed syllables is largely based on
seventeenth-century orthographic insertion of <r>.
(15) Early <r>-insertion:

<winder(e)s/wynders> ‘windows’ (1601, 1613)
<feller> ‘fellow’ (Vernay Letters, 1639)
<pillars> ‘pillows’ (1673)
As Britton (2007b) argues, such spellings are just as likely, or even more
likely, to represent the analogical insertion/sandhi of an unetymologi-
cal rhotic in high-frequency common-core words in unstressed <-ow>
[-ə] words: arrow, borrow, fellow, narrow, and so on. The analogy is based
on the perceptual similarity of [-ə] and [-ɹ]. Britton points out that
equating the occurrence of early sandhi to early loss is problematic on
two scores: first, hyper-rhoticity in PDE is most prominent in those
dialect areas in England where rhoticity is still strong, and second, there
are no accents where rhoticity is in decline and where /r/-loss is better
attested in unstressed final syllables than in stressed syllable codas. To
this we can add the findings in Hay and Sudbury (2005) regarding loss
of rhoticity in NZE, where the decline of linking /r/ lags significantly
behind the development of non-rhoticity – an asymmetry which sug-
gests that spellings such as the ones cited in (15) are not necessarily
prima facie evidence for early /r/-loss. :
After the seventeenth century the evidence for coda /r/-loss becomes
somewhat clearer. The dating varies depending primarily on the data
source. Dobson found that ‘only sources which reflect vulgar pronuncia-
tion give evidence of the change before 1700, and even they give little’
(1968: §332). In stressed syllables ‘the change could not have occurred
in good speech before . . . c. 1800’ (§427). There is clearly a very impor-
tant socio-linguistic component to the change, where literacy is a major
factor. This is to be expected; good parallels with many more social var-
iables from PDE are offered by studies of ongoing decrease or increase
of rhoticity. The dating of the vocalisation in SSBE is most helpfully
surveyed in Windross (1994), whose analysis of the orthoepists’ com-
mentaries 1791–1836 shows that there cannot have been codification
of /r/-loss in standard speech in the eighteenth century. Variable
/r/-vocalisation in SSBE continued as late as the 1870s, and its spread
continued in NZE during the twentieth century, with female speakers
lagging behind in developing /r/-sandhi (Hay and Sudbury 2005).32
The accents in which the fully consonantal realisation of /r/ became
restricted to the syllabic onset are ‘non-rhotic’. Along with SSBE, non-
32
For ongoing changes in rhoticity see Hay and Sudbury (2005) for NZE; Lawson
et al. (2008) for Scottish English; Nagy and Irwin (2010) for AE. For variable /r/-
vocalisation in SSBE as late as the 1870s see Lass (1999a: 115); Trudgill (1999);
Mugglestone (2003: 86–94).
rhotic English is spoken today in parts of the US eastern seaboard and

the Gulf Coast, South Africa, Australia and New Zealand. The main
rhotic varieties are spoken in Scotland, Ireland, South-West England,
extreme West Lancashire, most of the USA and Canada. This is the
overall picture; as noted earlier, many regional and social varieties
show that the state of rhoticity is not fixed and allows progressive /r/-
vocalisation or reversal of /r/-loss. :
In terms of articulation, the vocalisation of /r/ can be accounted for
as a gradual separation of the gestural components of constricted /r/,
as proposed in Heselwood (2009). If the palatal gesture is temporally
delayed until after the pharyngeal gesture, it could lead to its loss in
non-prevocalic contexts. The suggested path of realisation of coda ⁄r⁄
is from bi-gestural to uni-gestural; the change from a rhotic to a non-
rhotic state is from a bi-gestural [ɹ], to bi-gestural [əɹ], to uni-gestural
[ə]. The perceptual similarity between the sonorant /r/ to schwa based
on the F3 has already been discussed. Thus articulatory and acoustic
factors go in tandem in advancing the vocalisation.
The phonological representation of <r> in, for example, doctor,
summer, poor, sheer, is a matter of debate because of complications sur-
rounding the realisation of /r/ in pre-vocalic contexts, where it is not
vocalised. Arguably, following Heselwood (2009), even in the most
advanced non-rhotic varieties of PDE, /r/ can be analysed as present
in all positions as an underlying segment, and word-final schwas are the
surface realisations of a vocalised rhotic. :
The vocalisation of coda /r/ in non-rhotic varieties is blocked if a
word-final <r> is followed by a vowel-initial word in the same pro-
sodic domain. This ‘preserved’ /-r-/ is widely known as a ‘linking’ /r/,
illustrated by the pronunciation of the <r> in the first column of (16).
(16) Linking /r/ in non-rhotic English:

summer in Spain vs . . . that summer.
poor Andy vs poor Mandy
sheer awe vs sheer boredom
Another process related to the history of coda /r/ is the insertion of an
unetymological /r/ in words ending in non-high vowels when the next
word or even the next morpheme is vowel-initial. This type of /r/ is
known as ‘intrusive’ /r/, a feature of non-rhotic varieties which is still
regarded as ‘incorrect’ and avoided by the ‘speech-conscious’ (Wells
1982, II: 284).
(17) Intrusive /r/ in non-rhotic English:

Fa[r] a long, long way vs fa, sol, la
Emma[r] is vs Emma saw

aphasia[r] and vs aphasia problems
withdraw[r]ing vs withdrawnness
Intrusive /r/ is typical of non-rhotic varieties, but not exclusively so. Its
appearance has triggered many different analyses, all of them in some
way relying on the perceptual similarity of syllables ending in non-high
vowels [ə, ɔ
, ɑ
,
] and syllables ending in alternating /r/ ~ /ə/. In
an analogical account the avoidance of intrusive /r/ after high vowels
is predictable because there is no similarity between /r/ and the high
vowels /i/ and /u/.33 The analogical mechanism behind the historical
spread of /r/-sandhi in non-rhotic varieties is worked out in Sóskuthy
(2010), whose extensive eighteenth-century corpus demonstrates the
strikingly higher density of <r>-final words compared with neutral- or
central vowel-final words (Emma, idea). Phonetic similarity and fre-
quency of the source of alternation are shown to result in the extension
of the alternation to words which never had etymological <r>.
Whether /r/-loss in word-codas is the same type of change that
affected the rhotics in pre-consonantal position is a hard question. The
general phonetic properties of the rhotics – their gestural complexity,
their high sonority, and their acoustic similarity to central vowels – are
clearly involved in both the loss of /r/ in /-rC/ clusters and in the loss
of coda-/r/. On the other hand, the actual mechanism of the loss is dif-
ferent: assimilatory in the case of pre-consonantal loss, and analogical in
the case of post-vocalic word-final coda loss. The processes seem to have
originated at different times, with assimilation before coronals occur-
ring first, though they do overlap eventually – late pre-consonantal loss
and early post-vocalic loss are both active in the eighteenth century,
and the variability of their realisation continues to this day.
5.2.5 The other liquid: Bridgestowe-Bristol. Historical parallels

between /r/ and /l/
In some varieties of PDE, namely SSBE and South African English,
the realisation of the lateral liquid /l/ is a paradigm case of allophonic
distribution: it is velarised to [l̃] in the coda, as in pill [pil̃], while the
‘neutral’ [l] is restricted to the onset, as in lip [lip] (see 2.3.1). In addition
to velarisation, some varieties of English show ‘vocalised’ /l/ in word-
33
Specifically for /i/: high front vowels have high first formants, rhotics have lowered
F2, usually lower F3, which would predict that the transition in /i-r/ would be opti-
mised by the insertion of a schwa, so /i - ə - r/ >> /ir/. On the cross-linguistic avoid-
ance of adjacent high front vocoids and rhotics see further Hall and Hamann (2010).
final position and pre-consonantally: for example, feel is realised as [fi

w],
cool as [ku
w], knealt as [nεwt], and people as [pi
pu]. The process of vocal-
isation is ongoing; in Britain it is most pronounced in the south-eastern
dialects, but it is observed also in AmE, AusE, NZE and Falkland Island
English, and in ‘near’ RP speakers and speakers of SSE and non-regional
AmE. : Moreover, /l/ is also active as an ‘intrusive’ segment, as in paw
is [pɔ
liz], homophonous Saul: saw, spellings such as <falcet, falcit> for
faucet, <papal> for papaw ‘grandfather’, <seasowl> for see-saw in AmE
(Gick 2002). Wells (1982, II: 344–5) reports that for some Bristol speak-
ers idea: ideal, area: aerial, mango: mangle, and tango: tangle are homophones.
The etymology of the name Bristol itself is based on OE brycg ‘bridge’ +
stow ‘place’. Sóskuthy (2010) reports on the much higher density of
word-final /-ɔ:l/ tokens in PDE (110,874) than word-final /-ɔ:/ tokens
(12,219), and uses this as an argument supporting the idea of analogical
extension of the intrusive consonant: /r/ is intruded after schwa because
of the density of /-ər/# tokens, while the likelihood of /l/ insertion is
highest after /-ɔ:/ and lags behind after other vowels.
The parallels between the two liquids are obvious; indeed, the
similarity of the effects of /r/ and /l/ can be traced back to the earli-
est stages of Germanic, where the IE syllabic sonorants developed in a
parallel way (see 3.4.2).
(18) Syllabic sonorants in PIE and Germanic:
/r/ → <ur> PIE mr-ti-, Gmc *mur-þra, OE morðer ‘murder’
/l/ → <ul> PIE pl-no-, Gmc *fulla, OE full ‘full’
Like <-r>, the lateral /l/ in OE is associated with OE Breaking (see
5.2.2, 6.5.3).
(19) OE Breaking before <lC>:
Source West Saxon Gloss
Go. all-s eall ‘all’
*halb
¯ oz healf ‘half’
*selho-z seolh ‘seal’
Breaking before /-l/ in OE is even more restricted than before /-r/: it
does not occur after /i-/, it is sporadic after /e-/, and its occurrence
after /a-/ > /æ-/ is limited to West Saxon and Kentish. Anaptyxis
(see 5.2.2) involving /l/ is attested in, for example, wylif ‘she wolf’
< wylf, yleca ‘ilk, same’. Similarly, in heterosyllabic /-Cl-/ clusters
the transition from the first consonant to the lateral liquid is facili-
tated by an intervening consonant: PrG *ainlif-, OE WS endleofan,
OE Northumbrian ællefne ‘eleven’, bræmlas > bræmblas ‘brambles’, OE
þȳml- > ME thimble; compare the epenthesis in, for example, Goth.
timrjan > OE timbrian ‘to build’, or, going outside Germanic, OFr
cha(u)mbre < Lat. camera. There are also cases of metathesis involv-
ing /l/, as in seld ~ setl ‘seat’, aðl ~ ald ‘disease’, spatl ~ spald ‘spittle’
(Minkoff 1972: 92–3; Hogg 1992a: 256 :
Like /r/-loss, the first instances of /l/-loss date back to the early
post-Conquest records.
(20) Early instances of /-lC/-loss:
OE ME Gloss
hwylċ hwich ‘which’
ælċ ech ‘each’
wenċel wenche ‘wench’
myċel muche ‘much’
These are examples from non-Northern thirteenth-century texts: The
Lambeth Homilies, Poema Morale, St Katherine. The loss must be traced
through velarisation of /l/ to [l̃] in the coda, where its perceptibility
would have been minimised when adjacent to the highly salient [-tʃ].
The high frequency of these words could also have been a contributing
factor to the loss.
Coda-loss of /l/ + other coronals is recorded early in high-frequency
words of low prosodic prominence: as(e) < ealswa ‘as’; compare with
stressed als, also. By the fifteenth century spellings such as shud, sud
‘should’ and wud ‘would’ indicate /l/-loss in the modals.
Another indicator of the instability of /l/ comes from unetymological
insertion of <l> in the environment back vowel + /lC/.
(21) Early unetymological <l-> insertion (data from MED):34
<palker> for <packer> surname < ‘packer’ (1282)
<walke> for <wake> ‘wake’ (c. 1384)
<salme> for <same> ‘same’ (a. 1399)
<salke> for <sake> ‘sake’ (c. 1400)
The next step, fully covered in documents from and after the fifteenth
century, involves loss of coda-/l/ flanked by a back vowel in the peak
and a velar or a labial in the second coda slot (Wyld 1953: 297; Dobson
1968: §425).
(22) Extended /lC/-loss:
Spelling Gloss
34
The MED uses a. (‘ante’) before a date to indicate the latest presumed date of com-
position, as distinct from c. (‘circa’), which refers to the approximate date of the
manuscript.
hafpenny ‘halfpenny’, note [hei-] pronunciation35

fok(k)(e) ‘folk’
scowk ‘skulk’
‘Loss’ in these instances implies vocalisation of the dark [l̃], eliminating
the constriction component and preserving the backness of its secondary
component. This development is attested earlier in Northern sources.
Its spread southwards, where it was apparently originally stigmatised,
was still under way at the end of the sixteenth century.36
The same mechanism of loss of constriction leads to complete loss
of single-coda /l/; it is attested after back vowels in the seventeenth
century in Northern, Scots and Northern Irish English. Indeed, <a’>
‘all’ is ‘the current spelling in modern literary Scots’ (OED), with pre-
cursors in seventeenth-century <pow, pou, pu> for ‘pull’, <faa, fawe,
fa> for ‘fall’.
An important, though often ignored contributing factor in the history
of /l/-loss in English is the fact that Old French and Anglo-Norman
had undergone an independent vocalisation of /l/. Vocalisation of [l̃C]
clusters, similar to [rC] clusters, started in OFr in the ninth century,
and although the change in AN seems to have been somewhat delayed
(Pope 1961: §§390, 1,179), the absence of /lC/ in bilingual speakers
would undermine the stability of the cluster in ME. Thus we find Pamer
(personal name, 1207) < AN palmer, paumer ‘palmer, pilgrim’, OE palm
‘palm’ < Lat. palma ~ OFr paume, ME sauder, sawder < OFr soud(i)er,
saudier ‘soldier’, sauf ‘safe’ < Lat. salvus. The retention of the /l/ in some
modern forms, as in fault, vault is an Early ModE reversal to the original
Latin form, and in some cases we find interesting alternations such as
Wat, Watson < Walter.
Again, as has often been pointed out (Jespersen 1909; Dobson 1968;
Lutz 1991, 1994; Lass 1992a; Sóskuthy 2010), the historical develop-
ments of the liquids have much in common, based on their shared
properties. Both liquids have high sonority and they are bi-gestural,
where the secondary dorsal component involves minimal obstruction
and is therefore perceived as more vocalic than consonantal, so that
some similarities in the processes triggered by /r/ and /l/ are phoneti-
cally and phonologically predictable. On the other hand, there are some
35
The loss of /l/ causes lengthening of the preceding vowel prior to its diphthongisation
to [ei].
36
In Love’s Labour’s Lost (V, i) the pedantic Holofernes says that he abhors ‘such fanatical
phantasimes, . . . such rackers of orthography, as to speak dout, fine, when he should say
doubt; det, when he should pronounce debt, – d, e, b, t, not d, e, t: he clepeth a calf, cauf;
half, hauf’.
notable differences: complete pre-consonantal loss of more obstruent-

like /l/ is less advanced historically, and it has not become the definitive
dialect marker that /r/-loss is considered to be; intrusive /r/ is more
advanced than intrusive /l/; and the environments in which intrusion is
most salient are different: schwa for /r/-sandhi (Emma[r] is) and /ɔ:/ for
/l/-sandhi (paw[l] is). :
5.3 Cluster simplification: KNIGHT-NIGHT, WRITE-RIGHT, IAMB-IAMBIC

Compared with OE, PDE is a language with relatively restricted sets of
consonantal clusters. Some early simplifications were discussed in 4.3:
LOE/EModE /sk/ > /ʃ/, thus OE sċip > ME ship, OE disċ, ME dish. In
5.1.3 we covered the loss of initial /h-/ when followed by sonorants: OE
hlot > ME lot, OE hring > ME ring, OE hnappian > ME nappen. Recall
also that OE was a language where geminate consonants were phone-
mically contrastive: OE hoppian ‘to hop’ vs hopian ‘to hope’ (see 2.1.4,
3.4.5, 4.1.2), but stem-internal geminates were lost in ME. The next two
sections will cover more instances of cluster simplification in English.
5.3.1 Initial <kn-, gn-, wr->: KNIGHT-NIGHT, WRITE-RIGHT

The tendency to simplify consonant clusters at word-edges, which
started with /h/-initial clusters and word-final geminates, spread to
other clusters in ME and after. Unlike the pre-fourteenth-century /hr-,
hl-, hn-/ simplifications, later simplifications are not reflected in the
spelling.
During the fourteenth century, initial <kn-> and <gn-> were prob-
ably still realised mostly as /kn-/, /gn-/. The evidence for that is
found in alliterative verse where [kn-] is matched to [k-], thus in Piers
Plowman (fourteenth century) Langland alliterates clothed: copes: knowen;
knytten: coler: commune.37 Orthographic anaptyxis is another diagnostic of
continuing separability of <kn-> in ME.
(23) Early anaptyxis in /kn-/ onsets (from LAEME):
<kinicht, cinth> ‘knight’
<kinf> ‘knife’
<keneleden> ‘kneeled’
Such spellings, though not very frequent, show that one way of resolv-
ing the non-cohesiveness of the cluster was to separate the consonants,
in the same way that the German surname of American popular culture
37
For more details see Minkova (2003: 330–9).
hero Evel Knievel became [kə ni

vəl]. Another strategy, the first signs
of which appear in the fourteenth century, is occasional simplification
by dropping initial <k->, or adding it unetymologically before [n],
suggesting uncertainty about the velar in the scribe’s speech.
(24) (a) Etymological /kn-/ spelled <n->:
<neyg> ‘knew’ c. 1300 (?c. 1225)
<(horse)nave> ‘stable-boy/-knave’ (1314)
<nobbes> ‘knobs’ (a. 1398)
(b) Etymological /n-/ spelled <kn->:
<knoynge> ‘noying, annoyance’ (a. 1425)
<know> ‘now’ (a. 1475)
The spread of the simplified forms continued in the fifteenth and six-
teenth century, and by the latter half of the seventeenth century, the
innovation was already recognised by the grammarians. Intermediate
stages in the simplification include [-tn], [hn], and possibly [nn] (see
Lass 1999: 123). ˚
The non-cohesiveness and, ultimately, the simplification of the
/kn-/, /gn-/ onsets is both phonetically and lexically grounded. In
terms of sonority, /kn-/, /gn-/ are well formed, but there are other
factors that undermine the stability of these clusters. A major factor is
the articulatory effort of producing a sequence of two non-continuants.
There are no other onset clusters in ME, or in PDE, that require both
of the adjacent onset consonants to be produced with full closure.
Distributionally, /kn-/ is isolated: after the loss of /hn-/, there were no
other velar + nasal onsets in the system. Perceptually, /kn-/ and /gn-/
were often confused, and the long transition of the velar into the nasal
allows alternative interpretations of the signal, as in (23) and (24). Not
least, the overall number of items with initial /kn-/ was not high, if one
excludes the verb to know, and the number of /gn-/ lexemes was vanish-
ingly small.
Two other initial clusters, /wl-/ and /wr-/, were also subject to
anaptyxis and simplification. Words with initial /wl-/ were very rare
in the OE and ME lexicon, and the only survival is lisp < OE *wlispian.
Early examples of splitting the cluster <wr-> can be found in OE
example wrohte ~ worohte ‘wrought’, as in (10a) above, early ME werangus
‘wrong’ (Jordan 1974: 148).
(25) Early anaptyxis in /wr-/ onsets:38
wreche ~ werche ‘wretch’ (13th c.)
38
For further examples and discussion see Minkova (2003: 365–8).
wartзe ~ wredзe ‘wrath’ (13th c.)

wernches ~ wrenches ‘wrenches’ (13th c.)
The other option of realising the original cluster was by simplification.

Some fourteenth-century evidence from southern texts suggesting
instability is shown in (26).
(26) (a) Etymological /wr-/ written <r->:

wrynkul ~ runkel ‘wrinkle’ (c. 1400)
(b) Etymological /wr-/ alliterating with /r-/:

riche: ryden: wrathe Piers Plowman (14th c.)
(c) Etymological /r-/ written <wr->:

wright ~ right ‘right’ (14th c.)
The process of simplification was more advanced in the southern varie-

ties and it probably proceeded step-wise. One possible path of change
involves an initial merger of the rounded labiovelar approximant [w-]
with the following coronal trill [-r-] into a doubly articulated, labialised
coronal *[rw-]. The complex articulation is then simplified to [υ], the
voiced labiodental approximant heard today in some RP speakers, in
popular London speech, and even further afield, making red and wed,
wreck and reck homophones (see Jespersen 1909: 354–5; Cruttenden
2008: 221). The [υ]-realisation must have persisted into the nineteenth
century, evidenced by substitutions of <r> by <w> as in Wichard,
Twinity, thwee. : The final step would be loss of the labial component
of [υ-] and a merger with the pre-existing variety of /r-/.
The simplification of /hn-, hr-, hl-, kn-, gn-, wr-/ in English results
in a consonantal system with clear phonotactic limits: except for the
sequence obstruent + liquid, initial clusters in English cannot involve
two non-continuants. Also, as noted in 2.3.1, except for clusters involv-
ing initial /s-/, onset clusters are well formed only if the sonority of the
cluster is rising, banning onsets such as */kt-/, */gd-/, */kf-/, */gv-/,
and so on. Deliberate preservation of original Greek /(f)θ-/, /sθ-/,
/tm-/, occasionally /pn-/, as in phthisis, sthenic, pneu-, tmesis (see 3.2), is
rare and does not affect the native phonotactics.
5.3.2 Final <-mb, -mn, -ng>: IAMB-IAMBIC, DAMN-DAMNATION, SINGER-

LINGER
Well-formed coda clusters involve a drop in sonority. All final clusters

in English, unless they involve coda /-s/ (apse, axe, ellipse, blitz, kibbutz,
larynx), show a falling sonority profile. Another property of English

coda clusters is that two voiced obstruents are disallowed at stem-ends,
adze being the single notable exception; this phonotactic constraint
does not apply across inflectional boundaries, thus sheds, legs with /-dz,
-gz/, robbed, obliged with /-bd, -dd/. Comparing the PDE with the OE
system, the most conspicuous difference is the simplification of final
<-mb, -mn, -ng>, reflected in standard speech. Like the initial clusters
<kn-, gn-, wr->, the history of these final clusters is revealed by the
spelling, see below. In both onset- and coda-cluster simplification, the
deletion affects the edge-most lower-sonority consonant and preserves
the sonorant adjacent to the syllable peak. :
The simplification of [-mb] is phonetically motivated by the fact that
it is a homorganic cluster: the pre-pausal labial stop would be realised
as an unreleased allophone [b#], with a weaker acoustic signal. The
identical place of articulation would favour assimilation, so perceptu-
ally the whole cluster is very close to the bare nasal [m]. The regularity
of <mb> in OE and early ME orthography suggests that the [-mb] was
more or less intact, primarily due to the fact that stem-final <-mb>
rarely occurred word-finally, which resulted in the cluster straddling
two syllables in the paradigm of the few eligible words: clim-ban ‘to
climb’, lom-bes ‘lamb, gen. sg. or pl.’, dum-be ‘dumb, pl.’ Even so, some
scribal uncertainty can be detected early: the DOE records instances
of <dum(e), duman> ‘dumb’; <ge-camde> ‘combed’ < OE cemban, and
inverse spelling such as <þumbes> ‘thumbs’ < OE þuma appear in the
twelfth century. 39 The paucity of rhymes and the limited number of
lexical items with stem-final <-mb> make the dating difficult, but it is
clear that by the beginning of the seventeenth century /-m/ was the
accepted coda for lamb, dumb, climb, womb (Jespersen 1909: 216).
The simplification of stem-final [-mb] in native words was matched
by the behaviour of the cluster in borrowed words. The base forms of
tomb (1225), succumb (1490),40 bomb (1588), rhomb (1575), hecatomb (1598),
aplomb (1828), iamb (1842) have lost the final [-b]. In the set of native
words the simplified shape of the base is completely stable and persists
in the derivatives, for example lambie, lambiness, lambly, lambskin, lamb’s-
wool; climber, climbing, climbable. In the loanwords the picture is more
complex: bomb and bombable have [-m], tomb, also entombing are always
39
The spelling <thumb> is predominant in the fourteenth century. For OE cruma
‘crumb’, the first <-mb> spellings in LAEME and the MED are fifteenth century. OE
lim ‘limb’ is not recorded with <-mb> spelling in LAEME or the MED; the modern
spelling spreads in the eighteenth century.
40
John Walker (1791–1826) still recommends that succumb should keep [-mb].
with [-m], but the adjectives tombic (1874), tombal (1900) retain the
[-b-]; rhomb and iamb have [-m] ~ [mb ¬], but rhomboid, rhombic, iambist,
iambic preserve [-b]. This is a case where multiple factors interact:
resyllabification of the stop into the next syllable, as in rhom.boid, iam.bic,
will favour preservation of [-b-], while lexical frequency will favour
preservation of the shape of the base in the derivatives, as in bombable
and entombing.
The cluster <-mn> in word-final position appears only in non-native
words: autumn, column, condemn, damn, solemn. In some instances of early
borrowings, the <-mn> was avoided from the start, thus Lat. hymnus,
Late Lat. ymnus, OE epenthetic ymen; in damn the input was Old French
dampn-er ~ damn-er, the ME adopted forms are predominantly with
<-mpn->, but <dam(e)> is already in evidence c. 1440; note also the
spelling <dambd> for ‘damned’ (1611). The phonetic nature of this
simplification is based on avoidance of similarity: the cluster [-mn] is
perceptually opaque because of the overlapping features of the two
adjacent nasals. As in the case of /-mb/-simplification, morphological
composition may affect the realisation: damner, damning have no [-n-],
but damnation (1300), damnable (1303) preserve it; hymner ‘a singer of
hymns’ is both [himə(r)] ~ [himnə(r)], but hymnal has [-n-]. We pre-
serve [-m.n-] in alumnus, alumna, but not in the shortened form alum
(1683). The realisation of [-n] requires a vowel-initial suffix, but it is not
sufficient: (rare) autumny, no [-n-], unlike autumnal, solemnity, where [-n-]
is resyllabified in the syllable onset.
Another case of historical cluster simplification affects the coda
cluster [-ŋg]. The phonemic inventory of the OE consonants in 4.1
lacks the velar nasal /-ŋ/ (called eng [εŋ], or agma [ægmə], or angma
[æŋmə]). The velarised allophone of the nasal is of course reconstructed
in words in which <n> was followed by a letter indicating a velar stop,
thus OE singan [siŋgən] ‘to sing’, tunge [tυŋgə] ‘tongue’, ðanc [θɑŋk]
‘thought’.
The historical loss of the voiced velar stop in [-ŋg] in word-final
position justifies the inclusion of the velar nasal /ŋ/ in the inventory
of contrastive sounds in ‘standard’ Present-Day Englishes. The distinc-
tive function of /ŋ/ is testable today in minimal pairs such as rang-ran,
thing-thin, sung-sun. The simplification of the cluster started in the north
of England, first in unstressed syllables, as attested by the rhyme fechtyn:
syn (Barbour’s Bruce 1375), gradually spreading south in the fifteenth
century (Jordan 1974: 162). The suffix -ing is at the forefront of the
change. The instability of [-ŋg] and the perceptual confusion between
/-n/ and /-ŋ/ in that suffix is shown in some early inverse spellings and
rhymes.
(27) (a) <ng> spelling for <n> (data from the MED):
<birthing> ‘burden’ a. 1400 (a. 1325)
<Hethyng (strete)> ‘heathen’ (1380)
<kelsyng> ‘keelson/kelson’ (1402)
<coming> ‘cumin’ a. 1450 (?c. 1421)
<chappinge> ‘chopin, half-pint’ (1455)
(b) <-ng> rhyming with <-n>41
<Mapyne: endinge> (2347:2349)
<serpentyne: endyng> (3171:3173)
<tyþinge: appolyne> (84:86)
The mechanism of the merger is similar to the development of /-mb/,

except that in this case the assimilation of the obstruent results in a
complex segment combining properties from both the nasal sonorant
and the obstruent, involving velar closure. The further option, namely
a complete loss of the velar coarticulation and retention of just /n/, has
been taken only partially. It is well attested with the unstressed suffix
-ing during the sixteenth century, when even Queen Elizabeth writes
<besichen> for ‘beseeching’, although teachers were urged not to let
pupils pronounce [in] for <-ing>. Well into the eighteenth century,
Wordsworth, Byron, Keats and Tennyson rhyme <-in>: <-ing> (Lass
1999: 120). In SSBE [-iŋ] began to be preferred in educated speech
by the end of the nineteenth century, and it is now the recognised SE
pronunciation. 42
Some forms which we have inherited in the simplified variant are
midden < ME midding ‘a refuse heap, dunghill’ and tarpaulin < tar + pall +
ing.43 A related development in late ME is the epenthesis of /-n/ or [-ŋ]
in the coda of unstressed syllables before the voiced velar stop /-g/ or
the affricate /-d/ in (mostly) loanwords: farthingale, harbinger, messen-
ger, passenger, similarly native nightingale. Dobson (1968: §438) calls the
nasal ‘a species of consonantal glide’, easing the passage between the
preceding unstressed vowel and the following consonant, but this is not
the sole trigger of the epenthesis which is possibly also analogical to
frequent native variability of [-iŋ] ~ [-iŋg].
The [-in] pronunciation of <-ing>, as in sittin’, walkin’, continues as
a variant in all of the central areas of England, plus peripheral areas of
41
From The Sowdone of Babylon (Yorkshire, end of fourteenth century).
42
The stigmatisation of [-in] both in British and American English is undoubtedly due
to orthographic prescriptivism. On the preservation of [-ŋg] in the suffix -ing in places
like Birmingham, Manchester and Liverpool see Wells (1982, II: 365).
43
The OED warns that this etymology must ‘remain conjectural’.
the north, the south-western peninsula, the eastern counties and Essex
(Trudgill 1999: 48). In the southern US [-ŋ] is replaced by [-n] in many
socially prestigious varieties, while [-ŋ] is the norm in many working-
class varieties. More typically, however, the [-n] pronunciation for [-ŋ]
in -ing is associated with lower-working-class speech habits; it is also a
typical marker of deliberately affected casual speech style.44
In stressed codas, the replacement of [-ŋg] by [-ŋ] is supposed to
have occurred later than in -ing, but in The Sowdone of Babylon we find
everychone: among, distruccion: wronge; compare with (27b). The existence
of both [-ŋg] and [-ŋ] in stressed syllables was recognised by contem-
porary commentators in the middle of the seventeenth century. As in
PDE, consistent [-g]-dropping was observed only word-finally (Luick
1964: §767; Lass 1999: 119).
The distribution of contrastive /ŋ/ prompts some interesting pho-
nological questions. Like PDE /h-/, which can appear only in onsets,
/-ŋ/ is a ‘defectively distributed’ phoneme: it can be distinctive only
in coda position. The restriction reflects its historical origin since it
is only through the loss of the voiced velar stop in the coda that /ŋ/
became contrastive: kin-king, ban-bang, run-rung. Before /k/, as in plank,
sink, hunk, [-ŋ] remains a positional allophone followed by the voiceless
stop. Thus, we get a three-way opposition in pin [pin]-ping, [piŋ]-pink-
[piŋk], sin, sing, sink, and so on.
Another peculiarity of contrastive /ŋ/ is that it has to be domain-
final, that is, the [-g-] is preserved stem-internally, rendering [-ŋ-] allo-
phonic, as in Bangor, bingo, tango, single, hungry, Hungary, all with [-ŋg-].
Note that the preservation of [-g-] does not depend only on syllabifica-
tion, because in forms derived with -ing or the agentive suffix -er, the
[-g-] of the stem is not realised, and the derived form copies the shape
of the base form: singing, singer with just [-ŋ-]. In addition to the most
frequent -ing and -er, the majority of the native suffixes such as -y, -dom,
-hood, -ness, -ish, -less, -ling, also -let (OFr), preserve the shape of the base:
slangy, kingdom, thinghood, youngness, strongish, fangless, kingling, ringlet have
[-ŋ-]. The addition of a comparative suffix, -er or -est, as in longer, strong-
est, youngly, however, results in heterosyllabic [ŋ].[g]. The addition of
Latinate suffixes generally preserves the [g], as in fungation, diphthongal,
but not always, as in ringette, nothingism. Then there is vacillation with
44
A study of the substitution of [-ŋ] by [-n] in words ending in <-ing> by Detroit
speakers shows that [-n] forms are used by 19.4 per cent of the upper middle class, 39.1
percent of the lower midddle class, 50.5 per cent of the upper working class and by
up to 78.9 per cent of the lower-working-class speakers (Wolfram and Schilling-Estes
2006: 174–5).
some derivatives: prolong and prolonging are always just [-ŋ], prolongation
varies between [-ŋ-] and [-ŋg-], and prolongate is always [-ŋg-]. There
are some idiosyncrasies: dinghy, hangar allow both pronunciations, and
so do English and England. The behaviour of [-ŋ] in clitic groups and
phrases is also variable; the variability of [-ŋ] is of continuous theoreti-
cal interest. : In our diachronic context, the deletion of the [g] and the
possibility of phonemic /ŋ/ in narrowly defined contexts in late ME
and EModE is most notable as an addition to the inventory of contras-
tive sounds in English.
Although there are no [ng-] or [ŋ-] words in the native Germanic
vocabulary, or in loans from Latin and its descendants, the OED lists
more than twenty <ng-> words, all of them first recorded in the last two
centuries: ngaio ‘evergreen shrub’, ngaka ‘doctor’, ngapi ‘Burmese pressed
fish’, Ngbandi, ngiru-ngiru, and so on. The accommodation of the initial
cluster in such words can result in three different pronunciations.
(28) Accommodating initial <ng-> in PDE:
[n-] ngaio ‘NZ shrub’, ngawha ‘hot spring’ (Maori)
<ng> [ŋə-] ngapi ‘Burmese pressed fish’ (unassimilated)
[əŋ-] ngwee ‘Zambian monetary unit’, ngege ‘fish’, Ng (name)
In borrowings from Maori, SSBE and AmE favour simplification to [n-].

Occasionally, the dictionary records unassimilated pronunciations, as in
ngapi [ŋə´pi
], ngaio [ŋaio], ngawha [ŋafa]. By far the most common strat-
egy, shown in the bottom row, is the insertion of a schwa to the left of the
cluster, making a new unstressed syllable, placing the cluster conveni-
ently in the syllable coda and therefore within the pattern of native [-ŋ].
A similar strategy of schwa insertion can be observed also with the initial
cluster <mb->: mbongo ‘a political stoooge’ (from Zulu and Xhosa), mbari
‘an extended family unit’ (from Swahili) pronounced [əm-].45
The processes discussed above present only a partial picture; the
focus has been on the main innovations separating ME from PDE.
There are a number of ‘smaller’ changes that affect a limited number of
items. In ME the final [-tʃ] of OE ānlic ‘only’ and OE ic ‘I’, -lic(e) -‘ly’
were gradually dropped. The bilabial approximant [w] was lost in the
onset cluster [sw-] followed by a back vowel: OE swā, ‘so’, OE swylc, ME
suche ‘such’, similarly OE sweord ‘sword’, twā ‘two’, where the spelling
does not reflect the change. Along with the simplification in nasal +
45
The initial <x> in Xavier [zeiviə(r)] is sometimes taken to represent [ks-] as in X-ray,
triggering prosthetic [ε-/ə]: [εgzeiviə(r)] (see 3.4.1 for [ks] > [gz]).
non-continuant coda clusters (-mb, -mn, -ng), -nd tends to lose the dental,
as in ME scande/scanne ‘scan’, v. < Latin scand-, woodbine < OE wudu-
binda, which gives rise to hypercorrect sound < OFr soun, pound, v. < OE
punian, similarly compound, expound, tyrant, peasant, ancient. :
Figure 5.4 presents a survey of the major consonantal changes
covered so far. It shows typical spellings for the consonants and conso-
nant clusters in ME, followed by ME examples, the reconstructed value
of the consonant(s) and the PDE realisation. :
ME Representative Spellings Examples 14th-c. PDE
Value Value
sigh(en) ‘sigh’ [-ç]/[-h]

Ø
g, h, gh, hg, ȝ, ȝh, ch, gȝ, ȝg bough ‘bough’ [-x]/[-h]
ynogh ‘enough’ [-x]/[-h] [f]
helpen ‘help’ [h-/Ø] [h-]
h-, Ø humour ‘bodily fluid’

(h)ure ‘hour’ Ø [h-]/Ø
whete ‘wheat’
wh-, w-, qh-, qw-, qwh-, ȝw [hw-]/[w-] [ʍ]/[w]
whil(e) ‘while’
gn- gnawynge ‘gnawing’ [gn-]

[n-]
kn- knyght ‘knight’ [kn-]
wr- wroght ‘wrought’ [wr-] [r-]
-mb dumb ‘dumb’ [-mb] [-m]
-mn autumn(e) ‘autumn’ [m(p)(n)] [-m]
-ng song ‘song’, yonge ‘young’ [ŋg] [-ŋ]
-lC- half ‘half’, almesse ‘alms’ [l] Ø/[l]
-rC-, -r# farm, ferther ‘further’ [r]/ [-r-]/Ø
Figure 5.4 Fourteenth-century consonants and their PDE reflexes

The consonantal system has moved from more complex to simpler
phonotactics at syllable edges. This contributed to the progressive
disjunction of written and spoken English in the last six centuries, most
often commented on in connection with the vowels. The changes are
driven by articulatory and perceptual factors, in spite of the persistence
and standardisation of the conventions of etymologically based orthog-
raphy. The only notable exception to this pattern is the reversal of the
ME /h-/-less realisations in Romance loanwords of the type humour,

humble, where the model of regionally stable /h/-fulness, and the ortho-
graphic preservation of <h> in (re)-borrowings from Latin, result in a
very restricted lexical set of /h/-less stems: heir, honour, hour, GA herb.
The reinstatement of /h-/ was facilitated by the large number of /h-/-
initial Renaissance Greek loans (see 5.3.1) and the intrinsic phonetic
weakness of /h-/, which makes the alternation Ø ~ /h-/ perceptually
innocuous. In all other instances, the realisations selected in PDE are
phonetically and phonologically motivated.
5.4 Other inventory changes: the adoption of //

As noted in 2.1.3, the palatal fricative [] was a late addition to the
consonantal inventory of English. It is also the rarest of all English
consonants (Cruttenden 2008: 232), and indeed of all English phonemes
(Ladefoged 2005: 91). Its distribution is limited to loanwords, except
across word-boundaries (see (31) below).
In ME borrowings from OFr, words which have Modern French //
appear regularly as /d-/: jargon, juice, jealousy, abridge, rage, cage, courage,
plunge, village are all early loanwords. The simplification of the OFr affri-
cate /d/ to // was in progress in the thirteenth century in Continental
OFr, but the parallel change appears to have been delayed in Anglo-
Norman (Pope 1961: 93–4, 450). Thus in the fourteenth and the fif-
teenth century the sound [] would have been familiar to a minority of
well-travelled speakers of French, but [] in loanwords would still be
assimilated to the pre-existing /d/. The potential for [d]- pronuncia-
tions is reinforced by the very robust historical presence of [d] in core
vocabulary words like bridge, edge, ridge, which go back to OE. Indeed,
as we will see below, the tendency to assimilate // to /d/ in post-
ME loanwords has continued to this day; compare PDE garage, prestige,
melange with both // and /d/ as recognised ‘standard’ pronunciations.
Exposure to French loans with // was only one of the sources of []
in English. The allophone [] could arise in borrowed words in which
the voiced alveolar fricative [z] was followed by a palatal glide [j], trig-
gering assimilatory palatalisation of [z] to [] with concurrent loss of the
palatal glide, as in division, treasure, usual. The time-depth of this assimi-
lation is hard to judge, because neither spelling nor rhymes are informa-
tive, but orthoepistic evidence from the middle of the seventeenth
century onwards indicates that the palatalisation was common, though
not stable. : Some examples illustrating the continuing variability of
the results of palatalisation are shown in (29). As usual, parentheses
indicate the first recorded entry date in the OED.
(29) Palatal assimilation: [z] + [j] > []

[] only [z] ~ [] [z] only
measure (1225) osier (1175) symposium (1586)
confusion (1290) glazier (1385) gymnasium (1598)
vision (1290) nausea (1425) paradisian (1615)
occasion (1382) Elysian (1579) trapezial (1681)
closure (1386) transient (1607) magnesium (1781)
usual (1387) amnesia (1786) nasion (1879)
In terms of the consonantal system as a whole, the addition of [] can be

seen as a process which would have been facilitated by the functionality
of the voicing contrast in the obstruent set: the stops /p/-/b/, /t/-/d/,
/k/-/g/ and the affricates /tʃ/-/d/ were paired in voicing from OE
onwards. In ME voicing became contrastive for the fricatives too (see
4.4), thus /f/-/v/, /θ/-/ð/, /s/-/z/. After the palatalisation of /sk/ to /ʃ/
in late OE, early ME, the voiceless palatal fricative was the only voiceless
obstruent unmatched by a voiced counterpart, except for /h/, which, as
shown in 5.1, is quite different from the other obstruents. It is hard to
argue that the balance of the system with respect to obstruent voicing
contrasts is the driving force behind the rise of //, but it undoubtedly
increases the likelihood that the speakers would perceive, produce and
learn the /ʃ/-// contrast. As we will see in 5.4.1, the palatal assimila-
tions in the first column of (29) are not isolated either – they are part of a
more general process of coronal dental palatalisations in English.
The integration of // in the consonantal system has been progress-
ing gradually in the last centuries as new borrowings continue to intro-
duce the speakers to items with [-] in initial and final position: zho ‘a
Tibetan bovine’ (1841), Gitane, Giselle, Zdanovism/-ist (1957), zhoosh
(1977); ménage (1325),46 cortège (1679), espionage (1793), beige (1858), cam-
ouflage (1917). The word-initial occurrence of [-] is more restricted and
it tends to be more prone to variation between the palatal fricative and
the affricate: genre, gendarme, gigolo, georgette show [] ~ [d-]. If periph-
eral vocabulary additions, for example gigue (1685), (au) jus (1865), (beau)
geste (1920),47 become frequently used items, they may slow the rate of
nativisation, strengthening the position of // in the system. It is likely
that the globalisation of English will bring in more //-initial words,
as suggested by a long list of <zh>-initial words on the (unmonitored)
46
N.E.D. (1906) gives only /meinɑ
/, though OED allows // ~ /d/. Retention of
word-final // is matched by the retention of the stress on the final syllable, reinforc-
ing a word’s ‘foreignness’; compare ménage with manage, v. (1561).
47
Compare jig, ‘often assumed to be identical with Old French gigue’ (OED), juice
(c. 1290), jest (1300).
online Urban Dictionary.48 The extent to which // signals ‘foreignness’

therefore varies greatly among the speakers of PDE.
5.4.1 More alveolar palatalisations and affrications: s-, t-, d- + -j.

GOTCHA, INJUN
The development of [z] + [j] to [] shown in (29) is part of a more

general pattern of palatal assimilation in EModE affecting all alveolar
obstruents in that environment.
(30) EModE palatalisations and affrications of alveolar obstruents +
/j/:
[s j-] > [ʃ]: mission, sugar, passion
[-z j-] > [ ]: occasion, derision, measure
[t j-] > [t ʃ ]: mature, mutual, rapture, sculpture
[d j-] > [d ]: soldier, verdure, procedure
Of the four parallel developments in (30), the palatalisation of the frica-
tives appears to have occurred first. For [s] + [j] it is attested in cor-
respondence already in the fifteenth century: sesschyonys (Paston Letters,
1422–1509), consederraschons, oblygashons (The Cely Letters, 1472–88). The
change would be more advanced in unstressed syllables. We can assume
that the stressed syllable onset sequence [sj-] in ME had not coa-
lesced into /ʃ/: a word like ME sure would have /sj-/ and it would be
homophonous with PDE sewer. In stressed position the preservation of
[s] is suggested by fourteenth-century alliterative pairings such as suren
‘to assure’: sithen ‘since’: serven ‘to serve’ (PP 5.540) and asoyled ‘absolved’:
surely: sette (SGGK 1883).
Tautosyllabicity of the alveolar and the glide was the first step
towards palatalisation; the completion of the process involves loss of the
glide /j/. The sequence alveolar obstruent + [ju-] in primary stressed
onsets, as in suit, pursue, zeugma, Zeus, tune, Tuesday, due, dune, can be sim-
plified either by the absorption of the palatal into the consonant, as in
sure, sugar with [ʃu-], and tune, Tuesday with [tʃu-], or by preservation of
the features of the alveolar and loss of palatality, as in suit, Sue, suet with
[su
], due, dune with [du-]. In the sequence [sju-], as Walker (1791–1826:
60) noted, the only two items in English with stressed stem-initial [sj-]
to [ʃ-] change are sure and sugar and their derivatives. :
In unstressed position the sequence [-sj-] occurred most commonly
in the French suffix sio(u)n/cio(u)n, where /s-/ was a singleton onset and
the suffix was disyllabic; the orthographic was [i] and not a glide,
48
<http://www.urbandictionary.com/> (last accessed 23 May 2013).
so in Chaucer condicion, devocion, nacion, proporcion still had [-si.ən], not

[-sjən]. In Chaucer and throughout the fifteenth century the suffix was
disyllabic, rhyming on -on, for example destruccion: person, proporcion:
upon, scorpioun: confusioun. Disyllabic <–ion> [i.ən] continues into the
Renaissance, though only as an archaic feature in verse. :
The affrication of the alveolar stops was first reflected in informal
sixteenth-century spellings sawgears ‘soldiers’, seventeenth-century teges
‘tedious’ (cited in Lass 1999: 121). It was still a matter of debate in the
eighteenth century. Walker (1791–1826: 61) makes a distinction between
the behaviour of the dentals in stressed and unstressed syllables: he
writes that nature, creature, feature, fortune ‘have the t pronounced like ch, or
tsh’, but he objects vigorously to word-initial [tʃ] in tutor, tumour, tumult,
which his contemporary, the actor and elocution guru Thomas Sheridan
(1719–88) respelled in his Dictionary (c. 1774) as <tshootur>, <tshoo-
mour>, <tshoomult> – what was respectable from the stage had not
reached acceptance in the more purist academic circles. Walker disap-
proves of [tʃ] before native suffixes: he expects [t] in (thou) pitiest, mightier,
twentieth, and so on. He recommends the affricate in righteous, plenteous,
piteous, where the last two items have reverted to /t/, yet he opts for [tʃ]
in bestial, celestial, frontier, where the PDE norm is [t] + [i]. The variabil-
ity of [tj] ~ [tʃ] continues, and the recommended pronunciations can
be puzzling, so the OED lists latitude, longitude only with [-tju
d], while
multitude is given as both [-tju
d] and [tʃ(j)u
d]. For non-word-initial
[-dj] Walker (1791–1826: 50–1) says that soldier, grandure, verdure are
‘universally and justly pronounced as if written j’ (= [d] in his system);
for stressed-syllable initial [dj-], as in duke, reduce, his verdict is that the
pronunciation with [-d] ‘cannot be too much reprobated’.
The degree of acceptance of the innovative palatals and affricates still
varies, and their realisation often differs in casual and in careful speech.
For over-cautious speakers the palatal assimilations can be reversed for
some lexical items, probably an influence from the spelling, so one can
hear appreciate, negotiate with [-s-], Parisian, Tunesia with [z-], literature,
overture, mature with [–t-], ordure, endure with [-d-]. In some instances
doublet forms are reflected in the spelling and one of the forms, not
necessarily the one showing palatal assimilation, has become a distinc-
tive stylistic marker, as in creature ~ critter (1815), Indian ~ Injun (1812),
idiot ~ eejit (1919).49
Regressive palatal assimilations can occur in PDE across word
boundaries.
49
In informal speech the dentals in initial <tr-> and <dr-> can undergo palatalisation,
for example [tʃrai] for try, [drai] for dry.
(31) Palatalisations and affrications across word boundaries:

[-s] + [j-]: [ʃ]: this year, unless you
[-z] + [j-]: []: as you say, as yet
[-t] + [j-]: [tʃ]: got you (gotcha), not yet
[-d] + [j-]: [d]: had you, said yes50
Above the word-level, the assimilation of the alveolar obstruents to
a following palatal is even more gradient than within the word. The
presence of [j-] at the beginning of the next word is not a sufficient
condition for the assimilation. If the two words do not form a unified
prosodic domain, a clitic group with a single stress, but preserve their
independent stress, palatalisation and affrication do not occur. Thus,
there is usually no change in slow speech to the final [-s] in paints
yellow, the [-z] in reads Yeats, the [-t] in liked yoga and the [-d] in rigged
yacht, though coarticulatory effects do occur, especially in fast, relaxed
speech.
5.5 The glottal stop [ʔ] and the alveolar tap [ɾ]
The history of [ð] ~ [d] variation in conjuction with /r/, as in PDE
father < OE fœder, PDE burden < OE *byrðen was covered in 5.2.3. This
section turns briefly to two other processes: glottal stop insertion and
substitution, and the intervocalic tapping of /t/ and /d/ to [ɾ]
5.5.1 The glottal stop

The glottal stop [ʔ], articulated with the vocal cords pressed together,
sometimes likened to the sound of coughing, is not part of the inven-
tory of contrastive consonants in English (see 2.1, 2.1.2). It is, however, a
sound of long Germanic lineage. It is commonly reconstructed as present
in the onset of orthographically vowel-initial stressed syllables in OE.
The OE alliterative practice of matching non-identical stressed vowels:
ǣnne ofer yˉ ðe / umborwesende ‘alone over waves / child-being’ (Beowulf 46),
suggests strongly that the insertion of a glottal stop before the stressed
syllables [ʔæ
nnə
ʔy
ðə
ʔυmbor $wεzεndə] was what satisfied the
50
Hence ‘round yon virgin’ in the Christmas carol lyrics is (mis)heard as ‘round John
Virgin’, or wouldn’t you becoming wooden shoe (wooden chew?) in the once popular chil-
dren song Mairzy Doats (Mares Eat Oats). Such reanalyses are popularly known as
mondegreens. The word mondegreen itself has its origin in such a ‘mishearing’. It is based
on ‘the name Lady Mondegreen, a misinterpretation of the phrase [they have slain the
Earl of Murray and], laid him on the green in the ballad “The Bonny Earl of Murray”’
(OED).
identity requirement in onsets, so strictly observed elsewhere in the OE

and the Germanic alliterative corpus. The phonological justification
for the insertion of a consonantal onset is syllable-structure optimisa-
tion – throughout its history, the vocabulary of English shows a strong
preference for onset-ful stressed syllables. :
The insertion of a glottal stop in stressed vocalic onsets became much
more variable in ME, allowing maximal onset resyllabification across
word-boundaries in clitic groups. A late fifteenth-century gloss of Lat.
spica ‘spike, an ear of corn’ reads Hec spica: a ner; compare also the history
of tother < þet oþer, nuncle < an uncle, newt < an ewt. Such resyllabifica-
tion could occur only if the onsets of ear, other, uncle, ewt were empty.
Absence of glottal stop insertion is also associated with the development
of allomorphy: the loss of [-n] in the OE numeral ān ‘one’, weakened to
an before vowels and a before consonants. Loss of [-n] in the OE forms
mīn ‘mine’ and þīn ‘thine’ is involved in the alternation of my-mine and
thy-thine. The selection of the adjectival pronominal forms in EModE
was based on the nature of the following sound, as in my sister vs mine
uncle.
The more recent history of pre-vocalic glottal-stop insertion is
related to style of delivery. In SSBE and GA the glottal stop is realised
in stressed vocalic onsets after a syntactic break, as in the exclamation
spelled uh-uh, meaning ‘no’. It is a feature of emphatic pronunciation
of stressed vowel-initial syllables, as in is this art [ʔɑɹt]?, the word’s ample
['ʔæmpl], not sample. In RP the insertion is more widespread, occurring
between an unstressed and a stressed vowel in hiatus as in co-opt, reac-
tivate, pre-aortic. Since there is no Roman letter for [ʔ], no glottal stop
in Latin or French, and the sound is not contrastive, the testimony of
the orthoepists is scant, but it was apparently noted in the seventeenth
century. Glottal ‘reinforcement’ is implicit in descriptions of singing
techniques referring to ‘hard attack’ in vowel-initial words (Cruttenden
2008: 181).
Another process involving the glottal stop is glottalisation, the sub-
stitution of a coda voiceless stop by a glottal stop: that [ðæʔ], lap [læʔ],
sack [sæʔ], similarly intervocalic [t] in bitter [biʔə], matter [maʔə],
whereby the oral closure is weakened to a very short glottal closure.
The glottal stop realisations are more common in British varieties
of English. In RP the strongest tendency for replacement affects /t/
before homorganic consonants: great table, sat down, not now. The envi-
ronments for the replacements of [t] are broader for some speakers,
and there is evidence that [ʔ] for [t] is no longer stigmatised in London
Regional RP.
The chronological depth of glottal substitution is hard to assess since
there are no contemporary commentaries before the second half of the

nineteenth century. The replacement is considered primarily British,
and of recent vintage, since it is absent from AusE and not present in
older speakers of NZE, though it is showing up in the pronunciation
of younger NZE speakers.51 Recent reports on AmE also point to a
trend towards word-final glottalisation of /t/, more pronounced before
consonant-initial words (submit five, fault because), but occurring also
before vowels (lot of, that any, shut up, that a). This innovative tendency
is most strongly associated with the western US, with younger and/or
female speakers (Eddington and Taylor 2009; Eddington and Channer
2010). The full diachronic dimension of this change has yet to be
researched, but as Wales (2006: 176–7) points out, it probably originated
much earlier in the north and could have been latent in many speech
communities.
5.5.2 Voicing of [t] and tapping of [t] and [d]: MATTER-MADDER52

Another pattern of replacement, characteristic of AE, found also in some
Northern Irish varieties of English, is the realisation of the dental stops
/t/ and /d/ as the alveolar approximant tap [ɾ] between (a) a stressed
and an unstressed vowel word-medially, as in ladder [læɾ], waiter
[weiɾ], bidder [biɾ], setting [sεɾiŋ], (b) between two unstressed
vowels as in abnormality [$æb$nɔrmæləɾi], capitalise [kæpiɾə$laiz], or (c)
between vowels across a word boundary, as in at ease [əɾi
z], get úp! [gεɾ
p]. Phonetically, tapping involves voicing and a single rapid contact
of the tongue with the alveolar ridge. Since the replacement affects
both /t/ and /d/, the allophonic [ɾ] can neutralise the surface forms
of lexically distinct entities, resulting in homophonous pairs such as
matter-madder, both [mæɾ], kitty-kiddy, both [kiɾi] or [kiɾi].
The change to [t] to [ɾ] may be reconstructed as proceeding stepwise
from intervocalic /t/ > [d] > [ɾ]. Possible very early indications of
voicing are found in OE botm, ME variant spellings boþom, boddom, bodme,
boden, PDE Northern English and Scots bodome, bodom, bodum, boddem
(OED); LOE witter ‘wise’ spelled wiþer (MED, c. 1275); water spelled
warter, wader (MED, 1156–7, 1463); federyn ‘fetter’ (1440).53 Jespersen
(1909: 340–1) cites pottage > porridge from the early sixteenth century,
51
See Cruttenden (2008: 180–1) for AusE; Trudgill (1999b: 236) for NZE; Wales (2006:
175–7) for Northern English and the glottalising influence of urban Scottish English.
52
Although in some languages flaps and taps are functionally distinct, there is no such
distinction between them in English. The process described in this section can be
referred to as either tapping or flapping.
53
The MED labels <wader> an ‘error’.
also neverberrer ‘never better’ in Dickens, nor a bir of it ‘not a bit of it’
in Meredith, wasermarrer ‘what’s the matter’ in Jerome, representing
speech under the influence of alcohol or sleep. Variability between
[d] ~ [ɾ] is the most likely account for the current form of paddock <
OE pearroc ‘park, enclosed place’; the [d] and [r] forms have coexisted
since the middle of the sixteenth century. Although tapping is com-
monly considered an innovation in AmE, such data justify the assump-
tion that tapping in AmE is a feature which was indeed inherited from
the mother country, as suggested by Haugen (1938) and Montgomery
(2001: 139). Tapping is also attested in Australian and New Zealand
English. Its robustness is due partly to the current lack of social stigma
associated with it, which must have facilitated its codification. :
5.6 Recent trends: [ts-, ʃm-, ʃl-, ʃt-], MASH POTATO, MANAGE CARE,
STAIN GLASS
Our diachronic survey of the English consonantal system will con-

clude with observations on some recent trends of enrichment and
simplification.
Like <ng->, <mb->, the homorganic onset <ts-> is not found in any
native vocabulary items, nor in loanwords from French or Latin. Initial
/t/ in PDE can only be followed by /-r, -j, -w/ in the native vocabulary,
so /ts/ is phonotactically non-native but not universally unattested. The
earliest <ts->-initial borrowing in English is from Slavic: tsar (1555);
the voiceless alveolar affricate onset [ts-] in this word and its many
derived forms is most commonly assimilated to a singly articulated [z-],
though the affricate [ts-] pronunciation is also recorded. Only three
more [ts-] words were added in the sixteenth to seventeenth century,
nine in the eighteenth century, and twenty-three between 1901 and
1975, with sources from languages from all over the world: Burmese,
Chinese, German, Greek and Japanese. All of the most recent borrow-
ings preserve the [ts-]; apparently the phonotactic constraint which
generated the [z-] in tsar is no longer part of the system: no one would
say [zεtsə] or [zu
nɑmi]. In other words, [ts-], although of more recent
lineage, is likely to join [] (see 5.4) as an addition to the consonantal
inventory. The still marginal acceptability of a [ts-] onset can be related
to its relative complexity and lower frequency of occurrence, though
tsunami is hardly a rare item in English after 2011.
The situation with initial <s(c)hm->, <-s(c)hl> <sht->, as in
schmooze, schmeer, shlep, shlock, shtick, shtetl is somewhat different, in that
these clusters identify the origin of the words not just as non-Anglo-
Saxon or Romance, but as specifically Yiddish. After World War II,
Yiddish is spoken by more people in English-speaking countries – US,

Canada, Australia and Great Britain together – than in Continental
Europe or Israel. Large Yiddish communities in New York, Los Angeles,
Melbourne and Montreal have contributed to the recognition and inte-
gration of new vocabulary and new consonant clusters: [ʃm-, ʃl-, ʃt-],
which have become productive phonaesthemes in PDE.54 With at least
a quarter of a million Yiddish speakers in North America, and a strong
presence of Yiddish culture in film, TV and literature, familiarity with
these clusters is to be expected and there is no attempt to assimilate
them to some native sequence. What is more, the onset <schm-> [ʃm-],
treated as a ‘combining form’ since 1929 by the OED, is clearly produc-
tive in playful reduplication, generating mildly disparaging or comic
attention-getters: madam-shmadam, able-shmable, holy-shmoly, the Libros
Schmibros lending library in Los Angeles. Additions to the phonetic and
phonological inventory are predictable in a language with a constantly
expanding vocabulary. Onset [ts-, ʃm-, ʃl-, ʃt-] are well on their way to
being integrated into the PDE consonantal system.
The final process, reminiscent of the series of cluster simplifications
described in 5.3.2, is known as dental or, more broadly, coronal stop dele-
tion. In PDE it is noticeable and much discussed for AmE, but there are
parallel processes in BrE (Tagliamonte and Temple 2005). The essence
of the ongoing change is that coda clusters of consonant + dental stop
are often simplified, resulting in homophony: packed-pack, paste-pace,
lost-loss, generating spellings such as left lane close for closed, Wes Coast
for West Coast, manage care for managed care. Dental stop deletion is now
legitimised in forms such as ice cream (first OED record 1744), mash potato
(1797) and even stain-glass (2012). :
The simplification is phonetically ‘natural’ in the sense that the
dental stop in the coda is unreleased – the allophonic realisation is
[-t#] or [d#]. The absence of a release stage renders further lenition
and eventual deletion perceptually likely, as the difference between an
unreleased stop and zero is minimal. Historical records of this deletion
are scarce, but occasional spellings such as bes ‘best’ (Havelok, c. 1325)
and rhymes such as wirschip ‘worship’: Egipt ‘Egypt’ (The Northern Homily
Cycle, c. 1315); frek ‘quick’: effec ‘effect’ (Castelford, Chronicle of England,
fourteenth century) are informative. Fifteenth-century loss of coda
<t>, especially in Scotland, is found in excep(t), stric(t)ly, nex(t), and
feck (1488), feckful (1568), feckless (1586) < effect. The addition of a non-
etymological <-t> in the coda is also indicative of insecurity about the
54
A phonestheme is ‘a phoneme or group of phonemes having recognizable semantic
associations, as a result of appearing in a number of words of similar meaning’ (OED).
realisation of consonant + dental codas: OE behœ̄s > ME bihese ~ biheste

‘behest’. Among the unetymological additions of <-t> that have become
standard are behest, against, amidst, amongst, all attested in ME, and later
whilst, hoist, tuft, graft, tyrant, peasant, pheasant, pennant, pageant, ancient. :
In PDE the rate of dental stop deletion varies according to the nature
of the preceding consonant, the morphological affiliation of the dental
– whether it is part of the stem or inflectional – and the presence or
absence of a following consonant. The deletion is most pronounced for
stem-final dentals after liquids or coronals and before another conso-
nant.55 In the US, the rates of reduction vary along social and ethnic lines:
in the position of strongest vulnerability, that is, non-inflectional [-d]
or [-t], followed by a consonant (behind the, best movie), it is estimated at
66 per cent in ‘standard’ AmE. It is extremely common among speakers
of AAVE (88–97 per cent), New York City Puerto-Rican (93 per cent),
Native American Puebloan English (98 per cent), Vietnamese English
(98 per cent), Chicano working class (91 per cent) and Appalachian
working-class speakers (74 per cent).56
55
Further commentary on the hierarchy of the environments for stop deletion is
addressed in Labov et al. (1968), Côté (2004), Schreier (2005), Tagliamonte and
Temple (2005).
56
Data from Walt Wolfram & Natalie Schilling-Estes (2006: 253).
6 The vowels in Old English:
spelling, pronunciation. PDE
alternations traced back to OE
FOOT-FEET, FULL-FILL, MAN-MEN, CHILD-CHILDREN, HOUND-HUNDRED
6.1 The vowel inventories of PDE and OE: a comparison

As this is the first of three chapters that survey vowel histories, we start
with a brief review of the PDE vowel inventory. Figure 6.1, combines
Figure 2.4 and the diphthongs of 2.2.2 (2).
Recall from 2.2.1 that in GA and SSBE vowel length by itself is
not phonemically distinctive: long and short vowels differ in height,
backness and peripherality, and these parameters are sufficient to
define the phonemic contrasts in the inventory. It is nevertheless the
case that functionally, and above the level of the atomic segment,
the PDE vowels fall into two major types: short/lax and long/tense.
The parameter of length or tenseness is essential for the prosodic
behaviour of vowels: (C)V syllables are light, while (C)VV syllables
are heavy.
BACK
FRONT CENTRAL DIPHTHONGS
Unrounded Rounded
Upper high i: (FLEECE) u: (GOOSE)

––––––– –––––– –––––– aI (PRICE)
Lower high I (KIT) U (FOOT)
aU (MOUTH)
Upper mid eI (FACE) /ɜ: (FUR) oU (GOAT) OI (CHOICE)
––––––– –––––– @ (COMMA) –––––– ––––––
Lower mid E (DRESS) ö (STRUT) O: (THOUGHT)
Upper low { (TRAP)
Low A (LOT, Á (LOT)

PALM) (GA) (SSBE)
Figure 6.1 An overview of PDE vowels and diphthongs
151
The ‘short’ vowels in Figure 6.1 are transcribed in an IPA form, while
the ‘long’ vowels are marked with the length diacritic (
). This is a typo-
graphical convenience, which allows one to identify more readily the
lineage of short and long vowels and their eligibility for specific histori-
cal processes. PDE contrastive diphthongs are functionally the same as
the long vowels.
Turning back to OE, we follow the long tradition of separating the
short from the long vowels and diphthongs. Wherever relevant, the
descriptive parameters in Figure 6.1 – height, backness and rounding
– are also preserved. As in the idealised system shown for PDE, the OE
vowel system is idealised in that it is based on classical and late West
Saxon, the variety from which the most written materials survive. It
is also the variety of OE that is most commonly used for etymological
references. Since the focus of this book is to trace the major patterns
of correspondences between OE and PDE and to place the ‘outliers’
in a historical context, details regarding the OE dialects are avoided.
A note of warning is due, however: the West Saxon OE, on which
the Anglo-Saxon grammars written in the last two centuries draw,
does not have a direct and localisable linear descendant. With this
disclaimer in mind, Figure 6.2 presents an overview of the (late) OE
vowels:
Short vowels Long vowels Diphthongs
Front Central Back Front Back
Unround Round Unround Round
Upper i: y: u: iə
High
Lower i y ə υ ej
Upper e: o: eə
Mid
Lower ε ɔ j
Low ɑ : ɒ: ə
Figure 6.2 An overview of late Old English vowels
All vowels except for the central [ə] appear as nuclei of stressed syl-
lables. The IPA symbols used for OE are familiar from Fig. 6.1, except
for [y] and [y
], which are not found in SSBE and GA. The ‘small capital
y’ stands for a lower high rounded vowel, a counterpart of unrounded
‘small capital I’. The pronunciation of [y] in OE is as the <ü> in German
küssen ‘to kiss’, Münster ‘minster’, or as in French début ‘debut’. The
pronunciation of the long front rounded vowel [y
] is as in German kühl
‘cool’ or French juge ‘judge’.
THE VOWELS IN OLD ENGLISH 153
The choices for categorizing the low back vowels are based on subse-
quent history: short [ɑ], whose contrastiveness with respect to [æ] in OE
is a matter of uncertainty, must be assumed to have been at least quite
central in late OE, since in early ME it merged with [æ], except before
nasals (7.3.1). Positing a long back rounded [ɒ
] is similarly a projection
of later history: except in the northern dialects, OE [ɒ
] was raised to
[ɔ
] in ME (7.3.2).
The final column in Fig. 6.2 shows the diphthongal nuclei in late
OE. The entries [iə], [eə], and [æə] are important for the phonology of
Germanic and OE, but they cannot be traced forward to PDE, which
is why they are highlighted. All three were monophthongized in ME
and they merged with pre-existing vowels. To avoid any confusion, it
should be borne in mind that PDE/RP [iə] as in HEAR, or [εə] as in HAIR,
are innovations based on the historical rhotic in such words; there is no
direct association between them and the OE [iə] and [eə].
The discussion of vowel histories will follow the approach used in the
outline of consonantal histories: we compare the two admittedly highly
variable end-points, PDE and OE, and we isolate and focus on processes
which have contributed to the shape of the modern system. While some
of the vocalic processes, such as I-Mutation (6.3.1) are period-specific,
their phonetic underpinning and their structural triggers and effects are
not. It is therefore important to read the rest of this chapter both as a
selective description of the synchronic components of the phonology of
OE, and as the diachronic fabric giving rise to to the phonology of ME,
EModE, and PDE.
6.2 Orthography and the reconstruction of OE vowels

PDE spelling is largely etymological. Among the languages using the
Roman alphabet, English is notorious for violating the sound-letter
correspondences of Classical Latin. In many instances the five vowel-
letters and their combinations are more likely to reflect historical
sources and morphological affiliations than actual sounds. In contrast,
it is generally held that OE spelling was more or less phonemic, and that
inconsistent or innovative spellings indicate deviations from a pronun-
ciation ‘norm’. Using hedge-phrases such as ‘more or less’ is justified
because reconstructing the pronunciation of vowels in OE is even more
problematic than the reconstruction of consonantal values. Though not
fully systematically, the contrastive consonants in OE were associated
with specific letters (see 4.1). In addition to orthographic consistency,
the identity of consonants can be tested in the verse, where stressed
alliterating syllables have matching onsets. The situation is more
complex with the vowels, where we do not have the evidence of rhymes
to supplement the orthographic evidence. Ælfric, writing about letters
in 995, identifies five vowel-letters:
Five of those [letters] are vowels, that is, clypiendlīce ‘callings out’: a e i
o u. These five letters are by themselves their own names, and without
them no word may be written. Therefore they are called the five vowels.
To these is added the Greek y for the sake of Greek words and y is very
usual in English writing. All the other letters are called consonants . . .
(Throop 2008: 19) :
Ælfric’s testimony that the ‘five letters are by themselves their own
names’ supports the assumption that the values of the vowel-letters in
OE match roughly the range of the phonetic values of the correspond-
ing letters in Latin, or, from our perspective, the values of the same
letters in languages like Italian, German and Polish. The capital Greek
upsilon <Y> had the high front rounded value [y] in the Attic-Ionic
dialect, and later in classical Greek.1 The ‘ligature’ of Roman <a> and
<e>, ash <æ>, was a rendition of a runic symbol, named helpfully after
the first sound in the OE word æsc ‘ash(-tree)’, IPA [æ].
Even adding <y> and <æ>, seven vowel-letters are inadequate for
the representation of the twenty vowels in Figure 6.2. Like all other
early Germanic languages, OE had short and long vowels, yet the
scribes did not mark vowel length: the letter sequence <for> can stand
for [fɔr] ‘for’ or [fo
r] ‘travelled’; <metan> represents both [mεtən] ‘to
measure’ and [me
tən] ‘to meet’. Note that in this book, as well as in
many edited texts and textbooks, macrons are inserted to help lexical
recognition, for example <god> [gɔd] ‘God’ vs <gōd> [go
d] ‘good’. :
The doubling of vowels to mark vowel length is used in some early
OE manuscripts, but it is not a regular feature of the main body of OE
texts; it became a common practice only after the end of the thirteenth
century.
Some diacritics, a superscript curl, a circumflex, and an extra-long
macron marking short vowels, do appear in the eleventh century (Hogg
1992a: 17), but the credit for the first consistent attempt to represent
vowel length in English belongs famously to the monk Orm, whose
10,000-line autograph translation and commentary of the Scriptures,
c. 1180 uses double consonant letters to mark short vowels in closed
syllables, thus <goddspelless> ‘Gospels’, <heffness lihht> ‘heaven’s
light’. Orm’s innovative system was an isolated example; there were no
followers. :
1
Gnanadesikan (2009: 214–16).
In addition to using the same letter to mark different vowels, the

OE scribes sometimes used different vowel-letters in the same word;
for example, in early WS ‘giver’ occurs with equal frequency as both
<gifa> (x44) and <gyfa> (x45). The vowel-letters are used in both
stressed and unstressed position. The letter <e>, for example, stands
for [ε] and [e
], and it is also the commonly used letter representing
any unstressed vowel in late OE. <e> is also used as a diacritic marking
the palatal quality of the preceding consonant, producing the impres-
sion that the vowel was diphthongised, for example ġieldan ‘yield’,
sċieran ‘shear’. Figure 6.3 shows the most frequent correspondences
between the vowel-letters and the simple vocalic nuclei posited for
late OE.
SHORT LONG
Letter
Value Example Value Example
 ɪ binn ‘bin’ i rīdan ‘to ride’
<y> ʏ fyllan ‘to fill’ y fȳr ‘fire’

ε settan ‘to set’
<e> e swēte ‘sweet’
ə dǣlere ‘dealer’
<æ> æ þæt ‘that’ æ mǣl ‘meal’
<a> ɑ bannan ‘to ban’ ɒ bāt ‘boat’
<o> ɔ (ge)boren ‘born’ o dōn ‘to do’
 υ drunken ‘drunk’ u hūs ‘house’
Figure 6.3 Letter-sound correspondences in OE
Unlike PDE, where single vowel-letters can be silent, it is generally

held that all single vowel-letters in OE represent syllable peaks: swēte
‘sweet’ is a disyllabic word, and dœ̄lere ‘dealer’ has three syllables. The
horizontal line separating the and <y> rows is dashed both because
they were often graphic variants, as in <gifa> ~ <gyfa>, and because of
the later merger of short [y] and [i] and long [y
] and [i
] in large areas
of ME (see 7.3.1).
The dashed line separating [æ] and [ɑ] is justified by their ambigu-
ous phonemic status. At least originally, they were in complementary
distribution: they were both reflexes of West Germanic /a/, which was
raised to /æ/ in OE unless it was followed by a nasal, /w/, or a back
vowel in the following syllable. Moreover, the subsequent history of

the two vowels also points to a merger (see 6.3.1). However, an argu-
ment in favour of phonemic /ɑ/ comes from cases of <a> + nasal, as
in OE mann ~ monn ‘person’, fann ~ fon(n) ‘fan, winnow’, gram ~ grom
‘hostile’ (Hogg 1992b: 86, contra Colman 1983). The <a>~<o> vari-
able spellings suggest a shift to the back vowel area, possibly pre-nasal
[ɑ̃~ɒ̃]. OE Merican and Northumbrian texts regularly represent PrG
/a/ + nasal as <o> + nasal, for example monn ‘man’, gongan ‘go’, lomb
‘lamb’. In late West Saxon the <a> spellings are the more common,
though in ME the Northern and West Midlands spelling favours <o>.
:
The spelling-pronunciation correspondences for the OE digraph spell-
ings are very complex. We will get to some details of interpretation in
6.5.3; for now we list the most common digraph spellings and their recon-
structed values in two rather vaguely defined ‘early’ and ‘late’ stages. As in
Figure 6.3, the editorial macron indicates etymological length.
Value
Example
Digraph Early Late
<ēa> υ ə strēam ‘stream’
<ea> ə heall ‘hall’
<ēo> eυ eə sēon ‘to see’
<eo> εə ε ġeolu ‘yellow’
<īe> iυ iə, i: (ġe)līefan ‘to believe’
<ie> i ə i ġiefan ‘to give’
Figure 6.4 OE digraph spellings and reconstructed realisations
Since our focus is on late West Saxon, omitted from Figure 6.4 is the
spelling <io>, found primarily in Northumbrian and Kentish. Only
three of the six realisations of digraph spellings in the third column –
[æə], [eə] and [iə] – are uncontroversially considered contrastive. The
phonetic values assigned to these contrastive diphthongs are different in
the early and late stages of OE, and the exact phonetic interpretation of
the diphthongal end points have been and can be debated. Although the
matching orthographic sequences with supplied macrons and without
the macrons – <ēa>-<ea>, <ēo>-<eo> and <īe>-<ie> – seem to
invite a parallel interpretation of long and short diphthongs, we will see
in 6.5.3 that positing such a parallel is neither necessary nor compelling.
The next two sections will address two important OE processes

that contributed to the shaping of the system in Figure 6.2: the early
process known as I-Umlaut, and a selective vowel lengthening known
as Homorganic Cluster Lengthening.
6.3 I-Mutation: FOOT-FEET, FULL-FILL, SELL-SALE

I-Mutation, also known by its German moniker I-Umlaut, is the most impor-
tant coarticulatory process affecting the vowels of early Old English. It had
a far-reaching effect on the structure of the late Old English vowel system
and produced paradigmatic alternations in the lexicon which survive to
this day. Deservedly, I-Mutation takes pride of place in all phonological
histories of English; excellent discussions can be found in Campbell (1959:
71–85); Hogg (1992a: 121–38); Lass (1994: 59–71) among others. We will
not cover all of the specifics here; the goal of the section is to present a sim-
plified account of the mechanism and scope of the process, with particular
emphasis on patterns whose survival accounts for otherwise ‘irregular’ or
‘unexpected’ inflectional and derivational alternations of the type FOOT-
FEET, not *FOOTS, FULL-FILL, not *TO FULL, MAN-MEN, not *MANS.
The triggers and the mechanism of I-Mutation are well defined: a high
front vowel /i/, or a palatal approximant /j/ in a post-tonic syllable
cause anticipatory, or regressive, assimilation of the vowel in the pre-
ceding syllable: PrG *ubilaz, Goth. ubils > OE yfel ‘evil’; PrG *dōmjan
> OE WS dēman ‘to judge’, compare OE dōm ‘judgement, doom’; WG
*taljan > OE tellan ‘to tell’, compare OE talu ‘tale’. The resulting vowels
are more similar to the conditioning factor in terms of tongue-height/
palatality, hence the term ‘assimilation’, but since there is no good argu-
ment for contiguous spreading of palatality, the process is more appro-
priately classified as a type of vowel harmony (Hogg 1992a: 121). In most
cases the harmonisation triggers /i, j/ are not preserved in the written
records; their historical presence is reconstructed from comparative
evidence. Because of the loss of the triggering environment, some new
mutated vowels acquire independent phonemic status (Lass 1992b). :
The shifts of the monophthongs involved in I-Mutation are shown
in Figure 6.5.
The parenthesised length mark in Figure 6.5 indicates that the
changes affect both the short and the long monophthongs. It is assumed,
along with the standard representations of the pre-Old English vowel
systems (Campbell 1959; Reszkiewicz 1973; Hogg 1992a; Lass 1994),
that the difference between the short and the long vowels at that early
period was based crucially on duration, so that the contrasts were
/i/ - /i
/, /e/ - /e
/, and so forth. If any height or backness differences
Front Back
Unround Round
High i(:) y(:) u(:)
Mid e(:) (œ/ø:) o(:) . . . before /-i, -j/
Low æ(:) ɑ(:)
Figure 6.5 OE I-Mutation of monophthongs
existed in these pairs in pre-OE, they cannot be tested in the records.

Therefore, no qualitative differences between the short and the long
vowels are shown in Figure 6.5.
The parenthesised symbols in the mid round set (œ/ø
) are new and
call for a comment. The ‘slashed o’ [ø], commonly used for the I-Umlaut
of both short and long /o/, is the IPA symbol for the upper/close mid front
rounded vowel. The [œ] ‘o-e ligature’ is the IPA symbol for a lower/open
mid front rounded vowel, as in French œuf or German göttlich. This symbol
is similar to the ligature <œ> found in some non-West Saxon texts
(Kentish and Anglian). Since the assumption is that there was no height
difference between the short and the long input to I-Mutation of the mid
back vowel [o], the choice of either [ø] or the ligature [œ] is arbitrary.
The use of dashed arrows linking the high and mid rounded [y(
)]
and [ø(
)] to the front [i(
)] and [e(
)] is justified by diachronic insta-
bility within OE. The rounded vowels [y(
)] and [ø(
)] combined the
frontness of the trigger and the rounding of the input vowels. Such front
rounded nuclei were innovations resulting from I-Mutation, and they
were ‘intermediate’ in a long-term perspective. Their existence is not
in question, but crucially, they end up merging with the corresponding
high and mid front unround vowels. The merger between [ø(
)] and [e]
occurred earlier, and by the time of late West Saxon only the unrounded
mid front vowels are attested. The unrounding and high-vowel merger
of [y(
)] and [i(
)] was well under way in late West Saxon, but as we will
see in 6.5.1, the [y(
)] can be realised differently in different dialects.
The dashed vertical arrow in Figure 6.5 – between the low front
vowel /æ(
)/ and the mid front /e(
)/ – flags internal differences in
this set. West Saxon /æ
/, ultimately going back to PrG /ai/, remained
unchanged, so that it was only short /æ/ that participated in I-Mutation,
but not in all environments. The dashed vertical arrow connecting
/e(
)/ and /i(
)/ is there to indicate another complexity: long /e
/ in
Pre-I-Mutation Post-I-Mutation
Examples (editors’ spellings)
Spelling Value Spelling Value
OE ġe-lēafa ‘belief’ < PrG

2 3 *galauon-
<ēa> [æυ] <ie, i, y, e> [i, e]
ġelīefan ‘to believe’ < Goth.
galaubjan
wearm ‘warm’ < PrG *warmo-

<ea> [æə] <ie, i, y, e> [i, ε]
wierman ‘to warm’ < Goth. warmjan
fēond ‘fiend’ < Goth. fijands

<ēo> [eυ]4 <ie, i, y> [i]
fȳnd ‘fiends’ < PrG *fiund-iz
feorr ‘far’ < WG *ferro-

<eo> [eə] <ie, i, y> [i]
fyrran ‘to drive away’ < PrG ferrian
Figure 6.6 OE I-Mutation of diphthongs
WS was primarily the result of I-Mutation of PrG /o

/, so there are no
further shifts. Paradigmatic alternations between short /e/ and /i/ – as
in OE beraþ < PrG *ber-a-nþi ‘they bear’ vs OE bireþ < pre-PrG *ber-i-þi
‘he bears’ – could be due either to the earlier raising of /e/ before high
vowels, or to mutation of the original /e/ vowel in pre-OE.
The triggers associated with I-Mutation, /-i, -j/, have a fronting and
palatalising effect on the diphthongs of OE too. Once again, we omit
some details and focus on the forms that feed into the phonology of ME.
Figure 6.6 shows the shifts posited for the diphthongs subject to I-Mutation. 234
In summary: all back vowels, short and long, are subject to the harmo-
nising process, and so are the diphthongs. The front vowels undergo
I-Mutation much more selectively: the long vowels /æ
/ and /e
/ are
2
This diphthong comes from Germanic /au/, for example PrG. *strauma- > OE stream.
3
The spellings with <e> are found in Kentish and Anglian texts. The reconstructed
[e
] in these varieties is the basis of the ME forms with <-e-, -ee-, -i-e->, from which
we get PDE belief.
4
In PrG, original IE */eu/ + /i, j/ was raised to /iu/. We simplify further by glossing
over the fact that <eo> digraphs and <io> digraphs (as well as the corresponding
vowels) coalesced when subject to I-Mutation. For details see Hogg (1992a: 121–2).
never affected, the results for [e] are uncertain and the shift of [æ] to [ε]
is not fully carried out.
As for the timeline of I-Mutation, it must have been under way
prior to the earliest OE texts. It was not fully completed until the late
eighth to the early ninth century. Some early Latin borrowings show
I-Mutation: glesing ‘gloss’ < Lat. glossa ‘marginal note’; OE cælic ‘chalice’
< Lat. calyx, WG *kalik; OE ynce ‘inch’ < Lat. uncia. Late Latin borrow-
ings are unaffected: OE aprilis ‘April’ < Lat. aprilis; OE paradis(e) < Lat.
paradisus; OE organe ‘marjoram’ < Lat. (herba) origanum.
Under the rubric of ‘scope’ we should also include another topic of
interest: where does I-Mutation show up in the grammar, that is, what
are the morphophonemic implications of the change? As noted above,
the triggers of the change are most often missing even in the earliest
texts, though /-i-/ may survive in some suffixes, notably -il, -ir, -in, -ing,
for example yfel ‘evil’, compare Goth. ubils; cyning ‘king’, compare Goth.
kuni ‘tribe’. Many of the orthographic <-i->’s found in the classical OE
texts can be misleading because they do not represent /i/’s that were in
existence at the time of I-Mutation, and therefore the preceding vowel
remains unchanged. Such is the case with the derived class 1 weak verbs
in <-ian>, where the original form of the suffix was *ōj-an, as in OE lufu
‘love’ - lufian, without I-Mutation. Another relevant factor can be the
prosodic domain: typically, the harmony operates within the domain of
the prosodic word, which implies that compounds will behave differ-
ently. This is indeed the case: the presence of /-i-/ in the second half of a
transparent compound is insufficient to cause mutation. Some examples
of unmutated OE forms are shown in the second column of (1). :
(1) Late OE unmutated forms before /i/
Input Late OE
blōd ‘blood’ blōdig ‘bloody’
lufu ‘love’ lufian ‘to love’
– arabisc ‘Arab, adj.’
āc ‘oak’ + rind ‘rind’ āc-rind ‘oak-bark’
blōd ‘blood’ + drynce ‘drink’ blōd-drynce ‘bloodshed’
tūn ‘town’ + scipe ‘-ship’ tūnscipe ‘township’
The projection of an originally allophonic process into the morphology
is accompanied by the loss of the trigger and the phonologisation of the
original allophones. Once this happens, mutated forms become associated
with specific functions. A paradigm that shows the effects of I-Mutation
and that has prominent survivors in PDE is the ‘root’ or ‘athematic’
declension. In that set the PrG inflections were added directly to the root
with no intervening ‘theme’-vowel. Consequently, if the inflection con-
tained /i/, as it did in the dative singular (*/-i/) and nominative plural
(*/-iz/), an eligible root vowel would be mutated, as illustrated in (2).
(2) I-Mutation in PrG athematic noun stems:
Nom. sg. Dat. sg. Nom. pl.
PrG *fōt-s *fōt-i *fōt-iz
OE fōt ‘foot’ fēt fēt ‘feet’
PrG *mannz *mann-i *mann-iz

OE man(n) ‘man’ men(n) men(n) ‘men’
PrG *mūs-s *mūs-i *mūs-iz

OE mūs ‘mouse’ mȳs5 mȳs ‘mice’
About two dozen nouns, mostly feminine but also some high-frequency
masculine nouns, belong to the OE root declension. Although the type-
frequency of root-stems is low, and the triggering /i/ not in evidence
in OE, this fossilised group included common core-vocabulary items of
high token-frequency, which may have allowed them to resist the pres-
sure of analogy for the formation of an alternative plural in -(e)s. Other
survivals in this group are OE tōþ ‘tooth’, OE gōs ‘goose’, OE lūs ‘louse’
and the now obscure compound woman-women < OE wīf-man(n), where
the second element replicates the plural of man(n). :
In pre-OE, PrG inflectional /i/ could also affect masculine -nd
stems (frēond ‘friend’, fēond ‘fiend’) and -r stems (mōdor ‘mother’, dōhtor
‘daughter’, brōdor ‘brother’), which show I-Mutation in the dative singu-
lar and occasionally also in the nominative plural: frȳnd, mēder, dēhter,
brēþre(n).6
Another set of I-Umlaut fossils in PDE is exemplified by FULL-FILL.
These are noun or adjective vs verb pairs, where the base preserves
the vowel in its original form in the noun or adjective, but in the verb
the same vowel shows I-Mutation triggered by a historical derivational
suffix *-j-an for the infinitive and a connecting vowel *-i- in the preterite.
5
The form mȳs was also gen. sg. fem., where I-Mutation was triggered by PrG *-iz.
6
Breþ(e)ren, brethren was the standard plural until c. 1600. Brothers, after an isolated
appearance in the thirteenth-century Lagamon’s Brut, does not reappear until the end
of the sixteenth century. Shakespeare uses brothers ~ brethren indiscriminately. Starting
in the seventeenth century, brothers became the plural of the literal sense and brethren
was retained in reference to spiritual or ecclesiastical relationship (OED). For more
details on the transition of mutated plurals to -es plurals in the nouns of relationship
see Krygier (1996).
(3) OE I-Mutation in weak class 1 verbs in *-j-an:

Base Class 1 weak verb
(a) full ‘full’ *fulljan > fyllan ‘to fill’
lust ‘desire’ *lustjan > lystan ‘to desire’7
fūl ‘foul’ * fūljan > fȳllan ‘to defile’
cūþ ‘known’ *kunþjan > cȳðan ‘announce, kithe’8
(b) gold ‘gold’9 *gulþjan > gyldan ‘to gild’
(c) blōd ‘blood’ *blōdjan > blēdan ‘to let blood, bleed’
dōm ‘doom, judgment’ *dōmjan > dēman ‘to deem, judge’
fōd ‘food’ *fōdjan > fēdan ‘to feed’
(ġe)mōt ‘meeting’10 * mōtjan > metan ‘to meet’
(d) sala ‘gift, sale’ *saljan > sellan ‘to give, sell’
talu ‘tale’ *taljan > tellan ‘to tell’
hāl ‘whole, healthy’ *hailjan > hǣlan ‘to heal’
hāt ‘hot’11 *haitjan > h&¯tan ‘to heat’
The pattern in (3) was robust in pre-OE, but after the loss of the front-
vowel triggers, in many instances the base forms were analogically
restored in the previously mutated verbs. Had it not been for analogy,
presumably to the more frequent form, for OE hunger, n. - hyngran, v.
(< *hungrjan) we would have hunger, n. - *hinger, v.; for sprūt, n. ‘sprout’
- sprȳtan, v. we would have sprout, n. - *sprite, v.; for OE rūm ‘room, space’
- rȳman ‘make room’ the PDE verb would be rime. There were many
instances in OE in which I-Mutation is recognisable within the noun
and verb paradigms, where unmutated and mutated forms can coexist
depending on the presence of /i, j/ in the PrG inflections, for example
2nd and 3rd pers. sg. present tense indicative and the past subjunctive of
strong verbs: cuman ‘to come’ – cymest, cymþ, cyme(n). :
Of interest from the point of PDE is the history of present and pret-
erite forms in English as in sell-sold, tell-told. As seen in (3d), the infini-
7
Compare PDE listless ‘without desire or spirit’.
8
The long vowel in the base is due to the loss of the nasal /n/ before the fricative in
pre-OE; compare Goth. munþs - OE mūþ ‘mouth’, Goth. uns - OE ūs ‘us’.
9
The /o/ in this word is a reflex of PrG /u/ which was harmonically lowered to /o/ in
West Germanic before a non-high vowel, thus PrG *gulþam, Goth. gulþ, OE gold.
10
Compare the adjective ‘moot’ – originally a legal term for a case to be discussed at a
meeting, generalised to ‘debatable; uncertain, doubtful; unable to be firmly resolved’
(OED).
11
The expected PDE reflex of the OE long vowel in this word is the vowel of GOAT, but
already in ME it was shortened before a dental (see 8.3.2).
tives, and also the present tense forms of these verbs show I-Mutation,
but the preterite of a subset of class 1 weak verbs was formed by adding
the dental suffix /-d/ directly to the root, leaving the base vowel
unchanged. Such verbs are considered ‘irregular’ by PDE criteria, yet as
shown in (4), they reflect a pre-OE pattern of I-Mutation in the present
tense paradigm and no I-Mutation in the past.
(4) OE I-Mutation in weak class 1 verbs present and preterite forms:
Base Present (I-Mutation) Preterite (no I-Mutation)
*bug- *bugjan > byċġan ‘buy’ *boh-ta12 > bohte ‘bought’
*wur- *wurkjan > wyrċan ‘work’ *worh-ta > worhte ‘wrought’
*sōk *sokjan > sēċan ‘seek’ *sōh-ta > sōhte ‘sought’
*sal- *saljan > sellan ‘sell’ *sal-da > sealde ‘sold’
*tal- *taljan > tellan ‘tell’ *tal-da > tealde ‘told’
Other such pairs in OE were (Kentish) brengan-brohte ‘bring-brought’,
þenċan-þōhte ‘think-thought’,13 tœ̄ċan-tāhte ‘teach-taught’, and the now
‘regularised’ dwellan-dwolde ‘dwell’, streċċan-streahte ‘stretch’.
Another set of morphophonemic alternations due to I-Mutation
was found in the comparative and superlative forms of some adjectives
before the suffixes -ir-, -ist- < PrG *-iz-, *-ist. Once again, the high
vowel trigger is not in evidence in OE, where the comparative marker is
-ra, and the superlative -ost or -est. Some mutated adjectival forms in OE
are: eald ‘old’ - (i)eldra/yldra ‘elder’, (i)eldest ‘eldest’; hēah ‘high’ - hīerra
‘higher’ - hīehst ‘highest’; lang - lengra - lengsta, ‘long-longer-longest’.
The comparative forms ELDER ~ OLDER were interchangeable in ME; the
semantic differentiation started only in late ME; the superlatives can
be interchangeable to this day. The <o> forms (older, oldest) are clearly
copies of the unmutated base form on the analogy of the dominant
model of -er, -est comparatives. :
The PDE derivational pairs STRONG-STRENGTH, FOUL-FILTH, LONG-
LENGTH are another echo of OE I-Mutation. The nouns were formed
with the PrG suffix -iþō, used for feminine abstract nouns from adjecti-
val bases. Although only a few sets have survived, the pattern was quite
frequent in OE, including pairs such as mōd ‘courage, pride’ - ofermettu <
ofer-mōdiþ(o) ‘pride, insolence’; þēof ‘thief’ - þīefþ ‘theft’.
As noted already, derivational suffixes containing /-i-/ could also
cause I-Mutation. The vowels in the PDE pair FOX-VIXEN differ because
fox descends from WG *fuhs, Skt puccha, while vixen is the derivative
of *fuhs + PrG -inī, a suffix forming feminine nouns from masculine
12
For Germanic /o/ and /u/ before /a/ in *boh-ta, *worh-ta see n. 9 on gold above.
13
The PDE form of think with is from the cognate verb þynċan ‘seem’.
bases, OE fyxen ‘she-fox’; compare OE god ‘god’ - gyden ‘goddess’,

munuc ‘monk’ - mynecen ‘(obsolete) minchen, nun’. Other suffixes where
I-Mutation can be encountered are -il, -ing, -ig, -isc, where analogical
levelling of forms within the paradigm often results in unmutated forms,
for example ānlic/œ̄ nlic ‘only’. The strongest I-Mutation triggers are the
historical inflections and stem-formative elements in pre-OE, which are
subsequently lost, while ‘live’ suffixes such as the diminutive -incel are
more ‘word-like’ and are typologically more likely to behave like the
second elements of compounds. Whether the typological prediction is
borne out by the OE data remains an open question. :
We have described I-Mutation as an anticipatory segmental assimi-
lation of height and frontness within the domain of the prosodic word.
The process adds complexity to the vowel system by generating pre-
viously unattested contrastive front rounded vowels. Height harmo-
nisation parallels exists in the earlier history of Germanic. Why did
I-Umlaut cease to be productive in OE and why has it not recurred
later in the history of English? One obvious factor is that the preser-
vation of the properties of the base, ‘faithfulness’ to the base, tends
to inhibit the operation of phonetically driven changes, but there can
be other factors. Seeking an explanation for something that has not
happened is of course an ontological problem, yet knowing that such
coarticulatory processes are widespread in the languages of the world,
we can take their presence or absence as diagnostics for other properties
of the system.
In the case of OE I-Mutation, the valuable secondary information
we can extract refers to the properties of the triggers. Ideally, in har-
monic processes, the trigger either has to be prominent, or it has to have
contrastively specified features (Hyman 2002). For I-Mutation these
features were frontness and height. The prominence and the contrastive
features of the trigger interact dynamically with the domain of applica-
tion: inflectional suffixes or thematic elements, as in (2) and (3), have a
higher probability of influencing the base, but as the domain extends to
include derivational suffixes, the results become erratic. The picture is
further complicated by type- and token-frequencies. Still, a hypothesis
that can be entertained is that the loss of I-Mutation correlates with the
beginning of reduced prominence of the post-tonic syllables where the
triggers reside. A related fact here may be the fact that /i/ is the least
sonorous of the vowels. Reduced prosodic prominence is associated
with fewer contrasts; put differently, the phasing out of I-Mutation can
be seen as a diagnostic for the prosodic developments within OE leading
to weakening of the contrastive properties of the unstressed vowels.
While the suffixal ’s were generally not weakened, stem-internal
post-tonic ’s – as in the verbs lufian ‘to love’, lōcian ‘to look’, hopian
‘to hope’ – were not viable triggers, in spite of the fact that they are in
the right domain. As we will see in 6.5.4, this fits in with a more general
picture of unstressed vowel reduction in late OE.
6.4 OE homorganic-cluster lengthening: CHILD-CHILDREN, HOUND-

HUNDRED
The lengthening of final vowels in monosyllables in North-West

Germanic (see 3.4.4) was seen as an early manifestation of the universal
tendency for stressed syllables to be heavy. Already in pre-OE mono-
syllabic content words ending in a short vowel were disallowed. This
early lengthening was one of many historical adjustments of the weight
of stressed syllables, and has been analysed as the first link in a chain
of quantity changes in English cumulatively resulting in optimally
weighted stressed syllables. :
A prominent OE quantitative change whose initial stages are traced
to the latter half of the eighth century (Luick 1964: §§267–8), or even
earlier (Liberman 1992a), is Homorganic Cluster Lengthening. All historical
phonologies of English recognise the lengthening. We keep the tradi-
tional label for the sake of continuity, although it will become clear that
the label covers parallel, but not necessarily identical processes, whose
differences are as significant as their similarities.
Descriptively, the lengthening affected short vowels before clusters
of liquid or nasal plus a homorganic voiced obstruent. ‘Homorganic’
means that the two consonants forming the cluster have identical or
very close places of articulation. The clusters associated with this
lengthening are -ld, -nd, -mb, and possibly /r/ + coronal clusters: -rd,
-rn, -rs, -rl, -rð. While some ME spellings do indicate sporadic length-
ening before -rC – board, hoard, yearn, earl, earth had short vowels in OE
– the observed vowel length before these clusters may be due either
to pre-cluster lengthening or to /r/-vocalisation before a coronal (see
5.2.3). Because of the convergence of effects, we postpone the discussion
of lengthenings before /r/ + coronal clusters to 8.3.2. Some sporadic
lengthening effects are reconstructed in early ME before [-ŋg], but it
is ignored here because there are no traces of the lengthened forms in
PDE. :
The necessary segmental condition for pre-cluster lengthening was
the presence of a sonorant in the stressed syllable coda. Additionally,
lengthening required that the sonorant was followed by an obstru-
ent either as the second consonant of the coda in monosyllabic forms,
for example blind ‘blind’, or it could be the onset of a following weak
syllable, for example bin.dan ‘to bind’. Obviously, within the paradigm
of a single item the syllabic structure can vary between monosyllabic
and non-monosyllabic, thus blind, sg. - blin.de, pl., or bin.daþ, imperative
pl. - bind, imperative sg. After the loss of inflections, the paradigms show
uniform vowel length, so it is unclear how or whether syllable-structure
variability interacts with the lengthening. Until we have a fine-grained
statistical picture of the density of monosyllabic and non-monosyllabic
forms and the type of coda in the stressed syllable at the time of length-
ening, we can assume that tautosyllabicity of the cluster provided the
optimal condition for the lengthening. The assumption is dictated by
the absence of lengthening in forms in which a short vowel is followed
by a single sonorant, for example OE fell > PDE fell ‘skin’, but OE feld
> PDE field. :
The effects of Homorganic Cluster Lengthening are most clearly
traceable in early ME. Some typical examples of the process are shown
in (5).
(5) OE Homorganic Cluster Lengthening:
OE14 ME
ċild [cild] ‘child’ child [tʃi
ld]
mild [mild] ‘mild’ mild [mi
ld]
feld [feld] ‘field’ feld [fe
ld]
hund [hund] ‘hound’ hound [hu
nd]
(be)hindan [-hind ə n] ‘behind’ bihinde [hi
nd(ə)]
climban [klimb ə n] ‘climb’ climb(e) [kli
m(b)(ə)]
The lengthening was blocked if the homorganic cluster was followed by
a liquid word-internally, hence no lengthening in OE ċildru ‘children’,
OE hundred ‘hundred’, OE sund ‘sound, swimming water’, but OE sun-
drian ‘to sunder’. This is in line with the balancing of syllable weight
by shortening long vowels before -CCC- clusters, associated with late
OE, - early ME, that is, either simultaneously or somewhat later than
Homorganic Cluster Lengthening (see 7.5.1).
Another factor influencing the results of the lengthening is morpho-
logical structure. In suffixed words and compounds the addition of a
third consonant is not sufficient to change the shape of the base and the
lengthened forms persist, as in OE blindnes ‘blindness’, ċildliċ ‘child-
like’, grundleas ‘groundless’, goldsmið ‘goldsmith’. The details regarding
14
The [i], [e], [u] allophonic transcription for the OE short vowels /i/, /ε/, /υ/ is
prompted by the earliness of the process, the possibility that the sonorant-obstruent
cluster in the coda triggered a short tense allophone, and the post-lengthening results
[i
], [e
], [u
].
vowel height will be noted below, but here it should be added that the
template of long vowels before -ld, -nd, -mb continues beyond ME.
Thus -VCCC- words of later date, such as Cambridge (1580?), cambric
(1530),15 laundry (1533), foundry (1601), doldrum(s) (1811) preserve the
long vowel irrespective of their compositionality.
Focusing just on the non-rhotic clusters -ld, -nd, -mb, we now turn
to the specifics of the implementation and causation, and the contin-
gent issue of unity of the lengthening.16 The least problematic data
are presented by the small set of items in which the stressed vowel
was eligible for lengthening before <-mb>, a total of nine words: climb,
comb, coomb, dumb, December, lamb, timber, tumble, womb. Only four of these
show lengthening in PDE: OE climban ‘climb’, OE camb/comb ‘comb’, OE
cumb ‘comb/coomb’,17 OE wamb/womb ‘womb’. This is too small a set to
warrant any generalisations. The best we can say is that there could be
sporadic, item-specific lengthening, possibly compensatory for the loss
of [-b], developing simultaneously with the simplification of word-final
-mb (see 5.3.2). Recall that the evidence for [-mb]-simplification goes
back to late OE, as in dum(e) and duman ‘dumb’, ge-camde ‘combed’, and
inverse spelling such as þumbes ‘thumbs’ < OE þuma (1154), ember <
ON eim(y)rja. There is no compelling argument for classifying -mb as a
lengthening cluster.
The case of -ld as a lengthening cluster is much stronger. The length-
ening applied systematically to the low and back vowels: all surviving
OE <-ald-> and <-old-> items – words such as bold, cold, fold, hold, mold,
old, sold – had short vowels in OE, but the vowels were long by c. 1200.18
In addition, there are items in which front vowels were lengthened: child,
mild, field (OE feld), shield (OE sceld), but these are narrowly outnumbered
by forms in which the short nucleus of OE is preserved in PDE: seldom,
elder, build, gild, guild.19 The systematic nature of pre-[-ld] lengthening
of <a> and <o> suggests a phonetic basis: if the post-vocalic realisation
of the liquid was a dark ['], the transition from the vowel to the liquid
plus dental stop coda could be mediated by the insertion of a linking
15
The stressed vowel in this word, based on Flemish Cambray < Latin Camaracum, can be
both [-æ-] and [-ei-] (OED).
16
This section updates arguments presented in Minkova and Stockwell (1992).
17
The phonological history of the homophones comb/coomb [ku
m] ‘vessel’ and ‘valley’ is
identical; there is also a possible etymological link between the two nouns (OED).
18
The evidence for that is the special spelling system devised by Orm, who used double
consonant letters after short vowels (see further Anderson and Britton 1997; Fulk 1999).
19
The spelling of should, would, could suggests a long vowel; auxiliaries/modals are pro-
sodically weak, which accounts for the short vowel in PDE. The [-l] in them contin-
ued to be pronounced until the late seventeenth century (Minkoff 1972: 336).
l-coloured non-syllabic glide, reanalysed as an extension of the vowel.

The new bimoraic nucleus would then be identified with the phoneti-
cally closest pre-existing long vowel. The details of this hypothetical
development are tentatively reconstructed in (6), bearing in mind, as
always, that such narrow phonetic details remain untestable.
(6) Reconstructing pre-<-ld> lengthening in OE:
OE (Angl.) ald ‘old’: [ɑ'd] > [ɑ'd] > [ɒ
'd] > (ME) [ɔ
'd]
OE gold ‘gold’: [o'd] > [o'd] > [o
'd] > (ME) [o
'd]
OE mild ‘mild’ [mi'd] > [mi'd] > [mi
'd] > ME [mi
'd]
OE feld ‘field’ [fe'd] > [fe'd] > [fe
'd] > ME [fe
'd]
The only possible instance of lengthened <u-> before [-ld]in a PDE
word is shoulder, OE sculdor. As for the domain of lengthening, it has
to be restricted to -Vld forms in which the vowel and the cluster are
in the same syllable. This is crucial because there is no lengthening in
monosyllables ending in a single liquid; as noted above, OE fell > PDE
fell ‘skin’, but OE feld > PDE field. : The morphological domain of the
lengthening is also restricted: the entire cluster has to belong to the
stem. Thus inflectional -d(e) for the past tense of weak verbs attached
to V + [-l] stems has no lengthening effect: fill, swill, spill, all attested in
OE, preserve the short vowel in the past tense. Indeed, in some instances
the addition of the /-d/ past tense morpheme causes pre-consonantal
shortening, as in the past tense forms of deal, feel, kneel (see 7.5.1.1).
The third cluster associated with Homorganic Cluster Lengthening
is [-nd]. The long-term effect of this cluster is limited. It applies
systematically only to the high vowels [i] and [u].20
(7) OE <-ind> and <-und> lengthening:
Early OE ME
blind [blind] blind [bli
nd] ‘blind’
grindan [grind ən] grind(en) [gri
ndən] ‘grind’
(be)hindan [-hind ə n] bihinde [- hi
nd(ə)] ‘behind’
grund [grund] ground [gru
nd] ‘ground’
hund [hund] hound [hu
nd] ‘hound’
pund [pund] pound [pu
nd] ‘pound’
20
The association of <-nd> spelling and a preceding [u
] was extended to some [-n]
codas: AN soun ‘sound’ (compare Chaucer’s rhyme in The Miller’s Tale: And softe he
cougheth with a semy soun – / What do ye, hony-comb, sweete alisoun); similarly OE pūnian,
ME poun(en) ‘to pound’; ON būinn, ME boun ‘bound for, ready’; OE þunor, þunre, ME
thonder ‘thunder’; ME kin + reden (suffix) > ‘kindred’.
The pair wind, n. - wind, v. is a special case, probably best explained

on the grounds of homophony avoidance. Rhyme evidence suggests
that the forms were actually homophonous into the late seventeenth
century. The restriction of the lengthening effect to the high vowels
leaves a whole set of common OE lexemes such as band, hand, land,
sand, bend, blend, end, rend, send, spend, tender, wend with short vowels in
PDE. :
As with [-ld] lengthening, the mechanism of lengthening before
[-nd] is still under scrutiny. An important factor in the process must
have been the sonority of the first consonant in the cluster following
the short vowel and the possible gestural overlap between the nasal and
the voiced coronal stop. It is logical to assume with Phillips (1981a) and
Jones (1989) that the lengthening was the result of phonologisation of
coarticulatory lengthening. One could speculate that the presence of
a nasal + voiced stop would trigger allophonic lengthening before all
vowels, but because of the intrinsic shorter duration of the high vowels,
it was only the ‘longer’ [i] and [u] that were reanalysed as ‘long’, while
the lengthening remained allophonic with the mid and the low vowels.
Another possible direction of inquiry could be to look further into
the nature of the nasal: as shown in Lavoie (2009: 35), the intensity of
/n/ in the neighbourhood of a reference /i/ vowel is actually greater
than that of the vowel, and the intensity could be perceived as length.
Since the coarticulatory effects of nasals are language-specific (Cohn
1993), however, these hypotheses are not testable. The oddity of pre-nd
lengthening in OE remains to be accounted for. :
Summing up: the shared environment for OE Homorganic Cluster
Lengthening is a tautosyllabic sonorant followed by a voiced homor-
ganic stop. It is a more restricted version of a broader process of
glide insertion and lengthening before sonorants, singly or followed
by obstruents, that recurs in some varieties of later English, with or
without sonorant loss, as in art, cord, first, curse, false, bolt, boulder, aunt,
chance (see Chapter 8).21 Taken by itself, however, OE Homorganic
Cluster Lengthening is hardly a unified process. The scope, the targets
and the mechanisms of implementation of the OE changes are highly
specific, and it is only in the most general sense that we can classify
them as being different manifestations of the same process, where the
presence of a sonorant increases the probability of a glide transition,
21
Broadening further the chronological range of lengthening before nasal + obstruent
will include the pre-OE lengthening contingent upon loss of nasals before spirants,
thus PrG *gans- > OE gōs ‘goose’, PrG *fimfi > OE fīf ‘five’, PrG *munþaz > OE mūð
‘mouth’.
misperceived as a portion of the vowel, leading to reanalysis of the

vowel as long.
6.4.1 Are pre-cluster lengthenings prosodically incongruous?

Another aspect of the OE pre-cluster lengthenings has to do with the
overall prosodic shape of stressed syllables in the history of English.
Translated into units of prosodic weight, where a short vowel has
one mora (µ), and a long vowel or a diphthong has two moras (µµ),
the minimal major-class word in OE was bimoraic (see 3.4.4). OE -VC
syllables are bimoraic too; the coda consonant in OE can be shown
to add to the way the syllable weight is computed in verse (see 9.3.2).
This does not tell us what the upper weight limit of a stressed syllable
in OE was, and indeed OE stressed syllables could be light, heavy and
superheavy. :
Counting the first consonant in the coda as a mora, superheavy syl-
lables are syllables with long or diphthongal nuclei and a filled coda: OE
ēast ‘east’, āc ‘oak’, þīeft ‘theft’, ūt ‘out’, f ȳlfþ ‘filth’, prēost ‘priest’ can be
analysed as tri-moraic and superheavy. The superheavy grade can be
recognised in various quantitative processes in ME (see 7.5.1), leading
to the long-term vision of regularisation of syllable weight by elimina-
tion of superheavy syllables (Luick 1898; Lass 1992a: 70–83). For early
OE, however, the upper limit of weight of the stressed syllable was
not regulated: there is no evidence that -VC, -VCC, -VV, -VVC and
even -VVCC syllables were functionally different, though they clearly
represent a continuum of ‘heavyness’, with VVCC being longest/
heaviest. We know that all five types jointly were strongly favoured in
the stressed syllables of OE: for disyllabic words Getty (2002: 213–14)
reports 17.8 per cent of light stressed syllables; the other 82.2 per cent
are heavy syllables, though the counts do not differentiate the heavies
further. The universal bias towards co-occurrence of weight and stress
was thus in evidence already in OE, yet that does not tell us how much
weight is ‘good’ weight. Criteria such as the weight of a minimal word,
or the overall ratio of lights to heavies in stressed syllables are unin-
formative with respect to the difference between -VC, -VCC, -VV,
-VVC and -VVCC.
All OE pre-cluster lengthenings generate forms that are ‘superheavy’,
or at least heavier than the input -VR([C])- or -VRC syllables, where
R = sonorant. The lengthened syllables are -VVR([C]) ([bi
n[d]an])
or -VVRC ([bi
nd]). Since other VVC([C]) and -VVCC syllables
undergo shortening to VC([C]) and -VCC (see 7.5.1), one unexplored
line of inquiry would be to see whether the homorganic lengthening
clusters -RC- contribute an intermediate degree of weight between

that of a singleton coda and an obstruent-initial coda cluster. Put dif-
ferently, the question is whether the make-up and cohesion of -ld, -nd,
-mb renders them ‘lighter’ than other -CC clusters, bringing the weight
of the lengthened syllables closer to the very common -VVC stressed
syllable shape. The commonness of the -VVC shape can then be taken
as motivating shortenings in -VVCC(C) syllables or lengthenings in
-VC syllables, unifying all of the major quantitative changes in late OE
and ME. :
The hypothesis of unidirectional prosodic weight adjustments first
formulated by Luick (1898) treats all shortenings and lengthenings in
terms of an ‘optimal’ bimoraic syllable. In that context, the -VVCC
outputs of OE Homorganic Cluster Lengthening are anomalous and
require special explanations. Lass (1992a: 75) labelled pre-cluster
lengthening a quantitative ‘misfit’, and is content to leave it out of
the ‘conspiratorial’ picture of goal-oriented quantitative changes.
Within a theory of phonologically gradient weight, most recently
explored by Gordon (2006) and Ryan (2011a), the weight difference
between -V, -VC, -VVC and -VCC becomes a matter of phonetically
testable degree, not an abstract numerical correspondence between
a segment and a mora. The cut-off between heavy and light syl-
lables can be different for different phonological processes, and it is
also language-specific. Within the theory of gradient phonological
weight we should also reconsider the proposal (Phillips 1983) that
there was a discontinuity in the application of the ‘optimal syllable’
principle between early OE, when it was ignored, and late OE and
ME, when it was respected. As we will see in Chapter 8, the ideal of
a bimoric syllable continued to be ignored in early Modern English.
The question in the title of this section, ‘Are pre-cluster shorten-
ings prosodically incongruous?’, should therefore be answered in the
negative.
6.5 The late OE vowel inventory in detail

In preparation for the discussion of later developments, the final sec-
tions of this chapter provide more details on the OE vowel system at the
end of the eleventh century.
6.5.1 The short vowels in late OE

Figure 6.7 reproduces the short-vowel portion of the overview of the
OE vowel system in Figure 6.2.
Short vowels
Front Central Back
Unround Round
Upper
High
Lower i y ə υ
Upper
Mid
Lower ε ɔ
Low æ ɑ
Figure 6.7 The short vowels of late Old English
Except for [ə], to which we return in 6.5.4, the vowels in Figure 6.7
appeared in stressed position. Some typical forms illustrating them are
shown in (8).
(8) Examples of OE words with short stressed vowels:
Front unround Front round Back
siftan [siftən] ‘to sift’ fyllan [fyllən] ‘to fill’ lufu [lυvə] ‘love’
settan [sεttən] ‘to set’ god [gɔd] ‘God’
blæc [blæk] ‘black’ catt(e) [kɑtt(ə)] ‘cat’
From a PDE perspective, the front unround and the back short vowels
are straightforward: their realisations are identical or close to familiar
PDE vowels. Of note is that for the high and mid short-vowel inputs
to I-Mutation and pre-cluster lengthening in Figure 6.5, and in (5)–
(7), we used the same symbols as the long vowels: [-ild] > ME [-i
ld]
as in ċild ‘child’; [-eld] > [-e
ld] as in feld ‘field’; [-und] > [-u
nd] as
in grund ‘ground’, that is, they are treated as qualitatively identical,
differing only in duration. The rationale for that is that the vowels
lengthened in early OE merged with pre-existing long vowels at the
same height.
The use of the high and mid non-peripheral vowel symbols [i, υ, ε, ɔ]
in Figure 6.7 assumes a qualitative difference from the corresponding
long vowels, implying lowering and centralisation of the short vowels
between early and late OE. Positing a qualitative shift of the high and
mid vowels is defensible on the grounds of later quantitative processes,
whereby lengthened [i, υ, ε, ɔ] merged with long vowels at a lower
height: so that [i] > [e
] as in OE wicu, ME wēk(e) ‘week’; [υ] > [o
] as in
OE duru > ME dōr(e) ‘door’, and so on (see 7.5.2). In addition, ON /ε/ is
identified with OE and ME [ε] and is consistently found as <e> in ME
spelling.22 The exact date of the lowering cannot be established, and

some unresolved questions remain, yet it is reasonable to accept, with
Hogg (1992a: 199–202), that late OE West Saxon had non-peripheral, or
‘lax’ high and mid short vowels. :
The high front rounded [y], as in fyllan, which was most commonly
the result of I-Mutation of /u/ (see 6.3), is the only OE short vowel
which does not survive contrastively in SSBE and AmE.23 Initially a
fronted allophone of the high back vowel spelled , it had become
phonologised in all dialects of OE by the mid ninth century (Lass 1992b:
103). The loss of the I-Mutation trigger in the paradigms shown in (2)–
(4) makes the new sound an exponent of morphological alternations.
Phonetically [y] must have been perceived as being much closer to the
unrounded [i] of late West Saxon than to the original back vowel: the
spelling records show and <y> used interchangeably (<dyne ~
dine> ‘din’ (PrG *dunjan), <dysiġ ~ disiġ> ‘foolish, dizzy’), while there
are no records of ‘confusion’ between and <y>. By the end of OE,
[y] was treated differently in the different dialect areas: it merged with
[i] in the East Midlands and the Northern dialects, it was lowered to
[e]/[ε] in Kentish, and the rounded realisation became characteristic of
the West Midlands (see 7.3.1).
6.5.2 The long vowels in late OE: STREET-STRATFORD

Except for the front rounded [y
], the values of the long vowels in OE
are also familiar sounds. Unlike the short stressed vowels in Figure 6.7,
however, which have been more resistant to qualitative changes, not a
single OE long vowel survives in PDE with all of its original features.
Figure 6.8 shows the long-vowel portion of the OE vowel survey of
Figure 6.2.
Rounding is shown only for the front vowels. The origin of the high
front rounded [y
] parallels that of [y] – it is most frequently the result
of I-Mutation. As we will see in 7.3.1, the subsequent history of the two
vowels is also parallel. All back vowels are round. Keeping the editorial
macrons over the long vowel-letters, (9) shows examples of words with
long stressed vowels in late OE.
22
Dance (2003: 117, 122).
23
This is a simplification. For other sources of the high front rounded vowel see Hogg
(1992a: 167–70). SSE has a front rounded vowel for the vowels of GOOSE and FOOT.
Long vowels
Front Back
Unround Round
i: y: u:
Upper
High
Lower
e: o:
Upper
Mid
Lower
Low æ: ɒ:
Figure 6.8 The long vowels of late Old English
(9) Examples of OE words with long stressed vowels:

Front unround Front round Back
mīn [mi
n] ‘mine’ mȳs [my
s] ‘mice’ nū [nu
] ‘now’
dēma [de
mə] ‘judge’ dōm [do
m] ‘doom’
mǣl [mæ
l] ‘meal’ hāliġ [hɒ
li] ‘holy’
Some information on the sources and the allophones of the low vowels
will help us link the spelling and the etymology of otherwise puzzling
ME and PDE forms. In pre-OE, West Germanic /a
/ – using <a> as a
generic symbol for a low non-front vowel – was raised and fronted to
[æ
], for example *slāp- > OE slœ̄p ‘sleep’. The raising was paralleled by
short /a/ > /æ/ (see 6.2); in both instances the low front vowel is a new
addition to the pre-OE vowel inventory. The most notable aspect of the
fronting was that in Anglian the fronted /æ
/ was raised further to /e
/,
so we get WS <&¯> - Angl. <ē>, a difference not just in spelling but in
vowel height: WS slœ̄ p - Angl. slēp ‘sleep’, WS dœ̄ d - Angl. dēd ‘deed’,
WS sœ̄ d - Angl. sēd ‘seed’. The split survived in ME. This is relevant to
the history of the pair STREET-STRATFORD: Lat. strāta, an early loanword,
is recorded as strœ̄t in West Saxon, strēt in Anglian. When shortened
(see 7.5.1), [æ
] became [æ] ~ [a], while [e
] became [ε
] (see 7.3.2);
the OE compound Strœ̄ tford became Stratford, but compare Stretham
(Cambridgeshire). The spelling <ee> for PDE street goes back to the
Anglian form with [e
].
The OE dialect split in the treatment of pre-OE /a
/ has a long-
term effect on its orthographic representation in PDE. There was no
necessary continuity between WS and ME, and the incipient late ME
standard inherited mostly the Anglian forms with the upper mid front
[e
]. After the two long mid front vowels of ME, [ε
] and [e
], began to
be differentiated in spelling, <ea> for [ε

] and <ee> for [e
], the system
ended up representing identical etymological entities with either
<ee> or <ea>, depending on whether the spelling was based on the
continuation of the WS or the Anglian form.
The fronting of /a/ was prevented by a following nasal: WG /-a
n/
in OE appears to have retained its backness, that is, we can posit an
intermediate stage [ɒ̃
], which subsequently merged with pre-existing
/o
/, thus Gmc *mān- > OE mōna ‘moon’, Gmc *sān- > OE sōna ‘soon’.24
Coarticulation with a following labial approximant [-w] also blocks
the fronting, so that WG -āw- can be reconstructed as a low rounded
[ɒ
], thus Gmc *blā- > OE blāwan ‘blow’. These developments are
summarised in (10).
(10) Germanic /a:/ in OE and ME: 2526
Input OE ME Examples
WS [æ:] [ε:] meal, year25
/a:/ ---- [æ:]
Angl. [e ] [e:] sleep, street26
(+ nasal) [ɒ̃ ][o ] [o:] moon, soon
(+ [w]) [o ] [oυ] blow, know
The Long Vowel Shift (see 8.2.2) obscured the differences between the
West Saxon and the Anglian forms for the standard varieties, but the
dialectal split shown in (10) is of considerable etymological and ortho-
graphic interest nevertheless. It is also a major test for the identification
of dialect features of ME texts.
As (10) indicates, Germanic /a
/ did not survive in late OE, which
raises the question of the source of the low back vowel [ɒ
] in our
inventory in Figure 6.8. That vowel results from monophthongisation
of the West Germanic diphthong /ai/, held to have occurred after
the fronting of /a
/ to /æ
/. Examples of this process are Gmc *stain-,
*haim-, *aigan- > OE stān ‘stone’, hām ‘home’, āgan ‘own’ (see the first
row in (11) below). The presence of the high front /-i/ in the WG diph-
thong suggests that initially the monophthongal long vowel would have
been more central than [ɒ
], but since the next step in the history of that
24
Regressive nasalisation can cause either raising or lowering depending on the height
and backness of the vowel. Low back vowels are typically subject to raising (see
Beddor et al. 1986).
25
Compare German Mal/Mahl, Jahr.
26
Compare German schlafen, Lat. strata.
vowel in the southern dialects of ME shows definitely a back rounded

vowel, reconstructed as [ɔ
], we can posit [ɒ
] in late OE.
6.5.3 Diphthongs and diphthongoids

The correlation between orthographic representation and recon-
structed values for the OE diphthongal nuclei is one of the most
controversial topics in the study of English historical phonology.27
This is unsurprising in view of the instability of complex nuclei in
general, the difficulty of perceiving and representing the transitions of
the stressed part of the nucleus to the syllable coda, and not least, the
confounding variability of digraph spellings found in the OE dialects.
The effect of I-Mutation on diphthongs complicates the system further
(see Figure 6.6). Agreement exists, however, that OE inherited a set
of diphthongs from pre-settlement Germanic. Focusing just on late
West Saxon, (11) shows the West Germanic diphthongs and their OE
counterparts.
(11) West Germanic diphthongs in OE:
WG Late WS Examples
/ai/ [ɒ
] hām ‘home’
/au/ [æə] strēam ‘stream’
/eu/ [e ə ] þēof ‘thief’
/iu/ [i ə ] ġe- līefan ‘believe’
Since WG /ai/ was monophthongised (see 6.5.2), late WS [æə], [eə],
[iə] are the core inherited diphthongs. They are not ‘long diphthongs’
(though that is their traditional name), just simply diphthongs. The
reconstruction shown in (11) assumes evolution of an original high
back glide to, most probably, a schwa-like glide. The plausibility of the
values [æə], [eə], [iə] is based on their later pronunciation: they merged
with the corresponding long vowels [æ
, e
, i
]. Indeed, the diphthongal
nature of the reflexes of WG /au/, /eu/ in early OE is clear, but varia-
tion between [æə] ~ [æ
] and [eə]~[e
] must have come into existence
during OE, because <æ> ~ <e> spellings for [æə] as in <stræm>,
<strem> ‘stream’, and <e> for [eə] as in <tre> for more common trēo
‘tree’ are documented in late OE. :
By virtue of their compositionality, diphthongs can be falling, when
the first element is more prominent, as in PDE [ai], [aυ], or they can
27
The controversy goes back to at least the first half of the twentieth century, and it con-
tinues today. Good surveys of the earlier literature are found in We'na (1978: 21–7);
We'na (1987: 44–56); Hogg (1992a: 16–24); Lass (1994); White (2004).
be rising, when the ‘stressed’ part is in second position, as in [jɑ

]. The
compositionality of the unit allows a shift of the prominence within
the two elements of a diphthong, a switch known by its German name
Akzentumsprung. In some OE words the second element of <ēo>became
more prominent at the expense of the first element; the switch was
triggered by the properties of adjacent consonants. More specifically,
palatal onsets palatalised and absorbed the first element, which presum-
ably passed through an intermediate stage of being a palatal glide, while
the second element remained a bimoraic monophthong: OE sċēotan >
ME shōte(n) ‘to shoot’. A following /-w/ could also result in shift of
the prominence onto the second element, thus OE fēower > ME four(e)
‘four’, OE trēowþ > ME trowþe ‘troth’. :
At least two other falling diphthongs belong to the diphthongal set
of late OE (see Figure 6.1). The diphthongs in question are [æj] as in
dæġ ‘day’, whose uninflected form is also spelled <dæig, deig, daig,
deih, dæi, dei, dai> (DOE), and [ej] as in Angl. grēġ ‘grey’, also spelled
<greig, grei> (DOE). The glide [-j] in these diphthongs was originally
a palatalised allophone of the voiced velar consonant representing PrG
/γ/ (see 4.2.1). The difference between the inherited diphthongs and
the ones arising by glide vocalisation in OE is that the [æj], [ej] preserve
their diphthongal character much longer, most likely due to the more
perceptually distinctive end point [j].
The formation of [æj], [ej] presupposes that the vowel and the glide
are in the same syllable; it is therefore logical to expect that diph-
thongisation would be delayed in forms realised predominantly as
disyllables. This is one of the reasons why these diphthongs are gen-
erally treated under the rubric of ‘New ME diphthongs’ (Lass 1992a:
49–53) – the loss of the final unstressed vowel in many forms in ME
increases the incidence of tautosyllabic vowel + palatal glide. Another
reason why these diphthongs are chronologically (dis)placed in ME is
that contextual diphthongisation due to the emergence and vocalisa-
tion of post-vocalic glides is broader in the transitional period and in
ME, involving not just the original palatal [j], but also the voiced velar
fricative [γ] (see 4.2.1), the labial approximant [w] and the different
realisations of the voiceless fricative /h/ (see 5.1.4). Another reason to
postpone the discussion of a significantly enriched diphthongal system
is loan phonology: Old Norse had a set of [-j, -w] diphthongs, but the
ON phonological influence is not clearly traceable before early ME
(Dance 2003). In addition to the ON influence, ME borrowed /oi/ and
/ui/ from Anglo-Norman. We will revisit these new and not so new
diphthongs in 7.4.
One final comment on the OE contextual diphthongs: vocalisation
of the glide obscures the etymological length of the original vowel;

whether the input was a short vowel, as in dæġ ‘day’, Gmc *dag-, or a
long vowel, as in grēġ ‘grey’, Gmc *græ ˉ-, the result is a falling diph-
thong of the same type as the inherited [æə], [eə], [iə]. The neutralisa-
tion of the original vowel length in favour of simply ‘a diphthong’ is
an argument against positing a fully developed system of diphthongs
contrasting in length in OE: had there been a pattern of well-integrated
monomoric diphthongs (see below) vs bimoric diphthongs (all of the
inherited diphthongs), it is hard to explain why the length distinction of
the input would not be maintained in [-j, -w] diphthongs.
This last point takes us to one of the greatest uncertainties in recon-
structing the OE vowel system: the interpretation of the digraphs
<ea>, <eo>, <ie> representing historical short vowels as in <earm>
‘arm’, <geolo> ‘yellow’, <seolh> ‘seal’, <hliehhan> ‘laugh’.28 The
digraphs, which are appealingly similar to the digraphs for the inher-
ited diphthongs in (11), show up before coda /-h/, coda liquids,
heterosyllabic back vowels and in identifiable palatal contexts. The
spellings <ea>, <eo>, <ie> are commonly taken( as representing
( ( inno-
vative monomoric phonemic short diphthongs /ea/, /eo/, /ie/ specific
to OE. However, the reconstruction of underlyingly contrastive short
diphthongs faces a series of objections.
First, typological comparison calls in question an analysis of coexist-
ing mono- and bimoric diphthongs which are otherwise identical. In
addition to the general paucity of languages with contrastive long and
short diphthongs, phonemic short diphthongs would make OE unique
within its own language family. No other Germanic language, past or
present, has contrastive long and short diphthongs that do not differ
in some other feature. Another objection, based on the OE system of
contrasts, was mentioned above: if diphthongal length was contrastive
elsewhere in the system, there is no reason why the etymological length (
in [ej] as in OE hēġ ‘hay’, also spelled <heig>, and (presumably) [ej ]
as in OE weġ ‘way’ < Gmc *weg-, should not have been preserved as
/e
j/ in hay and /ej/in
( way, in conformity
( with the hypothesised length
contrast /ea/ and/ea/, /eo/ and /eo/. :
The third, and probably
( ( the (
most damaging argument against the
phonemic status of /ea/, /eo/, /ie/, comes from subsequent history: in
late OE and ME these ‘short’ diphthongs merged with short monoph-
thongal counterparts: <ea> merges with <æ>, and <eo> merges with
<eo,
( e, o, ue>, the latter representing [œ/ø] ~ [e] (see 6.5.1). Ignoring
/ie/ because of scribal instability (see Figure 6.6), (12) shows some
28
This section follows the arguments in Minkova (2013).
characteristic examples of the orthographic mergers in late OE–early

ME.
(12) Merger of <ea>, <eo> with short vowels:
OE Early ME (based on LAEME)
mearc; æppel mark(e), ‘mark’; appel ‘apple’
healf; bæþ half(e), halue ‘half’; bað, bath ‘bath’
deorc; berie derk(e), ‘dark’; berie ‘berry’
ġeolu; elles yelewe, geleu, geluwe ‘yellow’; elles ‘else’
Finally, the evidence from actual minimal pairs is quite limited and
open to alternative accounts. Frequently cited minimal pairs are uncon-
vincing: the orthographic distinction between ( bearn ‘child’ vs bær ‘bore,’
p.t. sg., is not necessarily interpretable as [-ear-]: [-ær] because the form
bær could be influenced by the past plural form bæ ˉ ron [bæ
r-], and by the
related form -bæ ˉ re as in leohtbæ
ˉ re ‘light-bearing, luminous’, lustbæ ˉ re ‘joy-
bearing’. We also find the spellings <barn>, <bærn> for ‘child’. Another
pair, ærn ‘house’ vs earn ‘eagle’ (Smith 2009: 56–7), is dubious because
the spellings <earn> for ‘house’ and <ærn> for ‘eagle’ are firmly in
evidence; they could have been homophones. : ( (
The strongest defence of the phonemic status of /ea/, /eo/ is con-
sistency of spelling, but this does not apply to all dialects. The digraph
spellings are fairly uniform in the ‘focused’ variety of late West Saxon
OE, where some orthographic homogeneity can be attributed to the
strong normative tendencies characteristic of the Winchester school
and the stability of the Ælfrician texts. Scribal codification notwith-
standing, it appears that since the diphthongal spellings occur in well-
defined phonetic environments, it makes good sense to treat them as
representing allophones of the respective non-diphthongal vowels with
which they later merge.
The ‘digraph controversy’, one of the most ‘complex and acrimoni-
ous debates in the history of OE scholarship’ (Lass 1994: 45), was con-
ducted in terms of a theory of phonology that separates phonemes from
allophones categorically. Current phonological theory recognises that
segmental inventories are comprised of units that can range from fully
contrastive to fully allophonic, with intermediate stages (Goldsmith
1995: 12). In principle, the OE situation parallels the gradience illus-
trated by the tensing of [æ] to [e
] in AE. The process is gradual in
some northern dialect areas, categorical in the Mid-Atlantic cities,
and lexically idiosyncratic before /d/ in Philadelphia.29 Seen in the
context of scalar contrastiveness, the OE short digraphs never stood for
29
The three words that have the tense vowel in Philadelphia are bad, mad and glad. See
( (
stable and fully phonemic entities. The realisations /ea/, /eo/ or, more
likely phonetically [ε], [e], are best analysed as ‘not-yet-integrated
semi-contrasts’, a status half-way between a phoneme and an allophone
(Goldsmith 1995: 12).30 From the point of view of the subsequent
history of the sounds spelled with the digraphs <ea>, <eo>, <ie>, then,
calling them ‘diphthongs’ is justified only for the vowels marked here
(but not in the manuscripts) with the macron diacritic.
6.5.4 Unstressed vowels

So far our survey has focused on vowels in stressed position; it is only
in prosodically prominent syllables that one expects to find the full
set of vowels shown in the inventory in 6.2. Under stress, phoneti-
cally associated with the degree of loudness, length and pitch in the
pronunciation of a syllable, the vowels are most likely to be realised
with their full set of features. Phonologically, stress plays a decisive
role in the maintenance and survival of contrasts. Stress is perceived
in relation to adjacent syllables: we identify [-d ei-] in a.dja.cent and
[si-] in sy.lla.ble as more prominent than the neighbouring syllables. In
non-monosyllabic words one syllable carries the primary stress; other
syllables may have reduced stress, or they can be completely unstressed.
The lower the degree of stress in a syllable, the more likely it is that
the nuclear vowel will lose some of its characteristics, resulting in vowel
reduction.
Recall from 3.4.3 that the system of free pitch accent characterising
Proto-Indo-European was replaced in Germanic by stress on the first
syllable of the word root. The entailment of prominence fixed on the
first root syllable is that the vowels in inflectional syllables, post-tonic
root syllables and commonly in affixes (see 9.4.2), were pronounced less
forcefully. This leads to a series of vowel reductions in unstressed posi-
tion in all Germanic languages. The details in the branches of Germanic
differ, but taking Gothic as a model, all final vowels lost a mora: mono-
moric vowels except *u were lost, long vowels were shortened and
overlong vowels became long/bimoric (Ringe 2006: 75).
Further, reduced force in the pronunciation of vowels in unstressed
syllables in Germanic led to the elimination of length distinctions.
Thomas (2001: 19–23) for further details on the split, which he describes as ‘not truly
phonemic in the traditional sense’.
30
The schwa glide is a short cut. The reconstruction of a rounded glide allophonically
induced by the following consonant, with the degree of rounding variable depending
on the height of the vowel is convincingly presented in Stockwell (2002a).
Some of the original short vowels were lost, and all long vowels were
shortened. The reconstruction of the set of unstressed vowels in early
OE (Campbell 1959: §§369, 371; Reszkiewicz 1973: 89–99; Hogg
1992b: 119–22) shows only four vowels in the so-called ‘little’ vocalic
system.
(13) The unstressed vowels of early OE (pre-c. 700):
Front Example Back Example
/i/ cyning ‘king’ /u/ duru ‘door’
/æ/ endæþ ‘ends’ /ɑ / nefa ‘nephew’
A general lowering of /i/ to /e/ (or possibly /i/ to /ε/) in all
unstressed syllables – except in the derivational suffixes -iġ, -isc, -ing,
which were heavy syllables and were possibly carrying a degree of
stress – created an unstable three-way contrast in the front set: /i/
- /e/ - /æ/. As Crosswhite (2004) has shown, vowel reduction is not
a unitary phenomenon and it can be driven by different factors. The
scale of reduction may be related to accentual prominence, syllable
weight and to vowel sonority. One way in which lack of prosodic
prominence affects vowels is loss of sonority; in the case of the early
OE unstressed /e/ and /æ/ this resulted in merger of these two front
vowels into a single mid vowel represented by <e>. The next stage in
the elimination of front vowel contrasts is their convergence into the
low-sonority /ə/.
As for the back vowels, /u/, when followed by a consonant, tended to
be lowered to /o/ unless the following consonant was a labial, as in the
dat. pl. inflection -um: stānum ‘stones’, heofenum ‘heavens’. The merger of
the back vowels, like the earlier lowering of fully unstressed /i/ to /e/,
may be attributed to the phonetic phenomenon of ‘undershoot’, ‘a situ-
ation in which a given speech sound is articulated in a manner that does
not fully instantiate the canonical realization of that sound’ (Crosswhite
2004: 216). The articulation time for the high vowels is short, which can
prove insufficient for the full realisation of the sound. From the point of
view of perception, too, the lowering of /i/ and /u/ in unstressed posi-
tion can be related to the nature of the vowels: high vowels are shorter
than low vowels; the quality of the shorter vowels would be harder to
perceive.
Still, in early OE the unstressed vowels maintained some distinctive
qualities, and one can assume that vowel-letters in unstressed syllables
represent syllabic peaks. Starting in the ninth century, and definitely
by the end of the OE period, however, there is ample evidence that
the contrasts in unstressed final syllables were obscured. Thus late
OE–early ME orthographic records indicate that most unstressed
vowels were subject to blurring and confounding of phonetic contrasts.

Phonologically, this means that the vowels were losing distinctive
height and backness feature specifications, merging into a single schwa
/ə/, using /ə/ as an umbrella identification consistent with a wide range
of articulations in a wide range of contexts. : The diagnostic of vowel
reduction is the replacement of earlier vowel-letters by <-e>. Such
replacements occurred both in the unstressed syllables of monomor-
phemic words and in inflections. Some examples are <nosu> ~ <nose>
‘nose’, <nama> ~ <name> ‘name’, <fana> ~ <fane> ~ <fanu> ‘vane’,
<heofon> ~ <heofen> ‘heaven’, <biscop> ~ <biscep> ‘bishop’, <lufian>
~ <lufien> ‘to love’, which were probably pronounced with a type of
/ə/-compatible vowel in the second syllable. Such variant spellings are
found in pre-Conquest documents, in spite of the conservative nature of
late OE orthography and the expected time lag between sound change
and its written representation. Indeed, scribal evidence for vowel reduc-
tion is found in literary manuscripts as early as the second half of the
tenth century, prompting a reconsideration of the dating of the OE
to ME transition period: ‘The transition period from Old to Middle
English is not the twelfth century, as the grammarians used to think,
nor even the eleventh, as most of them think today, but rather the tenth’
(Malone 1930: 117).
Not all contexts were equally conducive to reduction. Inflectional
suffixes are more prone to reduction and even syncopation. Initially
syncope was prosodically conditioned by the presence of a stressed
heavy syllable, but by the end of OE syncopation of inflectional
unstressed vowels was more general. : This is particularly noticeable
in the 2nd and 3rd pers. sg. present indicative of both strong and weak
verbs, where dropping the <-e-> in the inflections -(e)s(t) and -(e)þ
was the dominant pattern in West Saxon texts. There is also frequent
confusion in the paradigm of originally monosyllabic stems ending
in a sonorant, as in wœ̄ pn ~ wœ̄pen ‘weapon’, in originally disyllabic
stems such as engel ~ englum ‘angel’, and even syncope of internal
vowels in derivatives as in dyslic ~ dysiglic ‘dizzy’, eft-selenes ~ eft-selnes
‘reward’.
Heavy derivational -VC(C) syllables appear to preserve the quality
of the unstressed vowel distinct into early ME: the suffixes in, for
example, (ge)bēorscipe ‘beership’, brosnung ‘corruption’, dœ̄dliċ ‘deed-ly,
active’, manig ‘many’ do not show orthographic variation of the vowels.
A difference should be made between the pre-tonic and post-tonic
vowels. An unstressed initial syllable in OE was prefixal by default, but
only the light prefixes: be-, ge-, a-, uniformly unstressed in the lexicon,
can be treated as fully reduced, for example <be-> ~ <bi->, <ge-> ~
<gi->. Other prefixes, such as un-, mis-, wiþ-, which had reduced stress
only in verbs and adverbs, preserve their spelling unchanged.31
The next instalment in the history of final vowels in English is 7.4,
where we look at the details and the profound phonological and ortho-
graphic consequences of the reduction and ultimate loss of unstressed
vowels. As this section has shown, however, the amalgamation of vowel
characteristics in many unstressed contexts was a necessary first step in
the later process of schwa loss. In addition to the scribal testimony that
fully unstressed vowels were similar, there is also a strong probability
that in faster and more relaxed speech final schwas were ignored. Some
lexical items have doublet uninflected forms with or without a final
vowel: adela ~ adele ~ adel ‘filth’, (a)dun ~ (a)dune ~ (a)duna ‘downward’,
bæc ~ bece ‘stream’, cat ~ catte ‘cat’, cœ̄ġ ~ cœ̄ġe ‘key’, drync ~ drynca ‘drink’,
eall-rihte ~ eall-riht ‘exactly’, earc ~ arc ~ earce ~ arce ~ earca ~ arca ‘ark’,
ēare ~ ēar ‘ear’, fœ̄ hþ ~ fœ̄ hþu ‘hostility’. Though tentative, there is some
evidence from verse suggesting that inflectional final vowels could be
elided before another vowel, as in Beowulf 1997b Gode ic ðanc secge ‘to
God I thanks say’, a verse which would be a vanishingly rare type if the
sequence gode is . . . is scanned as trisyllabic. : It is also the case that
the final vowel of the first elements of compounds, generally held to be
stable until ME, is in fact missing in many OE compounds; thus we find
cēap-mann ~ cȳpe-mann ‘merchant’, dūne-stīgan ~ dūn-stīgan ‘descend’,
dyrn-licgan ~ dyrne-licgan ‘commit adultery’ (DOE). There can be no
doubt that the roots of the pervasive ME schwa loss are in evidence
in OE. Therefore, if we associate the ‘middle’ period in the history of
English by levelling of the vowels in final unstressed syllables, which
existed already in the second half of the tenth century, and their incipi-
ent loss, ‘the beginning of the Middle English period . . . must be put at
A. D. 1000 or thereabouts’ (Malone 1930: 110).
31
The patterns of prefixal stress in OE are discussed in Minkova (2008a).
7 The vowels in Middle English.
Dialects. Spelling innovations.
Vowel quality and quantity. PDE
alternations traced back to ME
DIZZY-BUSY, FURY-BURY, MOON-MONDAY, STEAL-STEALTH,
GAME-GAMMON, GRASS-GRAZE
7.1 ME dialects
Linguistic change is continuous and uneven, and putting an exact
date to an innovative pronunciation or a variant grammatical form is
impossible. The beginning of the ‘middle’ period in the history of the
English language is a stretchy notion (see 1.3, 6.5.4). Bearing in mind
that the entire eleventh century is ‘transitional’, we still talk of pre- and
post-Conquest English, taking a major historical event, the Norman
Conquest of 1066, as the symbolic start of the ‘middle’ period.
The reference variety that we used in describing the phonology
of late OE was based on clusters of texts reflecting the scribal tradi-
tions in major monastic centres at the end of the first millennium.
While there was no ‘standard’ OE, there was at least a template of
accepted and codified forms, a ‘focused’ variety, commonly referred
to as Ælfrician English, which could serve as a model for other scribes
and authors. Geographically, that variety of OE is associated primar-
ily with Winchester. Turning to ME, we find that the input to what
will eventually be recognised as Southern Standard British English
(SSBE) is rooted in forms characteristic of documents produced in the
Midlands and East Anglia towards the end of ME (Samuels 1963). The
discontinuity is even more striking if we add that the documents identi-
fied as Mercian in OE, one potential ancestor of ME Midlands and East
Anglian varieties, come from Worcestershire and Warwickshire, and
are thus essentially West Midlands in origin. This leaves us in the dark
about the provenance of Standard English – all we can say is that it is not
based on any particular variety of OE.1
This perplexing labyrinth of dialectal tracks is unsurprising in view
of the historical and linguistic context. The Norman Conquest led to
1
This is a summary of the fully argued agnostic conclusions in Hogg (2006b) and Laing
and Lass (2006).
184
THE VOWELS IN MIDDLE ENGLISH 185
new administrative and cultural practices: the use of the vernacular for
legal records was replaced by Latin and to some extent French, though
some copying of texts in the late OE-focused model continued for about
a century. There is virtually no surviving original writing in English for
several generations after the Conquest. The second continuation of The
Peterborough Chronicle (1154/5) is considered the first document in ‘truly
Middle English language – that is, language that reflects how the spoken
language of this region had developed in the preceding century’ (Laing
and Lass 2006: 419). When the metaphorical curtain on English is
lifted in the second half of the twelfth and the early thirteenth century,
the records are very patchy: the majority of the extant texts are from
the West Midlands, the East Midlands, and the Southeast. For early
ME, the period covered by LAEME (c. 1150–c. 1325), there are large
geographical areas for which the information is either lacking or very
sparse: almost nothing from the North or the North Midlands, from the
Central Midlands or from the extreme South-West. It is this patchiness
and discontinuity that makes the designation ‘Middle English’ even
more of an agglomeration than ‘Old’ or ‘Early Modern’ English: Middle
English is a convenient composite of a range of unevenly documented
dialects, often remarkably unlike each other. The textual coverage
after the middle of the fourteenth century is fuller, but even LALME
(c. 1350–c. 1450) is thin on more northern sources. A special problem is
presented by multilingual/mixed-language texts, where there is code-
switching between English, Latin and French (Schendl 2002). There is
also the complication of multiple copies of lost originals, where the lan-
guage of the manuscript may show a mixture of authorial forms, forms
native to the copyist and forms reflecting the standard(s) that the scribe
was following.
Interesting as the detective work in this historical area is, it is not our
remit in this book, so we will take a very general view of the periodi-
sation and varieties of Middle English. We will follow the traditional
chronological designations of ‘early’ and ‘late’ ME, and as for dialect
areas, we will refer to the divisions in the map in Figure 7.1: Lowland
Scots, Northern, West Midland, East Midland, South-Western and
South-Eastern.
Put crudely, then, the linguistic situation in post-Conquest England
can be characterised with reference to a shift from records primarily in
English, to records in Latin and Anglo-Norman, the suppression and
subsequent re-emergence of English as a language used in the produc-
tion of literary and administrative documents, and the fragmentisation
of English into dialects, followed by incipient formation of a national
written standard in the fifteenth century.
Figure 7.1 Schematic map of the dialects of Middle English, Figure 4.4 in
Horobin and Smith (2002: 51), An Introduction to Middle English, Edinburgh
University Press
7.2 Notes on ME spelling (vowels): <TAXI>-<ATAXY>, <TYRE>-<TIRE>,

<COME>-<CUT>, <SEE>-<SEA>
The reconstruction of sound histories from the surviving ME written
records poses a myriad of challenges: scarcity of twelfth- and thirteenth-
century records in English, lack of a relatively unified orthographic
practice, the scribes’ exposure to Latin and French writing, their own
uneven command of the three languages, and the fact that in most cases
they were copying not just original English compositions but transla-
tions and copies made by others. Until at least the middle of the four-
teenth century the basic training for the scribes would have been for
them to write Latin or French and not in the vernacular. :
The expanding use of English in the fourteenth century makes it pos-
sible to identify clusters of texts which are less obviously idiosyncratic,
and which can serve as the basis for establishing the lineage of the incip-
ient fifteenth-century written standard. Two of these clusters – Type
I, Wycliffite texts, mostly stemming from the Central Midland coun-
ties, and dating from the mid fourteenth century on, and Type IV, the
‘Chancery Standard’ used in government documents after c. 1430 – are
‘focused’ varieties. It is important to bear in mind that there is as yet, and
for more than three centuries to come, no ‘fixed’ standard, but at least
by the middle of the fifteenth century ME was a language represented
more reliably by scribes trained to follow a more rigid orthographic
tradition, in principle not unlike the Ælfrician tradition of late OE.:
7.2.1 Letter-to-vowel correspondences in ME (single letters):

SUN-SON, CONE-COME
Acknowledging that there is no single stable spelling system that can be

taken as a continuation of the late OE system, and that ‘middle’ English
comprises four hundred years of phonological variation and change,
one can still draw hypothetical connecting lines between the letter-
sound correspondences of ME. Some of the recurrent correspondences
between vowel letters and the sounds they represent most commonly in
ME are summarised in Figure 7.2. :
The values shown in 7.2 are neither static nor exhaustive – the chart
only highlights differences between the OE representation of vowels
and ME, in order to provide a link to the values of vowel-letters in
EModE.
The letters and <y> (and also the digraph <ie>) were used
interchangeably already in OE (see 6.5.1), and indeed the choice of 
or <y> in PDE can still be unstable, as in tyre ~ tire, especially in names:
Robin ~ Robyn, Judi ~ Judy, Cindi ~ Cindy, Kelli ~ Kelly, Joni ~ Jony. In
spite of the convention to use <y> word-finally – ataxy, namby-pamby –
 shows up in clippings and prefixes as in taxi, maxi, combi, ambi-. The
potential for ~ <y> variability is exploited commercially as in the
attention-getting Citi(bank), Cristal (champagne), Infiniti (car), Sereniti
(records). The phonetic values [y] and [y
], preserved mainly regionally
in ME (see 7.3), could also be represented by <y>, , <ui>, <yu>.:
The letter <y> shape-shifted with <þ> in ME – this is familiar from
SHORT LONG
Letter
Value Example Value Example
 bin ‘bin’ rid(en) ‘to ride’

i i:
<y> synn(e) ‘sin’ fyr(e) ‘fire’
e: swet(e) ‘sweet’
<e> ε sett(en) ‘to set’
ε: del(e)‘deal’
<æ> æ/a bæð/bað ‘bath’ æ:/ε: ræd(en) ‘read’

<a> ɑ fall(en) ‘fall’ ɑ:/ɔ: bat, bot(e)’boat’
ɔ (i)bor(e)n ‘born’ o: mon(e) ‘moon’

<o>
υ son(e) ‘son’ ɔ: bot(e) ‘boat’
drunken ‘drunk’ hus(e) ‘house’

 υ u:
vnder ‘under’ vt(e) ‘out’
(<v>)
y furst ‘time’ y: duk(e) ‘duke’
Figure 7.2 Letter-to-stressed-vowel correspondences in ME (single letters)
archaising modern spellings of <the> as <ye> or, more frequently,

<ye> as in Ye Olde King’s Head. : On the shape-shifting of <y> ~ <þ>
see Laing and Lass (2009), who describe the paleographic basis of the
confusion and define the importance of the different scribal practices
with respect to <y> ~ <þ> in establishing the regional provenance of
texts. The runic þ fell into disuse in the fourteenth century though the
use of <y> instead of <th> continued in hand-writing into the nine-
teenth century.
The letter <e> was the most common letter for unstressed vowels,
and it was also used for at least three contrasting stressed mid vowels in
ME: the front upper mid [e
], the front lower mid short [ε] and long [ε
].
In the fifteenth century and after, historical [e
] and [ε
] began to be dif-
ferentiated in spelling, using <ee> for [e
] as in flee, see, sweet, and <ea>
for [ε
] as in flea, sea, seat. The reliability of the <ea>-<ee> spelling as
an indicator of etymological [e
] and [ε
] respectively is much lower if
the vowel is followed by /-r/ (see 8.3.2).
The ligature ash, <æ>, is preserved only in the earliest documents;
by the end of the thirteenth century the scribes were no longer using
it. It was replaced by either <a> or <e>. The letter <o> represented
three contrasting stressed mid vowels in ME: the back upper mid [o
],
the back lower mid short [ɔ] and the corresponding long [ɔ
]. As with
[e
] and [ε
], the upper mid [o
] and the lower mid [ɔ
] began to be
spelled differently in late ME, with [o

] represented by <oo> and [ɔ
]
represented by <oa>.
The letter continued to be used for [υ] and also [u
], though see
7.2.1 for <ou> and <ow> replacing for [u
]. An ME innovation is
the addition of <v> to the inventory of letters. It was used both for the
consonant /v/ and for the vowels [υ] and [u
], usually in word-initial
position, as in vnder ‘under’, vt ‘out’. :
Of particular interest because of its effect on the spelling of PDE is
the scribes’ practice of replacing etymological with <o> when
adjacent to <i, u, m, n> within the same word: OE <sunu> > ME
<son(e)> ‘son’, OE <cuman> ME <com(en)> ‘come’, OE <druncen>,
ME <dronk(en)> ~ <drunk(en)>, similarly some, cover, hover, honey, monk,
tongue, wonder, worse, shortened month, dove, glove. The substitution was
intended to avoid a confusing sequence of downstrokes in the writing
of letters with vertical strokes: combinations such as <-um-, -un-> are
harder to read than <-om-, -on->. This is a purely visual stratagem:
the spelling change does not indicate a shift in the pronunciation of
the vowel, so son, come have the same PDE vowel as luck, cut (see 8.2.1).
Recall from 6.2 that vowel length in OE is not represented in the
orthography. This situation began to change in ME. The most famous
innovation in this respect was the system introduced by the monk Orm
in his extensive verse translation and exegesis of scriptural texts. He
devised a system whereby short vowels were marked by doubling of
the consonants: follc ‘folk’, þiss ‘this’, þennkenn ‘think’. Long vowels can be
marked by acute accents, as in fór ‘fared’, tíme ‘time’. The system was not
perfect, but as the earliest attempt at a ‘spelling reform’ in the history of
the language, it was remarkably forward looking. :
Orm’s highly informative spelling conventions were not adopted
by other scribes. Vowel length continued to be ignored by thirteenth-
century scribes, though some alternative ways of marking long vowels
gradually came into being in the fourteenth century. One way of
indicating length was by doubling the vowel letter: <aa> ‘aye’,2 <see>
‘sea’, <stoone> ‘stone(s)’, <ook> ‘oak’, are some early examples found in
LAEME. Another innovative way of marking long vowels, related to
the lengthening of stressed vowels in disyllabic words in the environ-
ment <-Ce> (see 7.5.2), was to code length by preserving or adding the
final <-e>, as observed in uninflected forms of ME wife (OE wīf), stone
(OE stān), hale/hole (OE hāl) ‘whole’. This is, of course, the ‘silent <-e>’
familiar to any schoolchild – <-e> is now established as the diacritic
2
This adverb, OE ā ‘always, ever’, is recorded as <aa> already in OE (DOE).
Occasional doubling of for [i
] in OE is found in wiif ‘wife’, tiid ‘time’, liif ‘life’.
that distinguishes the vowels in pal – pale, pet – Pete, pin – pine, rob – robe,
cut – cute, lyric – lyre (see further 7.5.3).
7.2.2 Letter-to-vowel correspondences in ME (digraphs): BEAT-BEET,

ROAD-ROOD
Yet another way of distinguishing vowel length was by using digraph

spellings. Some digraph spellings were inherited from OE: <ea>,
<eo>, <ie>. They continued in use in OE documents copied until
the thirteenth century. Subsequently, the same graphs were adopted
for the representation of new values: <ie> for [e
], <eo> for [e
] or,
perhaps, for a front rounded [œ]/[ø
] in the West Midlands and the
South-West.:
It must be kept in mind that the PDE use of the digraphs <ea>,
<oa> does not go directly back to OE. Towards the end of ME, when
[ε
] and [e
], and [ɔ
] and [o
] were still contrastive (see 7.3), the letter
<a> was employed as a diacritic to mark the low/open mid vowels. As
noted above, <ea> was associated with [ε
] (beat, sea, meat). A parallel
innovation was to use <oa> for [ɔ
] (boat, coal, road). Doubling of <e, o>
marked the high mid vowels [e
] (beet, see, meet) and [o
] (boot, cool, rood).
One further regional peculiarity in the marking of long vowels is that
the letter , in addition to marking the second element of diphthongs,
as in dai ‘day’, pleinten ‘complain’, boi(e) ‘boy’, was also used as a length
diacritic. In that function too, alternates with <y>. In ME this
scribal practice is characteristic of northern dialects, including Lowland
Scots, for example <ai> for [a
] as in baith, bayth ‘both’, laith, layth ‘loath’,
stain, stayn ‘stone’; <ei> for [e
] as in keip ‘keep’, heid ‘head’, weill ‘well’;
<ui> for [y
] as in bluid ‘blood’, buik ‘book’, guid ‘good’, puir ‘poor’.
The repertoire of the post-Conquest scribes was expanding; it was
also remarkably fluid, allowing multiple variant spellings of the same
lexical item. To illustrate: the OE word <prēost> ‘priest’ is recorded
in the MED as ‘preste, presth, priest, prist, (late) preast, (chiefly N)
preist, (SWM) pruest (chiefly early or S or W) preost, prust, prost, (early
SWM) proest, (early SW) præst . . . and (in names only) pres, presse, preos,
pris-, presti-, purs-.’:
Figure 7.3 presents an overview of some of the most frequent vowel
doublings and digraphs in ME and their main sound correspondences.3
3
Digraphs with <y> and <w> as a second element are not included in the chart. The
letter <w>, which was introduced by Anglo-Norman scribes to replace the OE wynn
(<W>), was also interchangeable with and <uu>: hous(e) ~ hows(e) ‘house’, fouel
~ fowel ~ fouul ‘fowl’. Recall from 2.2.2 that the rationale for transcribing diphthongal
Letter Some typical values Examples
<aa> [ɑ:], [a:] baas ‘base’, n. and adj.
<ai> [aj], [ej], [a:] dai ‘day’, þai ‘they’, stain ‘stone’
<au> [ɑw] clau ‘claw’, drau ‘draw’
<ea> [ε:], [e:] sea ‘sea’, beand ‘being’
<ee> [e:] or [ε:] been ‘be’, eet ‘eat’
<ei> [ej], [aj], [e:] dei ‘day’, þei ‘they’, theif ‘thief’
<eo> [e:] feond ‘fiend’
<eu> [εw], [ew] reume ‘realm’, deu ‘dew’
<ie> [e:], [iə] trie ‘tree’, fiet ‘feet’, fier ‘fire’
<iu> [iw] niuwe ‘new’, bliu ‘blue’
<oa> [ɔ:] woa ‘woe’, toa ‘toe’
<oi> [oj] boi(e) ‘boy’, boil(en) ‘boil’
<oo> [o:], [ɔ:] good ‘good’, stoon ‘stone’
<ou> [u:], [ow] hous ‘house’, stou(e) ‘stow’
<ui> [uj] (AN) puint ‘point’, puisun ‘poison’
Figure 7.3 Doubling of vowels and some digraph spellings in ME
Of relevance to the spelling of PDE is that for the value [u

] the <ou>
spelling was ‘borrowed’ from Anglo-Norman, where <ou> was the
normal digraph for [u
]. In the course of the adaptation the <ow>
spelling was preferred word-finally and when the [u
] is adjacent to a
‘vertical strokes’ letter such as <l, n>; otherwise the spelling is kept as
<ou>; compare OE <nū>, <tūn> with ME now, town; OE <(a)būtan>,
<mūþ> with ME about(e), mouth. As noted above, continued in use
for [υ], occasionally also [u
], but the introduction of <ou, ow> ‘freed’
 for the representation of the sound written <y> in OE (see 7.3).
glides as [i] for the front glide and [υ] for the back glide can be debated. We have kept
these for PDE, but we switch to [j] and [w] in the context of the ‘new’ ME diphthongs
because of the clear association between the consonantal sources and the new glides.
7.3 Qualitative changes in ME
7.3.1 Short vowels: DIZZY-BUSY, FURY-BURY, MERRY-MIRTH

This section addresses the major qualitative shifts of the short vowels
from Old to Middle English, summarised in Figure 7.4. 4
Late Old English Late Middle English Examples4
i y υ i υ OE [synn] ME [sin] ‘sin’
ε ɔ ε ɔ
æ ɑ æ/a OE [θæt] ME [θat] ‘that’
Figure 7.4 Short vowel changes from late OE to late ME
Three of the stressed short vowels, [ε], [υ] and [ɔ], remained relatively
stable in most environments: OE lettan – ME let(en) ‘hinder, let’, OE botm
– ME bot(te)m ‘bottom’, OE full – ME ful ‘full’. The hedge ‘in most envi-
ronments’ is because coda [-ŋg], or single or covered nasal codas, had the
effect of raising the nuclear mid vowels, so LOE <-eng> and <-ong>
were raised to <-ing> and <-ung>. Thus we find early fourteenth-
century rhymes such as OE genge, ON gengi ‘troop’, ME gyng, rhyming with
coming, kyng; EModE weng(e) < ON vœ̄ ngr ‘wing’ rhyming with gerunds in
-ing.5 Only the front vowel raising is attested reliably in the orthography
because of the ambiguity of <o> and spellings before <n>, so OE
(on)ġemang ~ ġemong ‘among’ appears in ME spelling with <-ong>, rarely
<-ung>, but the [-υŋg] value can be inferred from its later pronunciation:
among rhymes with lung, OE lungen, not with strong < OE strang ~ strong.
The most notable example of the raising of the front mid vowel is the
adjective and noun English, for which there are no <ing-> spellings in
OE, but in LAEME’s thirteenth-century database we find twenty-two
tokens of Inglis(s). Other items in this set are ON vœ̄ngr > ME wenge ~
winge ‘wing’, ON flengja > ME flengen ~ flingen ‘fling’, AN *vencir, OFr
guenci(e)r > ME wincen ‘wince’, OFr enque > ME enke ~ ink ‘ink’. The
raising of <-en(C)> > <in(C)> started in early ME, continued into
EModE (late ME lenger > EModE linger), and is mirrored by the PIN-PEN
merger throughout the American South, in southern California, central
Ohio, Kansas and elsewhere in the US. :
4
The examples cover only vowels which underwent some qualitative changes from OE
to ME.
5
For more examples of rhymes showing raising before nasals see Ikegami (1984: 330–1).
The major changes, enclosed in the dashed ovals in Figure 7.4, are
mergers of pre-existing vowels. In the high front portion of the vowel
space the unrounding of [y] and its merger with pre-existing [i] for a
large portion of the input items is paralleled by the unrounding of [y
]
to [i
], so we can follow the practice of previous studies in treating them
together in this section. :
Recall from 6.5.1 that the OE spelling records show and <y>
used interchangeably. Even before the end of OE, [y] and [y
], the
output of I-Mutation (see 6.3), were treated differently in the differ-
ent dialect areas: they were lowered to [e]/[ε] and [e
] in Kentish
and they were in the process of merging with [i] and [i
] in the East
Midlands and the Northern dialects. For forms such as OE cyss(te)
‘kissed’, the recorded spellings for the vowel in OE are <y> ~ ~
<e>: cys-te ~ cis-te ~ kes-te, and for OE hȳdan ‘to hide’ we find hyden,
hiden, heden. To these the early ME records add a new spelling 
for the short vowel and , <ui> for the long vowel, as in cus-te,
hude(n) ~ huide(n). :
The interchangeability of and <y> makes it hard to judge
whether <-y-> spellings in early ME represent a rounded vowel or
not, but the use of <-u-> in words which had the front rounded vowel
in OE has traditionally been considered an indication that the vowel
preserved its rounding, so that <custe> ‘kissed’ is reconstructed as rep-
resenting [kystə], and hude(n), huide(n) ‘to hide’ as [hy
də(n)], parallel
to the pronunciation of French loans, such as bugle ‘bugle’, duc(e)/duk(e)
‘duke’. The spellings of OE <y> with , suggesting preservation
of front rounded vowels, are characteristic of texts originating in the
West Midlands. It is commonly held that the rounded quality of the
vowels descending from the OE front rounded vowels persisted into the
fourteenth century; the [y] and [y
] realisations were supposedly rein-
forced by the existence of the same vowels in French loans. The most
common description in the textbooks is that OE <y> had three regional
manifestations in ME, roughly as shown in Figure 7.5.
Examples
OE fyrst ‘first’ OE cȳ nde ‘kind’, n.
[i], [i:] (N, East Midl.) <first> <kind(e)>
OE [y], [y:] [y], [y:] (West Midl.) <furst> <cund(e)>
[ε], [e:] (South-East) <f/verst> <kend(e)>
Figure 7.5 Commonly posited realisations of OE [y], [y

] in ME
This tripartite division is entrenched in the canonical studies of ME, but

it is also debatable on several scores. As shown in Lass and Laing (2005),
a fresh examination of the presumed sound-spelling correspondences
in West Midland ME belies the claim that a front rounded realisa-
tion [y], [y
] was systematically associated with particular spellings.
Instead of positing continuing [y], [y
] in the Western dialects, which
previous researchers link to scribal practices based on French, Lass and
Laing suggest that in those areas the OE front rounded vowels were
retracted, merging with pre-existing [υ] and [u
]. The realisation of
front rounded vowels in French loans did exist, but the textual evidence
fails to confirm the ‘neat geographical tri-partition for /y(
)/’. There
are further aspects of the reconstruction that need clarification. For the
short vowel it is indeed the case that the retraction of [y] to pre-existing
[υ] in specific environments goes back to the twelfth century (Jordan
1974: 70–2); many of the forms surviving in PDE reflect this merger.
(1) West Midlands form of OE [y] in PDE:
OE ME PDE
clyċċan clicchen/clucchen clutch
ċyriċe chirch(e)/church(e) church
dyncge dung(e)/ding(e) dung6
myċel mich(e), mech(e), much(e) much
For the set in (1), which includes also PDE blush, bush, crutch, churn,
cudgel, furze, hurdle, shut, shuttle, such, sundry, thrush, thud, trust, the merger
with [υ] was sufficiently early and widespread to affect the embryonic
early fifteenth-century standard. Even some French loans such as OFr
buf(f)et, iug(g)e ‘judge’, iust ‘just’, punish(en) ‘punish’, ruser, ruiser ‘rush’ join
the backing to [υ],7 so there is no compelling argument from orthog-
raphy or later forms in support of OE [y] continuing as phonemically
contrastive in the West Midlands areas of ME.
On the other hand, positing a parallel backing of the long [y
] of OE
to [u
] based on -spellings in the Western area is much more prob-
lematic: there are no surviving lexical items reflecting the presumed
change – one would expect the nucleus OE brȳd ‘bride’, dȳfan ‘dip, dive’,
hȳdan ‘hide’, and so on to emerge as [u
] > [-aw] in some lexical items in
some PDE varieties, but it is not the case. : This is surprising in view
6
For this and the following item spellings with <o> for (see 7.2) are ignored.
7
Similar to the native high front rounded vowel, the [y] in French loans could be
unrounded to [i], for example OFr escumer > ME skimmen ‘skim’. One interesting
etymological pair is brisk-brusque. The OED notes suggest that ‘brisk is identical with
French brusque (which appears as bruisk in Scots c1560, and as bruske as early as 1600);
. . . the words appear to have influenced each other in early use.’
of the remarkably high rate of adoption of the West Midlands reflexes

of the corresponding short vowel in PDE as shown in (1). It is therefore
probable that the spellings of the etymological long front rounded
vowel and in the French loans in the West Midland dialects represented
a range of realisations: conservative [y
], possibly [ju
], and also [u
] and
[i
] – the latter based in the eventual absorption of OE [y
] by [i
] in
all areas except the South-Eastern areas, where the reflex was [e
], as
in OE bȳtel > ME (South-East) <be(e)tel> ‘beetle’. It is relevant that
the French front rounded vowel [y
]/[ju
] as in endure, fume, use did not
merge with the vowel of OE brȳd, dȳfan, hȳdan. So we know that in the
West Midlands OE [y
] was kept separate from French [y
], we know
that it was commonly spelled (also <ui>), we know that it joined
the [i
] of the neighbouring regions, but we are not close to a convincing
reconstruction of a single quality of that vowel. The identification of the
reflexes of OE <y> in ME in Figure 7.5 is thus open to more research,
which should take into consideration the uneven distribution of surviv-
ing texts in the various geographical points, the effect of the different
consonantal contexts on the merger with the front or the back vowel,
the token and type frequencies of loanwords with the relevant sound,
and the numerous lexical idiosyncrasies of spelling and survival.
The great majority of lexical items that had a front rounded stressed
vowel in OE do end up as [i] or [i
] in late ME, as in OE synn > ME sin
(see 7.4). This is in line with the observations that the emerging southern
standard was based primarily on Midlands and East Anglian forms (see
7.1). There are, however, a substantial number of items that do not fit the
OE synn > ME sin pattern in spelling, pronunciation or both. For these
items the OE <y> vowel can emerge with unexpected West Midlands
 or South-Eastern <e> in spelling, and a possibly unmatched [i],
[ε], [υ] in the pronunciation. This apparently unmotivated mixture of
ME developments of OE [y] is illustrated by the pairs in the title of this
section: dizzy-busy, fury-bury, merry-mirth. Dizzy and busy are a perfect
rhyme, and the identity of the vowel goes back to OE: dysiġ – bysiġ, but
in busy the spelling is West Midlands. Fury < French furie and bury < OE
byrian are only an eye-rhyme – the [-ε-] in bury is South-Eastern. In merry-
mirth the adjective has the South-Eastern spelling and pronunciation, but
the noun is East Midlands. Post-ME standardisation has obscured much
of the variability of the ME reflexes of OE <y>, but place names, which
commonly resist regularisation, are a strong reminder of the options in
ME. The third and the fourth columns in (2) show the orthographic and
phonological sources of these and of some other related items in PDE.8
8
All place-name etymologies are from Ekwall (1960).
(2) Mixed regional forms of OE <y>:

OE form Gloss ME spelling ME pronunciation
source source
(be)byrian ‘bury’ burie [-ε-]
bysig ‘busy’ busie [i]
myriġ ‘merry’ meri(e) [ε]
myrġð ‘mirth’ mirth(e) [-i-] or [-υ-] (8.3.2)
ċyriċe ‘church’ church(e) [-υ-] or [-i-] (8.3.2)9
byriġ ‘-bury’ bir- [-i-] or [-υ-] (Birstall, W. Yorks)
ber- [ε] (Berry Pomeroy, Devon)
bur- [ε] (Bury St Edmunds, Suffolk)
hrycg ‘ridge’ rug(ge) [-υ-] (Rugeley, Staffordshire)
hyll ‘hill’ hull(e) [-υ-] (Bishop’s Hull, Somerset)
h&¯þ-hyll ‘heath-hill’ Hethella [ε]/[ə] (Hethel, Norfolk)
mylen ‘mill’ melle [ε] (Mellis, Suffolk)
There is clearly no seamless uniform development of the OE front
rounded vowels in ME, but the common denominator in all of the out-
comes is loss of rounding. The observed instability raises interesting
theoretical questions. Including both the short and the long vowel, why
should the high peripheral [i] and [υ] be preferred over [y], and [i
]
preferred over [y
]?
The causes of unrounding of [y] to [i]/[y
] to [i
], or retraction of
[y] to [υ] can be addressed from different phonetic and related phono-
logical angles. In terms of phonological feature typology, one approach
could be to treat [y] and [y
] as intrinsically more complex because
they involve the additional phonological feature of rounding, which is
considered more ‘marked’ for non-back vowels (for example, Chomsky
and Halle 1968: 409ff.). Across the languages of the world front rounded
vowels are less common than the ‘primary’ vowels such as [i]/[i
] and
[υ]. Typically, backness and rounding are mutually enhancing, and the
disassociation of backness and roundness in [y] and [y
] adds to the
motivation for their avoidance. :
Phonetically, lip-rounding for the high back vowels maximises
articulatory ease; the flip side of this is that maintaining lip-rounding for
the high front vowels requires more effort. Perceptually, the oppositions
based on height, manifested in the acoustic frequency of F1, are more
basic and salient than the backness oppositions whose acoustic manifes-
tations involve the frequency of F2; F2-based contrasts are inherently
9
Compare the place name Cheriton (Devon) with the South-Eastern spelling and
pronunciation.
weaker (see Kaun 1995: 5.3.1 and the references there). In discussing F2
contrasts and vowel dispersion, Flemming (2004: 4.1) also finds that F2
contrasts involving non-peripheral vowels such as front rounded vowels
are ‘sub-maximal’ and ‘dispreferred’. In his dispersion framework there
are markedness relations over contrasts as well as over sounds – and the
contrast between [i] and [y] is more marked than the contrast between
[i] and [υ], that is, [i] and [y] are less distinctive than [i] and [υ]. Thus
the instability of ME [y] and [y
] can also be seen in terms of the vowel
inventory as a whole. An additional contrast in the set of high vowels
makes for a crowded vowel space while the loss of rounding optimises
the perceptual distance between the vowels.
Two other points related to the history of the front rounded vowels
need further clarification. What is it that determines whether the input
[y] will be unrounded to [i], as in OE synn > ME sin, or whether [y]
will emerge as [υ] as in OE blyscan > PDE blush? It has been suggested
(Jordan 1974: 70–2) that there is an association between the ME [υ]
realisation and adjacency to /tʃ, d , ʃ/, possibly before a rhotic. The
correlation is not stable, however, because we find divergent results
from the same input, so OE bryċġ > PDE bridge vs OE cyċġel > PDE
cudgel. It is possible that the affricates and the sibilant had a secondary
articulation involving lip-rounding which would enhance the probabil-
ity of a [-υ-] outcome. There is as yet no detailed study of the entire
corpus of OE [y] in ME which takes account of the exact phonetic envi-
ronment and the frequency of the relevant items. It is not clear whether
the selection of the back vowel over the front vowel can be related to
the later simplification of [y
]/[ju
] to [u
] (yod-dropping) as in sue, chew,
Jew, rue (see 8.2.2.3). :
Another area which will profit from further study is the different
behaviour of short and long vowels. Why do all of the OE long [y
]’s
end up as [i
] (OE hȳdan > ME hīde(n)), while the phonetically quite
similar [y
] in early loanwords merged most frequently with /iu/, as
in duke (1129), use (1225), huge (1275), jewel (1290)?10 Is it a historical
accident that West Midlands variant forms in presumably early and
conservative [y
], later [u
], were ignored in the selection of forms that
survived locally or in the standard language? Finally, how do we address
the fact that while the front rounded [y
] was ultimately unrounded in
the Southern varieties of ME, late thirteenth- and early fourteenth-
century Northern [o
] was first fronted to [œ] and then raised to [y
],
which is the reconstructed ME reflex of OE [o
] in words such as foot,
soon, spelled <fut>, <sun(e)>, recognisable (after shortening, see 7.4.)
10
One notable exception is PDE trifle from ME trufle < Old French trufle.
in the modern Scottish realisation of the vowel in the FOOT lexical set
as [y] (Wells 1982, II: 401–2). Clearly, the newly generated Northern
front rounded vowels are incompatible with accounts which rely on
the intrinsic markedness of such vowels without reference to the whole
system, the density of lexical items with front rounded vowels and the
influence of the phonetic environment. Furthermore, front rounded
vowels, both [y] and especially [œ], are very much in evidence in PDE:
rounding of the GOOSE vowel may be a change in progress in RP, and
the rounding of the NURSE vowel is attested in South African English,
Australian English, New Zealand English, and in the UK in Tyneside,
Liverpool, London, the West Midlands and South Wales (see Mayr
2010: 94 and the references there). The inconsistencies of survival,
assimilation and (re)introduction of front rounded vowels in the history
of the language and its current varieties is open to further study.
Moving on to the changes of the low vowels: the circle around [æ]
and [ɑ] in Figure 7.4 indicates another possible merger in ME.11 In
6.1 we noted that the contrast between OE [ɑ] and [æ] is a matter of
uncertainty, though following Hogg (1992a: 14, 98) we assumed that
they were contrastive, at least in West Saxon. The presence of a coda
nasal presents a special case: <a> ([ɑ]) is characteristic of West Saxon,
while Anglian has <o> ([ɔ]) in words such as OE bana ~ bona ‘killer’,
camp ~ comp(e) ‘battlefield’. This distinction continues into ME, but the
area of <-on-, -om-> spellings is confined to the West Midlands, with
<-an-, -am-> spellings much more broadly distributed; it is from the
East Midlands, the South and the North that PDE has emerged with the
<a> forms in man, hand, lamb, camp, and so on (see Jordan 1974: 50–4). :
The letter ash <æ> was used only in the earliest ME documents
(see 7.2). In Anglian OE as well as in Kentish OE [æ] shows raising and
fronting to [e]/[ε].12 Elsewhere the [æ] must have started lowering to
[a] in late OE; moreover, the phonological basis of the historical differ-
ence between [æ] and [ɑ] was often obscured in the paradigms of the
same word, and variant <æ> ~ <a> spellings are commonly attested,
for example OE bæð ~ bað ‘bath’, fæder ~ fader ‘father’, fæst ~ fast adj. ‘fast’.
During the thirteenth century the letter <æ> was completely replaced
by <a>, or <e> in the areas where the Anglian and Kentish forms were
preserved or adopted. Based on the orthographic substitution of <æ> by
<a>, the usual textbook accounts posit a general lowering and retrac-
tion of OE [æ] to [a]. This would cover the additional merger of [ɑ],
11
The complexities of the merger of the OE dipthongoid [æ] with [æ]/[ɑ] as in OE
feax~fæx ‘hair’, heall~hall ‘hall’ (see 6.5.3) are ignored here.
12
Compare OE bæc ‘back, ridge’ in Backbarrow (Lancashire) with Beckhampton (Wiltshire).
spelled <a> in OE, and [a] spelled <æ> in OE. The assumption that
the result of the merger is phonetically [a] is not necessary, however;
the merger could be anywhere in the [a] ~ [æ] range and it can be seen
as allophonic variation based on the phonetic closeness of the two low
vowels. Spellings such as beth ‘bath’, sed ‘sad’, weter ‘water’ are easy to
find in the ME records. OE [-ɑ-] and [-æ-], if subject to open-syllable
lengthening (see 7.5.2), emerged as a long vowel reconstructed as [a
]
in late ME: OE cnafa > ME [kna
və] ‘knave’, OE hatian, ME [ha
t(ən)]
‘hate’, OE hæsel > ME [ha
zəl] ‘hazel’. Reconstructing low front [a
]
is supported from later history: it is raised and diphthongised and
eventually merges with pre-existing [ej]. The intermediate steps for
lengthened OE low vowel [a] ~ [æ] to EModE [æj]/[ej] are a matter of
conjecture and one should be sceptical about the accuracy of the <a>
spellings as representing a uniformly low vowel if there was no <æ> in
the scribes’ repertoire, and <e> stood for a contrastive sound. The point
to take away is that there is nothing in the phonetic or phonological
history of the ‘generic’ short /a/ phoneme in ME to exclude continuing
allophonic [æ] in ME in words such as OE æppel ‘apple’, OE glæd ‘glad’,
OE hæt ‘hat’. This is a case where neither spelling, nor rhymes, nor
system-internal evidence can make the reconstruction testable. Still,
positing continuity of allophonic [æ] in ME is appealing in view of the
long-term instability of the ‘main’ allophone of the low short vowel in
the history of the language, as shown in (3).
(3) Long-term perspective on the instability of short /a/:
PrG/WG OE ME EModE PDE
*/a/ /æ/ [a] ~ [æ] [æ] ~ [a] /æ/
*apl- /æp-/ [ap-] ~ [æp-] [æp-] ~ [ap-] /æp-/ ‘apple’
*baco- /bæk/ [bak] ~ [bæk] [bæk] ~ [bak] /bæk/ ‘back’
*þat /θæt/ [θat] ~ [θæt] [ðæt] ~ [ðat] /ðæt/ ‘that’
Seen from this macro-perspective, the fully open pronunciation [a] has
been in competition with the neighbouring higher or backer vowels
throughout the history of the language. The realisations of PDE /æ/
and its most recent fluctuations are quite telling: Gimson (1970: 106)
identifies the allophones of /æ/ for British English as [ε ~ æ ~ æ ~ a
~ ä ~ ɑ)]; one generation later Cruttenden (2008: 112–14) adds an even
higher diphthongoid [e] for ‘refined’ RP, and marks both [æ] and [a]
as standard RP. In New World English the vowel of BACK and THAT
also shows a variety of realisations: while generic /æ/ is widespread,
one finds also a slightly raised [æ], strongly raised [eə ~ eε ~ iə] and
slightly lowered [æv ~ æ] (Thomas 2001: 19–21). In view of OE <æ>
~ <a> ~ <e> dialectal and paradigmatic alternations – for example,
bæc ~ bec(es) ~ bacu(m) ‘back, n.’ – and continuing variability attested

by forms such as ME appel ~ eappel ~ eppel ‘apple’, there is no reason to
assume that there was an across-the-board lowering of OE [æ] to [a]
in ME. The ‘middle’ period is not different from EModE; crucially, a
stable phonemic contrast between [æ] and [a] cannot be reconstructed
for any dialect at any time. The low front vowel is extremely suscep-
tible to contextual, regional and social variability, so for ‘generic’ ME
/a/ it is reasonable to reconstruct realisations ranging from a higher
[æ] to a retracted realisation [ɑ], the latter most likely to occur before
liquid codas, as in stark(e) ‘stark’, warnen ‘warn’, barbour ‘barber’, salven ‘to
annoint’. The relevance of this to a PDE speaker is that the coexistence
of [a] and [æ] as non-contrastive realisations of the same phoneme can
be projected all the way back to OE.
A final short-vowel merger, part of the processes discussed in 8.3.2,
has to do with the effect of a tautosyllabic rhotic on late ME <-er>,
which begins to appear as <ar> in the fourteenth century, starting in
the North (Jordan 1974: 98–9, 234): OE/early ME (ge)beorc ‘bark of a
dog’, herġan ‘harry’, sterr(e) ‘star’, teru ‘tar’ > late ME bark(ing), harye(n),
star(re), tar(re). By the end of ME the lowering had begun to appear also
in the Southern dialects. It has been suggested that the path of change
was via [&] (Dobson 1968: §64, n.2), but positing an intermediate front
value is not necessary, especially in view of the well-attested centralis-
ing effect of /r/ on the preceding vowel. The spread of the lowering
was enhanced by the parallel change in OFr and Anglo-Norman: AN
merveille ~ marvaille ‘marvel’, (h)erber ~ arber ‘arbour’, persil ~ parsel
‘parsley’. The orthographic representation of the change can be erratic;
the word parsley in the various manuscripts of Chaucer’s Cook’s Tale
appears as <parsel, percely, persle, persele, percelly, parselee, persly>. The
[-ar] pronunciation in native words was more stable and advanced, and
was fully established in EModE (Dobson 1968: §64, §§67–9). However,
in many loanwords the Latin spelling with <-er-> was preserved and
a spelling-based pronunciation with [-εr] eventually replaced earlier
[-ar] in loanwords such as OFr and AN <mercant ~ marchand ~ marchant
~ merchand, merchant ~ mercheant> ‘merchant’, but notice the surname
Marchand, or OFr and AN parfit ~ perfit < classical Latin perfectus,
‘perfect’; compare PDE parfait, a recent (1884) borrowing from French.
7.3.2 Long vowels: GAE-GO, OAKE (Somerset) - AIKE (East Riding of

Yorkshire)
The OE set that served as an input to the ME long-vowel set (see
6.5.2 and Figure 6.8) included two high vowels [i
] and [u
], two upper
mid vowels [e
] and [o
], two low vowels [æ
] and [ɒ
], and the high
front rounded [y
], discussed along with its short counterpart in 7.3.1.
Figure 7.6 shows the correspondences between the OE and the ME
long-vowel sets.13
Old English Middle English Examples
i< –y u i u OE mȳs [mys] ME [mis] ‘mice’

e o e o
ε ɔ ¯ dan [l
OE l ¯ dən] ME [lεd-] ‘lead, v.’
æ ɒ a OE āc [ɒk] ME (N, Sc.) [ak] ‘aik/oak’
Figure 7.6 Long vowel changes from late OE to ME
A comparison of the inventories in 7.4 and 7.6 shows a reduction of the

number of short-vowel contrasts through mergers, while the member-
ship of the long-vowel system remains numerically the same as in OE
in spite of the unrounding of [y
] to [i
]. Except for the high front vowel
merger whereby the nuclei of OE mīn [mi
n] ‘mine’ and OE mȳs [my
s]
‘mice’ were no longer contrastive, the high and upper mid long vowels
remain relatively stable: OE mīn [mi
n] > ME [mi
n] ‘mine’, OE nū [nu
]
> ME [nu
] ‘now’, OE dēman [de
mən] > ME [de
m(ən)] ‘deem’, OE
dōm [do
m] > ME [do
m] ‘doom’. Monophthongisation of OE [eə] as in
þēof ‘thief’, ċēosan ‘choose’ (see 6.5.3) also resulted in either [e
] or [o
].
The diphthongised and raised fourteenth-century allophones of these
four
vowels – [i
] ~ [ij]/[ ii], [u
] ~ [uw]/[ υ u], [e
] ~ [e (ə)], [o
] ~
[o (ə)] – are at the core of the changes associated with the initial stages
of the chain shift, known commonly as the Great Vowel Shift, described in
Chapter 8.
The low front [æ
], which had different sources and different reali-
sations in OE (see 6.5.2), appears to have been raised a notch to [ε
] in
most dialects. : The raising, evidently allophonic in its initial stages,
can be seen as a precursor of the more general tendency of the long
monophthongs of ME to shift upwards, ultimately resulting in an
across-the-board upward shift of all non-high long monophthongs (see
8.2.2). In late OE the diphthong [æə] as in strēam ‘stream’ had started
to monophthongise, and so we find both <æ> and <ea> spellings for
[æ
] in early ME texts. Yet another source of ME [ε
] is the lengthening
13
The examples in the third column are only of vowels that changed from OE to ME.
The ME forms are given only in the reconstructed pronunciations that provide the
input to the EModE long-vowel changes.
of OE [ε] in open syllables, as in OE stelan [stεlən] ‘steal’ > ME stel(e)

[stε
l(ə)] (see 7.5.2).
After the end of the thirteenth century, the spelling with <e>, later
<ee>, is used to represent both the inherited [e
] and the new [ε
]
vowel, so that in Chaucer’s manuscripts one can find both [e
] < OE
[e
] or [eə], and [ε
] < OE [æ
] or [æə], spelled <ee>: OE drēam, ME
dreem ‘dream’, OE mētan, ME meete ‘meet’. The identical spellings can
be deceptive: Chaucer, whose rhyming practice is an important test for
late fourteenth-century London/Southern pronunciation, tends to keep
ME [ε
] and [e
] apart, rhyming OE [æ
] and [æə] with themselves, for
example meene ‘mean’: clene ‘clean’, and OE [e
] and [eə] also with them-
selves as in queene ‘queen’: ysene ‘seen’. For someone reading Chaucer it
is important to keep in mind that in spite of the uninformative, identical
spellings, and the coalescence to [i
] in later English, the open and close
front-mid vowels have different histories. :
The two types of mid vowels were kept separate long enough for
them to be represented differently in fifteenth-century spellings: as
noted in 7.2, the lower mid front vowel [ε
] was associated with <ea>
as in mean, clean, meal, meat, while the upper mid vowel [e
] was repre-
sented by <ie> as in the French loanwords brief, piece, and in native thief,
fiend, and by <ee> as in queen, seen, heel, seek. In practice this makes the
second letter in the digraph <ea> a diacritic for openness of the vowel
(see 7.3.2); as we will note again below, the diacritic use of <-a> to mark
openness is mirrored by the pair <oa> for [ɔ
] - <oo> for [o
] for the
back vowels. :
Metaphorically speaking, the raising of late OE [æ
] to [ε
] adds a
new rung to the ladder of vowel heights in the set of long front vowels.
Looking at Figure 7.6 again, we see that the new lower mid vowel in
the front set is matched by a lower mid back vowel [ɔ
]. This inno-
vation results from the raising of the OE low back [ɒ
] to [ɔ
] in the
non-Northern dialects, as in OE āc [ɒ
k] > ME [ɔ
k] ‘oak’. The shift
of [ɒ
] to [ɔ
] started in the latter half of the eleventh century in the
West Midlands, the Berkshire/Oxfordshire area, the South Midlands
and the South-East (Liebl 2006; Stenbrenden 2010: 531). It was a slow,
lexically specific process, not fully developed before the end of the
thirteenth century. Raising of OE [ɒ
] was one of the sources of [ɔ
]; the
other source was lengthening of OE [ɔ] in open syllables, as in OE nosu
[nɔzu] ‘nose’ > ME nose [nɔ
z(ə)] (see 7.5.2). The new long vowel in the
back series is represented initially by <-o-> or <-oCe->, and, after the
middle of the sixteenth century, as <oa> (Scragg 1974: 77) as in oak (OE
āc), boat (OE bāt), coal (OE col(u/a)), toad (OE tāda).
In the North the OE low back [ɒ
] vowel was fronted to [ɑ
] >
[a
], represented in the spelling as <a>, <aCe> and <ai>, where the
<-Ce> and <-i> are the Northern/Scots markers of vowel length (see
also 7.5.3). The dotted arrow in Figure 7.6 indicates the fronting; it
is a regional development recognisable in Scots forms and Northern
place names in PDE, as in both ~ baith, do ~ dae, go ~ gae, home ~ hame,
more ~ mair, toes ~ taes, whole ~ hail. The raising and fronting of the
long low monophthong [a
] possibly to [ε
] in the North, pre-dates the
raising of the low vowel in the South, and it is arguably the first step
in a front-vowel chain shift, identified as the ‘Northern Shift’ (Smith
1996: 99–101), a set of changes comparable to, but not identical with
the (Southern) Vowel Shift. The preservation of the low vowel, its
fronting and raising and later diphthongisation to [ej] was most likely
influenced by the existence of parallel ON forms with [ej] < *[aj].
Examples of the different outcomes of the split of OE [ɒ
] are shown
in (4).
(4) Regional developments of OE [ɒ

] in ME:
OE ON PDE Southern PDE Northern
āc eik oak aik
Oake (Somerset) - Aike (East Yorkshire)
brād breiðr broad braid
Broadway (Somerset) Braithwaite (Cumbria)
stān stein stone stane
Stonehenge (Wiltshire) Stainborough (Yorkshire)
Folkestone (Kent) (Dwarfie) Stane (Orkney)
The addition of an extra level of height contrasts for the front and back
long mid vowels is a systemic innovation in ME. As noted at the begin-
ning of this section, one should also consider the strong probability of
variable purely monophthongal and slightly diphthongal realisation of
inherited long vowels, which is tantamount to initial long vowel shift-
ing. The earliest orthographic evidence for diphthongisation of [i
],
namely <ei/ey> spellings for the vowel in abide, betide, appearing also
in rhyme position, goes back to the first half of the thirteenth century,
especially in the West Midlands. More such spellings are found in the
next century, and there is no doubt that by the early fifteenth century
the change was well under way. The evidence for the diphthongisation
of [u
] is confounded by the introduction from French of <ou> and
<ow> spelling for [u
] (see 7.2), but since the raising of [o
] is attested
from the thirteenth century, and there was no merger with [u
], the
initial steps in the direction of diphthongisation to [uw] can be dated
to approximately the same time. For this vowel too, the shift of OE [u
]
to PDE [əw]/[aυ] is firmly rooted in Middle English. We will return to

these issues in 8.2.2. :
The premise of a fourteenth-century start of diphthongisation of the
high vowels means that Figure 7.6 is quite conservative. It is a represen-
tation of the long-vowel system found in all introductions to Chaucerian
English; however, it conceals the option of initial diphthongisation,
which would bring the vowels closer to their PDE values. A less archais-
ing reconstruction of the system for reading Chaucer aloud would allow
the vowel of OE mīn [mi
n] > ME [mi
n] ‘mine’ as [ij]/[ij]/[əj], and the
vowel of OE nū [nu
] > ME [nu
] ‘now’ as [uw]/[υw]/[əw].
7.4 The diphthongal system of ME

Recall from 6.5.3, also 7.3.2, that the core inherited diphthongs of late
WS [æə] and [eə] merged with the corresponding long vowels [æ
]/
[ε
] and [e
] in early ME. Ignoring the instances of diphthong-internal
prominence shifts resulting in rising diphthongs (OE ċēosan > ME
chesen ~ chose(n) ‘choose’), the mergers of diphthongal and simple long
vowels are shown in Figure 7.7.
OE Example ME Example
[e] scēp s(c)hep ‘sheep’
[e]
[eə] cnēo kne(e) ‘knee’
[æ] mœ̄l me(e)l(e) ‘meal’
[ε]
[æə] strēam stre(a)m(e) ‘stream’
Figure 7.7 Monophthongisation of the OE diphthongs in ME
The removal of the diphthongs [æə] and [eə] from the system does
not mean that ME had no diphthongs. This would be a strange situ-
ation, given the rich diphthongal presence in the vowel systems of
OE and all varieties of PDE. A set of simple long vowels without any
diphthongal realisations would also be peculiar in view of the func-
tional unity of diphthongs and steady-state long vowels throughout
the history of the language. Diphthongal vowels can be reconstructed
for OE both from inheritance and in cases where the nucleus was
followed by the palatal glide [-j] based on an earlier palatal fricative.
Thus Gmc *daœ̄-, OE dæġ ‘day’, with recorded OE spellings <dæig,
deig, daig, deih, dæi, dei, dai>, is interpreted as representing [æj],
and spellings such as <greig, grei> for OE (Angl.) grēġ ‘grey’ are
interpreted as [ej] (see 6.5.3).
The pattern of glide vocalisation and diphthong formation illus-
trated by day and grey is not an isolated change – it is just the change
most clearly discernible in the orthographic records. The palatal glide
[-j] was drawn into any front-vowel nucleus when it was in the same
syllable or in the same stem: early OE *briġdel > late OE brīdel ‘bridle’,
drȳġness ~ drīġness > late OE drinesse ‘dryness’, flȳġan ~ flȳan, p.t. fliġde
‘(cause to) fly’, all with [-ij]. :
The trajectory within a diphthong involving a high front vowel + [j]
is minimal; it is perceptually difficult to keep [ij] apart from the simple
long [i
]. While the merger of [-ij] with [i
] already in OE is beyond
doubt, there is no principled way of deciding whether the dominant
post-merger realisation was [i
] or a slightly diphthongal [ij] – the latter
could easily be the input to the vowel shift that occurred later. The fact
that OE scribes commonly use both and <ig> as representing ‘pure’
[i
] in word-final position as in <bi ~ big ~ bii> for OE be/bi ‘by, prep.’,
<hwy ~ hwig> for hwȳ ‘why’, <sie ~ sy ~ sig> for sȳ ‘be, pres. subj.’
confirms the merger and allows a reconstruction of [ij] as the primary
allophone of the long high front vowel in early ME.
A process parallel to the diphthongisation of front vowels + [j]
involves also the OE voiced velar fricative [γ] (see 4.2.1), the labial
approximant [w] and the different realisations of the voiceless glottal
fricative /h/ (see 5.1.4). The voiced velar fricative [γ] in OE was
always preceded by a back vowel. The further lenition of [γ] to [w]
and subsequent merger under [w] was under way in late OE/early ME,
evidenced by variant spellings such as gnagan ~ gnawen ‘gnaw’, lage ~
lawe ‘law’, boga ~ bowe ‘bow’. When the approximant [-w] appeared in
the coda, increasingly likely at a time of general weakening of the final
unstressed vowels, it was vocalised and the glide became the second
element of a diphthong: OE bo.ga ~ bo.we (with possibly ambisyllabic
intervocalic consonant) > ME bow(e) > bow [bow] ‘bow’, OE būgan, ME
bouen ~ buwe(n) [buwən] ‘bow down’. The result for the vocalisation of
[γ] after back vowels is indistinguishable from the vocalisation of origi-
nal [w], as in OE stōw > ME stou [stow] ‘spot’. Vocalisation of etymo-
logical [-w] occurs also after [æə], [eə] and [i
]: OE dēaw ~ deau > ME
deu [dεw] ‘dew’, OE brēowan > ME breu ~ brew [brew] ‘brew’, OE nīwe >
ME niw(e) ~ new(e) [niw] ‘new’.
It is evident that although these diphthongs are commonly referred to

as ‘new’, they were rooted in the phonology of late OE. They are ‘new’
only in terms of the nature of their second elements: the diphthongs
[æə] and [eə] inherited from WG have a rather short and flat trajectory,
ending in the non-distinctive [-ə], while the [-j] and [-w] endpoints of
the ‘new’ diphthongs give them a longer trajectory and make them per-
ceptually more salient. As we will see in 8.3.2, a return to, or perhaps a
replication of, diphthongal nuclei with [-ə] as their second element, the
so-called centring diphthongs, occurs again in the environment of a coda
/-r/, lenited to [-ə] as in RP beer, dear, there, moor, cure.
Another source of diphthongisation in ME is the presence of the
voiceless glottal fricative [-h] in the coda, originally realised in agree-
ment with the nature of the nucleus: [-ç] after front vowels, as in EME
niht ‘night’, or [-x] after back vowels, as in sohte ‘sought’; the details
were discussed in 5.1.2, 5.1.4. The most common spelling for the frica-
tive in non-initial position was <gh>. Already at the beginning of the
fourteenth century forms such as <mit> for OE miht ‘might’, <brit>
‘bright’, <nit> ‘night’ testify to variable [-ç] ~ Ø in these words. The
coarticulation of front vowel + [ç] results in the epenthesis of a transi-
tional palatal glide [-j-], so that phonetically <miht> ~ <mit> ‘might’
can be reconstructed as [mijçt] ~ [mijt]. Inputs with the mid vowels
[-εç], [-e
ç] result in [-εjç] ~ [-εj] (later [ej]), and [-ejç] ~ [ej] (later
raised, merging with [ij]): OE ehta > ME ehte ~ eihte ‘eight’, OE nēh >
ME neh, neih, nih ‘nigh’.
A parallel process of glide epenthesis occurs between a back vowel +
[-x]; in this case the inserted glide is [w]/[υ]: OE drūgoþ > ME druзt
~ drout(e) [drυxt] ~ [druwt] ‘drought’, OE tahte > ME tahte ~ tauhte
‘taught’, OE bohte > ME bohte ~ bouhte [bɔυht(ə)] ‘bought’. The initial
evidence for the back glide epenthesis before the velar fricative is also
dated to the fourteenth century with the full-blown developments in
evidence during the fifteenth century. :
Figure 7.8 presents an overview of the diphthongal vowels emerging
from glide vocalisation in late ME.
Input OE/EME LME Examples
(a) V (front) + [-j] [-ij-] [-ij-] OE *brīġdel, ME bridel ‘bridle’
[-e:j] [-ej] OE (Angl.) grēġ, ME grei ‘grey’
[-εj] [-εj] ~ [-æj] OE weġ, ME wei ~ wai ‘way’
[-æj] [-aj] ~ [-æj] ~ [-εj] OE dæġ, ME dai ‘day’
Figure 7.8 The diphthongal vowels in late ME

(b) V + [w] [-i:w] [-iw] ~ [-ju] OE nīw(e), ME niwe ~ newe ‘new’
[-eəw] [-ew] OE brēowan, ME breu ~ brew ‘brew’
[-æəw] [-ew] OE hēawan, ME hew(e) ‘hew’
[-ɑw] [-ɒw] OE þawian, ME thaw(e) ‘thaw’
[-ɒ:w] [-ɔw] ~ [-ow] OE cnāwan, ME cnowe(n) ‘know’
[-o:w] [-ow] OE stōw, ME stou ‘spot’
[-υw] [-uw] OE fugol, ME foul ~ fowel ‘fowl’
[-u:w] [-uw] OE būgan, ME bouen ~ buwe(n) ‘bow down’
(c) V (front) + [ç] [iç] [ijç] ~ [ij] OE miht, ME might(e) ~ mit ‘might’
[eç] [ejç] ~ [ij] OE nēh > ME neh, nei(h) ‘nigh’
[εç] [εjç-] ~ [ej] OE ehta, ME ehte ~ eihte ‘eight’
(d) V (back) + [x] [υx] [uwx] ~ [uw] OE druhþ(e), ME droute ‘drought’
[u:x] [uwx] ~ [uw] OE rūh, ME ruhe ~ rowe ‘rough’14
[ɔx] [owx] ~ [ow] OE sohte, ME sohte, soute ‘sought’15
[ɔ:x] [ɔwx] ~ [ow] OE dāh, ME do(u)h ‘dough’
[ɑh] [ɒwx] ~ [ɒw] ON *slahtr, ME slauhter ‘slaughter’
Figure 7.8 continued
Glide vocalisation is a function of syllabification. In OE [j], [γ] and [w]

1415
were consonantal and syllabified as onsets, possibly as ambisyllabic,

but not as pure codas. Thus OE cnāwan ‘to know’ syllabified as [knɑ
.
wən] ~ [knɑ
[w]ən], where the second syllable would be unstable; in
some paradigmatic forms ending in a single deletable vowel in ME, the
14
The short vowel in PDE rough has to be attributed to an early, pre-Long Vowel Shift,
shortening of [uw]/[u
] to [υ] (see Luick 1964: §525; Dobson 1968). Other items in this
subgroup are enough, slough ‘snake skin’ (early ME slohu/slouh), but slough ‘[slaυ] ‘muddy
ground’ (OE sloh), also slew [slu
] ‘body of water’, clough ‘ravine’ (OE *cloh), both [klf]
and [klaυ]. In cough (ME cowhen) and trough (ME trou) the back glide is unstable and
the result is [ɔ].
15
The LME [owx]~[ow] in sought, also brought, bought, thought, fought, daughter must have
been lowered to [ɒw] in late ME-EModE.
syllable division changed to strengthen the [w]/[υ] coda position: 1st

p. sg. pr. tense or imp. [knɔ(
)w.(ə)] > [know]. The parentheses around
the length diacritic and the final vowel indicate variable pronunciations:
vowel length matters only if the word is pronounced as disyllabic. The
original vowel length was neutralised in the process: the nuclear vowels
of ME <dei> from an original short vowel + [j] and ME <grei> from a
long vowel + [j], are both bimoraic. There is no indication in the metri-
cal treatment or the later history to suggest that diphthongs originating
in short vowel + glide were treated differently from those originating
in long vowel + glide. : Moreover, the adjacency of [j] and [w] to
the original non-peripheral short vowels neutralises the height differ-
ence between the original vowels: OE fugol [-υγ-] ‘fowl’, and OE būgan
[-u
γ-] ‘bow down’ merge into [-uw-] in ME.
The great diversity of inputs shown in Figure 7.8 can make one’s
head spin, but if we focus on the outputs in late ME, we will see imme-
diately that the resulting diphthongs are quite straightforward. The sets
in 7.8(a) and in 7.8(c) overlap, and so do the sets in 7.8(b) and 7.8(d).
We can describe the general pattern as vocalisation of [j] and [w] or
epenthesis in the sequence V + [ç/x], resulting in [Vj] or [Vw].
The diphthongs formed through glide vocalisation merge with the
diphthongs found in borrowings from Scandinavian: [aj], [ej], [ɑw], [ou],
[uw] as in ON waive ‘turn aside’, ON hreinn ‘rein(deer)’, ON sweinn
‘swain’, ON þei-r ‘they’, ON vindauga ‘window’, ON þōh ‘though’, ON
*drugna, ME droun(en) ‘drown’.16 The model of diphthongal nuclei in [-j]
and [-w] was augmented by the presence in ME of Romance borrowings
with the same diphthongs: OFr [aj] as in grain, vain, AN [ej] as in obey, prey,
(h)eir, OFr [ɑw] as in faut(e) ‘fault’, assaut ‘assault’, cause ‘cause’, [ow] from
/-l/-vocalisation (see 5.2.5), as in ME sauder, sawder < OFr soud(i)er, saudier
‘soldier’; OFr soud-, saud-, soldure ‘solder’ > PDE [sɒldə(r)] ~ [sɑ
də(r)];
[ew] as in beute ‘beauty’, dew, du ‘due’; and [iw] ~ [ju] as in duc ‘duke’, frut
~ fruit ‘fruit’. All of these were absorbed in the native set of diphthongs.
In addition, two new diphthongs were added from the French: OFr
[oj] and its AN counterpart [uj], as in choys ‘choice’, destruie ~ destroi
‘destroy’. The practice of avoiding the letter in conjunction with
other vertical-stroke letters, including (see 7.2), makes the dif-
ference between [oj] and [uj] hard to trace. The complexities of the
dialectal and lexical distribution of the two diphthongs in ME are not
relevant to their later history; they merge under [oj]. One reason why
the end result was [oj] may be that [oj] was a possible native diphthong,
16
See further Dance (2003: 126–30) for evidence of merger of the ON and ME
diphthongs.
as in OE croġen-denu, Croidene (Domesday Book 1086) ‘saffron-valley’

(Ekwall 1960: 134). Some Dutch loanwords also had [oj]: loytren ‘loiter’,
toye ‘toy’, possibly boy.17 The number of items with [oj] is higher; the
relative sparseness of [uj] items could also have contributed to the loss
of [uj]. As noted above, the <oi> spelling for [uj] loanwords could also
be a factor, especially for ‘learned’ words such as foison < AN fuisun. The
dating of the merger is uncertain, but it must have been under way in
the second half of the fourteenth century. :
Ignoring the chronological depth and the diversity of sources,
Figure 7.9 summarises the late ME diphthongal system and shows
the incipient mergers which will be the input to the EModE changes
involving diphthongal vowels.
Front-gliding diphthongs Back-gliding diphthongs
[-ij-] (bridle) [-iw] ~ [-ju]

[-ju] (new, dew)
[-ej] [-ew]
[εj] ~ [-ej] (way, grey) :

[-æj] ~ [aj] [-ɒw] (thaw)
[oj] [ow] (know)
[oj] (voice, point)
[uj] [uw] (fowl)
Figure 7.9 Late ME diphthongs
The long vowel changes and the substantial enrichment of the diph-
thongal system created a set of bimoraic nuclei quite different from the
late OE system. The interaction between long vowels and diphthongs
will be addressed in Chapter 8. Now we turn to changes which affect the
length of vowels in ME.
7.5 Quantitative changes: why do they matter?

Any description of PDE recognises the existence of two sets of
vowels: short/lax/non-peripheral and long/tense/peripheral (see 2.2.1).
17
The etymology of boy remains uncertain. The two most likely sources are Anglo-
Norman or Dutch (see the entry for boy, n. in the OED and the references there).
Although vowel duration is an unstable predictor for the taxonomy

in PDE, from a historical point of view the distinction between the
two sets is best expressed in terms of quantity. This section covers the
major quantity shifts in Middle English and their role in shaping the
pronunciation and orthography of PDE.
A literate speaker of English can easily correlate the quantitative
properties of the vowel in an unfamiliar word to its spelling. The
made-up words in (5) illustrate this correlation.
(5) Spelling cues for vowel length in PDE:
Spelling Presumed pronunciation
<sklipe/sklype> [klaıp] but not *[klip]
<splone> [sploυn] but not *[splɑn]
<clunzer> [klΛnzər] but not *[claυnzər]
<strecked/streckt> [strεkt] but not *[stri
kt]
<brout> [braυt] but not *[brt]
<bleem> [bli
m] but not *[blim]
How do we know that clipe, splone, brout, bleem have long vowels, but
clunzer and strecked have short vowels? What suggests that the stressed
vowels in restilence, switter, utmer, hosked would all be short, even though
we have not seen those nonce-forms before? Alternatively, for clanes,
dyter, lote, hoiged, we are likely to guess that the stressed vowels belong
to the set of long or diphthongal vowels. This is not a silly spelling-
pronunciation game: we know how to translate spellings into sounds
and sounds into reasonable spellings because we make analogies to
words that we already know, but that does not tell us much beyond the
fact that we have learned what the most frequent patterns of English
spelling are. Our goal in the next two sections will be to take that
primary school-acquired knowledge one step further and explore the
historical rationale behind the choices. :
Let us start by rehearsing the factors involved in the enhancement or
loss of vowel length. Long vowels in unstressed syllables are shortened
and further reduced (see 3.4.3–4, 6.5.4). Recall from 2.3.3 that long or
diphthongal vowels in the peak make a syllable heavy, and English
aligns with the well-attested cross-linguistic preference for heavy syl-
lables to attract stress. The co-occurrence of stress and weight goes back
to OE, where stress was placed on a heavy syllable in approximately 80
per cent of the lexicon. The relation is commutable: the ‘ideal’ stressed
syllable is heavy. We can expect therefore that short vowels in light
stressed syllables (C)V- will be under some pressure to lengthen, heavy
syllables of the type -(C)V(V)C would be stable and the long vowel in
(C)VVCC syllables might get shortened. Syllable weight is determined
both by the length of the vowel and the presence or absence of a syl-
lable coda: codas add to the syllable weight, so that a long or diphthon-
gal vowel plus a tautosyllabic consonant will have some extra weight
that could be shed without damaging the overall distinction between
light and heavy syllables. Moreover, the contribution of different types
of peak vowels and coda consonants, that is, rhyme duration, to the
weight of the syllable can be gradient, although the full effect of the
gradience for all of the quantitative processes discussed here remains
under-studied (see 9.2). :
Yet another important parameter is the syllabic composition of the
entire word: numerous studies have confirmed the observation (Lehiste
1970) that there is a negative correlation between the number of syl-
lables in a word and their duration: syllable durations shrink as their
number increases, for example [] in luck is phonetically shorter in lucky,
luckily. Projected onto vowel length, this means that an original long
vowel can be under pressure to shorten in longer words.
A fourth factor influencing stress is the syntactic function of a word:
function words such as prepositions, articles, conjunctions, pronouns
and auxiliaries are weaker prosodically and of shorter duration than the
major class/content words: nouns, adjectives, adverbs and verbs.
In summary: the main linguistic factors involved in the enhancement,
maintenance or loss of vowel quantity in the history of English are:
• prosodic prominence
• syllable structure
• word length
• syntactic specification.
The list is not exhaustive. Stress, syllable structure and word length
intersect with foot-structure. The changing membership of the short–
long vowels sets in English is also influenced by vowel quality and
the type of coda consonant. Lexical frequency – base alone, or base
vs derivative(s) – can also play a role in quantitative changes. These
numerous factors are in competition. Apart from showing how histori-
cal phonology can enlighten us about English spelling, discovering the
conditions for vowel shortenings and lengthenings is an enterprise of
serious theoretical consequence. The perception and recategorisation
of vowel length bears on every aspect of the sound system: the featural
composition of individual segments, the principles of their distribu-
tion, their interaction on higher linguistic levels and the role of vowel
quantity in verse meter.
The following two sections address the vowel quantity changes
responsible for the main sound-spelling correspondences in PDE.
7.5.1 Shortenings: CREEP-CREPT, MOON-MONDAY, MOUTH-PORTSMOUTH,

LEEK-GARLIC
The PrG system is reconstructed as having a three-way quantitative

contrast: vowels could be short, long and overlong (Ringe 2006: 214). The
distribution of vowel length was unpredictable. Long or short vowels
were found in both stressed and unstressed syllables. By the time of the
split of North-West Germanic into its dialects, the quantity contrast for
vowels became binary: short vs long/diphthongal. As discussed in 3.4.3
and 6.5.4, the reduced stress on syllables following the first root syllable
in pre-OE led to recursive reduction of vowel length in that position. As
a result, long vowels were no longer found in fully unstressed syllables
in OE. This is the first instance of vowel shortening in the history of the
language. It sets up a correlation between lack of stress and vowel short-
ness – the first step in a long process of vowel reduction and loss that
changed the shape of the lexicon in ME (see 7.6). The elimination of
long vowels from fully unstressed syllables was the initial move towards
predictability of vowel length in the English lexicon. :
Two vowel-shortening changes, first occurring in OE, contributed
further to the transparency of vowel length: pre-consonantal shorten-
ing and shortening in some trisyllabic words. The rationale of dealing
with these quantity shifts in the context of ME rather than with the
OE changes in Chapter 6, is that the patterns of shortening in OE are
sporadic and difficult to test. The shortening straddles Old and Middle
English, and even early ME shows a great deal of variability. The results
are stabilised gradually in ME and it is only in later ME that one can
muster sufficient orthographic and metrical evidence to reconstruct the
process as categorical.
7.5.1.1 Pre-consonantal shortening: FEED-FED, WIDE-WIDTH, SHEEP-

SHEPHERD
Beginning in the seventh century, and continuing through Old and
early Middle English, long vowels in stressed syllables were shortened
before three or before two consonants, other than the homorganic clus-
ters such as -ld, -nd which caused lengthening (see 6.4). The process was
probably initiated in the environment of three consonants, or geminates
followed by another consonant, that is, -VVCCC-, thus *brœ̄mblas >
bræmblas ‘brambles’, nœ̄ ddre > næddre ‘adder’. A less restrictive environ-
ment for the shortening is the sequence of a long vowel followed by two
consonants, not necessarily in the same syllable. In addition to syllable
structure, the shortening is sensitive to word-length: the occurrence of
shortening in monosyllables in OE is rare. :
In OE, as in PDE, coda clusters show falling sonority (see 2.3.1). Like
PDE, in the earlier stages of English the most common first consonant
in coda clusters in simplex words was a sonorant or [s], and the second
consonant typically a stop, naturally limiting the subset of monosyllabic
lexical items eligible for shortening. Only about a dozen OE monosyl-
lables ending in -VVCC show a short vowel in PDE and half of them
end in -st, as in OE blœ̄st ‘blast’, brēost ‘breast’, dūst ‘dust’, fȳst ‘fist’, pos-
sibly PDE mist, rust. As we will see with the lengthening data in 7.5.2, the
cluster [-st-] behaves inconsistently – in some words, especially those
with high vowels, coda [-st] triggers shortening, as in OE dūst ‘dust’, fȳst
‘fist’, but in the overwhelming majority of monosyllabic content words
in -st the original vowel length is preserved: Christ, east, least, priest, ghost,
roost. ME borrowings from Anglo-Norman, as well as later borrowings,
also preserve the original vowel length before [-st]: beast, feast, host,
coast. :
The two most stable and productive sub-patterns of pre-consonantal
shortening emerge in two quite specific morphological sets: before the
past tense dental [–d/-t] suffix (over forty verbs), and before the deri-
vational -th or -t suffix (about twenty nouns). These two patterns are
illustrated in (6) and (7).
(6) Vowel shortening in [-d/-t] preterites:
Early OE ME Gloss
m&¯n-an [mæ
nən] menen [mε
n(ən)] ‘intend, mean’
m &
¯ nte [mæ
ntə ] ment(e) [mεnt(ə)] ‘meant’
crēop-an [kreəpən] crepen [kre

p(ən)] ‘creep’
crēpte [kre
ptə] crept(e) [krεpt(ə)] ‘crept’
The pattern in (6) is robust – shortening in the set of weak verbs, such
as mœ̄ nan ‘intend, mean’, also dœ̄lan ‘deal’, lœ̄dan ‘lead’, mētan ‘meet’,
hȳdan ‘hide’, is practically exceptionless, though in some verbs the long
vowel is later restored by analogy with the present stem, as in PDE
heal, deem, need < OE hœ̄lan, dēman, (ge)nēodian. By the beginning of the
fourteenth century the regularity with which the dental suffix of the
past tense causes shortening is shaken – along with healed, deemed, we
get new weak forms without shortening, such as older bereft ~ bereaved,
cleft ~ cleaved, crept-creeped (out). Since only about a quarter of the OE
verbs were strong, that is, forming their past tense by changing the
root vowel, some strong verbs tended to develop ‘weak’ past tense
forms, whereby the root vowel was shortened, as in OE crēopan in (6),
for which had the strong preterite crēap competed with the weak form
crepte. This is especially true of dental-final strong verb stems, where
the back-to-back dentals result in geminate [-tt] or [-dd], simplified

word-finally to a singleton: thus bītan-bāt ~ bitte ‘bite-bit’, slīdan-slād ~
slidde ‘slide-slid’. :
Following the principles of syllable division, the originally disyl-
labic forms such as OE mœ̄nte, crēpte (syllabified as mœ̄n.te, crēp.te) would
have a heavy stressed syllable by virtue of the filled coda: the preferred
matching of syllable weight to stress would not be damaged by shorten-
ing the vowel. In fact, the pattern in (6) shows a phonologically con-
ditioned vowel shortening reinterpreted as a signal of a grammatical
distinction: long vowel in the present, short vowel in the past tense. This
creates the apparent ‘irregularity’ in the paradigms of sleep-slept, meet-
met, lead-led, and so on. Pre-consonantal shortening is also the source
of etymological doublets such as screed-shed, based on OE scrēadian,
past tense (ME) s(c) hredde. The evidence that this pattern of shortening
became ‘morphologised’ is of course strengthened by the addition of
instances of originally strong verbs which developed dental preterites as
alternative forms and the long vowel of the present stem was shortened,
as in attested rid, strid for rode, strode.
All other pre-consonantal shortenings of surviving words occur in
derived words, but not all affixes have the same effect on the stem. The
addition of the suffix -th ~ -t to a VVC stem produces another system-
atic set of shortenings; indeed, all pre-1250 eligible forms undergo the
shortening, a very different picture from that in forms derived by other
affixes. The examples in (7) illustrate the effect of this type of affixation
on the quantity of the input vowel.
(7) Vowel shortening in -th ~ -t derivatives:
Early OE ME Gloss
fūl [fu
l] foul [fu
l] ‘foul’
*fūl + (iþ) >fȳlþ filth [fılθ] ‘filth’
dēop [de
p] dep [de
p] ‘deep’
*dēop + þ > *dēopþ depth(e) [dεpθ] ‘depth’
The early productivity of this shortening gives us pairs such as dear-
dearth, five-fifth, heal-health, steal-stealth, slow-sloth, thief-theft, weal-wealth
with -th, and drive-drift, heave-heft, shrive-shrift, weave-weft with -t. In
this set, as in the preterite forms in (6), shortening of the vowel aligns
the syllable weight of the derived forms with the majority of bimoraic
monosyllabic content words. The shortened derivatives in (7) may be
reinterpreted as independently stored lexical items: if we test PDE
speakers on the association between filth, shrift, sloth, wealth and the
base forms, only speakers with training or interest in etymology would
connect them to foul, shrive, slow, weal. The suffix continued to be mar-
ginally productive in ME: only mild-milth (c. 1300) and the rare dialectal
side-sidth (1855) show shortening (OED), while the majority of more
recent forms preserve the long vowel: coolth (1547), growth (1557), blowth
(1602), gloomth (1753), greenth (1753), zeroth (1896). Note that -th with
ordinal numerals causes shortening in fifth, tenth but not in ninth, nor in
any -eenth forms. :
The results in (6) and (7) prompt the observation that shortening
occurred with regularity in -VVC stems followed by a non-sonorant
dental-initial suffix in Old and early ME. The addition of other
derivational suffixes generally leaves the quantity of the stem intact.
Except for OE wīs [wi
s] ‘wise’ – wīsdōm(e) [wi
sdo
m(ə)] ‘wisdom’,
ME [wızdəm] – there are no other examples of shortening before -dom,
neither before -ful, -hood, -less, -ship, -some, -ster, though many eligible
derivatives are recorded in OE; words like cheerful, priesthood, lifeless,
township, loathsome go back to the pre-Conquest records. :
Compounds are another set of words where the shortening of the
vowel with primary stress is attributed to the sequence -VVC followed
by a consonant-initial syllable as in (8).
(8) Vowel shortening in historical compounds:

OE ME Gloss
gōd [go
d] [go
d] ‘good’
gōd + spel(le) [gɔ(d)spəl] ‘gospel’
OE hwīt [hwi
t] [(h)wi
t] ‘white’
hwīt + Sunnandæg [(h)wıtsυndæj] ‘Whitsunday’
stān [stɒ
n] ‘stone’ [stɔ
n] ‘stone’
stān [stɒ
n] + ford [stanfərd] ‘Stanford’
Other pairs illustrating the same process are bone-bonfire, sheep-shepherd,

house-husband, shire-sheriff, moon-Monday, cheap-Chapman, goat-Gatwick,
down-Dunton, deep-Dep(t)ford, foul-Fulford. No regularity can be estab-
lished in compounding in general, however: chapman but footman,
Eastman; shepherd but cowherd, goatherd, swineherd, also sheepskin; husband
but housewife (1225), later hussy (1530).
In concluding this section: the period when shortening was most
active was late OE–early ME – the results are in evidence in the later
twelfth-century orthographic system of the Ormulum. After that, identity
with the base in derived forms becomes a powerful factor inhibiting,
and even reversing the shortening. The clearest phonological trigger
of pre-consonantal shortening is the presence of a dental obstruent in
the coda.18 In the verbal tense system, where the model of present-past
stem allomorphy is already in evidence with the strong verbs, vowel
shortening before the dental suffix becomes reinterpreted as (one) of
the possible ways of signalling grammatical tense. This is the part of
the grammar where the otherwise fossilised shortening continued as a
marginally productive rule, a case similar to the plural fricative voicing
(roof-roofs ~ rooves) discussed in 4.4. Thus from the very start, describing
the shortening as occurring in extra-heavy stressed syllables, -VVCC,
is insufficient to define the conditions for the loss of vowel length, nor is
‘closed syllable shortening’ a productive phonological rule in PDE. The
probability of shortening in the historically inherited cases depends on
the nature of the coda consonants, and quite possibly the interaction of
all factors named in 7.5, plus the rather elusive, but potentially signifi-
cant factor of density of occurrence – why [-st] clusters cause shorten-
ing in some items and not others, why wisdom but leechdom, why meet-met
but greet-greeted, feed-fed but need-needed? Once again we find ourselves
asking more questions than we have answers to; once again the hope is
that the new electronic corpora will come to the rescue. :
7.5.1.2 Trisyllabic shortening: HOLY-HOLIDAY, SOUTH-SOUTHERN

Trisyllabic shortening means that original long vowels were shortened
in a stressed syllable followed by two unstressed ones. The process
is featured most prominently in accounts of the type of paradig-
matic alternations found in loanword pairs such as divine-divinity,
cone-conical, prime-primitive, nation-national, profound-profundity. To the
extent that such alternations are regular, they are commonly linked
to the nature of the derivational suffix attached to the borrowed root;
compare prime-primary, legal-legalise, frequent-frequency, favour-favourable,
account-accountancy.
How the divine-divinity pattern of shortening the stressed vowel of
the base before some suffixes came to be one of the features of PDE
is a complex story, often used to illustrate and test specific theoretical
paradigms. We will not rehearse the positions and the polemics regard-
ing this aspect of the phonology of PDE; our focus is to trace the roots
of this process in the earlier history of the language prior to the influx of
Romance loanwords in early Modern English.
Since the classic historical quantitative study by Luick (1898), all
phonological accounts of English include trisyllabic shortening as one
of the major prosodic changes in English. Moreover, it is regularly
18
Dentals are also the most frequent environment for the sporadic shortening of vowels
in monosyllables, as in PDE head, dead, death, foot, blood (see further 8.3.1).
integrated into a macro-perspective teleological ‘conspiracy’ aimed at

regularisation of the stressed syllable weight in English (see Lass 1992a:
70–6). The two most frequently cited examples of the change are the
pairs in (9).
(9) Early instances of trisyllabic shortening
OE ME Gloss
hāliġ [hɒ
lij] [hO
li] ‘holy’
hāliġ + dæg [hɒ
lijdæj] [hOli] ‘holiday’
sūþ [su
θ] [su
θ] ‘south’
sūþ + erne [sUðərnə] ‘southern’
However, a solid empirical premise for positing trisyllabic shortening
as an active shortening process at any time in Old or Middle English
is missing. The presumed effects of trisyllabic shortening can often
be attributed to other causes, such as pre-consonantal shortening, as
in Whitsunday in (8), swine - Swinburne, stepfather (OE stēop ‘orphan’),
Tadmarton (Oxfordshire) < OE tād-mere-tūn ‘toad-pool farm’, Shepperton
(Middlesex) < scēap-hierde-tūn ‘shepherd’s farm’. :
Since OE did not have uninflected and underived words of more than
two syllables, the eligible pool of items would be very small anyway.19
Including inflected disyllabic forms in the set of inputs, as is sometimes
done, is problematic. Inflected forms of disyllabic words with a long
stressed vowel, if not syncopated (hēafod – hēafdu ‘head, nom. pl.’) typi-
cally resist shortening: nom. pl. sċōtunga ‘shootings’, īdelu ‘idle’. Another
empirical difficulty arises from the lengthening of short vowels in
disyllabic base forms: OE æcer ‘acre’, beofor ‘beaver’, cradol ‘cradle’, hæsel
‘hazel’. The number of such items is also limited for reasons addressed
in 7.5.3 below, but the meagreness of the shortening data combined
with the lengthenings is sufficient to throw doubt on the assumption
that trisyllabic shortening was indeed a prosodic optimisation strategy
characteristic of the phonology of Old or Middle English. This is not a
rejection of the more general principle (see 7.5) that word-length, or,
more precisely, the way the word is ‘prosodified’ can interact with the
weight of the stressed syllable and/or the length of the vowel in it. All
that is argued here is that it is impossible to generalise the shortening
to the lexicon of pre-Renaissance English: the isolated examples cited
in the literature are insufficient for positing a trisyllabic shortening rule
going back to Old English.
19
The restriction applies to the native vocabulary of OE and excludes early loans such
as basilica ‘basilisk’, baptista ‘baptist’, comēta ‘comet’, cucumer ‘cucumber’, December.
Nevertheless the pattern noted at the beginning of the section –

divine-divinity, cone-conical, humane-humanity, profound-profundity – is
pervasive and productive: vine-vinegar, chaste-chastity, court-courtesy,
compare-comparison, vile-vilify, line-linear, Spain-Spaniard, vain-vanity,
abound-abundance. As argued in Minkova and Stockwell (1998), the loan
vocabulary where the paradigmatic alternations are exclusively located
entered the language piecemeal, so that in many cases the earlier ‘base’
long form and the shorter derivative just kept their original vowel
length. The dates of entry could be centuries apart.
(10) Date of entry and vowel length in Romance loanwords:
Long Short
brief (1330) brevity (1509)
cave (c. 1220) cavity (?1541)
grade (c. 1400) gradual (?1541)
profound (1300) profundity (?1475)
Immediate source could also be a factor: Latin had vowel quantity
contrasts which would have been familiar to the educated population
who were the first, and also the prestigious, users of the loanwords. Pre-
consonantal shortening characterises many of the Latin loans, whether
they are recycled through Anglo-Norman or not, so that pairs such as
redeem-redemption, resume-resumptive, deduce-deductive, borrowed with the
allomorphy already in place, reinforce the model. Old French had no
functional vowel-length distinctions after c. 1400. However, many of
the items borrowed directly from French were interpreted as having a
long stressed vowel at the right edge: degree, delight, profound, while inside
the word the vowels were assigned to the short set: gradual, delicious,
profundity. Ultimately, the sheer bulk of Renaissance borrowings and the
concurrent adoption of stress-sensitive and vowel-shortening suffixes
such as -ity, -ic bring about the establishment of allomorphy labelled tri-
syllabic shortening in PDE. Productive trisyllabic shortening in English
is thus an early Modern English innovative pattern confined to forms
derived with a subset of borrowed Romance suffixes.
Although the idea of historical continuity of trisyllabic shortening
in English cannot be supported on empirical grounds, the process is
an important aspect of the PDE loan phonology, whose interpretation
requires reference to some universal prosodic principles. Eschewing
complex theoretical notions, there is a very general linguistic prin-
ciple which favours the surfacing of a short vowel in a stressed
antepenultimate syllable followed by two unstressed syllables. This
relates to the word-length factor listed in 7.5: long vowels are much
more stable in mono- and disyllabic words than in words where two
or more unstressed syllables need to be ‘packed in’ before the end of

the word. It is also part of a complex interplay of the other quantity-
regulating factors discussed before: degree of stress and syllable
structure.
7.5.1.3 Vowel shortening in unstressed syllables: BUT-ABOUT, LEEK-

GARLIC
Unstressed vowel reduction was already under way in Old English (see
6.5.4). Recall that ‘reduction’ involves both loss of distinctive height
and backness specifications, and loss of length – fully reduced vowels
are short/lax by definition. In OE and ME unstressed syllables are typi-
cally inflectional, or the second syllables in monomorphemic words.
Lack of stress is associated also with morphological and syntactic
specification: verbal and adverbial prefixes were always unstressed; the
prefixes ge-, be-, for- were always unstressed; and prepositions, conjunc-
tions, pronouns and auxiliaries, especially if they were monosyllabic,
would also be weakly stressed. Some notable historical shortenings
attributed to the weak prosodic position of the lexical item are shown
in (11).
(11) Vowel shortening in weakly stressed words:
OE ME Gloss
&¯ fre [æ
vər] [εvər] ‘ever’
būtan [bu
t ə n] [bυt(ən)] ‘but’
(on)būfan [ ə bu
və n] [əbυv(ən)] ‘above’
mōste [mo
st ə ] [m ɔst(ə)] ~ [mυst(ə)] ‘may, must’
tōdæġ [to
dæj] [t ə dæj] ‘today’
ūs [u
s] [υ s/ əs] ‘us’ :
The shortening is untestable for (late) OE. However, had the input long
forms not been shortened in ME prior to the vowel shift, the PDE forms
would be *[i
vər] for ‘ever’, *[əbaυv] for ‘above’, *[tu
dej] for ‘today’
and *[aυs] for ‘us’ – the way that the stressed vowel in the OE disyllabic
adverb onbūtan [əbu
tən] ‘about’ developed.
Vowel shortening can also occur in the unstressed parts of com-
pounds. There is good evidence that already in OE compounding pre-
served primary stress on the initial element, while the second element
would have a lower degree of prominence (see 9.4.1). If the semantic
link between the original components of the compound and the end
product became obscured, that is, if the compound was lexicalised, orig-
inal long vowels were subject to shortening. Lexicalisation accompanied
by vowel reduction is particularly common in the onomastic data, but it
is also found in ordinary nominal compounds. :
(12) Vowel shortening in historical compounds:

OE ME Gloss
dæġes ēage dayesye ‘day’s eye, daisy’
scīrġerēfa scherive ‘shire reeve, sheriff’
gār-lēac garlic(e) ‘spear-leek, garlic’
upp-tūn Upetone ‘higher-town’, Upton
grāf-hām Grafham ‘grove-home’, Graffham (Sussex)
swīn-dūn Swindone ‘pig-down’, Swindon (Staff.)
For more on vowel reduction in unstressed syllables see 7.6. Now we

turn to another set of quantitative changes in the history of English:
lengthenings.
7.5.2 Lengthenings: GAME-GAMMON, GRAZE-GRASS

Throughout the history of English stressed syllables have behaved in
line with the universal tendency towards co-occurrence of stress and
weight. The hierarchy of weight: CV < CVC < CVV < CVVC, where
CV is lightest and CVVC heaviest, is confirmed by the changes we have
already examined. : The lengthening of final vowels in stressed mono-
syllabic words in North-West Germanic, whereby the final vowels of
lexical CV monosyllables became uniformly CVV: Goth. nu, OE nū
‘now’, Goth. swa, OE swā ‘so’, was covered in 3.4.4 This process is the
inverse of the reduction of final vowels in unstressed syllables: having a
long vowel in a stressed monosyllabic word increases the word’s percep-
tual salience. The constraint on final vowels in lexical monosyllables is
a feature of PDE too. It excludes stressed CV monosyllabic words: buy
[baj], low [loυ], see [si
], rue [ru
] are well-formed, but bi *[bı], lo *[lɔ], se
*[sε], ru *[rυ] are not. The prosodic constraint on the minimal word for
content words in English is therefore defined as either (C)VC, as in sit,
it, cup, up, or (C)VV, as in sigh, eye, row, owe.
The lengthening of final vowels in stressed monosyllables is the only
quantitative process in the history of English which has been carried
through systematically. All other processes have the effect of increasing
the predictability of vowel quantity in PDE, but they apply selectively to
subsets of the original long or short vowels, depending on their position in
the word and their quality, more specifically on their height. We saw in 6.4
that the results of pre-cluster lengthening were constrained by both the
nature of the vowel and the nature of the coda, and a word-level -VCCC-
sequence blocked lengthening, thus child but cildren, hound but hundred. We
now turn to another lengthening whose results are also variable depend-
ing on the type of vowel and the prosodic structure of the entire word.
7.5.2.1 ME open-syllable lengthening

The process known as open-syllable lengthening in ME is illustrated
in (13).20
(13) Middle English open-syllable lengthening (non-high vowels)
OE ME Gloss
talu [tɑlə] tal(e) [ta
l(ə)] ‘tale’
nosu [n ɔ z ə ] nos(e) [nɔ
z( ə)] ‘nose’
medu [mεdə] med(e) [mε
d(ə)] ‘mead’
Here are the main descriptive facts: roughly between the end of the
twelfth century and the end of the fourteenth century the stressed
short mid and low vowels in open syllables were lengthened in disyl-
labic words when only one consonant intervened between the first and
the second syllable of the word. The lengthened [ε] merged with [ε
]
< OE [æ
] or [æə] (see 7.3.2), the lengthened [ɔ] merged with [ɔ
], the
reflex of OE [ɒ
] in the non-Northern dialects. The lengthened [ɑ] in
the Southern dialects, or OE [æ], emerged as [a
], which fills the low-
vowel slot in the long-vowel set vacated by the raising of OE [ɒ
] (see
7.3.2). :
The lengthening was very inconsistent with high vowels, which are
inherently shorter in duration. Some frequently cited examples are
given in (14).
(14) Middle English open-syllable lengthening (high vowels):
OE ME Gloss
wicu [wıkə] [we
k(ə)] ‘week’
yfel [ y v əl] [e
vi l] ‘evil’
wudu [w υd ə ] [wo
d( ə)] ‘wood’
High-vowel lengthening is chiefly Northern, and its results in PDE are
very limited, so that we find OE wicu > PDE week, but OE hype > PDE
hip, OE dile ~ dili ~ dyle, PDE dill; OE wudu > PDE wood, but OE cuman
–> PDE come, OE hulu > PDE hull.
The subset in (14) is of considerable interest nevertheless, because
its results show the lengthened short vowels merging not with the high
vowels [i
] and [u
] but with the upper mid vowels [e
] and [o
]. Paired
with the reverse development, the sporadic shortening of [e
] and [o
]
resulting in (lower) high [i] and [υ] (OE wēoc(e) > ME week(e) ~ wik(e)
‘wick’, OE mōnaþ, ME mon(e)th(e) ‘month’, OE flōd > ME fludmarke ‘high-
water mark’ (1291), the qualitative change of the vowels in (14) is a
20
For a complete list of OE lexical items eligible for open-syllable lengthening surviving
in PDE, the results of the lengthening and an account of its causes see Minkova (1982).
strong argument in favour of positing a pan-chronic English distinction

of height for the long and short vowels, supporting the reconstruction
shown in Figure 6.2. :
A precise dating of the lengthening is impossible to establish because
of inconsistent spellings, dialectal differences, possible coarticulatory
effects and lexical diffusion. It is generally accepted that the length-
ening started earlier in the Northern dialects (Luick 1964: §§391–6;
Jordan 1974: 47–8). Thirteenth-century poets rely predominantly on
self-rhymes, but in the late thirteenth-century Northern Cursor Mundi
we find rhymes indicating length such as name: blame (OE nama, OFr
blā(s)me), fare: mare (OE faru ‘journey, fare’, OE māra ‘more’). The south-
ward spread of the change in the first half of the fourteenth century is
not fully documented, but pairings such as blame: name, trone: anon (AN
tron(e) ‘throne’, OE on ān ‘anon’), before: sore (OE beforan ‘before’, OE sār
‘sore’) are found in The King of Tars. The chronological beginning and
end points of the change are thus fuzzy and fall roughly between the end
of the twelfth and the end of the fourteenth century, allowing for lexical
idiosyncrasies. :
The very name of the process, open-syllable lengthening, seems to define
a prerequisite for its operation: in an onset-maximal syllabification
one and only one consonant can intervene between the short stressed
vowel and the second syllable peak: (C)V.CV-. In any weight hierarchy,
input (C)V- syllables are at the bottom of the scale; only open syllables
whose peak is filled by a syllabic sonorant (r$ , l$, m $ , n$ ) could be lighter,
but in English such syllables are, and have always been, stressless.
Cross-linguistically, vowels tend to be longer in open syllables than in
closed syllables, everything else being equal (Maddieson 1985: 213–14).
Given the universal tendency of co-occurrence of weight and stress, the
vowels in the first syllables in, for example, OE no.su [nɔ-] ‘nose’, me.du
[mε-] ‘mead’ would be likely to have longer realisations, where vowel
duration goes hand-in-hand with the extra prominence required for
stress. The lengthening must have started as allophonic. To illustrate:
the first syllable in early ME nose would be [nɔzə] ~ [nɔ*zə] ~ [nɔ*z] ~
[nɔ
zə] ~ [nɔ
z] (where the symbol * is the IPA diacritic for half-long).
Indeed, there are good arguments that the short-long allophony for the
low vowel [ɑ] ~ [ɑ*] ~ [a*] was already in place in late West Saxon
(Hogg 1996). :
Important as it is, the occurrence of a short vowel in an open stressed
syllable is not a sufficient condition for the operation of the lengthen-
ing. The input for the lengthening has to be an original disyllabic base.
Occasionally, the disyllabic base is provided by an inflected form, as OE
stæf (nom. sg.), stæfe (dat. sg.), stafas (nom. pl.) ‘staff, letter, stave’, hwæl
- hwæle(s) ‘whale’. The restriction of the input to disyllables indicates

that the process requires reference to a prosodic domain larger than
the individual syllable – it is a foot-based phenomenon (Minkova 1982,
1985). However, the most interesting aspect of open-syllable length-
ening in Middle English, which makes English different from other
Germanic languages where similar processes are observed, is the strong
correlation between the type of input and the survival of the length-
ening. A database including all OE and ON lexical items eligible for
lengthening of the mid and low vowels, ((C)VCV(C)) words, shows a
striking imbalance of the results in favour of lengthened forms that also
become monosyllables by word-final schwa loss. In 90.5 per cent (220
out of 243) of the OE input forms which become monosyllables due to
schwa loss – the tale, nose, mead type – the long vowel is stabilised and
survives in PDE (see 7.5.3 for the orthographic consequences of this
correlation). On the other hand, only 15.2 per cent (22 out of 126) OE
disyllabic forms which remain disyllabic in ME and PDE – the acorn
(OE æcern), over (OE ofer) type – have lengthened forms. Put differently,
84.8 per cent of the disyllabic input forms with stable second syllables
resist the lengthening.21 This disproportionate preference for length-
ening in conjunction with historical weakening and loss of the second
syllable provides a strong empirical base for describing the process in
compensatory terms: the chances of incrementing the stressed syllable
weight are much enhanced by concomitant reduction of the weight of
the second syllable.
Additional factors at play are the height of the vowel and the nature
of the unstressed syllable. Lengthening of #(C)VCV# words, where the
stressed vowel is low and the unstressed syllable is coda-less [-ə] –
the name, tale type of word – occurs in 95.8 per cent of the cases, while
the mid vowels show greater instability. On the other hand, 97.6 per
cent of the items in which the second syllable is closed by an obstruent –
the type represented by OE ganot, PDE gannet, haddock, radish – retain
the short vowel. Unstressed syllables with potentially variable sonorant
peaks r$ ~ ər (acre), l$ ~ əl (navel), m
$ ~ əm (besom) and n$ ~ ən (open) are sig-
nificantly more likely to undergo lengthening: nearly one-quarter of the
eligible items (39 out of 166) do in fact show lengthened forms in PDE
(see Bermúdez-Otero 1998, who extends the compensatory account to
these cases).
When the number of syllables in the inflectional paradigm varied:
stæf-stæfe, and in cases of syncope which closes the initially open syl-
lable: hefen-hefnes ‘heaven’, note the <-ea-> spelling in PDE, one can
21
See Minkova (1982). The numbers cited here are very slightly modified in Kim (1993).
expect variable results. In most cases only one of the forms survived,
though it is still easy to find PDE pairs showing the early variability of
the lengthening. :
(15) Variable open-syllable lengthening results in mono- and
disyllables:
OE ME Gloss OE ME Gloss
gamen gam(e) ‘game’ gamen gamen ‘gammon’
hræþe rað(e) ‘rathe’ hraþor raþer(e) ‘rather’
læt(e) lat(e) ‘late’ lator later(e) ‘latter’
sceadu shad(e) ‘shade’ scead(u)we shadew(e) ‘shadow’
blæc blak ‘black’ blæce blak(e) Blake

dæl dal/del ‘dell’ dalu (pl.) dale ‘dale’
græs gras ‘grass’ grasian grase ‘graze’
stæf staf ‘staff’ stæfe stav(e)(s) ‘stave’
The addition of consonant-initial derivational suffixes or second ele-
ments of compounds can preserve the short vowel: OE þrote, PDE
‘throat’, but late ME throtel < throt(e) + -le suffix, ‘throttle’ with the
short/shortened vowel; OE stelan PDE ‘steal’, but ME stel + -th suffix
‘stealth’; similarly OE nosu ‘nose’, but nose + þyrl ‘hole’ > ‘nostril’;
early ME dal(e) ‘dale’, but Dalwood (East Devon); compare Staffield,
Cumberland (< staf- ‘boundary mark’) with Staveley Derbyshire (< ME
stave-ley ‘wood where staves were got’) (Ekwall 1960: 435, 440). Absence
of stress can also affect the lengthening; compare auxiliary have vs
behave, are vs care.22
A considerable body of scholarship on ME open-syllable lengthening
reveals further complexities. The interested reader should consult Lass
(1985, 1992a); Liberman (1992b); Kim (1993); Ritt (1994); Fulk (1998);
Bermúdez-Otero (1998); Lahiri and Fikkert (1999); Murray (2000); and
Page (2006, 2012), who addresses all previous scholarship, treats ME
open-syllable lengthening as compensatory, and proposes an account
based on listener-based hypo-correction. Along with the shortenings
discussed in 7.5.1.1 and 7.5.1.2, open-syllable lengthening is a central
anchor for teleologically oriented accounts of the quantitative changes
in the history of English, addressed in all of these studies, a fertile
research area.
22
The lengthened, also so-called ‘strong’ forms of the auxiliaries, are used in verse
until much later. In EModE we find such rhymes as spare : are (Donne), have : grave
(Herbert), were : bear (Shakespeare), examples from Barber (1997: 129), and Dryden
rhymes are : pair.
7.5.3 Orthography and quantitative changes

As noted in the opening of 7.5, the quantitative changes of the stressed
vowels (see 7.5.1, 7.5.2) have made the distribution of vowel length
transparent in many environments today. With some exceptions, for
example before /ʃ/, vowel quantity remains predictably contrastive
in PDE only in #(C)VC#: #(C)VVC# non-function words: bin-bean,
cot-coat, mat-mate, fit-feet, fuss-fuse, SSBE bud-bird. Even in this relatively
limited lexical scope of unpredictable vowel quantity, the spelling of a
word can be a fairly solid clue as to how a vowel should be classified.
The historical association between vowel shortness and the presence
of two consonants after the vowel goes all the way to West Germanic
gemination which occurred only after short vowels, as in OE tellan ‘to
tell’, settan ‘to set’, cynn ‘kin’ (see 3.4.5). In OE (see 7.5.1), long vowels
followed by geminates could be shortened: early OE fœ̄tt ‘fat’, ME fat;
also the past tense forms of weak verbs with a dental geminate: bīten-
bitt(e)-bit ‘bite-bit’, mēten-mett(e)-met ‘meet-met’, slīden-slidd(e)-slid ‘slide-
slid’ (7.5.1.1). Pre-consonantal shortening strengthened the association
between shortness of the vowel and two consonants to the right of it.
Already in OE orthographic doubling of consonants appears, albeit as
a ‘not wholly consistent indicator of a preceding short vowel’ (Hogg
1992a: §2.78, n. 1). A much more systematic recognition of the useful-
ness of double-consonant spellings to mark short vowels goes back to
the last decades of the twelfth century, when the monk Orm became the
first spelling reformer in the history of English. In his autograph versi-
fied homilies covering special calendar days, known as The Ormulum,
he devised a system of marking a short vowel in a closed syllable by
doubling the consonant after it. The first lines of Orm’s Preface illustrate
his practice:
Þiss boc iss nemmnedd Orrmulum
forrþi þatt Orrm itt wrohhte
‘this book is named Ormulum
because Orm created it’ (Preface, ll.1–2)
This system is a good diagnostic of length: the double consonants in þiss
‘this’ nemm.nedd ‘named’, wrohh.te ‘created, wrought’ indicate that the
preceding vowel is short; boc ‘book’ has a long vowel, and the second
syllable of Orr.mu.lum is open, so the length of the vowel is ambiguous.
Although there is no direct continuation between the Ormian system
and PDE spelling, the phonological changes associated with pre-
consonantal vowel shortening feed into an important orthographic con-
vention in PDE: within the paradigm of a single word, the orthographic
doubling of consonants is expected only after short vowels, as in

PDE <spite-spiting> but <spit-spitting>, <coat-coating> but <cut-
cutting>, <brave-braved> but <rev-revved>. In monosyllables the
letters <f, l, s, z> are usually doubled after short vowels.
Orthographic geminates can also be associated with specific proper-
ties: while <c> in PDE can be both [s] and [k] (city-college), double-c
<-cc-> never stands for [s]; it can only represent [k] (accord, hiccup), [ks]
(success, vaccine), or [tʃ] in Italian loans (capriccio, fettuccine). Double-s,
<ss>, shows restrictions on voicing: a single <s> can represent both
[s] and [z]: (mouse-rise), while <-ss> is associated with voicelessness, a
pattern mirroring the behaviour of OE fricatives (see 4.4), thus lesser, bliss,
grass. This pattern carries over into the sound-spelling correspondences
of loanwords in which [s] is palatalised to [ʃ] as in mission, session, making
the spellings of Aussie, scissors, dessert, hussar, Missouri, possess – this is an
exhaustive list of the items where <ss> stands for [z] – distinct oddities.
Yet another orthographic convention which mirrors phonological
change is the use of the <-e> to mark vowel length, the so-called ‘dis-
continuous’ V-e (Lass 1992a: 38), a direct link between open-syllable
lengthening and PDE spelling conventions. Apart from affecting vowel
quantity and therefore subsequent quality through the long vowel shift
(see 8.2.2), as in game vs gammon, Blake vs black, open-syllable length-
ening paved the way for the ‘silent -e’ convention of English spelling.
The logic of marking long vowels followed by a single consonant with
a final -e is evident: if /-VCə/, orthographically <-VCe>, produces
/-VVCØ/, the letter sequence <-VCe> can be adopted for the rep-
resentation of long vowels; <-Ce> becomes a letter sequence associ-
ated both with historically lengthened vowels and any long vowel, of
whatever origin.
A note on the chronological ordering of the <-Ce> spellings: open-
syllable lengthening occurred earlier in the Northern dialects (see
7.5.2.1); not surprisingly, the special diacritical use of <-Ce> is also
earlier in the North, so that the base forms of words such as OE wīf
‘wife’, OE stān ‘stone’, OE œ̄r ‘ere’, OE þœ̄ r ‘there’ start appearing with
<-e> (Smith 1994: 436). The practice of adding ‘inorganic’ <-e>’s
spread southwards in late ME, and it is part of the EModE spelling
conventions. In his widely used and influential Elementarie (1582), the
spelling reformer Mulcaster advocated the use of <-e> as a marker of
length. Today<-Ce> is a fairly solid indication of length, hence the
intuitions that written <sklipe/sklype>, <spone>, <dake>, <bume>
correspond to [sklaip], [spoυn], [deik], [bju
m].23 Combined with the
23
One notable exception is the <-ve> sequence, which follows both long (drive, save)
two-vowel spellings (see 7.2), the representation of etymological long

vowels in PDE is unproblematic. The qualification ‘etymological’ is
needed to exclude cases of shortened long vowels not picked up by the
spelling, as in read, pr. t. - read past t., breath-breadth, mouth-Monmouth.
Conversely, the expectation for a <(C)VC(C)> sequence is that the
vowel in such a word is short: pit, pet, pat, pot, put; this makes pronuncia-
tions of orthographic troth, Job with [oυ] striking and unexpected. The
loanword pattern of trisyllabic shortening (divine-divinity, cone-conical)
is also predictive in terms of spelling: <sklipical, sponity> are more
likely to have short vowels. Further discrepancies are addressed in the
next chapter, but barring those, the historical phonological information
covered here provides some justification for vowel quality marking, and
good justification for vowel quantity marking in PDE spelling.
7.6 Reduction and loss of unstressed vowels in ME

Typically, in languages distinguishing stressed and unstressed syllables,
the inventory of the vowels that can appear in unstressed syllables is
much smaller than that of the vowels in stressed syllables. The pho-
netic basis for the reduced vowel inventory in unstressed position is
well studied: vowels in such positions are very short; this makes the
maintenance of F1 and F2 difficult and characteristically results in the
neutralisation of contrasts. In PDE the inventory of vowels occurring
in stressless syllables is severely constrained: only [ə, i,] come close to
being contrastive in that position. All other realisations of the peaks in
unstressed syllables alternate with [ə]. How far back does this situation
go? :
The elimination of contrast is a function of lack of stress. In early
Germanic the position of the stress in the word was fixed on the first
syllable of the root (see 3.4.3), but unstressed syllables could still have
long or short vowels. During the next chronological stage, early OE,
the vowels in unaccented syllables underwent progressive reduction
accompanied by loss of vowel length distinctions. In 6.5.4 we listed the
unstressed vowels found in early OE, repeated here as (16).
(16) The unstressed vowels of early OE (pre-c. 700):
Front Example Back Example
/i/ cyning ‘king’ /u/ duru ‘door’
/æ/ endæþ ‘ends’ /ɑ/ nefa ‘nephew’
and short vowels (live, breve, above), though recently the tendency is to dispense with
the <e-> after short vowels: rev (1851), derv (1948), marv (1964), improv (1979), chav
(1998).
This inventory is subject to further reduction. Orthographic inter-

changeability of the unstressed vowels in late OE suggests that the
spoken realisation of the unstressed vowels converged to a neutral
unstressed vowel, some kind of schwa [ə]. The most common repre-
sentation of the unstressed vowels was <e>, which was particularly
common in the spelling of inflections. In Ælfric’s Lives of Saints (c. 1000),
for example, one finds <-e> for <-a>, <-es> for <-as>, <-en> for
<-an, -on> and inverted spellings of <-a> for <-e>. Word-medially,
vowel reduction is attested throughout OE in the widespread syncope
and epenthesis of word-internal <-e->, especially when the etymo-
logical unstressed vowel is next to a sonorant, thus <myc(e)le> ‘much’,
<heof(e)nes> ‘of heaven’, <feaxnæd(e)l> ‘hairpin’. In the next two
hundred years <-e> was established as the default spelling for post-
tonic final vowels in monomorphemic words too: <bane> ‘killer’ (OE
<bana>), <deme> ‘judge’ (OE <dema>), <nose> (OE <nosu>).
Reduction and loss of vowels in final unstressed syllables proceeded
unevenly in ME. Among the system-internal factors instrumental in
the rate of reduction and ultimately apocope, are the nature of the
unstressed syllable: part of the stem, inflectional or derivational; the
syntactic specification of the word: content or function word; the pho-
notactic interactions between the unstressed vowel and adjacent seg-
ments, both within the same word and across a word boundary in cases
of cliticisation.
Another important aspect of the process is the strong presence
of bilingual English–Old Norse speakers in the transitional period
between OE and ME. Language-contact situations typically favour
simplification of the morphological system: the use of two Germanic
languages side by side would have created communication pressures
to ignore the language-specific inflections and concentrate on the
semantics of the stem. Although not actually testable, this language-
contact factor must be recognised in any attempt to track the course of
inflectional reduction and loss in ME. :
In ME the grammatical distinctions of case, gender and in some
instances number for nouns and adjectives, were no longer signalled by
different inflections; homophonous, or nearly homophonous inflections
were used for more than one function. Abstracting from a much more
complex system, (17) illustrates the treatment of the ‘ideal’ Classical OE
inflections, shown in their orthographic forms, by the end of ME.
(17) Homophony of OE inflections in ME:

Expected (Classical OE) Late ME
<-e> for dat. sg. (masc., fem., neuter) <-e>[-(ə)]
<-a> for gen. pl. (masc., fem., neuter) <-e>[-(ə)]
<-u> for nom. acc. sg. (fem., neuter) <-e>[-(ə)]
<-es> for gen. sg. (masc., neuter) <-es>[-(ə)s]
<-as> for nom. acc. pl. (masc.) <-es>[-(ə)s]
<-an> for weak nominal declensions <-en>[-(ə)n]
<-um> for dat. pl. (all nouns) <-en>[-(ə)n]
<-on> for pret. ind. pl. <-en>[-(ə)n]
<-aþ> for pres. indic. imp. pl. <-eth>[-(ə)θ]
<-od> for past part. (weak verb) <-ed> [-(ə)d]
A look at the second column in (17) makes it obvious that the distinc-
tive features of the vowels in the OE inflections have been neutralised.
The set of inflections is reduced to [-(ə)], [-(ə)s], [-ən], [-(ə)θ], [-(ə)d].
Along with the loss of phonological distinctiveness, the grammatical
distinctiveness of the inherited inflections was in jeopardy.
The presence or absence of a coda in the unstressed syllable is impor-
tant. A bare inflectional schwa is least stable: the <-e>, <a>, in
the first three rows in (17) were joined by <-e> in the verb paradigms,
marking pres. ind. sg., 1st pers., or subj. pres. sg, all persons, or subj.
pret. sg, all persons, or 1st and 3rd pers. sg. pret. indic. of weak verbs,
or 2nd pers. pret. ind. of strong verbs. The situation was unsustain-
able: reliance on [-ə] as a grammatical marker was no longer a viable
option; phonological and morphological change reinforced each other,
triggering further changes in the grammatical system: increased use of
prepositions and word-order changes. In late ME, the functions of the
final schwa in native lexical items were considerably reduced: if not in
elision environment (see below) or in trisyllabic words, <-e> continued
to be written and possibly variably realised, in the dat. sg. of nouns, the
plural of adjectives and 1st pers. sg. of verbs, and as adverbial markers.24
An independent factor, which serves to undermine further the shaky
status of final unstressed vowels, is the variable realisation of [-ə] in
hiatus with another vowel (earlier) or before an initial [h-] (later): the
<-e> is occasionally dropped/elided already in OE, thus <wene ic>
‘I hope’ ~ <wen ic>, <sæġde ic> ‘I said’ ~ sæġd ic (Luick 1964: §452).
24
One regional detail concerning the orthographic representation of unstressed vowels
is the use of or <y> in final (closed) unstressed syllables in ME. John Barbour’s
The Bruce, a Northern poem composed in 1375 but copied later (1487 and 1489), has
hundis ‘dogs’, lordis ‘lords’, fayis ‘foes’, askyt ‘asked’, lufit ‘loved’, nakit ‘naked’, evir ‘ever’,
mekill ‘much’, othir ‘other’, and so on.
Elision of an unstressed vowel before another vowel in speech is a

well-attested cross-linguistic phenomenon. In principle, the evidence
for elision can be orthographic or based on verse. OE verse tends to be
rather conservative and it is not a good testing ground for elision (see
10.2.1), but the earliest ME texts, both prose (The Peterborough Chronicle)
and syllable-counting verse (The Ormulum, The Owl and the Nightingale),
provide ample evidence for the instability of [-ə] in elision environ-
ments. In all of ME verse, elision of -e before another vowel or a weak
[h-] is the ‘rule’, as the following examples show. :
(18) Elision of -e before another vowel or a weak [h-] in ME verse:
Ic em nu alder þene ic wes, a wintre ent a lare Poema Morale 1
Þe wurrdenn swiþe offdredde Ormulum 3343
vor rite niþe & for fule onde O&N 1096
To the clepe I, thow goddesse of torment Troilus I, 8
That Love hem brynge in hevene to solas Troilus I, 31
The syntactic specification of a word is crucially related to its prosodic
prominence; function words and pronouns are weakly stressed and the
instability of the final vowels in them is expected. In Chaucer <-e>’s in
such words are ‘silent’ even before consonants:
(19) Apocope of <-e> in words of low prosodic prominence:
That by his fortune hath hire faire ywonne KT 2659
Upon thise steedes, that weren grete and white KT 2892
The overall syllabic structure of a word is also of consequence. Final
schwa loss occurs in trisyllabic words. The earliest evidence for that
loss is orthographic: in the Peterborough Chronicle (1132–54), weak pret-
erites occur regularly without <-e>: macod he ‘made he’, henged up,
‘hanged/hung up’, bebyred him ‘buried them’. The metrical evidence for
loss in words such as goddess(e), maner(e), answer(e), wepyng(e), almes(se),
countes(se), hostes(se), lover(e), and so on is also solid.
(20) Apocope of <-e> in trisyllabic words:
To the clepe I, thow goddesse of tormént Troilus I, 8
Lest it were wist on any manere syde Troilus I, 321
And of som goodly answere yow purcháce Troilus II, 1125
He somwhat is fro wepynge now withdrawe Troilus IV, 886
The loss of schwa in word-final position, both inflectional and stem-
final, was more advanced in the Northern dialect areas, where by the
middle of the fourteenth century its realisation would have been an
archaism. In the Southern dialects the apocope must have occurred
somewhat later. Chaucer, whose pronunciation as reconstructed from
verse shows other conservative features, often relies on final schwa for
metrical purposes. By the beginning of the fifteenth century consist-
ent pronunciation of schwa in word-final position was becoming a
marker of ‘foreignness’, as it is today. For a fifteenth-century speaker
of English, OE loans such as Noah, manna, Messiah, charta, podagra, Ursa
and later loans such as aroma (1220), (Ave) Maria (1225), hyena (1340),
contra (1362), alpha (1382), chimera (1382), cholera (1382), asthma (1398),
aura (1398), mania (1398), omega (1398), delta (1400), kappa (1400), uvula
(1400), santa (1450) would have been recognisable as non-native by the
presence of [-ə]. In PDE this can still be the case for rhotic speakers,
for whom the unstressed syllables of fauna (1771), koala (1808), retsina
(1920), bazooka (1935) carry the mark of foreignness.25
Final schwa loss proceeded gradually across the lexicon, occur-
ring first in the environments shown in (18)–(20). It was probably the
norm in the spoken language in all dialects by c. 1450. With the excep-
tion of [-i] in the affix <-y> (> OE -iġ), all uncovered vowels in final
unstressed syllables in native words were dropped.
Parallel to schwa apocope in open final syllables was the reduction
and eventual syncope of schwa in the inflectional suffixes: <-en, -es,
-eth, -ed>. The process was blocked in stems whose final consonants
would create phonotactic incompatibilities: [-s, -z, -ʃ, -, -tʃ, -d] for
the <-es> suffix and [t, d] for the <-ed> suffix; this gives rise to the
PDE patterns rates vs aces, faked vs lauded – recall the discussion of the
allomorphy of the morpheme /-d/, spelled <-ed> in 4.4.
Unlike coda-less final unstressed syllables, where all final vowels
were subject to pre-1400 apocope, including OFr and AN words (cause,
cure, noise, rule, obey), unstressed vowels in closed syllables were synco-
pated only in inflectional endings, not in syllables which are part of the
stem: so lettuce, foetus, common, orchid, mattress, tepid are disyllabic. The
spread of inflectional syncope is uneven across the various word classes,
and a full coverage of the process in ME is still outstanding. In PDE
the verb and noun inflections, if surviving, show phonotactically driven
allomorphy, while superlative -(e)st, which used to be syncopated, is
now restored to [-əst/-ist].
With respect to the weak verbal preterite <-ed>, there is ME verse
evidence that the /-əd/ realisation is least favoured in vowel-final
stems, thus di.en ‘die’, p.t. dey.d, plei.en ‘play’, p.t. pley.d. This is a predict-
able consequence of avoidance of hiatus – an optimal syllable will have
25
For non-rhotic speakers final schwa is associated with the spelling <-er> (see 5.2.4),
so that the author of this book has been addressed in writing by SSBE speakers as both
<Donker> and as <Minkover>.
a filled onset. The next most frequent attestation of /-əd/ syncope

occurs in -VR stems (callen, dwellen, flouren) – a -VRd coda would be
well tolerated in terms of the sonority profile of the cluster since ideally
sonority rises in onsets and falls in codas. Most resistant to syncope are
verb stems ending in consonant clusters (thanken, busken, trompen) and
stops (clepen, knokken). This type of loss was very gradual: syncope can be
traced in Northern texts already in the thirteenth century, yet syllabic
inflections in environments that would support syncope can be found
until the eighteenth century. :
The positioning of a lexical item within a larger prosodic domain
should also be considered. In the case of final -e, monosyllabic adjec-
tives preceding initially stressed nouns have been shown to hang on
to the [-ə] much longer than disyllabic adjectives in the same frame,
as in a clene maydyn, a riche feste vs a wrongful thing.26 The preservation of
the extra syllable in that frame is a matter of prosodic well-formedness:
apocope would create a stress-clash, and at least for some time towards
the end of ME, this was a factor maintaining the otherwise obsolete
inflectional syllable. Similarly in closed unstressed syllables, there is
evidence that the <-ed> of the past participle remained syllabic longer
than in the preterite, most likely due to its frequent adjectival use where
a syllabic suffix <-ed> acts as a prosodic buffer between an initially
stressed noun and the preceding adjective. In early Modern English syl-
labic <-ed> was used in adjectival participles which have a non-syllabic
suffix in PDE.
(21) Prosodic preservation of syllabic <-ed> in adjectival participles:27
With fixed Anchor in his skaly rind 206
Invests the Sea, and wished Morn delayes 208
And leave a singed bottom all involv’d 236
Of that inflamed Sea he stood, and call’d 300
Thus drooping, or with linked Thunderbolts 328
Indeed, the pattern persists: aged, beloved, blessed, crabbed, crooked, cursed,
learned have different pronunciations for the preterite and the adjectival
participle in PDE.
Unstressed vowels appear also in prefixes. By far the most frequent
fully unstressed OE prefix was ġe- [jə-], used both in the verbal system
and as a nominal and adjectival prefix. Its ME continuation is <i->, prob-
ably [i-/ə], increasingly replaced by zero, though deliberate archaising
continues in EModE, especially with Spenser: yclad, ytold, yglanced. OE
26
Minkova (1991: 171–91).
27
All examples are from Book One of Milton’s Paradise Lost.
ġenoh > ME inough ‘enough’ is an exception to the general loss of the

prefix, and so are some a- reflexes of OE ġe- in alike, along, aware. OE
and-/on-/a- survives as [ə-] in answer, along < OE and-swaru, and-lang.
Here belong also adjectives from prepositional on + a nominal base: abed,
afield, ashore, aglow, athrob, asleep, aghast, afloat, aslant, and so on. In this
connection it is interesting to observe the phenomenon of a-prefixing
in PDE, an archaic feature surviving in East Anglia, Welsh English,
South-East of England, South-West of England, Southeastern AmE,
Ozark English, Appalachian English, Earlier AAVE, Newfoundland
English and Cameroon English (Kortmann and Szmrecsanyi 2004:
1151–2). In those varieties there is a tendency to retain a schwa initial
syllable, spelled <a->, the historically reduced reflex of the preposi-
tion on before -ing forms (I was a-minding my business). The occurrence
of a- is a morphological relic, but its realisation is tightly constrained
by phonological conditions: it can only be used before an initially
stressed -ing form, making *I was a-relying on them ungrammatical and the
schwa cannot be used before a vowel-initial verb, excluding *a-opening,
*a-eating (Wolfram 1976; Wolfram and Schilling-Estes 2006: 77, 334).
This pattern provides a good parallel to the case of final -e preservation
in specific prosodic frames in late ME discussed above (a clene maydyn),
and to the early schwa dropping in hiatus. Many more instances of
interplay between phonological reduction, rhythmic alternation and
morphological form in the history of English are discussed in Schlüter
(2005).
8 Vowel quality and quantity in
Early Modern English and later Vowel quality and quantity in EMode and later
MOTH-MOTHER, DULL-BULL, LOST-POST, FEAR-BEAR, HERE-THERE,

MOOD-STOOD-BLOOD, DEAD-BEAD
8.1 Information glut. New sources of phonological reconstruction

The closer we get to PDE, the more daunting the task of describing any
aspect of the history of the language becomes. All of the surviving OE
manuscripts have been digitised and they are accessible in searchable
form from the OEC. For ME, the ratio of digitised to surviving materi-
als is in favour of the latter, but the very extensive samples included in
the MED, LAEME and LALME give us some confidence that the gen-
eralisations we come up with have some basis in the historical reality.
Tracing the development of the language after the sixteenth century
is both easier and more challenging in view of the exponential growth
of available information. In addition to the analysis of written sources,
spelling deviations and rhyming practice, the student of EModE pho-
nology is both assisted and hampered by ‘metalinguistic’ evidence:
numerous contemporary commentaries and manuals on spelling and
pronunciation, homophone lists, puns, informal private letters, court
transcripts and literary references, all of which can be both illuminating
and contradictory.
The seeds of a fixed spelling system are recognisable already in late
ME, and during the fifteenth century documents written in London
show clear signs of orthographic regularisation. Regularisation brings
about a separation of spelling and pronunciation. In the early sixteenth
century the gap between the visual and the phonetic form of sounds
widened. Printing made books available to a broader readership; the
ensuing increase in literacy prompted the necessity of explaining the
connections between letter and sound, and resulted in many attempts
to describe the sounds of English in relation to spelling both for native
and foreign learners. Prominent grammarians and literary figures, com-
monly referred to as orthoepists (Gk ortho - ‘correct’ + -epi < epos ‘word’),
strove to create a consistent spelling system. Their descriptions are
a major source for the study of the sound structure of EModE. The
234
VOWEL QUALIT Y AND QUANTIT Y IN EMODE AND L ATER 235
orthoepists were aware of the disjunction between sound and spell-

ing and they were troubled by the inconsistencies of the orthography;
writers on matters of pronunciation were also strongly biased in favour
of specific socially prestigious pronunciations, which they endorsed. As
Richard Mulcaster writes in his famous Elementarie (1582):
it is pretended, that the writing thereof [of English] is meruellous vncer-
tain, and scant to be recouered from extreme confusion, without som
change of as great extremitie. I mean therefor so to deall in it, as I maie
wipe awaie that opinion of either vncertaintie for confusion, or impos-
sibilitie for direction, that both the naturall English maie haue wherein to
test, & the desirous stranger maie haue whereby to learn.
. . . letters can expresse sounds withall their ioynts & properties, no
fuller then the pencill can the form & lineaments of the face, whose
praise is not life but likenesse: as the letters yeld not alwaie the same,
which sound exactlie requireth, but allwaie the nearest, wherwith custom is
content. (Mulcaster [1582] 1925: 35, 99)
Mulcaster himself had little to say about pronunciation, but the General
Table in his Elementarie lists 7,000 of the most common words in EModE
and sets the stage for further attempts at codification. The information
that can be extracted from contemporary observations is invaluable; it
is also conflicting and controversial. None of the orthoepists had the
benefit of the kind of training and knowledge that only became avail-
able after the introduction of instrumental phonetics. Many of the early
commentators confused sounds with letters and were inclined to con-
flate pronunciation and spelling. Their philological training was in the
Classics, they tended to be pedantic and conservative, and they often
tried to adjust their descriptions to fit the norms of Latin pronuncia-
tion. Not surprisingly, codification of spelling and pronunciation and
the next stage, prescription, take until the second half of the eighteenth
century. :
Another difficulty in reconstructing EModE comes from the imbal-
ance of information on the variety of English spoken by educated
upper-class Londoners and the varieties spoken in other parts of the
country. It was the former that became codified and prescribed; the
latter is traceable mostly in correspondence, local records, and poetry
in a local dialect, though regional pronunciations can be ‘borrowed’ by
the standard, as has happened with uncouth, bury, break, great. A com-
prehensive coverage of regional forms and the mechanism of dialectal
transfers is beyond the scope of this book. Constructing a coherent nar-
rative requires that much of the available material has to be kept outside
the narrow focus of the emergence of a ‘national’ language. So, bearing
in mind that an all-encompassing coverage of the available records is

unrealistic within the limits of a book chapter, or even of a whole book,
we will deal with the individual changes selectively, focusing on some
more salient features of PDE that can be illuminated historically.1
8.2 Qualitative changes

The phonology of Old and Middle English provides numerous examples
of the interaction between vowel quality and quantity. Open-syllable
lengthening of OE [i] and [υ] involved lowering of the vowel: OE
wicu [wıkə] > ME [we
k(ə)] ‘week’, OE wudu [wυdə] > ME [wo
d(ə)]
‘wood’. The reverse process, shortening of [e
] and [o
] accompanied by
raising to [i] and [υ] – OE wēoc(e) > ME week(e) ~ wik(e) ‘wick’, OE mōste
> ME moste ~ muste ‘must’ – also shows the possibility of simultaneous
change of quantity and quality of the vowel. This is to be expected in
view of the intrinsic durational difference between high and low vowels,
but sometimes the variability remains allophonic, and sometimes, as in
the cases cited above, the length difference triggers qualitative recat-
egorisation. In EModE the realisation of height and backness continue
to show the interaction between vowel quality and vowel quantity.
8.2.1 Short vowels: QUASH-QUACK, WASH-WAX, GOD-EGAD, PUTT-PUT

Section 7.3.1 covered the development of the OE short vowels up
to, roughly, late ME. Figure 8.1 shows the necessarily idealised and
homogenised five short vowels of late ME and their corresponding PDE
reflexes.
Now we turn to the individual histories, starting with the front
vowels. The high front non-peripheral /i/ continued unchanged. It
is the reflex of OE [i] as in OE fisc ‘fish’; OE [y] as in OE cyning ‘king’,
OE bysiġ ‘busy’; the raising of <-e(ng)> to <-i(ng)> as in England (see
7.3.1). EModE [i] can also be the result of ME shortening of either [i
],
as in fifteen, wisdom, or of [e
], as in sheriff, garlic (see 7.5.1). Saying that
the /i/ ‘continued unchanged’ is of course a statement only about the
stability of the phonemic distinctiveness of the /i/-type vowel in PDE.
Like every other phoneme, /i/ has different allophonic realisations
depending on the phonetic and prosodic environment; for example, the
typical duration of the vowel of KID is twice as long as that of the vowel
1
A survey of the sources of study in EModE is found in Nevalainen (2006: ch. 2).
For the history of standardisation see Nevalainen (2006: ch. 3) and Nevalainen and
Tieken-Boon van Ostade (2006).
Late ME PDE Examples
I I fish, king, England, busy
E E bed, kept, tread, bury
æ: ~ a happen, sad, man, land
a~æ Á ~ O(:) wand, waft (SSBE), quarter, quartz

a: ~ A: SSBE after, last, bath, car, half, dance
æ: ~ e@ AmE after, last, bath, dance, glad
O~A~Á hop, rotten, jolly, dollar
O
O: ~ Á(:) soft, lost, cloth, port
ö~@ hunt, love, us, rush, flood
U pull, bush, full, foot, hook
Figure 8.1 The short stressed vowels from Late Middle to Modern English
of KIT (Cruttenden 2008: 96); unstressed [i] is shorter than stressed [i].
Recall also the effect of the nasal coda, the PIN-PEN merger observed
in AmE (see 7.3.1). Then, there are regional and social differences not
linked to phonotactics. To sample just a few: in different parts of the
English-speaking world today the allophonic realisations of [i] can
show a tendency towards centralisation and lowering. In North America
(Thomas 2001: 16), a centralised [-i ] is reported for the North and in
old-fashioned Southern speech. Centralisation to [-i ] is a twentieth-
century NZE innovation which is on the increase (Trudgill et al. 1998).
Lowering of [i] in the direction of [ε] is part of the Northern Cities
Shift (Labov 1994); it also occurs in California and in Canada. Scottish
English also has realisations of /i/ as a lower or central vowel in the
direction of [ε)] ~ [], with the possibility in some varieties that /i/
and // are not distinct (Wells 1982, II: 404). Fronting to [i] and even
a glide insertion [iə] is characteristic of the Southeastern states, AAVE
and Caribbean English. Fronted [i] is found also in AusE and SAE.
Here belongs also the phenomenon of ‘HAPPy-TENSING’ (Wells 1982, II:
294–319), which refers to the realisation of the second vowel of happy as
a more peripheral vowel, a compromise between [i] and [i
]. :
The other short vowel that has remained stable is [ε], the continuation
of OE [ε]: OE bedd ‘bed’, OE helpan ‘help’. Other sources are the short-
ened OE [e
] as in OE cēpte ‘kept’; OE [eə] as in OE stēopcild ‘stepchild’;
OE [æ
] as in OE flœ̄sc ‘flesh’; OE [æə] as in OE mœ̄ nte ‘meant’. Among
the survivors of the ME Southeastern regional forms of OE [y] (see
7.3.1) are bury, merry, left. AN and OFr <e> also joins the set: accept, debt,
medal, tenor, sever, except that when followed by a nasal, [ε̃] is adopted
either as [ε], as in defend, gentle, tense, or as [æ], as in example, pansy; this
is parallel to the historical variation of [æn] ~ [εn] in some very high-
frequency native lexical items: then (OE þanne, þo˛nne, þænne, þenne); when
(OE hwanne, hwo˛nne, hwenne, hwœnne); any (OE æniġ, ęniġ, aniġ, ani, eniġ);
many (OE mæniġ, moniġ, meniġ). For the behaviour of [ε] + tautosyllabic
/r/ see 7.3.1 and 8.3.2.
As with /i/, the historical ‘stability’ of the short mid front vowel
becomes suspect if one looks into the PDE realisations of /ε/. A backing
to [] or lowering towards [a] occurs as part of the Northern Cities
Shift; it is also reported for the San Francisco Bay Area and Canadian
English. Younger RP speakers are also producing more open allophones
of [ε] (Hawkins and Midgley 2005). A strong tendency for raising of [ε],
especially in the younger generation, characterises NZE (Trudgill et al.
1998), illustrated by the confusability of, for example, check-in-counter ~
chicken counter. Raising is also common in the Southern US dialects
and in north-eastern Newfoundland (Thomas 2001: 18–19). Another
allophone involves the insertion of a diphthongal glide, especially in
stressed monosyllables before a voiced coda, so that in popular London
bed, leg can have [-ei-], and in the American South it can be a fully
diphthongal vowel, so that dead is pronounced [dejəd] (Thomas: 2001:
18–19).
The two possible outcomes of OE short [æ] in ME were [a] or [æ]
(see 7.3.1), hence the ‘dual’ input for the low vowel in the ME column
in Figure 8.1. Recall also that orthographic <a> was subject to regional
variation when followed by a nasal, resulting in ME West Midlands
forms such as mon ‘man’, onswere ‘answer’; that difference was levelled
out in favour of <-an> in the fifteenth century. In EModE <a>, pho-
netically [a] ~ [æ], and possibly [ɑ], underwent further phonotactic
changes. (1) reproduces the relevant section from Figure 8.1.
(1) Orthographic <a> and its reflexes in PDE:
æ~a happen, sad, man, land
a~æ ɒ ~ ɔ() wand, waft (SSBE), quarter, quartz
a ~ ɑ SSBE after, last, bath, car, half, dance
æ ~ eə AmE after, last, bath, dance, glad

The default reflex of ME [a] ~ [æ], highlighted by the heavier connec-

tor line, is a continuation of the variable ME pronunciation as in PDE
happen, sad, man, land. The representation of the sound with <a> in late
ME, the merger of the OE [æ] with ON and OFr [a], as well as the con-
tinuity of [a] in the modern varieties of Northern English, suggest that
the input to EModE is /a/. The evidence for [a] or for a more fronted
and raised allophone is mixed and the orthoepists’ testimony allows
conflicting interpretations. The dating of the categorical raising in the
literature ranges between the fifteenth century (Kökeritz 1953; Wyld
1953) and the eighteenth century (Jespersen 1909). Lass (1999: 85–6)
dates the ‘stabilisation’ of [æ] to about the 1650s. As argued in 7.3.1,
however, the possibility of continuing allophonic [æ] in ME cannot be
dismissed. Therefore, our account replaces the traditional formulation
of the ME change /a/ > EModE /æ/ by ME [a] ~ [æ] > EModE [æ] ~
[a]; the assumed raising is reinterpreted as a switch of the dominant
allophone from [a] to [æ], where ‘dominant’ is defined in terms of the
size of the geographical area where a particular variant was used.2
The variability of /æ/ in PDE was addressed in 7.3.1. Further, a
lowered allophone [a] occurs in Caribbean and Gullah; it used to be
considered ‘old-fashioned’ in the US, but now seems to be undergoing
a sudden spread in some varieties of North American English, reported
for western Pennsylvanian, central Ohioan, ‘young’ Texan, Californian
and Canadian (Thomas 2001: 20–1); the lowering is also associated with
‘younger, urban, or innovative’ RP speakers. :
A different type of allophony, based on the phonotactic effect of the
onset or the coda on the low vowel, resulted in recategorisation. The
first shift, from [a] to [ɒ], as in wand, quarter, quartz, is attributed to
the presence of a /w/-onset, which triggers backing and assimilative
rounding.
Chaucer rhymes freely <wa->-words with other <Ca->-words:
can: wan, warm: harm, so the rounding of /a/ cannot be dated before c.
1400. It probably started in the fifteenth century; some early evidence
for rounding can be found in The Cely Letters (1472–88), where we find
w(h)as ~ w(h)ose ‘was’, and inverse <a> spellings in warsse ~ worsse
‘worse’, w(h)arke ~ w(h)orke ‘work’. The variable pronunciation con-
tinues through the sixteenth and seventeenth centuries. Shakespeare
rhymes ward: guard, can: swan, Milton rhymes wand: land, and as late as
2
Whether the reflex of the historical <a> is phonemicised as PDE /æ/, on the basis
of number of speakers, or as /a/ as in Giegerich (1992: 72), ‘for reasons of simplicity
and cross-accent uniformity’, is not important; in either case regional allophones will
deviate from the ‘basic’ vowel specifications.
the first half of the nineteenth century one finds rhymes such as wand:
expand: land and war: far (Byron 1788–1824). The spread of the rounded
pronunciation was gradual, starting with prosodically weak words (was,
what). Walker (1791–1826) cites variable pronunciations for waft, wan,
wasp, quality, and the stressed vowel in water is still the vowel of MAT.
The rounding is inhibited by a following velar, hence quash but quack,
wash but wax, similarly wag, whack, swagger, twang, wrangle. When the
vowel is flanked by a [w-] in the onset and a fricative in the coda, the
outcome may be a long vowel in BrE as in waft, quaff, wrath. In those
dialects of AmE that have merged the vowels of words like cot, pod, Don,
knotty with the vowels of words like caught, pawed, Dawn, naughty (see
below), the result in words of the quart, quarter, quartz type is variable.
By normal historical development in the environment [Cw--rC], as in
quart, the resulting vowel in (General) AmE would be [ɔ]. But where the
merger has occurred, some further adjustment is required, and in fact
the vowel of the quart words is closer; the word ports rhymes with quartz
in such dialects. :
The next two realisations of the input low vowel in (1) illustrate the
difficulty of separating qualitative from quantitative processes. Here
the original short open vowel was lengthened when followed by a tau-
tosyllabic voiceless fricative: /-f/, /-θ/ or /-s/. Two factors seem to be
contributing to the lengthening: longer vowel duration before fricatives
as compared with stops (House 1961), and the intrinsic duration of low
vowels. Both factors are relevant to the pre-fricative lengthening of
the ME [ɔ] vowel, as in soft, lost, cloth (see (3) below). Some questions
remain, of course: vowels ‘gain’ length before voiced codas, so we should
expect jazz, pizzaz or chav with a long vowel, but it is not the case. Why
is lengthening before [-ʃ] restricted to AmE? If [æ] and [ɔ] could be
lengthened in pre-fricative contexts, why was/is [ε] not affected: we
get AmE [dejəd] for dead, [hejəd] for head, but not *[bejəst] for best, or
*[bejəθ] for Beth, or [lejəft] left?
Pre-fricative lengthening of /a/ ~ /æ/ has an important distribu-
tional consequence: recall from 7.3.2 that Northern ME maintains [a
] <
OE [ɒ
] (OE āc ‘oak’, Scots and Northern Irish aik), while the Southern
varieties lack a long unrounded open vowel (Figure 7.5). Today SSBE,
associated southern hemisphere Englishes and AmE show some variety
of a long unrounded open vowel in the range of [æ
~ a
~ ɑ
] in words
which had a short [a ~ æ] in ME: after, last, class, bath.
Another environment in which a new long open vowel emerged in
EModE was a rhotic in the coda, as in car, bar, hard; the long [ɑ
] ~
[a
] is characteristic of non-rhotic varieties of BrE, as well as New
England AmE, New York and some AAVE. A rounded [ɒ
] is domi-
nant in Southern r-less AmE speech, while r-ful speakers in the same
areas have [-ɒɹ], so that car and core are homophones (Thomas 2001:
45–7). Pre-consonantal loss of the liquid in /-'C/ codas, as in half,
calm, can also result in a lengthened nucleus; if the lengthening occurs
early (see 5.2.5), the vowel will undergo the expected long vowel shift,
as in halfpenny [heip(ə)ni], Ralph [reif]. The intermediate stage for the
lengthening before /-lC/ involves the replacement of the dark ['] by
a back glide, resulting in a diphthong [-ɑυ], later monophthongised to
a long back vowel [-ɑ
] or [ɔ
]. Lengthening before [-'C] is unstable,
and it is hard to draw dialect boundaries for the variants because of
high sensitivity to additional phonotactic factors and lexical inconsist-
ency; compare half, calf, calm, walk, talk and salve-valve [sɑ
v]-[vælve].
Generally, the lengthened variant is not the expected realisation in
North American English. :
Yet another environment where we get a mixed output in PDE is
the string of etymological short /a/ followed by /-NC/, especially
in Romance loanwords, where the results may vary depending on the
source of the borrowing. One difference is based on the rivalry between
Northern OFr and AN <-aun> and Central/Parisian OFr <-an>, for
example launch < AN launcher (13. . .) vs lance < OFr lancier (1330). The
AN forms survive in gaunt, haunt, laundry, saunter, while aunt, grant,
slander, sample go back to the <-an> input forms. The nature of the con-
sonant following the nasal also matters: a following voiceless fricative,
as in /-ns/, appears to have the strongest lengthening effect: advance,
chance, dance, enhance, lance, stance, and so on, but notice sans, pansy with
[-ænz]. A nasal followed by a tautosyllabic stop does not cause lengthen-
ing: lamp, champ, blank, flank, bland, grand (but grant, AN graunter, (1250),
with both [ɑ
] ~ [æ], also Alexander [-æ-] but Saunders). Finally, the
early diphthongisation of OFr and AN /a/ + (mostly) palatal obstruents
to [-ɑυ] as in ME daunger ‘danger’, chaunge ‘change’, raunge ‘range’, also
chaumbere ‘chamber’, results in late ME West Midlands and Northern
[a
], which then goes through the expected long vowel shift to [ei]. The
developments of /a/ + /NC/ are summarised in (2).
(2) Etymological /a/ + /-NC/ in the loan vocabulary:
Source Output Examples
/-NC/ (AN <-aunC>) [ɔ ~ ɑ] gaunt, haunt, laundry, saunter
/-NC/ (OFr <-anC>) [ɑ ~ æ()] aunt, grant, slander, sample, dance
/a/ +
/-NC#/ [æ()] lamp, champ, blank, flank, bland
/-NC (palat. obstr.)/ [ei] danger, change, range, chamber

It is evident that the trajectories and the results of the lengthening vary
greatly. The timing of the lengthening is of interest: the lengthened
vowels are kept separate from the vowels that constituted an input
to the long vowel shift, so that an [a
] resulting from open-syllable
lengthening, for example ME [ta
l] (see 7.5.2.1), continues on the path
to raising and diphthongisation, while the [æ
~ a
~ ɑ
] of, for example,
bath remains monophthongal. The lengthening must have started in the
seventeenth century and on the testimony of contemporary authori-
ties, it was still stigmatised as ‘bordering on vulgarity’ in the eighteenth
century.3
The pronunciation of the vowels in after, last, bath, car, half, dance is an
important dialect criterion today. The input [a] ~ [æ] is lengthened and
later backed and occurs as [ɑ
] in SSBE and East Coast AmE, though
that vowel is not the outcome in most of AmE. The low back vowel
characterises southern hemisphere varieties. New York City, many
New Englanders and Canadians have [æ
] in words of the after, fast
group. For further details on AmE see Thomas (2001: 21–3); Labov et
al. (2006).
The development of ME [ɔ] is another instance where AmE and BrE
differ; the difference is based on independent post-seventeenth-century
histories of the two varieties. The split is illustrated in (3), repeating the
relevant portion of Figure 8.1.
(3) EModE developments of ME [ɔ]:
ɔ~ɑ~ɒ hop, rotten, jolly, dollar
ɔ
ɔ ~ ɒ() soft, lost, cloth, port
The main sources of ME [ɔ] are: OE and ON [ɔ] (OE god ‘God’, ON
rotinn ‘rotten’); OFr [ɔ] (OFr jolif, ME jolie ‘jolly’); the shortening of OE
[o
] (OE gōd + spel(le), ME [ gɔ(d)spəl] ‘gospel’); the shortening of early
ME [ɔ
] from OE [ɒ
] (OE hāliġ + dæg, late ME [ hɔlidæj] ‘holiday’;
and the shortening of ME [ow] (ME cowhen ‘cough’) (see Figure 7.8(d)).
The PDE reflex of [ɔ] as [ɔ ~ ɒ] in dollar, hop, rotten may suggest an
uninterrupted realisation of the vowel with a certain degree of round-
ing. There is good evidence that the vowel was lowered to [ɒ] and
possibly unrounded to [ɑ] in the seventeenth century – it is found in
3
See Lass (1999: 103–8) for an excellent account of the complex variability of the out-
comes in EModE.
spellings such as <crass> for ‘cross’, <last> for ‘lost’, in Elizabethan

rhymes and, interestingly, the inverted <o> spellings of EModE
Continental loanwords: dollar (1553) < Dutch daler; boss (1653) < Dutch
baas ‘master’. Words borrowed from English into other languages also
suggest an unrounded pronunciation: ME box ‘punch’ > Low German
baksen, baaksen, German baxen, Danish baxe, Swedish baxas (OED), ME
frok(ke) < French froc, German Frack. :
The EModE lowering and unrounding of late ME [ɔ] can result in
further fronting to [a] in forms like <Gad> ‘God’; compare egad (< oh
God), <plat> for ‘plot’, <Tam> for ‘Tom’, drat (< ‘God rot’), the pair
strop-strap, both from ME stroppe. In Scots strap for ‘strop’, tap for ‘top’,
drap for ‘drop’ are the norm. The dating and the regional distribution
of the [ɑ] ~ [a] forms is complex, and there was a definite sociolin-
guistic dimension to the choices along the [ɔ] ~ [ɒ] ~ [ɑ] ~ [a] scale.
The frequent use of fronted [a] for ME [ɔ] in the seventeenth century
is described as a ‘mark of an affected style of speech’, exemplified by
the character of Lord Foppington in Vanburgh’s comedy The Relapse,
who uses <lard> for ‘lord’, <marning> for ‘morning’, <navelty> for
‘novelty’, <packet> for ‘pocket’, <rat> for ‘rot’ and <stap> for ‘stop’
(Barber 1997: 110). As late as the nineteenth century, William Barnes
(1801–86), whose Dorset dialect generally has [ɒ] for the LOT vowel,
allows drop: clap, lap, nap; beyond spelled <beyond>: hand; Johnny, spelled
<Jahnny>: Fanny; John also spelled <Jahn>: stan ‘stand’ (Burton 2010:
58–9).
In BrE the unrounding was reversed, possibly under the influence of
spelling. The more rounded pronunciation with [ɔ] in PDE is mostly
restricted to the inventory of BrE, though a great deal of variability con-
tinues to exist. The lowered [ɒ] ~ [ɑ] (but apparently not the fronted
[a]) vowel was exported and maintained in Newfoundland, most likely
carried by settlers from Ireland and south-western England (Thomas
2001: 25). Lexical idiosyncrasies are also common.
(4) Lexical idiosyncrasies of short vowels in the low back region:
Item OED Dictionary.com
LME mokk ‘mock’ Brit. /mɒk/, US /mɑk/ /mɒk/
ME dogg(e) ‘dog’ /dɒg/, US/dɔg/, /dɑg/ /dɔg/, /dɒg/
ME fogg(e) ‘fog’ /fɒg/ /fɒg/, /fɔg/
OE malt ‘malt’ Brit. /m ɔ
lt/, /mɒlt/, /mɔlt/
US /mɔlt/, /mɑlt/
ME salt(e) ‘salt’ /sɒlt/ /-ɔ
-/ /sɔlt/
The situation in AmE is complicated further by a widespread merger of
[ɑ] and [ɔ
], the COT-CAUGHT merger, which results in a partially, or fully
unrounded [ɑ]. The merger occurs in large parts of North America (:

see Maps 1 and 2 in Companion, Appendix 8.3). This ‘reverts’ AmE to the
seventeenth-century BrE realisation of the vowel.
As with ME [a], late ME [ɔ] can be lengthened in the environment of
a tautosyllabic voiceless fricative: /-f/, /-θ/ or /-s/: off, cloth, lost, most
commonly pronounced with a rounded monophthong [ɔ
] ~ [ɒ(
)].4
Like the corresponding lengthening of /æ/ before fricatives, the process
is gradual and produces variable results. Walker slams the lengthened
vowel in broth, froth, moth as used ‘frequently by inaccurate speakers, and
chiefly those among the vulgar’ (1791–1826: 32). :
Tautosyllabic /-r/ also lengthens and lowers the preceding etymo-
logical [ɔ] as in horse, port, lord. Walker considers the vowel in for, former
‘perfectly equivalent to the diphthong au [as in taught]; and for and former
might, on account of sound only, be written faur and faurmer’ (1791–
1826: 33). The lowering to [ɒ
] was variable in the eighteenth century:
Walker cites borne, corps, force, forge, form, fort, port, sport, and so on, as
having the sound of bone, alone. The length- and height-neutralisation of
mid and low back vowels + /r/ is an ongoing process, so that in some
BrE varieties and in some earlier AmE we find the merger of the NORTH
and START vowels.5 A Northern English merger (‘broadest Geordie’) of
the NORTH-NURSE vowels (Wells 1982, II), earlier involving lowering
and centralisation, is currently modified to a front rounded [ø] ‘amongst
young women in particular’, as reported in Wales (2006: 173–4). For
AmE the merger of the HORSE and HOARSE vowels is the norm for most
speakers.
The etymological cluster <-ng>, PDE [ŋ], can also result in allo-
phonic lengthening of [ɔ], merging the vowels of lost and long. For the
influence of coda /-l/ see 5.2.5; the loss of /l/ and the attendant vowel
lengthening can be lexically specific, for example ME folk, golfe, bolt >
PDE [foυk], [gɒlf] ~ [gɒf], [boυlt].
The last short vowel in Figure 8.1 is the late ME high back rounded
[υ]. PDE reflexes are shown again in (5).
4
NYC older-generation and working-class speakers, also New Jersey and
Philadelphia, have raising and ingliding diphthongisation of [ɔ
] to (often stigma-
tised) [oυə ~ υə] in, for example, off, bought; compare also the NY and Mid Atlantic
states diphthongisation of [æ
] in, for example, fast to [fiəst] through these steps:
[æ
] > [æə] > [eə] > [iə]. For more details on the diphthongisation of [æ
] see
Thomas (2001: 21–3).
5
William Barnes (nineteenth-century Dorset) rhymes storm : harm, corn : barn, short :
heart (Burton 2010: 266). In AmE homophony of LORD-LARD is reported for the
lower Mississippi valley, the area from St. Louis to Evansville, Indiana, Texas,
Newfoundland and Utah (Thomas 2001: 47–8).
(5) EModE developments of late ME [υ]:

~ə hunt, love, us, much, rush, flood
υ
υ pull, bush, full, foot, hook
There are multiple sources of the short close non-peripheral [υ]. It is

the direct continuation of OE [υ], as in ME hunt(en) ‘hunt’, luve ~ love
‘love’ (see 7.2.1); it can be the result of shortening of [u
], as in dove, n.,
husband, us, but (see 7.5.1.1, 7.5.1.3). As noted in 7.3.1, late ME [υ] is also
one of the common developments of OE [y], as in clutch, much, or OFr
and AN <u(i)>, as in punish(en) ‘punish’, ruser, ruiser ‘rush’. AN and OFr
close [o] is adopted in English as [υ], thus colour, covet, glutton, stomach,
subtle, supper; unstable vowel length is suggested by the spelling country,
cousin (but coz), trouble, double, couple. :
The fifteenth-century input [υ] was unrounded, lowered and cen-
tralised, starting probably around the beginning of the seventeenth
century. The resulting vowel is a lower-mid back unrounded []. The
process must have been still ongoing in the first half of the seventeenth
century; the first solid contemporary evidence is from the 1640s (Lass
1999: 89). The association between the letter and a short non-high
unrounded vowel is shown by the different accommodation of Indian
loanwords spelled with <a> in English and other European languages:
curry (1598) < Tamil kari, Portuguese caril, French cari; pundit (1661) <
Sanskrit pan. d.ita, French and Italian pandit; bungalow (1676) < Hindustani
banglā; Punjab (1833) < Urdu Panjāb.
The lowering and unrounding is considered the ‘default’ develop-
ment, indicated by the heavier line in (5). The realisation varies in the
different PDE standards between [] (Southern BrE, SSE, Irish English)
and a somewhat higher and more central [ə] ~ [] (AmE). Indeed, an
exact characterisation of the phonetic properties of the reflexes of [υ]
is impossible; the use of [] ~ [ə] ~ [] here covers what Lass calls ‘a
rather vague range of opener centralised-to-central vowel qualities’
(1999: 90). :
Lowering and centralisation to [] is not the only reflex of ME [υ].
As shown in (5), the ME quality of the vowel can be maintained, as
in pull, bush, full, foot, hook. The PDE vowel in pull, also pulley, Pulman,
put, push; bush, bullion, bullet, bulletin, bushel, butcher; full, also Fulham; wolf,
wood is attributed to the influence of the labial onsets [p-], [b-], [f-],
[w-], which could inhibit the unrounding. The phonetic rationale is
clearly the maintenance of the labiality feature in the transition from
the onset to the nucleus. If the pattern was clean-cut, it might suggest
an allophonic distribution: [-] after all other onsets and [-υ] after
labials, but an elementary ‘minimal-pair’ test shows immediately that
the seventeenth-century shift of [υ] to [] introduced a new contras-
tive unit in the system. That // is an independent phoneme is evident
from the fact that the presence of a labial has been ignored in numerous
lexical items: pulse, pulp, punish, pulmonary, but, butter, buttock, bulb, bulge,
bulk, fumble, fudge, and so on. Some items show [υ] in other environments:
cushion, sugar, should, would, gooseberry, and some words show variability:
AmE pulpit [pəlpət] ~ [pυlpət], mush [məʃ] ~ [mυʃ]. Then, there
are the minimal pairs such as PUT-PUTT, PUSS-PUS, whose number is
increased by the items whose ME [o
] was raised and shortened to [υ],
as in TOOK-TUCK, LOOK-LUCK, ROOM-RUM, BOOK-BUCK, COULD-CUD. Some of
these shortened forms undergo further unrounding to [] in spite of the
labial onset as in FLOOD, BLOOD (see 8.3.1).
The seventeenth-century split of the input vowel [υ] was character-
istic of the Southern and South-Western dialects of English. It occurred
also in Scottish and Irish English. In a large dialect area south of the
Scottish Borders and north of an isogloss running from the Wash to
the Welsh border (The Wash-Severn line on FOOT-STRUT, Wells 1982,
II), the vowel has preserved its closeness to [υ], providing a salient
dialect criterion for separating Northern from Southern English. The
importance of this dialect marker was recognised early. Kirkby (1746)
comments on the vowel of skull, gun, supper thus: ‘This Sound is scarce
known to the Inhabitants of the North, who always use the short Sound
of the eighth Vowel [the vowel of too, woo, food, DM] instead of it’ (Kirkby
[1746] 1971: 7). Commenting on the preservation of [υ] in some words,
Walker calls it a:
whimsical deviation . . . sufficient to puzzle Englishmen who reside at any
distance from the capital, and to make the inhabitants of Scotland and
Ireland (who, it is highly probable, received a much more regular pro-
nunciation from our ancestors) not infrequently the jest of fools. (Walker
1791–1826: 34)
He reports that ‘some speakers have attempted to give bulk and punish
this obtuse sound [υ, DM], but luckily have not been followed’ and
opines that ‘we cannot be too careful to check the growth of so unmean-
ing an irregularity’. Variability and cross-dialectal borrowing continued
in the in the nineteenth century, when comparisons with the vowels
of other languages are very common in the description of this new
phoneme in English for which there was no corresponding letter. It is
likely that the lowering and opening of the vowel was very widespread
Figure 8.2 The []-[υ] isogloss in England, Map 11 ‘But’ in Trudgill (1999a),
The Dialects of England, Wiley-Blackwell
in the nineteenth century, affecting items which were later reverted to

[υ] in the south. Cockney is one variety where the increasingly lower
realisation of the CUT vowel results in (IPA [ɐ]).6
Figure 8.2 shows a map of the but vowel in England at the end of the
twentieth century.
6
Unexpected [] in nineteenth-century Dorset and southern Somerset is attested by
rhymes such as put : nut, shut, pudding : blood in, roof : buff, stuff, enough (see Burton 2010: 60).
For PDE, the []-[υ] divide remains an important regional identifier,

though, as can be expected, there is ample evidence of continuing vari-
ability: in the twentieth century, the RP variant of the CUT vowel has been
moving forward towards [a] (MacMahon 1998: 410–11, 456–7). Wales
(2006: 172–3) reports that Northern English speakers participating in
‘the growth of suburbanization’ and born since 1938, have pronunciations
‘which are neither RP phonemically or traditionally local: so for Hull
speech as well as Merseyside and Newcastle, a schwa-vowel in mud, . . .that
is neither /u/ nor //’.7 While unrounding may affect the upwardly
mobile speakers of Northern English, the opposite, a tendency for [υ] in
the vowel of cut, is reported in AmE for Detroit; backing and rounding is
also part of the Northern Cities Shift in AmE (Thomas 2001: 27–8).
8.2.2 The Great/Long/English Vowel Shift

La, a note to follow sol . . .
LA, The City of Angels
This section turns to the most emblematic of all phonological changes
in the history of English: the set of long-vowel changes that started in
late ME and led to the renaming of the vowel letters <a, e, i, o> from
labels identified with the Latin values to their current names [ei, i
, ai,
oυ]. : English is the only European language using the Roman alpha-
bet in which the reference to the vowel letters in FACE, FLEECE, PRICE,
GOAT differs significantly from the value assigned to them in Latin,
or German, Spanish, Hungarian, for example. This English-specific
naming started out as reflecting the pronunciation of the letters repre-
senting historically long vowels /a
/, /e
/, /i
/, /o
/. Their realisation
was in the process of changing in late ME, though the dominance of
Latin in the elementary language-education curriculum and in the
universities delayed the recognition and acceptance of the innova-
tions. The very popular didactic verse treatise The Prick of Conscience
(Northern, probably mid fourteenth century), in a section heavily
interspersed with Latin lines and even couplets, plays on the association
between <a> - /a
/, and <e> - /e/
.
(6) Rhyming <a>: <swa> and <be>: <e> (Northern):
And by þat cry men may knaw þan
Whether it be man or weman,
7
See also Lass (1987: 250): ‘Northerners who “standardize” without going all the way to
adopting RP or a general SBE [SSBE in this book] profile will normally “correct” the
foot/but identity (sometimes with a “compromise” [ə] in but) . . .’
For when it es born it cryes swa; 480

If it be man, it says ‘a, a,’
Þat þe first letter es of þe nam
Of our forme-fader Adam.
And if þe child a woman be,
When it es born, it says ‘e, e.’ 485
E es þe first letter and þe hede
Of þe name of Eue þat bygan our dede.
(Morris 1863: 14)
Similarly, it is commonly held that the realisation of the long vowels
in Chaucer was close enough to their Latin pronunciation. The evi-
dence cited for that is that he could rhyme the name of the letter A
with omnia.
(7) Rhyming <a>: <Lat. omnia> (Southern, Chaucer):
On which was first i-write a crowned A,
And after, Amor vincit omnia. CT Prologue 161
Rhymes can be a good source of evidence for phonological change, yet
when a rhyme clearly echoes a Latin sound or word, its reliability can be
questioned. In both instances the interpretation of strictly low monoph-
thongal [a
] is not the only option: the rhyme <a>: <swa> in (6) could
have been on [æ
] or even [ε
] in the North,8 and A: omnia in (7) could be
a spelling-based stretch and the final vowel in omnia could also be [æ
]
under secondary stress. : As for <be>: <e> in (6), it is hard to tell how
raised the [e
] of be was; the rhyme does not preclude a diphthongal [ej],
as is suggested already in OE, where the names of the consonant letters
of the alphabet are recorded as both be or bei for , ce or cei for <c>,
de or dei for <d> and ge or gei for <g> (DOE).
Before we look further into chronological details, it will help to take
a long-term view and clarify the scope of the events and the labels they
have been given. The change of the OE long vowels to their current
pronunciations affects all vowels in the system reconstructed for ME in
Figure 7.5, used as the ME base in Figure 8.3.
The vowel correspondences between late ME and PDE, which
nineteenth-century philologists reconstructed and described, are listed
in Figure 8.4.
The arrows in Figure 8.4 are a crude approximation of the recon-
structed input–output correspondences. Some version of this set of
8
On early raising of /a
/ in the North see Smith (1996: 99–100); Stenbrenden (2010:
80–1, 532).
Middle English Examples
i u [mis] ‘mice’ [mus] ‘mouse’
e o [se] ‘see’ [fol] ‘fool’
ε ɔ [sε] ‘sea’ [ɔk] ‘oak’
a (N, Sc.) [ak]‘aik/oak’, [bak(ə)] ‘bake’
Figure 8.3 Long-vowel inheritance in Middle English
Middle English PDE9 Examples
i: aI price, mice, bind, wife, sign
e: i: fleece, see, tree, be, fiend
E: i:/e sea, meal, break
a: eI face, bake, strange, save
O: oU goat, oak, boat, nose, stone
o: u: goose, fool, do, moon, scoop
u: aU mouth, mouse, how, cow
Figure 8.4 Long-vowel changes from ME to PDE
changes, known collectively as the Great Vowel Shift (GVS), a moniker

first found in Jespersen (1909: 231), appears in all histories of the English
language and all descriptions of Chaucerian English. 9
The term ‘shift’ with reference to these changes is used in a highly

specific meaning: except for the SEE-SEA homophony in most varie-
ties of PDE, the phonemic contrasts between the original entities
have been preserved, in spite of the change in their phonetic realisa-
tions. This kind of input–output maintenance of contrast is typical of
9
Transcription practices for the diphthongs resulting from the shift can vary, especially
for the vowels of FACE and GOAT. Some the common alternatives are:
[ei]: [e
] ~ [e] ~ [ej] ~ [ey]
[oυ]: [o
] ~ [o] ~ [ow] ~ [əυ] (BrE)
‘chain shifts’; the label implies a lock-step development of a whole

set of related sounds. The structural appeal of all seven long vowels
behaving in a very similar fashion, within the same time frame, with
comparable results, is undoubtedly strong and the last century has
generated a huge amount of scholarship on the construct of the ‘Great’
vowel shift.10 Starting with Karl Luick (1898) and Otto Jespersen
(1909), the changes in Figure 8.4 have been treated as a unified set
of phonological events, where each new value is supposed to be the
result and the trigger of a chain-like reaction affecting all ME long
vowels. The position is well represented by Jespersen’s summary
statement:
The great vowel-shift consists in a general raising of all long vowels
with the exception of the two high vowels /i*/ and /u*/,11 which could
not be raised further without becoming consonants and which were
diphthongized into . . . [ai, au]. (Jespersen 1909: 231)
The structural connectedness among the changes is represented in the
often-cited chart in Figure 8.5 from Jespersen (1909: 232), where <*>
marks vowel length:
All ME long vowels: the ones that were long in OE, the ones that
acquired length through homorganic cluster lengthening (see 6.4) or
open-syllable lengthening (see 7.5.2.1), participate in the raising and
diphthongisation shown in Figure 8.5. This all-encompassing, chain-shift
view of the GVS dominated the scholarship throughout the early parts
of the twentieth century. Hypotheses addressing the initiation, causation
ai i u au
e o
ε ɔ
a
Figure 8.5 Directionality of the Great Vowel Shift according to Jespersen

(1909: 232)
10
Placing ‘Great’ in scare quotes is now common in the literature. For the myth of
‘greatness’ and the history of its debunking see Watts (2003, 2011). Stenbrenden (2010)
avoids ‘Great’ in the title of her extensive research project on the shift, and keeps ref-
erences to the ‘Great’ vowel shift in scare quotes – a practice adopted here.
11
Jespersen uses the pre-IPA symbol [*] (= IPA [
]) for length.
and propagation of the putative massive chain shift have been proposed
in terms of numerous phonological theories, including structuralism,
generative phonology, lexical phonology, dependency phonology, parti-
cle phonology, lexical diffusion and optimality theory. With rare excep-
tions, a representation of the changes from Middle to Modern English,
with each long vowel linked to an arrow pointing upwards or outwards, it
is still repeated in textbook accounts of the history of English phonology.
The twentieth century also saw the ‘deconstruction’ of the events
associated with the shift. One shortcoming of the early discussions of
the shift as a unified chain, with the front and the back vowels advanc-
ing in harmony, is that the scrutiny of the end results was confined to
SSBE. Another difficulty with the construct in Figure 8.5 is that that it
presupposes a beginning and an end to long vowel shifting, although,
as we will see below, the written records of the various innovations
are a challenge to the traditionally assumed time span of 1400–1750.
Yet another serious source of discomfort is the treatment of the pro-
cesses affecting the long vowels in isolation, without reference to the
coexisting diphthongal nuclei.
The problem of regional variants and local sub-shifts is acknowl-
edged in all or most accounts of the ‘Great’ vowel shift; analysts are
aware that neither the ME input vowels nor the output vowels are
uniform across the varieties of English. It is still the case, nevertheless,
that the PDE values shown in Figure 8.4 and 8.5 represent roughly
the pronunciation of the long vowels in current ‘General’ AmE and
SSBE, and this is also the pronunciation recognised and used in ‘global
English’. It is therefore possible to confine the account to these varieties
and set aside the dialect differences. In the next sections the focus will
be on the chronology, the mechanism and the motivation of the changes
that produced the supra-regional long-vowel system of PDE. :
8.2.2.1 Chronology and dating

The term ‘chronology’ can be used both in reference to the arrange-
ment of ‘events in the order of time’ and to the ‘assignation of events
to their correct dates’ (OED). The two temporal dimensions are inter-
related; the ordering of the changes with respect to each other cannot
be divorced from the dating of those changes, but for the sake of clearer
exposition we will consider them separately.
Chronological ordering is essential in determining system-internal
causal relationships. Establishing the chronology of phonological
change is directly linked to the adaptation of loanwords in the language.
Loanword phonology is a useful test for the existence of a particular
vowel realisation at the time of borrowing. One can be fairly sure that
the words price, AN price, priese (c. 1225) and sign (?c. 1225), AN seigne,
sein(e) had vowels that were identified with the native vowels of mice,
bind, wife (see Figure 8.4) – all of them having been shifted. Similarly,
borrowings from French with etymological [e
] such as friar (c. 1290),
OFr frere; require (1312), AN requere, requerer; die, n. pl. dice (c. 1330), OFr
dé, pl. dés, must have had a vowel identical or very similar to the MICE
vowel in order to get through the shift. On the other hand, police (1450),
machine (1545), tambourine (1579), magazine (1583), caprice (1673), critique,
v. (1751), all with etymological stressed [i
], are either too late or too
rare, or both, to join the general diphthongisation of [i
]. :
For the shift in Figure 8.5, locating the initial impetus is central to the
overall account of the events. If the impulse for the diphthongisation
of the high vowels /i
/ and /u
/ came from the raising of the high-mid
vowels /e
/ and /o
/, the closeness of the raised allophones to the his-
torical /i
/ and /u
/ could be a trigger of diphthongisation of the high
vowels. This is the ‘push-chain’ theory which originates with Luick. If
the high vowels /i
/, /u
/ started to diphthongise first, one can imagine
that the resulting vacant vowel areas would enable the raising of /e
/
and /o
/; this is the ‘drag-chain’ theory which originates with Jespersen.
Further, if we look at the low vowels, the fronting and raising of OE
[ɒ
] to [a
] in the North, as in ‘aik/oak’, could have been the initiation
of a more general push-chain front-vowel shift in the northern dialects,
unrelated to the changes in the South (Smith 1996: 99–101). Yet another
chronological option is to consider the raising of OE [æ
] to [ε
] in
early ME (see 7.3.2, Figure 7.5) as a precursor of the whole push-chain
process of the long vowel shifting in ME and EModE. :
The metaphors of ‘pushing’ – avoidance of merger, or ‘pulling’ –
avoidance of large gaps in the phonological space, dominated the dis-
course on the shift in the last century, but their usefulness is limited:
the new data on the dating of the high and upper-mid vowel changes
examined in Stenbrenden (2010) show that the raising of [e
] and [o
]
and the diphthongisation of [i
] and [u
] started simultaneously in
parts of the East Midlands and the West Midlands in the course of the
thirteenth century.12 This leaves us with a chronology of the changes
in Figure 8.5 which bundles together the high and the upper-mid long
vowels /i
/, /u
/, /e
/ and /o
/ as the ‘leaders’ of the shift. Further iden-
tification of lexical frequency and phonetic and sociolinguistic factors
may lead to new insights on the chronological ordering, but the available
evidence makes both the ‘push’ and the ‘drag’ shift scenario suspect.
12
Earliness of the shift is argued for in Stockwell and Minkova (1988, 1997a); Minkova
and Stockwell (1997a: 33–5, 2003a passim); Johnston (1992).
Dating the events is a separate matter; it is of practical consequence

for editing, teaching and reading Middle English texts. While pegging
a date onto a particular sound change will always be open to dispute,
especially the initial stages, the availability of LAEME and LALME
allow the type of close scrutiny of the written records that was una-
vailable to earlier scholarship. Extensive studies of these records
by Stenbrenden, culminating with Stenbrenden (2010), show that a
hypothesis of a thirteenth-century incipience and fourteenth-century
vigorous spread of the new values of the historical long monophthongs
is robustly supported by irregular spellings.
As noted in 8.2.2, the assumption in most sources is that the begin-
ning of the shift should be dated to the late fourteenth century, hence
the common reconstruction of Chaucerian pronunciation with fully
unshifted /i
/, /u
/, /e
/ and /o
/. Dissenting opinions expressed early
come from Wyld (1953) and Kökeritz (1961), who famously described
the reconstruction of fully unshifted long vowels for Chaucer as ‘a pro-
nunciation [which] would probably sound old-fashioned to Chaucer,
could he hear it, perhaps reminding him of the speech of his grandpar-
ents’ (Kökeritz 1961: 9). Wyld proposed a bold early dating: ME [e
] was
[i
] before the end of the fourteenth century, ME [o
] approximated [u
]
in the early fourteenth century in the East, whence it spread to London.
Stenbrenden, who refers to Wyld’s hypothesis with approval (2010:
542), has now presented persuasive arguments and testable ortho-
graphic data which are the most thorough substantiation of the earliness
of the change. These conclusions should find their way into the manuals
on Chaucer’s language.13 :
It is clear that the initial stages of the restructuring of the long-vowel
system can be detected much earlier than the traditional ‘Early Modern’
dating in the literature. Another question regarding chronological
boundaries is, when does the shift end? As noted in 8.2.2, the date of
entry of a word in the language may correlate with the non-operation
of the change; recall caprice (1673), critique, v. (1751). The shifting of a
vowel interacts also with the quantitative changes covered in 7.5: if the
vowel was long in OE, but remains unshifted, it must have been short-
ened prior to the shift; meant, stealth, wisdom had long vowels in OE, but
they have not gone through the process of shifting (see 8.3.1). On the
other hand, the original short vowels in OE child, hound and ME tale,
13
The evidence for the early stages of the shift is primarily from orthography. Rhyme
evidence for early fifteenth-century diphthongisation of the high vowels is adduced in
Jordan (1974: 239–40); for the upper-mid back vowel we find OE dōm, ME dome ‘doom’:
meum Lat. pronoun stressed as meúm (Everyman).
nose were lengthened; these ‘new’ long vowels were subject to shifting.
Vowels which were lengthened in early Modern English: dance, bath,
lost, half (see 8.2.1) generally do not undergo the shift, except for isolated
cases such as halfpenny [heip(ə)ni], Ralph [reif].14 Thus in one sense,
the vowel changes discussed here do fit within a chronological frame,
albeit a rather flexible one. This is especially true of the native lexicon,
the definition of which must also be flexible in order to include early
loanwords.
As regards the local manifestations of the long-vowel changes in
the modern dialects, and the behaviour of post-ME borrowings, the long
vowel shifting is very much part of the phonology of PDE. An important
aspect of the historical shifting is that it created a pattern of allomorphic
alternations which can be applied to borrowed words. Thus the native
vowel-patterning in heal-health, knee-knelt, wide-width, five-fifth, bone-
bonfire (see 7.5.1) is replicated in such pairs of loanwords as serene-serenity,
please-pleasure, deprive-deprivation, resign-resignation, phone-symphonic, and
so on. The second words in these pairs are suffixed formations coming
directly from Latin and French. The pairing here is between the pres-
ervation of length in the borrowed stressed vowel, as in serene, please,
deprive, resign, phone, in which case the long vowel shifts like any long
vowel in the native vocabulary, and the lax vowel in derived forms,
either inherited as lax or laxed after borrowing. Shifted and unshifted
(and further reduced) vowels in the loan vocabulary frequently alter-
nate depending on stress. If the borrowed long vowel preserves stress,
it is shifted, but if the stress falls away from the vowel in a form derived
with a stress-shifting suffix, the vowel is not only shortened, but further
reduced to [ə]: able-ability, mason-Masonic, legal-legality, aroma-aromatic,
horizon-horizontal. These synchronic allomorphic patterns present many
challenges and are much discussed in the literature.15
Like every other phonological change, long vowel shifts stretch
over a considerable period of time, affecting different items at a dif-
ferent rate. Useful lexical diffusion information for some long vowels
in native words is presented in Ogura (1987), but many unknowns
remain. To frame the problem: a loanword such as police (1450) keeps
[i
] unchanged, while profile, n. (1638) is ‘anglicised’. Walker (1791–1826:
26) cites Pope rhyming besieg’d: oblig’d on [i
], where the vowel of besiege
14
There is variability in the adaptation of OFr/AN low back vowels before nasals, where
AN had the diphthong [au] and OFr [a], both merging in possibly nasalised [ɑ̃
]. The
latter tends to diphthongise to [ei], especially before palato-alveolars, as in angel,
change, ancient, danger, strange. The results may vary; compare ancient-pansy, and before
labials chamber-lamp.
15
See Minkova and Stockwell (1998) and references there.
(1297) is from OFr [e

], and the vowel of oblige (c. 1325) is from AN
[i
] and should have been diphthongised. Some of Walker’s ‘unshifted’
words with borrowed [i
] coincide with the PDE forms: antique, caprice,
critique, quarantine, magazine, and so on but not vertigo (1528).16 What is
the rationale behind the shifting, or lack thereof, in the combining forms
macro- (1821) (unshifted) vs micro- (1849) (shifted), while macron (1851)
allows both [æ] and [ei]? The word route (1225) can be both [u
] and
[aυ].17 The name Levi may be unshifted, but both vowels shift in Levis
(1926). Using examples with just the low vowel, doublet forms exist
for pater (1400), mater (1425), data (1646), gala (1625), qua (1647), but
not for aleph (1300), the note la (1325), lama (1654), façade (1656), lager
(1855). These examples show both the persistence and the apparent
haphazardness of the application of the shift to new vocabulary.
8.2.2.2 Mechanism and causation

The developments bundled together under the general umbrella of the
long vowel shift are not as unilinear or straightforward as Figure 8.5 may
suggest. The shift is a subset of a complex series of long nuclei changes
and mergers which do not occur at the same time, in the same regions
or for the same reasons. The great diversity of intermediary steps and
outcomes cannot be dealt with in every detail here; moreover, many
changes indicating continuity of the shift are still under way (see 8.4).
Limiting the focus on changes extracted as the ‘Great’ vowel shift,
the consensus is that the high and the upper-mid long vowels /i
/,
/u
/, /e
/ and /o
/ were the ‘leaders’ of the shift, with the high vowels
becoming diphthongal and the high-mid vowels reaching their PDE
height by c. 1550 in the variety of Southern English that was recognised
as ‘standard’ after the seventeenth century. There is also consensus that
there was as much as a whole century between these earlier changes
and the rest of the long-monophthong changes: the ‘peak’ period for the
shifting of the low-mid [ε
], [ɔ
] and the low [a
] was around the middle
of the seventeenth century. This chronological gap prompts accounts
which ‘set these later developments aside, and restrict the term “GVS”
to the original high/higher-mid chain shift of the 15th–16th centuries’
(Lass 1992a: 153). :
16
The fashionable pronunciation well into the nineteenth century was [vəti
gəυ]
(OED). For profile Walker allows both [i
] and [ai]. PDE cation (1834) is now [kætaiən],
most likely on the analogy of ion, but the earlier pronunciation was [kætiən] (OED).
17
‘The pronunciation with a diphthong is recorded from the second half of the 18th
c. and preferred by some, but not all commentators . . .; it disappears from standard
British English in the course of the 19th cent., but is still widespread in North America’
(OED).
əi i u əυ18
GVS (c. 1400–1550)
e o
Low-vowel changes (c. 1550–1750)

e ε ɔ o
ei æ/a
Figure 8.6 Phases of the long vowel shift in Southern Standard English
The point of Figure 8.6 is that the historical details do not justify the
18
reconstruction of a step-by-step, cause-and-effect unified ‘Great’ event.

Since initiation of the shift of the high and upper-mid vowels cannot
be separated chronologically (see 8.2.2.1), assumptions that high-vowel
diphthongisation ‘causes’ mid-vowel raising, or that mid-vowel raising
‘causes’ high-vowel diphthongisation are unwarranted: both processes
‘appear to have started in parts of the E Midlands and W Midlands in the
course of the thirteenth century’ (Stenbrenden 2010: 542). Nevertheless,
it is an incontrovertible fact that the letters , <e>, <a>, <o>, 
represented different long nuclei in the early thirteenth century and
today, and like all other changes surveyed here, this one invites inquiry
into the reasons for the selection, maintenance and codification of the
variants that emerged and spread in later English.
Starting with the early ME high vowels /i
/ and /u
/, the generalisa-
tion is that they were diphthongised. Focusing first on /i
/, recall from
7.4 that already in OE the palatal glide [-j] formed a diphthong with a
preceding front-vowel nucleus when it was in the same syllable or the
same stem. Examples with a high front vowel include early OE *briġdel
> late OE brīdel ‘bridle’, drȳġness ~ drīġness > late OE drīness(e) ‘dryness’,
flȳġan ~ flȳan, p.t. flīġde ‘(cause to) fly’, all with [-ij]. The perceptual
distance between the [ij] of, for example, brīdel ‘bridle’ and the [i
] of,
for example, wīf ‘wife’ would be minimal; it is therefore quite logical
to treat [ij] and [i
] as allophones already in late OE. Support for this
allophony is found in variable spellings of historical [-ij] forms as in
drȳġness ~ drīġness, flȳġan ~ flȳan, and in ‘reverse’ spellings where both
 and <ig> stand for etymological [i
] in final position: <bi ~ big ~
18
The diphthongal realisations [əi] and [əυ] represent intermediate pronunciations
of /i
/ and /u
/ – they are the closest to what philologists think of as the ‘authentic’
Elizabethan/Shakespearean pronunciation of these vowels.
bii> ‘by, prep.’, <hwy ~ hwig> ‘why’, <sie ~ sy ~ sig> ‘be, pres. subj.’
(see 7.4). This minimally diphthongal [ij] can be posited as the primary
allophone of the long high front vowel in early ME, that is, as the input
to the later change of OE /i
/.
The situation with the high back vowel was parallel: if an OE back
vowel was followed by the OE voiced velar fricative [γ] (see 4.2.1), the
result was a diphthongal nucleus with a second element [-w]: OE būgan,
ME bouen ~ buwe(n) [buw-(ən)] ‘bow down’; OE (for)suwod ~ (for)sugod
[suw-(ə)d] ‘silenced’; OE sugu, ME suw(e) ‘sow, female swine’, where
[uw] would be an allophone of [u
] in, for example, mūs [mu
s] ‘mouse’.
The number of lexical items with [uw] was increased in ME through the
loss of the voiceless glottal fricative /h/, as in OE druhþ(e) > ME dru t
~ droute [drυxt] ~ [druwt] ‘drought’ (see 5.1.4). This allophonic [uw]
serves as the proto-diphthongal realisation of the original long /u
/. In
the case of the high vowels, then, diphthongal and monophthongal vari-
ants merged toward a more perceptually salient diphthongal realisation.
These initial diphthongs were further optimised by differentiation of
their end points, reaching the values [əi] > [ai] for mice, bind, bridle, dry,
wife, sign, and [əυ] > [aυ] for mouse, how, cow, crown, sow, fowl.
The increase of the distance between the elements of the nucleus pro-
ceeds gradually, and each stage is attested historically. Schematically,
this part of the long vowel shift can be represented as in (8).19
(8) The mechanism of high-vowel shifting in ME:
[ij] [uw]
[ij] [əj] [əw] [uw]
[i] [u]
[j] [w]
[ai] [aυ]
This sequence of events, labelled the ‘Center Drift’ (Stockwell 2002b:

267), is not the only possible way of reconstructing the intermediate
historical phonetic values of the diphthongs.20 In all accounts, however,
19
It is assumed that diphthongal transcriptions using the IPA high front glide [-j] and the
high back rounded glide [-w] are functionally equivalent to transcriptions using [-i]
and [-υ] (see Cruttenden 2008: 94; also 2.2.2, n. 9).
20
A very good survey of the positions is offered in Wełna (1978: 184–7); Lass (1999:
80–1). The strikingly diverse dialectal detail on the realisations of OE [i
] (17 vari-
a monophthongal input becomes a diphthongal entity with the original

short trajectory within the nucleus getting extended. In the inputs [ij]
and [uw] the glides are at the same general height. The subsequent
lowering of the first element produces a ‘better’ diphthong in terms of
perceptual distinctiveness by height-dissimilation: the first element is
lowered and centralised while the end point remains at the same height.
For reasons which are still being studied, different varieties of English
today show ‘arrested’ development in maintaining the incremental dis-
similation of the diphthongs: for example, [əi] for PRICE and [əυ] for
MOUTH occur in Welsh, Irish and Dorset; [i] for historical [ij] before
voiceless codas is found in Scots (also [ɑ*e]) and famously in Canadian
English (see 8.4.1). In popular London the PRICE vowel nucleus is
retracted to [ɑi], and rounded to [ɒi] in Cockney.
The opposite tendency is towards assimilation of the end points of the
diphthong, in which case the underlying phonetic impetus is minimisa-
tion of effort, as in ‘extreme’ Cockney [ɑ,ə ~ ɑ,
] (Wells 1982, II: 306–10),
NZE [ɑe] (Bauer 1994a: 389), Northern US [æi ~ i]), or Southern US
[ae ~ a
], where the result is ultimately a monophthong (Thomas 2001:
34–8; see also Thomas 2001: 39–43 for the ‘bewildering array’ of /aυ/
realisations in AmE). These attestations of historically intermediate
stages and the ongoing reversals of diphthong optimisation make it
very clear that the assumption of an overarching directionality of the
long-vowel changes in the second millennium in English is problematic:
the stressed long vowels of English, as in other Germanic languages,
become diphthongal, and diphthongs become monophthongised, justi-
fying the metaphoric description of their state as ‘diachronic perpetuum
mobile’ (Minkova and Stockwell 2003a: 174).
At the same time as the high-vowel incipient diphthongs were becom-
ing less monophthongal and more perceptually salient, the upper-mid
long vowels [e
] and [o
] were raised, becoming more similar to [i
]
and [u
]. Two considerations should be included in the account of the
raising. First, the vowels which we reconstruct as [e
] and [o
], follow-
ing a long philological tradition, may have been allophonically higher
than the spellings with <e> and <o> suggest.21 Occasional spellings
ants) and [u
] (45 variants) in Ogura (1987: 30–109) leaves no room for doubt that the
realisations posited in (8) are realistic.
21
Paradigmatic alternations of [e
] ~ [i] and [o
] ~ [υ] also provide the speaker/learner
with a pattern of alternations that support the assumption of ‘raised’ upper-mid
vowels, as in Grimstead (Wiltshire, 1242) < OE grēn ‘green’ + -styde ‘-stead’, Brumstead
(Norfork, 1165) < OE brōm ‘broom’ + stede ‘place’. The shortening of ME [e
] as [i]
(see 7.5.2.1) is attested in such Chaucerian rhymes as fil (pret.sg.) < OE fēoll: wil, n. (KT
1103).
such as OE <fyt>/<fiet> for <fēt> ‘feet’ and early ME spellings such as

<shipe> for <shēp> ‘sheep’ (Cursor Mundi), <sliep> for <slep> ‘sleep’
(Vices and Virtues) point in that direction. :
Another consideration comes from the structure of the phonologi-
cal system. In principle, overlapping realisations of adjacent phonemes
are not surprising; the denser the overlap, the higher the probability of
merger. However, once the allophonic cloud of original [i
] and [u
] had
started shifting away, the allophonic realisations of raised [e
] ~ [ei]
and [o
] ~ [oυ] would not impinge on the distinctiveness of the origi-
nal entities. The boundaries of the clusters of allophonic realisations
remain crisp and the identification of raised [e
] and [o
] with original
/i
/ and /u
/ does not result in merger, as is indeed the case with the
vowels of MICE-FEET, MOUSE-MOON. Metaphorically speaking, the diph-
thongisation of the earlier [i
] ~ [əj] and [u
] ~ [əw] ‘vacates’ the high-
front and high-back slots in the vowel system and allows raising of the
upper-mid vowels. The logic of this correlation makes the upper part of
the changes in Figure 8.6 the only part of the GVS that can legitimately
be claimed to be a chain shift, a lock-step change of contrastive units
which preserve the original contrast but change their properties. This
is shown in (9), which combines the centre drift of the high vowels with
the raising of the mid vowels.
(9) The mechanism of high- and mid-vowel shifting in ME

(non-Northern):
[ij] [uw]
[i] [u]
[j] [w]
[e] [o]
[ai] [aυ]
The hypothesis about the correlation of the events in (9) is supported

by an interesting regional development in ME. During the thirteenth
century in the North the upper-mid [o
] was fronted and raised in the
direction of [y
], rhyming with OFr [y
], so we get rude < AN rud(e):
gud(e) < OE gōd ‘good’ (Jordan 1974: 86). It is also a fact that in the
same areas where [o
] was fronted, the [u
] remained unchanged, which
results in the Northern realisation of [u
] for the MOUSE vowel and
[iə] ~ [iυ] for the MOON vowel. This simplified description is instructive
in terms of the mechanism and structural relatedness of the changes in

the mid- and high-vowel part of the system. :
The mid- and high-vowel changes in (9) were pretty much complete
by the end of the sixteenth century. The other long-vowel changes in
Figure 8.6 involve early ME [ε
], [æ
], [a
] and [ɔ
]. As noted in 8.2.2.1,
the emergence in early ME of [ε
] and [ɔ
] from raising of [æ
] or [ɒ
]
(see 7.3.2), or from open-syllable lengthening of [ε] and [ɔ
] (see 7.5.2.1),
is also part of the larger picture of events. Acknowledging that, here we
focus on the subsequent history of [ε
], [a
] ~ [æ
] and [ɔ
] – the changes
that result in their PDE values also have their roots in ME, but appear to
have taken longer to spread and become incorporated into the incipient
standard.
The first thing to note here is that unlike the high vowels [i
] and
[u
] in (9), the lower-mid front [ε
] (OE slæˉ p, ME [slε
p] ‘sleep’, OE
medu, ME [mε
d(ə)] ‘mead’) and the lower-mid back [ɔ
] (OE nosu,
ME [nɔ
z(ə)] ‘nose’) do not behave in a parallel fashion. In the front
series [ε
], in the variety later codified as the Southern Standard, was
raised further to [e
] and merged with the historical [e
], so that it is
only the not fully reliable PDE spelling that signals that the vowels
of sea and see, meat and meet, heal and heel were originally distinct (see
7.3.2). The dating of this merger is difficult to test because it is dialect-
and item-specific. There is rhyme evidence for it in the fourteenth
century, though the merger was still not completed before the end of
the seventeenth century. The earliest examples of the merger are in
word-final position as in se(e) ‘sea’ (OE sæˉ ): me ‘me’ (OE mē), but there
are also rhymes involving [ε
] in closed syllables, as in del ‘deal’ (OE
dæˉ l): stel ‘steel’ (OE Angl. stēl), grete ‘great’ (OE grēat): fet ‘feet’ (OE
fēt). :
The historical merger of the front mid vowels [ε
] and [e
] can be
seen as functionally undesirable; the justification for it is often sought
in the avoidance of ‘crowding’ in the vowel space: a four-vowel height
system could be less stable than a three-vowel height system, which
minimises the allophonic overlap between adjacent phonemes and
guarantees greater perceptual salience. :
The merging proceeded gradually; the range of allophones for the
front lower-mid vowel was apparently quite broad. The picture is
extremely complex and involves cross-dialectal and cross-social class
and register variation. The ‘middle’ position of early ME [ε
] allows
reconstructed realisations ranging from [æ
] ~ [ε
] ~ [e
] ~ [e
] ~ [ei],
and indeed there is evidence well into the eighteenth century that all
these variants were available from the fifteenth and sixteenth century;
the final stage of codification had not been reached for Dryden and even
Pope.22 Potential confusion between historically distinct vowels exists

in words containing early ME [e
], [ε
] and [a
], further complicated by
the possibility of mergers with the diphthongal [-æj] ~ [εj] ~ [ej] (see
below). A sociolinguistic interpretation of the changes and distribution
of these vowels c. 1600 presents the pattern in (10) for OE mētan ‘to
meet’ with OE [e
], OE mete ‘meat’ with [ε
], and ME mate ‘defeat’ (AN
mater) with [æ
].
(10) Social dimensions of the long vowel shift c. 1600:23
Aristocracy Bourgeoisie Lower class
meet [i
] meet [i
] meet [i
]
meat [ε
] ~ [e
] meat [e
] meat [i
]
mate [æ
] ~ [ε
] mate [e
] mate [e
]
Even the most conservative pronunciation in the first column allows
homophony of meat and mate; that homophony is well attested in the
middle column. The dominant pattern is a two-way, rather than the
input three-way distinction, that is, the vowel space has been optimised.
Given the pre-existing diphthongs [εj] ~ [ej], which merge with [ε
],
the realisation of the MEAT vowel c. 1600 would still quite commonly
be [e
] ~ [ej]. In most cases the raising continues on to [i
], as in the
third column in (10), leaving only a handful of words in the Southern
standard preserving the earlier diphthongal realisation: great, break, yea.24
Again, this is only the outcome in the Southern standard, distilled from
a variety of competing realisations. The merger of the long mid vowels
did not occur in the North Midlands, where OE [æ
], as in OE læˉ st
‘least’, and the lengthened OE [ε], as in OE etan ‘to eat’, did not merge,
so that LEAST is [liəst] but EAT is [ej], neither of them joining ME [e
] in
the raising to [i
].25
The mechanism of the change of the long low [a
] ~ [æ
] and the
lower-mid [ɔ
] also involves raising and merger with highly similar
diphthongs, both old and new, discussed in 7.4 (see also Figures 7.7, 7.8).
22
Dryden (1631–1700) rhymes make: speak, great: seat, and Pope (1688–1744) rhymes
shade: mead (cited in Barber 1976: 293).
23
The schema was proposed by Leith (1983: 148–9). This allows Shakespeare the flex-
ibility of rhyming meat: mate either as [ε
] or as [e
], with lower-class characters more
likely to rhyme meat: meet. On the long front-vowel overlaps in early ModE see also
Barber (1976: 292–3/1997: 114–15).
24
The word steak is commonly included in this set, but it could be a continuation of the
ON form steik ‘steak’. Walker (1791–1826: §240) identifies the vowel of steak, break,
great with the vowel of bear, pear, swear.
25
The history and scholarship on the North Midlands shift is discussed in Stockwell and
Minkova (1999: 90–8) and Minkova and Stockwell (2003a: 171).
As a reminder, selecting only the relevant units, late ME had a set of

diphthongs that come from the sources presented in (11).
(11) Sources of ME diphthongs relevant to the history of the low long
vowels:
(a) Vocalisation or epenthesis in the OE sequence V + /-j, -w, -h/:
[-ej] (OE Angl.) grēġ, ME grei ‘grey’)
[εjç-] ~ [ej] (OE ehta, ME ehte ~ eihte ‘eight’)
[-æj] ~ [εj] ~ [ej] (OE dæġ, ME dai ‘day’)
[ow] (OE stōw, ME stou ‘spot, -stow’)
[ɔw] ~ [ow] (OE cnāwan, ME cnowe(n) ‘know’)
[ɔwx] ~ [ɔw] ~ [ow] (OE dāh, ME do(u)h ‘dough’)
(b) Scandinavian loans:

[ej] (ON megen ‘main, strength’, ON þeir ‘their’)
[ow] (ON *þōh ‘though’)
(c) French loans:

[ej] (OFr deis ‘dais’ (a. 1259), OFr delei ‘delay’ (1275)
The presence of these diphthongal allophones in the system is impor-
tant since it destabilises the perceptual difference between the pre-
sumably monophthongal [æ
] and [ɔ
] and the clearly diphthongal
nuclei [-æj] ~ [εj] ~ [ej] and [ɔw] ~ [ow]. English stressed long vowels
become diphthongal readily, and the availability of falling diphthongs
with peaks identical to the monophthongs makes the task of discrimi-
nation between overlapping realisations harder. In this scenario the
diphthongal vowels provide the basis of the mergers in (12).26
(12) Low and lower-mid vowel mergers with pre-existing diphthongs
in late ME: 2728
[-æj] ~ [εj] ~ [ej] [ɔw] ~ [ow]

[εj] ~ [ej] [ɔw] ~ [ow]
[æ]27 [ɔ]28
26
This part of the ‘traditional’ ‘Great’ vowel shift account is based on Stockwell and
Minkova (1988), Stockwell (2002b), and Minkova and Stockwell (2003).
27
The source of the low-front [æ
] is open-syllable lengthening of OE and OFr [a] as in
OE tale, ME [ta
l(ə)] ~ [tæ
l(ə)] ‘tale’, OFr age, blame, grace, place, scale (see 7.5.2.1).
28
The sources of [ɔ
] are the raising of OE [ɒ
] as in OE āc, ME southern [ɔ
k] ‘oak’
(see 7.3.2) and open-syllable lengthening of OE [ɔ] and OFr [o] as in OE nosu, ME
[nɔ
z(ə)] ‘nose’, OFr cloke ‘cloak’ (see 7.5.2.1).
Along with the possible merger of [ε

] with [εj] ~ [ej] (as in break, yea,
great), the continuum of lower-mid long front vowels allows the same
diphthongs to be identified with [æ
]. The raising and possible diph-
thongisation of [æ
] is attested only sporadically in ME. Occasional
early rhymes such as seyde: made (Guy of Warwick); hayte ‘hate’: fayt ‘dis-
semble’ < OE feġan: waite < AN waitier might be imperfect rhymes or,
alternatively, they are early indications of imperfect discrimination of
the vocalic entities in that corner of the vowel space. :
The appearance of the raised vowel in the north before 1500 is found
first before the anterior coronal /-s/, as in Scottish rhymes of place,
grace, case and others in -ness, -less (Jordan 1974: 238), which is what
one would expect: there is a well-established bidirectional interaction
between anterior coronals and vowel frontness.29 Ignoring possible early
indications and dialect differences, the raising of long [a
] to [æ
] is a
reliably attested feature of late ME. The identification of historical [a
]
with a mid vowel in the fifteenth century is common: the mid-fifteenth
century Paston Letters (Norfolk) show spellings such as <hest> ‘haste’
and <mek> ‘make’. The diphthongal allophone is recoverable from
the spelling of OFr male ‘mail’, appearing as <mayl/mail> in late ME,
rhyming with taylle ‘tail’, OE tæġl, tæġel (Townley Plays, Lincoln (?) a. 1500
(a. 1460)).
The history of late ME [ɔ
] is comparable. Here too the first inklings
of the raising are found in rhymes involving historical [o
] and [ɔ
]: so
‘so’, OE swā: to ‘to’, OE tō; wo ‘woe’, OE wā: do ‘do’, OE dōn. The raising
of [ɔ
] seems to be more advanced in word-final position, though it
is not clear why this particular environment should be conducive to
height neutralisation. : The merger of the raised [o
] with the diph-
thongs shown in (12) is recognised in the first half of the seventeenth
century, though earlier phonotactically conditioned diphthongisation
of [ɔ
] to [ɔw] ~ [ow] before tautosyllabic dark ['] is indicated by MED
spellings such as <owld> ‘old’ (c. 1376, a. 1400), <could> ‘cold, n.’
(1440).
One can thus say that both the upper and the lower half of the tra-
ditional ‘Great’ vowel shift underwent not just raising or independent
diphthongisation, but a series of mergers with the pre-existing diph-
thongs (shaded in 13), ultimately favouring diphthongal realisations as
the primary allophones which define the rising standard and the shape
of the PDE vowel system. Merging (9) and (12), (13) summarises the
main events discussed in this section.
29
The phonetic and phonological basis and the typology of the interaction between
coronals and vowel frontness is discussed in Flemming (2003).
(13) A summary of the major long-vowel changes in ME and EModE:30

[ij] [uw]
[i] [u]
[ej] ~ [e] [o] ~ [ow]
[e] [j] [w] [o]
[εj] ~ [ej] [ɔw] ~ [ow]
[ε] [ɔ]
[ai] [aυ]
[æ] [a] [ɒ]
In some ways the schema in (13) supports the traditional assumption

that the mechanism of the shift involves raising and diphthongisation
of the long vowels: but it crucially refers also to the ambient phono-
logical system: ‘pure’ long vowels and diphthongal long vowels coexist
and alternate historically. In the mid-vowel area a set of two or three
or more contrastive entities within a relatively restricted vowel space
merge to a single unit. This is in fact the opposite of a chain shift: there
is no displacement of contrasts. This means that the fundamental condi-
tion for a chain, namely the no-merger condition, can only be supported
for the high vowels.
In conclusion: it is true that all long vowels underwent some kind
of change in ME and EModE. Nevertheless, any representation of
the mechanism of the changes ignoring the coexisting long vocalic
nuclei in ME is too restrictive. The shaded diphthongal variants of the
long vowels in (13) existed in OE or arose in ME from coarticulatory
changes and from borrowing. The diphthongal and the monophthongal
variants of neighbouring allomorphic clouds merged towards either a
diphthongal or a monophthongal realisation. Some diphthongs were
optimised by differentiation of their end points: [ai] for bind, design,
wife, sty, and [aυ] for cow, crown, how, fowl. Thus [ai] and [aυ] are the
attested SSBE results of the gradual change of both pure long vowels
and minimal diphthongs of high vowel followed by a homorganic
glide; it was the diphthongal realisations that were formative in terms
30
This is an elaboration of the schema proposed in Stockwell (2002a: 267).
of the later history of these vocalic units. Similarly, the history of the
ME mid long vowels [ε
] as in OE brecan, ME [brε
k(ə)] ‘break’; /a
/
as in OE bacan, ME [ba
k(ə) ~ bæ
k(ə)] ‘bake’; and /ɔ
/ as in OE āc,
ME Southern [ɔ
k] ‘oak’ cannot be separated from the history of pre-
existing diphthongs as in day, play, weigh, blow, stow, dough. Any descrip-
tion of the reorganisation of the vowel system of ME should take
into consideration both the history of the long vowels proper and the
diphthongal entities with which the long vowels merged in the course
of the long vowel shift.
In this ‘deconstructive’ approach to the mechanism of ‘Great’ vowel
shift, the changes were early, often simultaneous, and fully integrated
within the vocalic system. Lexical diffusion is a main, and not yet fully
understood component of the process: some environments are more
conducive to change, and there is ample evidence that the frequency of
a lexical item is of consequence.
Addressing the causation of the changes is more challenging. Some
possible linguistic cause-and-effect factors in the initiation and diffu-
sion of individual changes have already been identified. There is no
single motivation that applies to all changes, but one can summarise
the processes in (13) with reference to the interplay of four factors that
jointly define the output. These four factors are:
1. Diphthong optimisation in perceptual terms.

2. Diphthong optimisation in articulatory terms.
3. Optimal spacing of adjacent entities – merger avoidance.
4. Optimal spacing of adjacent entities – merger.
The perceptual optimisation of diphthongs is illustrated by the progres-

sively longer diphthong-internal trajectory as in [ij] > [ai] and [uw] >
[aυ]. Diphthong optimisation in articulatory terms is the minimisation
of effort in terms of number of gestures, the necessity to coordinate the
gestures, economy of time and muscle energy, that is, the realisation of
diphthongs with shorter trajectories. An example of the minimisation of
effort is the assimilation of the end points of the diphthong in the PRICE
vowel; as noted above, it appears as [ɑ,ə ~ ɑ,
] in ‘extreme’ Cockney, as
[ɑe] in NZE and [ae ~ a
] in Southern AmE. Merger avoidance is the
principle that maintains the contrast between historically contrastive
entities: OE hȳdan ‘to hide’ and OE hēdan ‘to heed’ were not homo-
phones, neither are their PDE reflexes. Changes resulting in merger are
also attributable to optimal spacing of the vowel entities: phonemic dis-
tinctions based on small acoustic and articulatory differences are hard
to maintain – in the account presented here this was the case with the
overcrowded mid-vowel space in (13).
In addition to these linguistic structural forces at work, motivation
for the selection of a particular variant pronunciation can be sought
in the historical sociolinguistic setting of the process. From the start,
all scholarship on the ‘Great’ vowel shift emphasises that the system
described in, for example, Figure 8.4 can be applied only to a variety of
English that was spoken in the southern part of the country that even-
tually became codified as the standard variety. The regional dialects
of England and Scotland underwent their own long vowel shifts, some
of them very different from the ‘Great’ vowel shift. Moreover, many
current developments of the long vowels in American, British and
Australian English are also ‘shifts’ – we will return to some specifically
North American changes at the end of this chapter. The sociolinguistic
aspects of the long vowel shifting are outside the scope of this book,
but it must be acknowledged that the interaction among the regional
varieties in essentially ‘oral’, pre-literate and pre-standard times is an
important component of the historical record. The introduction of the
printing press, the rise in literacy and the involvement of the orthoepists
in the standardisation of written English is another key step towards
the selection and diffusion of particular pronunciations in London and
in the areas around London in the sixteenth and seventeenth centu-
ries. It was the establishment of a written standard that provided the
background for the codifying and prescriptive work of such writers
as Thomas Sheridan and John Walker that culminated in ‘Southern
Standard British English’. :
8.2.2.3 Further instability and enrichment: BREW-NEW, DO-DUE,

AUNT-HAUNT
The picture of the long vowel nuclei in EModE in (13) is incomplete: it
focuses on monophthongal vowels becoming diphthongal. The absorp-
tion of long vowels by pre-existing diphthongs suggests the potential for
the opposite process, historical diphthongs becoming monophthongs,
and it is indeed a type of change encountered in the monophthongisa-
tion of OE diphthongs in ME (see 7.4, Figure 7.6). Monophthongisation
is also attested during the early and late Modern English periods.
In addition, the English long vowel system was enriched with two
diphthongs from French, /oi/ and /ui/. Figure 8.7 outlines the main
diphthongal changes from late ME to PDE.
Figure 8.7 conceals some of the complexity of the processes involved
in view of the multiple sources of the merging diphthongs: Middle
English, Old Norse, Anglo-Norman and (Old) French. Some the details
iw new, music, rule
ju ~ u
ew dew, hue, brew
æj day, lay, maiden
ei
εj, ej grey, weight, veil, they
au ɑ ~ æ() dance, aunt
ɔ ~ ɑ haunt, thaw, bald, chalk, law

ɒw, ow bought, sought, slaughter
oυ flow, soul, bowl, dough
oi ɔi choice, joy, boy, royal
ui əi > ɔi boil, join
Figure 8.7 Diphthongal changes from Middle to Modern English
on <auNC> were touched upon in 8.2.1 and (11) lists the main sources
of the ME diphthongs, but some clarification is still in order.
In the case of the [iw] and [ew] merger at the top of Figure 8.7,
the native sources were joined by a large set of French words with
a high front rounded vowel close to [y], which was probably reana-
lysed and adopted as [iu]: due, duke, sugar, sure. A very similar diph-
thong was already in existence in ME both as a continuation of [i] +
[w] (as in Tuesday < OE Tiwesday) and from a raised ME [ew] as in
brew. :
The change from a rising diphthong to a monophthong – [iu] >
[ju
] > [u
] – has been ongoing since c. 1600. It is most advanced in
words in which the vowel is preceded by the sonorants [r-] and [Cl-]
and the affricates [tʃ-], [d-]: rude, clue, chew, June. In some other coronal
environments – [s-], [z-], [θ-]: suit, assume, resume, enthuse; after single
onset [l-]: lute, absolute; and after [t-], [d-] and [n-] there continues
to exist widespread variation across regional accents, registers and

individual lexical items. :
The merger of [æj], [εj], [ej], if they were ever really kept apart
(see 7.4), occurred probably already in Southern ME, though the
orthographic record is, as usual, more conservative and, indeed, quite
inconsistent. The raised realisations [εj] and [ej] for <ai>, though not
acknowledged until the middle of the sixteenth century, must have
been common in the fourteenth century. Chaucer’s rhyming practice
suggests that for his variety of English the two diphthongs were already
merged, allowing flexibility in the choice of rhymes.31
For the ME diphthong [au] the two outcomes are [ɑ
] ~ [æ(
)] as in
dance, aunt, or [ɔ
] ~ [ɑ
] as in haunt, thaw, bald, law, straw, sauce, laundry.
The split is often along etymological lines: Old French vs Anglo-
Norman (see 8.2.1). In the sixteenth century <a> + [-m, -n], as in dance,
lamp, angel, flame, danger, was pronounced as [ɑ
] (vulgar or dialectal)
or [aυ] (elegant). As in the case with the [ε
] and [e
] mergers in the
course of the long vowel shift, it appears that the originally less pres-
tigious, provincial pronunciation became codified; the variant with [ei]
(angel, strange, danger) was a shifted version of [ɑ
] after the seventeenth
century.
The [ɔ
] ~ [ɑ
] set is joined by words with the ME diphthong [ɒw],
[ow] before historical [-h], as in bought, sought, slaughter. Here belong also
some late ME forms with [ɑw] from <a> + <l>, where the diphthong
is due to the development of a back glide before the liquid, as in all, hall,
call, false, salt, and so on (see 5.2.5). Before velars the /-l/ is dropped,
for example chalk, stalk, talk. These forms were identical with the forms
containing /l/ + labial (calf, calm, half) until the seventeenth century.
The pronunciation with [æ
] in calf and half is recorded from the early
eighteenth century in British English; it is also the standard American
pronunciation.
The last pair of diphthongs in Figure 8.9 represent the only vocalic
nuclei in the late ME sound system that have no native pedigree. The
large majority of <oi> and <ui> words in PDE come from French (choice,
employ, loin, moist, turmoil, soil, join, poison) and Dutch (buoy, foist, loiter), and
occasionally from other languages: Greek hoi polloi, Algonquian Illinois,
Bengali poisha, Chinese hoisin, Dinka toich, imitative ahoy, boink, oink. The
original [ui] has an unusual history in that, like the more recent history
of /h-/ (see 5.3.1), it is one of the rare well-documented instances of
a sound change in progress inhibited and partially reversed by the
31
In the Canterbury Tales we find rhymes they [ej]: awey ‘away’ (ParT 541–2, The
Hengwrt ms), away: day ([εi] (?)) (MT 3553–4, The Cambridge ms).
influence of spelling. Many of the words with PDE <oi> – loin, boil, coy,
oil, join, point, choice, poison – had variant pronunciations with [oi] and
[ui], the latter commonly from Anglo-Norman. The ‘normal’ develop-
ment of [ui], involving lowering and centralisation of the first element
of the diphthong (see 8.2.1), was towards [əi], which in the seventeenth
and into the eighteenth century was also a possible realisation of his-
torical [i
]. That the two etymologically distinct entities were treated
as identical is shown by rhymes like loin: line, boil: bile, point: pint as late
as the second half of the eighteenth century. In spelling the [ui] > [əi]
words alternated between <ui> and <oi>. Eventually the centralised
pronunciation [əi] for historical [ui] was abandoned in favour of [ɔi],
no doubt supported by the spelling and pronunciation of the majority of
the loanwords in that group. :
The incorporation of the new diphthong in boy, joy, coin into the
English phonological system is commonly described as incomplete.
Vachek (1976: 162–7, 265–8) argued that lack of parallelism with the
other diphthongs ([ei-oυ], [ai-aυ], but not *[εi-ɔi]) and lack of mor-
phophonemic alternations involving [ɔi] in the PDE system, makes
[ɔi] a ‘peripheral’ phoneme in SSBE. He hypothesised that the survival
of [ɔi] is associated with its pragmatic function of differentiating syn-
chronically foreign words from native words, especially polysyllabic
words. Lass (1992a: 53) also emphasises the ‘foreignness’ of [ɔi] and its
structural isolation, and concludes that it [ɔi] ‘has just sat there for its
whole history as a kind of non-integrated “excrescence” on the English
vowel system’. While there are certainly asymmetries and restrictions
on the distribution of [ɔi], in PDE the diphthong is as variable as other
diphthongs, possibly in response to the ambient system. In AmE (as in
London and Cockney English, where [ai] is [ɑi]) there is widespread
raising of [ɔi] to [oi] and occasionally [ui] in the younger generation
of speakers. In the American South the diphthong can be realised as
triphthongal [ɔoi] ~ [ɑo]. African American speakers show lowering
of the second element to [oε] ~ [oæ] or [oγ] ~ [ɔ]; merger of [oi] and
[ai] is also recorded in Jamaican, Caribbean and Newfoundland English
(Thomas 2001: 38–9). The continuity of the foreign status of [ɔi] is
doubtful also in view of the fact that many of the original loanwords
were monosyllabic and thus fit the prosodic pattern of the native vocab-
ulary. The high productivity of the suffix -oid in the last two centuries
would be another argument against the special nature of the diphthong.
To what extent newly created items such as oik (1917), onomatopoeic
oink (1935), boing (1952), boink (1963), as well as catchy clippings such as
droid (1952), roid (1978), earmark [ɔi] as exotic is an issue that will profit
from psycholinguistic testing.
8.3 The effect of phonotactics on long vowel shifting

One of the most reliable signals of historical vowel length in English
orthography is the representation of a vowel with two vowel letters or
digraphs; the use of these in boat, read, keep, loud, moon, fear, and so on is
an indication that the vowel was long prior to the vowel shift (see 7.5.3).
Yet matching English spelling to pronunciation may appear erratic;
compare booth-book, bear-fear, mood-stood-blood, dead-bead. In all these
sets of words the input is a historically long vowel, but the outcome is
different.
Pre-vowel shift lengthenings are usually marked by a digraph or
by the sequence <-VCe>: OE medu ‘mead, OE talu ‘tale’. In contrast,
shortenings, especially in derived words, may not be orthographically
marked: south-southern, clean-cleanse; being short prior to the shift, the
vowels in southern, cleanse are exempt from it. In addition to such shorten-
ings (see 7.5.1), the nuclei of monosyllabic words tend to shorten before
specific codas. In these cases too, the orthography remains unchanged,
but the outcome is a short vowel, as in book, stood, dead. Tautosyllabic /r/
is also a special factor, responsible for, for example, bear vs fear. The next
two sections address these special cases.
8.3.1 Shortening in monosyllabic words: LEAD (Pb)-LEAD, v.,

DEAF-LEAF, MOOD-STOOD-BLOOD
The most noticeable environment for shortening in monosyllabic

items is before dentals, as found in, for example, PDE bread, dead, deaf,
death, head, lead (Pb), shred, thread, sweat, threat.32 As the <-ea-> spell-
ings suggest, the vowels in these monosyllables go back to ME [ε
]; the
shortening must have occurred prior to the shift to [i
] in the varieties
that were later accepted as ‘standard’. In ME the shortened variants
were in rivalry with the historical long variants; long variants of head,
bread, dead are attested into the seventeenth century, and they are amply
attested in the regional varieties of English today. For the front mid
vowels, statistics show that the shortening was most advanced before
[-d], spreading to [-t], [-θ] in the seventeenth century, while deaf con-
tinued to have a long vowel until the late eighteenth century. Within
a given set, the vowel of the most frequent word changed first, thus
32
Other monosyllables with long front vowel nucleus that went through pre-shift short-
ening are red < OE rēad, wet < OE wœ̄ t, ten < OE Angl. tēn(e), compare fifteen with the
original long vowel shifted; hot < OE hāt ‘hot’, compare hœ̄ tan ‘to heat’. Note that hot,
red, wet, ten would tend to be prosodically weaker in a noun phrase where the noun is
also monosyllabic.
dead, head were shortened earlier than spread, stead (Ogura 1987: 148–9,
185–90, 196–9). Many questions remain: the pairs death-lead, v. inf.,
sweat-treat, v., deaf-leaf have approximately the same frequency, so while
frequency is a factor within the group of items that did change, a full
account of the shortening is still needed.
Vowel shortening in monosyllables occurs sporadically also in words
with late ME–early ModE [u
], the raised reflex of ME [o
]. Thus in
the sixteenth and seventeenth centuries book, cook, look, took, nook, rook,
forsook, crook are all recorded with [u
], and so are good, stood, food, foot
(Dobson 1968: §§35–9). The centralised/laxed and shortened variants
with [υ] gradually gained ground in the seventeenth century. For some
items the change occurred early enough for them to participate in the
further qualitative change of [υ] to [ ~ ə], as in flood, blood, glove (see
8.2.1). Within the set of shortened items, a coda [-d] seemed to induce
the shortening first, followed by [-v], followed by [-t, -θ, -k]. Within
those subsets, lexical frequency and the presence of an initial cluster
also favoured the change (Ogura 1987: 145).
The shortening before dentals, both before the high back vowel
and the mid-front vowel, has been a puzzle for a long time, prompt-
ing explanations based on syntactic context, syllable well-formedness,
analogy, semiotics and coarticulation (see Ritt 1997; Phillips 2002; Ritt
2007). The probabilistic basis of all these hypotheses is still debated; the
leading trigger of the shortening, coda /-d/, should not be conducive
to shortening in view of the general tendency of the preservation of
phonetic duration before voiced obstruents. :
Shortening also occurs before the voiceless velar coda [-k], as in book,
cook, look, took, nook, rook, forsook, crook. The phonetic mechanism of this
seventeenth-century shortening can be related to the inherent likeli-
hood of a shorter vowel before a voiceless stop, but the consistency with
which this particular shortening occurred before the voiceless velar
in English is surprising. Note that monosyllabic words in PDE which
ended in [-u
k] in the seventeenth and eighteenth centuries and which
maintained the original length of the vowel are practically non-existent.
It is possible that once the shortening occurred on a single word, the
fashion for [-υk] in the eligible monosyllables spread and affected all
such words. How and whether this development relates to the complete
absence of [-u
g] or [-υg] codas in stressed monosyllables is difficult
to determine; the gap in the phonotactic system may be completely
accidental. :
The chronological sequence of shortening [u
] > [υ] presented here
is not the only logical possibility. In the items undergoing shortening of
the high back vowel, [u
] is the raised reflex of ME [o
], as in OE bōc
‘book’, OE gōd ‘good’, OE fōd ‘food’, OE glōf ‘glove’. Recall, however,

that if lengthened, OE [υ] > [o
], as in OE duru > ME dōr(e) ‘door’ (see
6.5.1), and that shortened of [o
] could result in ME [υ], as in moon (OE
mōna) - Monday (OE mōnandæġ) (see 6.5.1); similarly, the stressed [o
]
in OE mōnað ‘month’, mōðor ‘mother’ show EModE shortened [υ], later
non-Northern [].33 These examples suggest the analytical possibility
of OE [o
] being directly shortened to its height-partner [υ]. :
Figures 8.8 and 8.9 show how the items with shortened vowels
discussed in this section compare with items resisting the shortening. 34
OE dēaf ‘deaf’ OE lēaf ‘leaf’
Input [æ], [æə] [æ], [æə]
Shortening [ε] ---
Vowel shift --- [i]
PDE [dεf] [lif]
Figure 8.8 The DEAF-LEAF contrast in PDE
OE mōd ‘mood’ OE stōd ‘stood’ OE blōd ‘blood’
Input [o] [o] [o]
Vowel shift [-ud] [-ud] [-ud]
Shortening --- [-υd] [-υd]
[υ] > [ ~ ə]34 --- --- [-d]
Figure 8.9 The MOOD-STOOD-BLOOD contrast in PDE
The values of the vowels in the three items in Figure 8.9, mood-stood-
blood, were in place by the eighteenth century, but for some items the
variability continues to this day: hoof, roof, root, soot can be either [u
]
or [υ], groom and broom are also variable, but not gloom; soon and spoon
are variable, but not moon and noon; hoop and whoops vary, but not stoop.
These rather disparate data do not point to a particular pattern – a full
33
The pair moth-mother is sometimes cited as an example of erratic English spelling. Moth
is from OE moððe > PDE [mɒθ], [mɔθ], [mɑθ] (see 8.2.1). Mother belongs to the set of
common words such as other, brother, also smo(r)ther, for which Jespersen (1909: 332)
attributes the shortening to the cluster /-ðr-/ in inflected forms.
34
The lowering is dated after the first half of the seventeenth century (see 8.2.1).
set of regional variables may reveal the reasons behind the selection of a
particular realisation. : Why monosyllabic shortening targets ME [ε
]
and (late) ME [u
] < [o
] words but other long vowels are unaffected, is
yet another unanswered question.
8.3.2 Vowels in relation to /r/: PERSON-PARSON, TEAR, v. - TEAR, n.,

FLOOR-POOR
Figure 8.10 I Lost My Khakis (from <http://weknowmemes.com>
There are no sections on ‘Vowels in relation to [-p]’ or ‘Vowels in rela-

tion to [-z]’, so the heading of this section must be justified. Our starting
point is the properties of /r/ discussed in 5.2.1. In PDE the consonant in
question is a member of the class of rhotics, a complex class of sounds
collectively represented as /r/, a continuant with variable realisations
of which the alveolar approximant [ɹ] is most common. Additionally
we find the alveolar coronal trill [r] in some accents of Scots, alveolar
tap [ɾ] intervocalically in ‘Refined RP’, also the most common rhotic in
Scots, the retroflex [] (South-West BrE, North American, similar to

[ɹ]), the uvular continuant [r], or the uvular fricative [ʁ] (North-East
England, Lowland Scots), and a labiodental continuant [-] or [], the
velar approximant similar to an unrounded [w] in London/Estuary
English.
The most likely allophones of /r/ for late OE are the central approxi-
mant [ɹ] or the coronal trill [r] in the syllable coda (see 5.2.2). The early
changes involving short vowels plus coda /r/: Breaking, metathesis,
epenthesis and the very early ME pre-consonantal loss point to the
significant coarticulatory effect of /-r/ codas, but they did not trigger
a restructuring of the vowel system (see 6.5.3). The possible effect on
long vowels can be illustrated as in (14), acknowledging that some of the
scribal variants are rare, or specific to a particular set of texts.
(14) Scribal variants for OE long vowel + /-r/ coda:

OE dēor ‘deer’: <deor ~ dior ~ dyor ~ dear ~ dær>
OE fȳr ‘fire’: <fyr(r) ~ fyer ~ fir ~ fur ~ fer ~ feer ~ fier ~ fares
~ fære>
The early pre-consonantal loss of /-r/ (see 5.2.3) can be seen as the
beginning of a process stretching over more than six centuries and
affecting different dialects and different lexical items unevenly. The
nature of the rhotic, the most sonorous of the consonants, contributes to
its weakening in coda position, leading to its confusability with schwa
and eventual reanalysis to schwa, or loss in the non-rhotic varieties of
English. As this section shows, loss of coda /-r/ emerges as one of the
most important triggers of phonotactically conditioned vowel lengthen-
ing and diphthongisation in EModE, changing the composition of the
vocalic system.
Before discussing lengthening and new types of diphthongs in rela-
tion to /-r/, we turn to a process covered briefly in 7.3.1 in the context
of late ME short vowel changes. In the course of the fourteenth century,
the sequence <-er-> [-εr-] became <-ar-> [-ar-], as shown in (15).
(15) Late Middle English [-εr] > [-ar]:

OE EME LME PDE
herġan [hεrjən] [hεrj(ən])] [hari] ‘harry’
steorra [stεrrə] [stεr(rə)] [star] ‘star’
[fεrm] < Fr. ferme [farm] ‘farm’

EME LME PDE

[arbər] ‘arbor’ (1300)
[(h)εrb] < AN (h)erb-
[(h)εrbəl] ‘herbal’ (1516)
[parsən] ‘parson’
[pεrsən] < OFr persone
[pεrsən] ‘person’
The innovative forms were first observed in the North:

ĕ before r began to change to ă in the early fourteenth century in the
North, and by the end of the century the new sound had begun to appear
in the South. . . . The lowering was probably in the first instance to [&],
and subsequently to [a]. (Dobson 1968: §64, n.2; §§67–69 and notes)
The change occurs later than pre-consonantal shortening: OE dēor
‘dear’ > darling, OE fēorþing > farthing, OE stēor ‘steer’ + bord > starboard.
In some of the French loanwords the spelling with <e> contin-
ued to be used after the fifteenth century, as in sergeant (c. 1200, OFr
sergent). Under the influence of spelling, [-εr] (subsequently [-ər] ~
[-
] ~ []35) was reinstated in many words, quite likely with reinforce-
ment from Latin, as in perfect, serve, with late ME variants parfit, sarve.
The existence of competing pronunciations in EModE accounts for
Shakespearean rhymes such as starve: deserve, desert: part. This incom-
plete change created interesting pairs such as university-varsity, Berkeley
[-] - Berkshire [ar-], clerk [-] (AmE) - clerk [-ar] (BrE). Interestingly,
err, v. (1303, OFr errer), recorded as <arr(en)> in the fifteenth century,
shows two spelling pronunciations, the earlier [
] and both [] and
[εr] in AmE, the latter possibly supported by more frequent error (1300
< OFr error). :
(16) Orthography and the variable results of fourteenth-century [εr] >
[ar]:
Source Expected [-ar] Unchanged/restored
[εr]/[
] ~ []
OE beorc ‘birch’ Barkham (Berkshire) Berkeley (CA)36
37
OE dēorby ‘deer-by’ Derbyshire derby (AmE)
35
The rhotacisation right-hook on schwa is very similar to the IPA right-hook reversed
epsilon [˘]. We use only the former for the NURSE vowel in PDE AmE.
36
Compare Berkley (Somerset), UK.
37
Compare Darby Lodge (Lincolnshire), UK; Darby, MT, Darby PA in the US.
OE heort ‘hart’ Hertford (Herts) Hertford (NC)

Gmc *sker- shard sherd
OE clerec ~ clerc clerk (BrE) clerk (AmE)
OFr errant arrant errant
OFr mercant Marchand merchant
OFr sergent sergeant
varmint vermin
The lowering and centralisation of retained, reinstated or newly bor-
rowed [-εr] > [-ər]/[] was under way during the seventeenth century.
It seems to have been preceded by a parallel change of the high vowels,
a merger of [-ir] and [-υr] into [-ər] in progress already in the sixteenth
century. In (Southern) EModE words such as fir < OE *fyre ‘fir’, circle
OE < Lat. circul ‘astronomical circle’, further < OE furðra, ME cursor
‘runner’ had the same stressed vowel. The three-way merger of histori-
cal <-ir, -er, -ur> is attested in Shakespeare’s rhymes of the type birds:
herds, curse: first. The merger is illustrated in (17).38
(17) <-ir, -er, -ur> merger in EModE:
Source ME EModE
OE gyrdel ‘girdle’ gir(r)del
AN surgien ‘surgeon’ surge(y)n [-ə(r)-]
OFr certain ‘certain’ certe(i)n
A coda /-r/ is a neutralising environment. For the short vowels neutral-

isation of [i], [υ], [ε] results in merger into the featurally neutral schwa
outcome. The phonetic rationale behind this is the coarticulation of the
short vowel + /-r/ which affects both segments: it lowers and centralises
the vowel, while the adjacent sonorant coda becomes more schwa-like;
its further weakening can eliminate the consonantal cues, leading to
complete loss of the consonantal properties of /r/. The difficulty of
perceiving the distinctive features of pre-/r/ vowels is attributed to
the acoustic similarity of the first two formants of [ɹ], [], and [ə ~ ]
(see 5.4.2). The next step, associated with the vocalisation of /-r/ in the
38
The difference between the original short vowels is preserved in Scots, where [-r] has
not had the same neutralising effect. This follows from the fact that Scots post-vocalic
[-r] is a tongue-tip trill, which is incompatible with simple retroflex coloration of the
preceding vowel. Irish English merges [-ir] and [-υr], but keeps [-εr] distinct (Lass
1987: 259, 264).
Southern varieties, is compensatory lengthening to [

] in SSBE, and
merger with the possibly unshifted [ε
] in, for example, earl (OE eorl),
earth (OE eorðe) (see below).
Note that unlike the [-εr] > [-ar] change, where /-r-/ has the same
effect whether it is tautosyllabic (farm, darling) or ambisyllabic (harry,
arrant), the mergers of the short vowels before /-r/ in (17) occur only if
the consonant is fully tautosyllabic. Thus words like terror (1375), terrace
(1575), serif (1830), mirror (1300), squirrel (1366), pyramid (1398), syrup
(1398), lyric (1581) preserve the quality of their original stressed short
vowels. :
Turning to the influence of /r/ on the preceding long vowel, recall
from 6.4 that some ME spellings indicate sporadic lengthening before
/-rC/ clusters, treated in the literature as part of the lengthening before
homorganic clusters: board (OE bord), hoard (OE hord), earl (OE eorl),
earth (OE eorðe), hearken (OE he(o)rcnian), heart (OE he(o)rte), mourn (OE
murnen). The extent of the lengthening is unclear; as long as /r/ was
realised with some friction, the tendency towards lengthening would be
counteracted by the tendency for long vowels to shorten when followed
by /-CC/ (see 7.5.1.1). The [-εr] > [-ar] change discussed above can be
a diagnostic of length: hearken, heart must have had the short [ε] for it to
be lowered to [a], while in earl, earth the input to EModE [-ə(r)-], PDE
[
] ~ [] could have been either [ε] or [ε
].
It was noted in 8.2.1 that loss of /r/ was one of the sources for the
introduction of ‘new’ vowels, [ɑ
] and [ɒ
] ~ [ɔ
], in PDE. In the long-
vowel subsystem the most interesting effect of a following /-r/ is the
rise of a new set of diphthongs in conjunction with the loss of /-r/ in the
non-rhotic varieties. The dating of /r/-loss (see 5.2.3) varies depending
on the social background of the data source. Dobson (1968: §332) found
that ‘only sources which reflect vulgar pronunciation give evidence
of the change before 1700, and even they give little’. The switch from
‘vulgar’ to accepted in ‘good’ speech took at least a century. In stressed
syllables ‘the change could not have occurred in good speech before c.
1800’ (Dobson 1968: §427). :
The loss of /-r/ after long vowels follows upon two analytically
separable processes, schwa-insertion (breaking) and pre-schwa-laxing
(Wells 1982, I: 214–16), though the exact timeline is hard to ascertain.
Bearing in mind that the chronology and the exact phonetic nature of
these transitional stages are hard to reconstruct, we can offer a schema
of the effect of coda /-r/ on the long vowels in EModE as in (18).39
39
See Lass (1999: 108–12) for a more detailed account of the dating of the transitional
stages with reference to the orthoepistic testimony.
(18) The effect of /-r/ loss on long vowels in non-rhotic Englishes:40

Late ME Schwa- Length /-r/ Examples
to EModE insertion adjustment -loss
-əir ~ -ir -aiər -aiər -aiə fire (OE fȳr)
-e
r ~ -i
r -i
ər -iər -ıə deer (OE dēor)

-ε
r ~ -e
r ~ -i
r -i
ər -iər -ıə ear (OE ēare)
-ε
r ~ -e
r -e
ər -εər -εə pear (OE pere)
-æ
r ~ - ε
r - ε
ər - εə r -εə hare (OE hara)
-ɔ
r -ɔ
(ə)r -ɔ(
)(ə)r -ɔ(
)(ə) oar (OE ār)
-o
r -ɔ
(ə)r -ɔ(
)(ə)r -ɔ(
)(ə) floor (OE flōr)

-o
r ~ -u
r -u
ər -υər -υə poor (AN pover,
pour)
-əur ~ -ur -auər -aυər -aυə bower (OE būr)
Schwa-insertion occurs also in the inherited diphthongs, for example

OE fæġer ‘fair, adj.’, OE leġer ‘lair’, where the palatal glide was lowered
before /-r/, resulting in merger with the new centring diphthongs
from etymological [-ε
r] as in pear, hare. As noted above, an important
consequence of the processes in (18) is the introduction of the new
diphthongs [iə], [εə], [ɔə], [υə] in the non-rhotic varieties of English.
Another observation about the data in (18) is that as with the short
vowels in (17), the presence of /-r/ in the coda can have a strong neu-
tralising effect, evidenced by the merger in the etymologically distinct
pairs deer-ear, pear-hare, oar-floor (and currently poor, pore in some varie-
ties). The mergers affect mostly the long mid vowels and they are not
carried through with consistency. While historical long [e
] as in deer
(OE dēor) generally participates in the raising of [e
] to [e
] ~ [i
] (see
8.2.2), <ea> spellings as in hear (OE Angl. hēran), dear (OE dēore) suggest
that the vowel was identified with historically lower-mid front [ε
] as
in ear (OE ēare). However, ME [-ε
r] can either emerge as the raised
[e
r] ~ [i
r] as in ear, or the /r/ acts as a deterrent to raising, so that the
outcome is [-εə], which merges with the reflex of ME [-æ
r] ~ [-ε
r] as
in hare (OE hara). Merger occurs with the mid back vowels too, produc-
ing homophones from historical [-ɔ
r] as in oar (OE ār) and [-o
r] as in
floor (OE flōr). Once again, the /-r/ may or may not inhibit the raising
of the high mid back vowel, so that an input [-o
r] can emerge as either
40
The glides [j] and [w] used for the second part of the reflexes of historical [i
] and [u
]
are replaced by [i] and [u] on the assumption of some positional lengthening before
/r/. The exact history of [j]-[i]-[i] and [w]-[υ]-[u] in this context needs further
clarification.
[ɔ
ə] as in floor or as [υə] as in poor. Going back to the key pairs in the
heading of this section, we can describe, if not explain, the difference
between tear, v. - tear, n. and floor-poor: in tear, v. (OE teren) and floor ME
[-ε
r] and ME [-o
r] remained at their original height, while in tear, n.
(OE tēar) and poor the same vowels participated in the long vowel shift.
These pre-schwa-insertion patterns are shown in (19).
(19) TEAR,v. - TEAR, n. and FLOOR-POOR:
tear, v. tear, n. floor poor
ME [-ε
-] ME [-ε
-] ME [o
] ME [o
]
Effect of [-r] [-ε-] --- [o
] ---
Long vowel shift --- [-i
-] --- [-u
-]
The possible articulatory basis of /-r/ vocalisation (see 5.2.4) is a sepa-
ration of the gestural components of constricted /r/. Phonetically, the
change from a rhotic to a non-rhotic state is from a bi-gestural [ɹ], to
bi-gestural [əɹ], to uni-gestural [ə]. An important contributing factor
is also the perceptual similarity between the sonorant /r/ to schwa.
In the non-rhotic varieties of English, the transition from [-əɹ] to [-ə]
is carried through, while rhotic varieties preserve bi-gestural [-əɹ],
usually rendered as []. It should also be added that the effect of /r/ is
not the same in all rhotic varieties. In SSE a coda /-r/ does not involve
schwa-insertion, thus bee and beer both have [i
], similarly hay and hair
both have [e
] (McMahon 2000: 232).
While loss of /r/ may be described as ‘natural’ in a phonetic sense,
it is still unclear why some communities of speakers preserved it when
others did not. One reason why the change may have taken off in the first
place, not usually considered in the textbook accounts, is loan phonol-
ogy. In Later Old French (eleventh to fourteenth century) and Middle
French (fourteenth to sixteenth century), pre-consonantal [r] was
assimilated to the following consonant and thereby lost in the spoken
language (it is retained in spelling to this day), producing rhymes such
as sage : large, fors : clos, ferme : meesme. Thus English orthography was
at odds with the functional factor of ease of articulation and with the
possibly prestigious pronunciation of recent loanwords in which pre-
consonantal [-r] had been lost. This may account for the considerable
lag time for the diffusion and codification of [r]-loss in early Modern
English. Rhotic and non-rhotic pronunciations must have coexisted for
over three centuries, even in the same dialects. As noted above, conserv-
atism based on spelling maintained rhoticity in the Southern standard
until just after the first quarter of the nineteenth century. :
A historical note on the spread of rhoticity: the main rhotic varieties
are spoken in Scotland, Ireland, South-West England, extreme West
Lancashire, most of the USA and Canada (see 5.2.4). The development
of North American rhotic accents is to a considerable extent a function
of settlement patterns in colonial times. Though the original settlers on
the East Coast were mostly from the south and east of England, which
are now non-rhotic, they arrived in the Colonies long before ‘drop-
ping the -r’ had become fashionable and codified there. The problem
is not explaining where American rhoticity came from, – it came from
Southern and East Midland England originally, and later waves of set-
tlers from Scotland and Ireland vigorously reinforced it – but rather why
non-rhoticity shows up anywhere in America. Since it only shows up in
centres of education along the East Coast, not including Philadelphia,
which has always been, and continues to be, rhotic, it may be that in
the major political and trading centres, namely Boston, New York,
Norfolk, Savannah, Alexandria and Charleston, the newly fashionable
British r-less accent took root and was reinforced by regular travel to
Britain. The Civil War is a watershed in the perception of rhoticity as
prestigious: prior to it r-lessness was associated with the Bostonian and
Virginian elite, but after 1870 New York was increasingly rhotic, the
British model was much less important and rhoticity gained prestige. :
8.4 Continuity or reinvention

The survey of the segmental histories of consonants and vowels in
earlier English revealed many parallels with ongoing changes in the
varieties of English spoken today. There is nothing surprising about
history repeating itself – language change is inherent in language use,
and language users are guided by the same principles of speech pro-
duction and perception. Before we move on to the description of stress
placement in the history of English, we go briefly over a topic which
illustrates the way in which one phonological pattern can straddle
historical and ongoing phonological change.
8.4.1 Canadian or long-vowel ‘Raising’: ICE [ΛIs] - EYES [aIz] ~ [aεz] ~

[aə] ~ [a
]41
In Canadian English (CE) the PRICE vowel has two realisations depend-
ing on the voicing of the coda consonant: [i] before voiceless codas,
as in price, right (referred to as ‘Canadian Raising’), and [ai] before
41
This section will focus only on the [i]-[ai] alteration, though the pattern is observed
also with [υ]-[aυ]. For details on the research history, the geographical range and
updates on the various findings see Chambers (2006).
voiced codas, as in prize, ride. As the reconstruction of the long-vowel

changes in (13) shows, [i] was an intermediate value in the history of
inherited [i
]. The logical question, then, is whether CE is conserva-
tive or innovative. Moreton and Thomas (2007) argue that the exist-
ing alternations associated with the so-called Canadian Raising ‘have
been independently re-innovated and show a systematic typology: The
voiceless environment selects the higher allophone.’ They focus on the
asymmetry between the first element and the offglide in response to
voicing: diphthongs assimilate to the first element before voiced codas,
hence [aiz] ~ [aεz] ~ [a] ~ [a
], and to their offglides before voiceless
ones, hence [is]. They find a modern parallel of the effect of the coda
consonant in CE on this vowel and the ongoing /ai/-alternation in the
Cleveland area in Ohio.42 This leads them to dismiss the hypothesis that
the alternations are a ‘more likely retentions of earlier values, related to
similar qualities in Scots and Hiberno English’ (Lass 1987: 285); recall
that the BITE vowel in southern Hiberno English is [i], in Standard
Scottish English [i]43 and in Mid Ulster English [əi]. Lass further
points out that [ə-] and [-] as first elements of the diphthong based
on ME [i
] were ‘normal’ in SSBE in the seventeenth and eighteenth
centuries (see 8.2.2.2), and the fully open [a-] was a later innovation.
In that scenario, the alternation is inherited and should be labelled a
non-lowering ‘archaism’ rather than ‘raising’.
Moreton and Thomas’s point that the innovations observed in PDE
that appear to replicate the ‘Canadian Raising’ and the phenomenon
in CE must be related by common phonetic motivation is beyond
dispute. In their account, that motivation is the tendency of diphthongs
to be dominated by the offglide before voiceless codas, and by the first
element of the diphthongs elsewhere. No matter what the motivation,
however, we have to assume that the same linguistic factors can be at
work at different times and places, and therefore dismissing the pos-
sibility that the [i]-[ai] alternation in CE is ancestral is not a logical
42
Further ‘Canadian Raising’ features are found in Rochester, New York, with extrapo-
lations to the Great Lakes basin on the American side, Ann Arbor, Michigan, Detroit,
Newcastle-on-Tyne and in the Fens of eastern England (see Chambers 2006 for full
references).
43
The literature on the issue of [i] vs [a*e], which is a sub-part of a more general
‘Scottish Vowel Length Rule’, or ‘Aitken’s Law’ is very rich, starting famously with
Aitken (1962). It refers to the process whereby in Scots before /r/, voiced fricatives
and before major boundaries there is also lengthening and lowering:
[i]: tight, fine, time, Fife, mice; tide
[a*e]: tie, fire, five, rise; tied
Note the exclusion of voiced stops from the lengthening environment.
necessity. Parallel, and innovative, developments in US English are

phonetically based, and so can be the ‘innovative’ lowering of the [i] >
[ai] before voiced codas in CE. It is important to recognise that ‘both
/ au/- and /ai/-raising were well established in Ontario among people
born as early as 1860’ (Chambers 2006: 111). Surveys of the settlement
history of Canada (Finegan 2006: 414) identify the Irish and the Scots as
the largest immigrant groups in Canada in the 1830s and 1840s, peaking
in 1851–61. Another wave of Scottish immigration occurred between
1901 and 1911. Evaluating the contribution of these demographic
factors to the formation of the CE vowel system against the likelihood
of independent innovation is far from straightforward. It certainly
allows for a reversal of the direction of change – from raised realisations
in all environments to lowered realisations before voiced codas. The
source, or more likely sources, of ‘Canadian Raising’ cannot therefore
be considered fully determined.
This brief excursion into the roots of a well-known regional char-
acteristic concludes the survey of the segmental histories of English
sounds. The next two chapters will address patterns of stress and the
early history of English versification in relation to phonological history.
: See further reading and appendices on Companion website.

9 The evolution of the English
stress system1
ALWAYS-CAUSEWAYS, PRÉSENT, N. - PRESÉNT, V. HARÁSS ~ HÁRASS
This chapter shifts the focus from segmental histories to the history of
word- and phrasal stress in English. We first define some terms used in
the description of the prosodic patterns of speech and revisit (see 2.3)
the ways in which syllable structure and syllable weight interact and
influence the assignment of stress. A brief sketch of the patterns of stress
assignment in PDE is followed by notes on the methodology of pro-
sodic reconstruction. Sections 9.4–9.6 turn to the description of Old and
Middle English word- and phrasal stress. The last two sections discuss
the effect of lexical borrowing from French and Latin on the prosody
of English and some interesting prosodic changes in post-Renaissance
English.
9.1 Preliminaries: definition of some terms

It is customary in linguistics to approach phonological issues from two
angles: the study of individual segments, or segmental phonology, or
the study of prosodic structures. The term ‘prosody’, as used here,
refers to the ‘suprasegmental’ domain of linguistic description and anal-
ysis, attending to the properties of individual segments only if relevant
to the structure and function of higher-level units, such as syllables.
Very importantly, syllables are the carriers of ‘stress’, the contrastive
intensity that marks some syllables as more or less prominent. The term
‘accent’ can also be used with reference to syllable prominence, so an
‘unaccented syllable’ and an ‘unstressed syllable’ can be synonymous. :
The goal of prosodic description is to identify the properties and the
organisation of syllables into words, phrases and utterances in speech.
This area of inquiry is also referred to as ‘metrical phonology’. In order
to avoid confusion between the ‘metrical’ properties of speech and the
properties of poetic meter, we will use ‘prosody’ for the structures in
1
Parts of this chapter are based on Minkova (2012) and Minkova (2013).
284
THE EVOLUTION OF THE ENGLISH STRESS SYSTEM 285
‘ordinary’ spoken language. For the structures found in verse, we will

use ‘poetic meter’, or simply ‘meter’. Both prosody and meter refer to
rhythm, the regularity of recurrence of stresses, which can interact with
the prosodic realisation of words and phrases. To illustrate: in PDE
rhythmic well-formedness accounts for the lack of word-internal three
unstressed syllables in a row, known as the ‘lapse’ constraint. Rhythm is
involved in stress-clash avoidance, as in ‘The Arab Rule’ (Ross 1972),
whereby a secondary stress in PDE is demoted after a light, main-
stressed syllable – the case of the alternation in Arab: [ei$ɹæb] vs [æɹəb],
or in the lack of secondary stress in [-riz-] in charismatic, though the
syllable is heavy and the input charisma is stressed on the penultimate
syllable. Rhythm is also instrumental in versification, where the distance
between prominent metrical positions tends to be tightly regulated.
Prosody is also the linguistic bridge between phonology and mor-
phology – linguistic units such as words and phrases have prosodic
structure which interacts with the unit’s morphosyntactic status. To
illustrate: présent, n. - presént, v., sóft pòrn, compound - sòft táp, phrase.
Stress is one of the main cues for speech segmentation into individual
words: a speaker of English will associate stress with the left edge of a
word, hence the familiar joke Be alert – your country needs lerts. The inter-
action between rhythmic constraints and the word-delimiting function
of stress in English is seen in the absence of words that have more than
a single fully unstressed syllable word-initially: saguáro [səgwɑroυ] is
fine, but saguarésque *[səgwərεsk] must have secondary stress on the
initial syllable, Japan [dəpæn] is fine, but Japanése has to have initial
secondary stress: [$dæpəni
z]; Magéllan [mədεlən], but Magellánic
[$mædəlænik]. :
Outside of linguistics, the term ‘prosody’ can also be used with refer-
ence to the study of verse and its properties; again, for the conventional-
ised rhythmic structures of verse we reserve the term ‘meter’, the topic
of Chapter 10.
Individual speech sounds are the building blocks of prosodic units
that are independently pronounceable, namely syllables.2 One impor-
tant property of syllables is that they have prosodic weight. The contri-
bution of an individual sound to the prosodic weight of a syllable can
be measured in terms of moras: recall from 2.3.3 that all short vowels
in English are associated with a single mora, all long vowels and
2
The unity and independence of the syllable in speech has been recognised for a very
long time. Ælfric’s pedagogically oriented Grammar (995) offers a definition which can
still be used today: ‘SYLLABA is stæfgefēg on anre orðunge geendod’ [A syllable is a
stave/letter-conjunction completed in one breath].
diphthongs are bimoraic, and consonants are moraic or non-moraic

depending on their position in the syllable. The ability to divide an
utterance into syllables, and, for English, the ability to perceive syllable
weight, is part of the intuitive knowledge that speakers have of their
language.
In addition to being the smallest independently pronounceable
units of speech, English syllables are the domains of stress-placement.
Phonetically, stress is associated with the use of increased respiratory
energy, increased tension of the vocal folds and loudness. Though
stress is technically not a phonological feature, it can be described
as binary in the sense that syllables are either [+ stress] or [- stress].
However, within the domain of an English word there may be more
than one syllable designated as [+ stress], and if so, those syllables will
have different levels of stress. The most prominent syllable in a word
carries primary or main stress, marked either with a superior vertical stroke
[] before the stressed syllable (IPA), or, informally in orthographic
forms, with an acute accent, ´ over the vowel, so [lætin] is the tran-
scription of Látin. Secondary and tertiary stresses, most common in
our loan vocabulary, here bundled together here as non-primary stress,
are marked either with the IPA inferior vertical stroke [$], or with a grave
accent, `, in orthographic forms, so [$tεrədæktəl] is the transcription of
ptèrodáctyl. Although informally we speak of ‘stressed’ and ‘unstressed’
vowels, and place the acute and grave diacritics over the vowel-letters
for typographic convenience, it is important to bear in mind that stress
is a property of the entire syllable.3
The relationship between vowel reduction and stress was discussed
in 2.3.3 and 7.6; in PDE lack of stress can be recognised by the presence
of unstressable peaks: [ə, n$ , m
$ ]; some degree of stress is always associ-
ated with [ei, ε, æ, ɑ, ɔ, υ, u
, ] peaks, and the remaining peaks can
appear in both stressed and unstressed syllables.4
9.2 Syllable structure and syllable weight

Recall from 2.3.2 and 2.3.3 that the syllable is a structurally complex
unit in that it is further decomposable. At the core of the syllable is its
3
Dictionaries vary in stress notation; stress diacritics can be placed before or after the
stressed syllable, thus for poster one finds OED: [poυstər] (IPA), The American Heritage
Dictionary: [pō ́stər], similarly Chambers 21st Century Dictionary. Alternative ways of
orthographic indication of stress are using bold (Dictionary.com. [poh-ster]), capitalis-
ing a particular syllable or word (POster, Was it THIS level we parked on?) or italicising.
4
This taxonomy follows Hayes (1995: 15). The OED transcription system allows many
of the ‘stressed only’ peaks to appear in unstressed syllables.
‘nucleus’ or ‘peak’, the segment of highest sonority in the string. The

nucleus is the only obligatory component of the syllable – it is usually
filled by a vowel or a diphthong, but sometimes also by a syllabic
sonorant /r$, l$, m
$ , n$ /. Consonants or consonant clusters to the left of
the nucleus constitute the syllable onset, and the consonants following
the nucleus make up the coda. The onset and the coda can be empty
(indefinite article a), filled by a single consonant (be, it, bit) or filled by a
consonant cluster (stay, east, crest). Together, the peak and coda, if there
is one, constitute the syllable rhyme. Universally, there is an asymmetry
between the onset and the coda: a filled onset is preferred to a filled
coda. The preference is known as Onset Maximalism, or the Maximal
Onset Principle.
Syllables vary in weight depending on the segment in the peak and
the composition of the syllable. Syllables with rhymes consisting only of
a short vowel V are light. All other syllables are heavy, but not equally
so: the higher the weight of the syllable, the more likely it is that it will
attract stress. Historically, syllables that carry stress are likely to become
heavy (see 7.5).
The usual binary division into light (V) and heavy syllables is useful,
but it is also insufficiently predictive. It has always been recognised that
heavy syllables attract stress; some studies of English have also recog-
nised -VVCC(C) syllables as superheavy (Lass 1992a, 1994). Current
studies show additional significant contrasts among the subtypes of
heavy syllables. As established by Ryan (2011a), the probability of a syl-
lable being stressed in English increases from a light syllable V (lowest
probability) to VC to VV to VVC (highest probability). Moreover, the
different weights are not distributed evenly on the continuum: the VC
to VV contrast is over twice as great in magnitude as the VV to VVC
contrast. Independently, the effect of coda size also increases from a
single consonant C to CC to CCC, where every contrast is statistically
significant.
English onsets have traditionally been considered weight-neutral.
While this has been the position in the handbooks, recent research
reveals some fine-grained distinctions between syllabic onsets and
stress. Kelly (2004), reports on tests both on existing lexicon and on
nonce/wug words which show a strong correlation between onset com-
plexity and stress attraction in English. His results are corroborated in
Ryan (2011a: 172–90).
In addition to syllable structure, the morphological structure of
a word can determine stress placement. Root stress is commonly
preserved after the addition of affixes: bóther-bóthersome, unbóthered;
válue-váluate, inváluable. Some borrowed affixes have special properties,
however: they can be always unstressed, always stressed or they can

attract stress on a particular syllable of the word. The effect of affixes
on stress may also be unstable, depending on conflicting factors, for
example -ic will normally attract stress to the immediately preceding
syllable, as in patriótic, symbólic, galáctic, but not in cátholic, lúnatic, rhétoric.
The morphosyntactic nature of a word – nominal vs verbal – can also
be involved in determining the stress placement, as in pairs of the type
áddict-addíct, présent-presént.
The description of the PDE prosodic system requires reference
to all of these factors, and even then there are numerous exceptions.
Accounting for English stress is a challenging undertaking, beyond the
remit of this volume. The goal here is to present the facts, to the extent
that they are recoverable, of the evolution of the prosodic system.
The description will be limited to word- and phrasal stress – the only
domains for which there is testable historical evidence. :
9.3 Historical sources of information for prosodic reconstruction

Without the help of speech-recording and -analysing technology,
without contemporary commentary, and without the visual props of
italics, bold or capitalisation in the older texts, any prosodic information
extracted from the surviving manuscripts and early printed materials
is by definition secondary. Among the important sources that can lead
to prosodic inferences are: typological comparisons, scribal evidence
for stress-sensitive segmental change, verse evidence and, after the
sixteenth century, contemporary descriptions.
English is a member of the Germanic language group. In all early
dialects of Germanic stress became fixed on the first syllable of the root
(see 3.4.3). The function of stress as a word-boundary signal continues
to be an important prosodic characteristic of all Germanic languages
today; it is therefore safe to start with the assumption that OE under-
ived words shared that pattern. Words inherited from OE – áfter, hállow,
blóssom, abróad, becóme – follow the Germanic Stress Rule (GSR). With
respect to OE, the validity of the GSR is uncontroversial, and if the
vocabulary consisted solely of underived words of Germanic descent,
the typological evidence would have been all that one needed to recon-
struct OE word-stress. This is obviously not the case – the OE deriva-
tional morphology is very rich and we have to account for the stress
of derived items, compounds and the higher-level prosodic contours
of phrases. Moreover, non-Germanic lexical items are well attested
already in OE, and for them typological comparisons are impossible or
unreliable.
9.3.1 Orthographic evidence for word-stress

For morphologically complex and borrowed words additional informa-
tion from spelling can help in the recovery of prosodic patterns in the
earlier stages of English. For the Germanic vocabulary, vowel reduction
and loss are expected in unstressed syllables. The spelling evidence for
that is ample in post-tonic syllables (see 6.5.4); similarly, the reduction
and loss of the prefix ge-: OE geriden > ME iriden > ‘ridden’. Alternative
spellings for the preposition for in OE are <for, far, fur, fær, fer, fr, f’>.
Such spellings support the assumption that ge- and for were unstressed
in OE.
The situation with non-Germanic words is more complex. The pro-
portion of loans in OE was low compared with PDE, yet the records
show close to a thousand borrowings from Latin, or Greek through
Latin, in Old English, about 3 per cent of the surviving vocabulary.
In the disyllabic and many of the trisyllabic loanwords the original
stress was serendipitously the same as the stress assigned by the GSR:
cánker, cránic ‘chronicle’, círcle, fénix, Jácob, Lúcifer, mártyr, órgan. With
some trisyllabic words spelling is a window into their adaptation to the
native pattern: the variable spelling of the second and third syllable of,
for example, Lat. cucúlla ‘cowl’ in OE: <cugele, cuele, cule, cuhle>,
suggests that the original penultimate stress was ignored in OE. The
evidence is not ample, but orthographic forms attesting lack of stress on
originally stressed syllables do exist, as in (1). :
(1) Scribal evidence for early stress-shifting in Old English loanwords:
Latin OE Gloss
abbā́dem <abud ~ abod ~ abed> ‘abbot’
altā́re <alter ~ altre> ‘altar’
acḗtum <eced ~ ecid> ‘vinegar’
coquī́na <cycene> ‘kitchen’
mortā́rium <mortere> ‘mortar’
sextā́rius <sester ~ seoxter> ‘a measure’
Such spellings can be compared with the more stable spelling of proper
names such as Abraham, Augustus, Babylon, Philippus, Saturnus, whose
lower frequency would be a factor in the slower rate of assimilation to
the native stress pattern.
Scribal evidence for stress-placement in ME loans is of the same type:
variable spellings or spellings with <e> of original long vowels in the
penultimate syllables, as well as loss of unstressed syllables. One can
infer that the word corúne ‘crown’ (OFr corone, corune, Lat. corōna) main-
tained stress on the penultimate syllable, as in Latin, because already
Orm (c. 1200) wrote <cruness> ‘crowns’; the initial syllable could be
syncopated only if it was unstressed.5
(2) Scribal evidence for stress-shifting in Middle English loanwords:
Loanword ME Gloss
AN caboge <cabache ~ caboch ~ cabish ~ cabech ~ cabush> ‘cabbage’
Lat. formālis <formal(l) ~ formel> ‘formal’
Lat. olīva <oleu(e) ~ olefe> ‘olive’
Lat. pictūra <picter ~ pictor ~ pictar ~ pictre> ‘picture’
The closer we get to PDE, the less likely it is that a new loanword will
be subject to orthographic and/or phonological adaptation: karakul
(1853), baccara(t) (1866), autogony (1870), taiga (1888) preserve the
original vowel letters in unstressed syllables in English.
The types of segmental changes in stressed and unstressed syllables
are very different. Shifting the focus from unstressed to stressed sylla-
bles, the orthographic records of processes such as vowel lengthening,
vowel shifting and consonant gemination are also useful, albeit self-
evident. The shifted vowels in PDE in silence (1225, < OFr silence, Lat.
silentium), libel (1297, < OFr libelle (fem.), Lat. libellus), mountain (1275, <
AN mountaine ~ muntaine, Lat. montāna) suggest that the initial stress in
these words was in place prior to the long vowel shift (see 8.2.2.1), unlike
loans such as machine (1545), tambourine (1579), toucan (1568), boulevard
(1769).
Stress patterns above the domain of the simplex word can also
be inferred from synchronic scribal variation, as with the shift from
phrase > compound > obscure compound > simplex word (see 9.5).
9.3.2 Verse evidence for stress

The deployment of words and phrases in verse is an important source
of information regarding the history of stress, yet verse texts are also
notoriously controversial as an evidential basis for phonological recon-
struction. The greatest challenge for the researcher is circularity: there
are no contemporaneous records of ‘rules’ governing early versification,
so we have to rely on templates extrapolated from the surviving poetic
corpus. The range of allowable templatic variation is founded on statis-
tical probabilities, which lead to inferences about the interplay between
language and meter with no possibility of speaker verification. The
way to avoid explaining the unknown by means of the more unknown
5
LAEME records only thirteen tokens of <cor-, cur-> against 109 tokens of <cr-> for
both the verb and the noun.
is by applying rigorous quantitative tests including comparisons with

non-verse material.
Parametrical verse features, such as rhyme and alliteration, are also
central, but must be used with caution. The data on OE alliteration, for
example, are strikingly consistent: in 26,088 verses of OE poetry, only
thirty-six, or 0.001 per cent, lack alliteration (Hutcheson 1995: 169).
The onsets of testably unstressed syllables, such as inflectional syllables,
never alliterate; therefore, we can use alliteration as a reliable indicator
of stress. No matter what theory of OE meter one adopts, there can be
no doubt that gefrǽtwade, fóldan, grímma, Gréndel, gerúmlicor and rǽste in
(3) have the main stress on the first root syllable (alliterating onsets are
in bold).
(3) Alliteration and main stress in OE verse:
ond gefrǽ twade / fóldan sceatas Beowulf 96
‘and adorned / earth’s regions’
Wæs se grímma gæst / Gréndel haten Beowulf 102
‘was the grim ghost / Grendel called’
gerúmlicor / rǽste sohte Beowulf 139
‘in roomier space / rest sought’
On the other hand, the very fact that alliteration as a metrical device
that keeps the two halves of the line together was so prominent in the
Germanic tradition suggests the possibility that composers and copiers
of verse could occasionally resort to alliteration on syllables that would
normally have lower prominence than the neighbouring syllables, as
in (4).
(4) Alliteration on non-primary stress in OE:
on þ&¯m dæge / þysses līfes Beowulf 1976
‘on that day / of this life’
Ðȳs dōgor þū / geþyld hafa Beowulf 1395
‘On this day you / patience have’
Elsewhere in the OE verse corpus determiners do not alliterate, so we
can assume that, as in PDE, the need for special emphasis in speech and
in verse can override other factors such as normal right-prominence in
the determiner-noun phrases ‘this day’, ‘this life’.
Violations in matching main stress to alliteration in the classical
6
Compare the treatment of the demonstrative in line 197 with the Beowulf line 1216:
Bruc ðisses beages, / Bēowulf lēofa
‘Enjoy this ring, / dear Beowulf’
OE poetic corpus are infrequent. The situation changed in ME: the

composers of alliterative verse became more liberal with the use of
visual alliteration, both because of the gradual shift from oral to written
composition and because of the greater density of non-initially stressed
words. The extent to which ME alliterative verse can be trusted with
respect to stress on prefixed native and borrowed words is contro-
versial: initial alliteration on native words as in behold or forsake is as
suspicious as initial alliteration on enthrone, rejoice, remove. While the
most common interpretation of such instances is that they are based on
orthography rather than on stress, there are arguments in the literature
that with some prefixes the alliterative practice reflects acceptable ME
pronunciations. :
In ME the use of end-rhyme became the dominant mode of keeping
verse lines together. Here again the requirement of prosodic identity
can be useful for the recovery of the stress-pattern. As shown in (5),
Chaucer’s rhyme practice indicates that the loanword philosóphre ‘phi-
losopher’ (c. 1330) had penultimate stress, which is confirmed by the use
of the same word line-medially, as in (6).
(5) Rhyme evidence for stress in ME end-rhymed verse:7
But al be that he was a philosóphre,
Yet hadde he but litel gold in cófre; GP 279–98:
In name of crist, to wexe a philosóphre.
Ther been ful fewe to whiche I wolde prófre CYT 1122–3
(6) Metrical evidence for stress supporting rhyme evidence in ME:
Was thér no phílosóphre in ál thy tóun? MLT 310
This wíse phílosóphre, thús seyde hée. PardT 620
A cautionary note is needed in this case too: because end-rhyme in this
tradition marks the line end, and because it keeps larger structures such
as couplets and stanzas together, ‘liberties’ can be taken with the actual
prosodic contour of a word: verse-specific promotions of unstressed syl-
lables occur commonly in ME even for native items whose stress in the
contemporary spoken language cannot be assumed to have been any-
thing but root-initial. Compare the contrasting prominence contours
in sweryng ‘swearing’ in lines 635 and 638 with line 643 in Chaucer’s
Pardoner’s Tale.
(7) End-rhyme as non-evidence for ME stress:
Of swéryng séith the hóoly jéremýe, 635
7
All Chaucerian citations and abbreviations are from The Riverside Chaucer (Benson
1987).
Thou shalt swere sooth thyne othes, and nat lye,

And swere in doom, and eek in rightwisnesse;
But ýdel swéryng ís a cursednesse.
...
Lo, rather be forbedeth swich sweryng
Than homycide or many a cursed thyng; PardT 635–44
Similarly, the rhymes felawe: ydrawe (GP 395–6); biddyng: thing (CYT
1144) are not good evidence that fellow or bidding had final stress in
speech. Compare the use of the same items in (8), where they are clearly
initially stressed.
(8) Mid-line metrical evidence for stress in ME verse:
Lat évery félawe télle his tále abóute KnT 890
His félawe wénte and sóughte hym dóun in hélle KnT 1200
Ay bíddynge ín hire órisóns ful fáste SNT 140
The literature on stress reconstruction in English often uses the evi-
dence of ME rhymes as a categorical indicator of the contemporary
realisation of stress in the spoken language. However, rhyme evidence
has to be evaluated carefully and compared with the evidence of the
placement of such words in line-medial position, where the matching
of prosodic to metrical prominences is much more reliable. The metri-
cal constraints on early English verse are covered in greater detail in
Chapter 10; here one should note that the unreliability of end-rhyme
as a source of stress reconstruction is paralleled by the positioning of
words at the left edge of the line, where prominence inversions (placing,
for example, réady, únder, máketh in the first iambic foot (W S) can be a
deliberate stratagem to avoid the monotony of repeated identical struc-
tures – it is not prima facie evidence of final stress in these native words).
On the other hand, if confirmed by mid-line behaviour, items such as
citées, justiće, poynáunt at the left edge of the early pentameter line are
good evidence for retention of the final stress. Decisions on the prosodic
history of loanwords have to be based on fine-grained and comprehen-
sive coverage of all data – another area in the history of the language
that remains underexplored.
This brief discussion of the methodology of stress reconstruction
focused on Old and Middle English sources. The task of tracing stress
patterns in early Modern English is facilitated by the appearance of
contemporary lists of rhymes. The first more systematic such record is
Peter Levins’s Manipulus Vocabulorum: A Rhyming Dictionary of the English
Language (1570). : Later attempts culminated in the descriptions of
the patterns found in Walker’s monumental dictionary (1791–1826),
including the first extensive comments on secondary stress (ibid.: 73–4).

There is much more to be discovered about the situation after the eight-
eenth century: stress alternations have continued to occur and variabil-
ity of stress is part of the complexity of the PDE prosodic system, but
at least for the last two centuries there are reliable first-hand records in
pronouncing dictionaries and, in the twenty-first century, practically
unlimited web-based opportunities for research on current prosodic
patterns.
9.4 OE stress placement

The typology of stress-assignment recognises languages that have free
or fixed word-stress. Further, stress can be determined by the morpho-
logical or by the phonological properties of the host syllable. Stress can
also be regulated by the proximity to a boundary or to another stress
(Hayes 1995: 31–2). Recall from 3.4.3 that Germanic stress became
fixed on the first syllable of the root; OE therefore is characterised as a
language with a word-bounded, morphologically assigned fixed stress,
though as will be seen below, the blurry distinction between prefix
and root could obscures the predictability and the ‘fixed’ nature of the
stress. In other systems, most relevantly Latin, it is syllable weight in
relation to the word-edge that governs stress-distribution. The types
of stress assignment are rarely found in ‘pure’ form. It is quite common
for morphological and prosodic factors to interact, and indeed English
has never belonged exclusively to one type. The rest of this chapter will
survey the interaction of the principles of stress placement in earlier
English.
9.4.1 OE word-stress
In OE the Germanic Stress Rule (GSR) (see 3.4.3) results in native words
having main stress on the first syllable of the root, leaving grammatical
affixes completely unstressed, and derivational affixes either unstressed
or weakly stressed. The stability of the primary stress on the first root
syllable was maintained in the entire derivational set, unlike PDE
where suffix-induced stress-shifts can leave root-initial syllables com-
pletely stressless: ídiot-idiótic, Málta-Maltése, sólid-solídity. (9) illustrates
the fixedness of main stress on OE root-initial syllables.
(9) The maintenance of OE root-initial primary stress in derivatives:

Base Derivatives
béalu ‘bale, evil’ béalufull ‘baleful’, béaluleas ‘innocent’
dḗma ‘judge’ dḗmend ‘God’, dḗmere ‘judge’, dḗmedlic ‘subject to

judgement’
dḗofol ‘devil’ dḗofolscipe ‘devil-ship’, dḗofollic ‘devilish’
eald ‘old’ éaldan ‘to age’, éaldordom ‘power, authority’
fǽder ‘father’ ġefǽdere ‘godmother’, fǽderlice ‘fatherly’
The weight of the root-initial syllable is irrelevant for the placement
of the main stress. Both heavy and light syllables attract stress root-
initially: in dēma [de
.mə], dēofol [deə.v(ə)l], ealdan [æl.dən] in (9) stress
is on a heavy syllable, and in bealu [bæ.lə], fæder [fæ.dəɹ] stress is on a
light syllable. This justifies a characterisation of OE stress-placement
as morphologically governed; however, syllable weight in OE is not
completely irrelevant. :
The addition of suffixes in Old English never affects the primary
prominence. Inflectional suffixes throughout the history of English have
always been unstressed. Some derivational suffixes, on the other hand,
can be assigned non-primary stress and appear in strong metrical posi-
tions (ictus) in verse, though such weakly stressed syllables never carry
the alliteration. It is likely that the prosodic prominence of derivational
suffixes in speech was gradient, ranging from secondary stress to com-
plete lack of stress, depending on phonological factors: syllable weight,
distance from the right edge of the word and distance from the main
stress. The examples in (10) illustrate the variability of stress on suffixes
depending on phonological factors.
(10) Variability of secondary stress on derivational suffixes in OE:
(a) wísdòme héold ‘with wisdom ruled’ Beowulf 1959b
(b) wórd ond wísdom ‘word and wisdom’ Andreas 569a
(a) of cíldhàde ‘from childhood’ Elene 914a
(b) mǽgðhad se mícla ‘maidhood great’ Christ A, B, C 86a in
In the (a) examples -dōm- ‘-dom’ and -hād- ‘-hood’ are word-internal and
have to be matched to strong verse positions, otherwise the verses will
have only three metrical positions – adjacent unstressed syllables in OE
verse fill a single weak metrical position (see 10.2.1). In the (b) examples
matching the same suffixes to weak positions results in scansions aligning
the verses with frequent types in the corpus: the pattern S w S w of wórd
ond wísdom (Andreas 569a) is attested very robustly, 16 per cent of the
entire corpus, while a potential S w S s (wórd ond wísdòm) is much rarer.8
When followed by an inflectional syllable, -lēas- ‘-less’, -dōm- ‘-dom’,
-fæst- ‘-fast’, -hād- ‘-hood’ are regularly placed in strong positions, but if
8
Hutcheson (1995: 175).
uninflected, they can be in weak positions. All of these suffixes are heavy
syllables, an additional factor facilitating stressability. :
Any full prosodic history of native suffixes has to take into account
rhythmic factors and frequency as well. The two equally productive OE
suffixes -hād and -dōm would be expected to emerge either both with a
full vowel or both with a reduced vowel in EModE. However, in ME
close to 70 per cent of the -dom derivatives followed a monosyllabic root
(earldom, freedom, kingdom, wisdom), where stress-clash avoidance resulted
in de-stressing of the suffix to [-dəm/-dm $ ], while during the same
period 73 per cent of -hood derivatives had a disyllabic stem (bishophood,
maidenhood, womanhood), allowing the preservation of secondary stress on
the suffix and raising of the vowel to [u
] prior to seventeenth-century
shortening to [-υ].9
The morphological status of the suffix is also of consequence. Quite
often in OE it is hard to assign suffixal status to morphological units
which are also attested as independent words: dōm, fæst, full, hād, lēas are
separate lexical entries and their autonomy elsewhere in the vocabulary
could factor in the preservation of stress; compare the divided spelling
of childhood in Elene 336a, 775b in cíldes hàd with Elene 914a: of cíldhàde.
Thus an array of factors: syllable weight, rhythmic preferences and
morphosemantic independence must be considered in the account of
OE suffixal stress.
Like derivational suffixes, prefixes can originate from independent
words. Within the larger family of affixes, suffixes are cross-linguistically
more likely to lose their independent word status than prefixes, and
therefore one would expect more root-like behaviour from prefixes.
Identifying the exact range of prefixes in OE is problematic, because
outside of the clearly prefixal bound forms – æf-, and-, be-, ed-, fær-, for-,
ge-, mis- – it is hard to determine whether forms such as ofer ‘over’, on ‘on’,
wiþ ‘against’, ymb ‘about’ are prefixes or roots. The prosodic treatment
of these forms in OE shows further complexity: in nouns and adjectives
most of the heavy prefixes exhibit root-like prosody, while light prefixes
(ge-, be-) remain unstressed. For verbs and adverbs prefixation does not
affect the main stress, so that we get minimal pairs as in (11). :
(11) OE prefixal stress on nominal vs verbal derivatives:

(a) ðæt ðu óndsware / mid óferhỳgdum Andreas 319
‘That thou an answer / with arrogance’
(b) Oferhógode ða / hringa fengel Beowulf 2345
‘Scorned then / the rings’ ruler’
9
Based on Minkova and Stockwell (2005).
(a) Sumes ýmbhògan, / úngemet gemen. MB 7.28

‘Of some consideration / excessive’
(b) Sylf in þam solere / ond ymbséteð utan Phx 204
‘self in that solarium / and surrounded from outside’
(a) swylce óncỳþðe / éalle gebette, Beowulf 830
‘also distress / all allayed’
(b) feond mid folme; / he onféng hraþe Beowulf 748
‘fiend with open palm / he took hold of quickly’
In the (a)-type examples in (11) the nouns óferhỳgd-um ‘arrogance’
‘ýmbhòga-n ‘consideration’, óncỳþð-e ‘distress’ have the stress contour of
compounds: leftmost primary stress and secondary stress on the first
syllable of the second element. Attached to verbs, the same prefixal ele-
ments in the examples in (b) – oferhógode ‘scorned’, ymbséteð ‘surrounded’,
onféng ‘took hold of’ – are non-alliterating; primary stress and alliteration
is on the first root syllable. Thus both syllable weight and the gram-
matical nature of the base are active in determining the stressability of OE
prefixes.
In compounds the stress on the first element is primary, marking off
the left boundary of the entire word, while the stress on the second root
is non-primary. In the verse, the obligatory alliteration is consistently
placed on the first stressed syllable onset. :
(12) Compound stress and alliteration in OE:
(a) wuldres wealdend, / wóroldàre forgeaf Beowulf 17
‘of the glory ruler / worldly honour gave’
ac he sígewæ̀pnum / forsworen hæfde Beowulf 804
‘but he victory-weapons / forsworn had’
(b) geond wídwègas / wundor sceawian Beowulf 840
‘beyond wide-ways / wonder to examine’
æfter déaðdæ̀ge / dom unlytel, Beowulf 885
‘after death-day / fame un-little’
héardhìcgende / hider wilcuman. Beowulf 394a
‘hard-minded / hither welcome.’
In the verse the second stressed syllable in a compound may alliterate only
if its onset is identical with the onset of the first syllable of the first root, as
in the (b) examples in (12). In the (a) examples the onsets of ār ‘honour’
in wóroldàr-e, and wæpn-um ‘weapon’ in sígewæpn- ̀ um do not alliterate. This
implies that compounds with identical root-initial onsets like wídwèg-as,
10
déaðdæg-̀ e will not appear in the second half-line in OE verse (see 10.2.1).
10
This restriction does not extend to suffixal elements; for example, láðlìce ‘hatefully’ is
Doubly alliterating compounds in OE verse present an interesting

dilemma for stress reconstruction: when positioned in the on-verse they
do not constitute evidence for left-edge prominence within the word.
Their behaviour is compatible with the reconstruction of equal levels of
stress on both elements. Compounding was a favourite stylistic choice
for the OE scops and it is not too far-fetched to assume that the ‘second-
ary’ stresses in these and possibly other compounds were on the level of
primary stresses, especially in recitation and especially for the compounds
in which the first part was disyllabic. In novel compounds both roots could
retain their semantic independence and strong prosodic prominence,
blurring the line between compounds and syntactic phrases. : Creativity
and language play lies behind the fact that many of the self-alliterating
compounds in the OE corpus are hapax legomena, single-instance forms.
(13) Self-alliterating hapax legomena in Beowulf:
béarn-gebỳrdo ‘child-bearing’ héoro-hòcyhte ‘savagely hooked’
éall-ìren ‘all of iron’ hílde-hlæ̀mm ‘battle crash’
fén-frèoðo ‘marsh refuge’ sǽsìð ‘sea voyage’
grýre-gèatwe ‘terrifying armour’ swát-swàðu ‘bloody track’
grýre-gìest ‘terrible visitor’ sýn-snæ̀d ‘huge cut’
héardhìcgend ‘hard-minded’ þéodþrèa ‘people-calamity’
The status of these constructions is an area deserving further inquiry.
Were the various names for the Danes in Beowulf – East(-)Dena, West(-)
Dena, Beorht(-)Dena and (uniquely) Healf(-)Dena – compounds or freely
formed phrases? As argued below, the alliterative tests in this case
are unhelpful, and there is indeterminacy between left-prominence,
right-prominence and equal prominence.11
9.4.2 OE stress above the word level

The prosodic properties of units above the word level in a dead lan-
guage are even more challenging to reconstruct than word-stress.
Morphosyntactic structures larger than the word form their own hier-
archy of prosodic domains: word, clitic group, phonological phrase,
intonational phrase and utterance (Hayes 1989). Attempting to recover
intonational phrase and utterance prosody for OE would be purely
found at the right edge of the off-verse, a further confirmation of the importance of
morphological status in stress assignment.
11
See further Giegerich (2009), who shows that end-stress on noun-noun compounds
in PDE of the type steel bridge, apple pie, Madison Avenue may reflect the syntactic prov-
enance of incompletely lexicalised forms and that nominals of the form attribute-head
can be both lexical and syntactic.
speculative, but some evidence for the historical treatment of clitic

groups and phonological phrases is available.
A clitic group is a combination of a stressed major-class headword and
one or more unstressed function words such as articles, prepositions,
conjunctions and pronouns – the bóok, at a schóol, lóved it, and so on – are
clitic groups. The adjunction of clitics to a host word, known as clitici-
sation, is not obligatory in English; it depends on the syllabic structure
of the function word, and on the strength of the syntactic boundaries in
the sequence. Cliticisation does not occur if there is special discourse
emphasis on a normally unstressed function word. Articles are the most
common types of clitics in English, but even they can be promoted from
clitics to fully stressed items, for example ‘You mean a [ei] gas station or
the [ði
] gas station on Highway 5?’ Cliticisation implies the presence of
a single strong stress in the clitic group.
Although prominence relations in clitic groups are typically not
reflected in our word-based orthographic system, we are familiar with
spellings such as fill’er for fill her, the flight’s for the flight is, a-milking,
a-leaping for historical on milking, on leaping. A good example of an origi-
nal clitic group where the clitic has lost its independent status is the PDE
form gonna (1913), for going to, also earlier Scottish ganna, gaunna (1806,
OED).12 Scribal evidence for clitic group formation is scanty in Old and
Middle English, quite possibly because of the learned nature of the texts
that we have to draw on. One place where clitics and hosts merge is in
the formation of lexicalised adverbs from prepositional phrases: on +
weġ ‘way’ > onwéġ ‘away’; tō + gædre, adv. > togǽdre ‘together’, where the
final stress is testable in the verse and persists to this day.
Another place where cliticisation appears with regularity in OE is in
negative contraction when the host word is vowel-initial, or [w-] initial,
as in (14a). In (14b) continuous spelling also indicates that there is no
internal boundary between the clitics and the hosts. :
(14) Scribal evidence for OE and ME clitic group formation:
(a) OE (b) ME
ne + ealles → nealles ‘not at all’ þe + oþre → þoþre ‘the other’
ne + wāt → nāt ‘not know’ to + eke(n) → teken ‘in addition’
ne + œ̄ nig → nœ̄ nig ‘none’ the + array → tharray ‘the array’
Cliticisation in OE is testable in the verse. Clitics, even when they are
orthographically merged with the host, do not affect the alliteration
patterns.
12
The OED describes gonna not just as ‘colloquial (esp. U.S.)’, but also as ‘vulgar pronun-
ciation of going to’.
(15) OE clitic groups and alliteration:

nealles swæslice / sið alyfed Beowulf 3089
‘not at all in a friendly way / a passage allowed’
in Caines cynne – / þone cwealm gewræc Beowulf 107
‘in Cain’s kin – / the killing avenged’
The next level in the prosodic hierarchy is the phonological phrase
(P-Phrase). In PDE P-Phrases have final prominence, which is ‘a
common but not invariant pattern across languages’ (Hayes 1995: 368).
The most easily testable type of P-Phrase historically is a syntactic
noun phrase, where the material to the left is adjoined to the head noun:
a fine teacher, on the village green, former student. Other common phrases
incorporating material to the left are adjective phrases: strongly sup-
portive, overly cautious; adverb phrases: very energetically, admirably well;
and verb phrases where subjects are adjoined to the left: Mary smiles,
the ship sailed, and complements are adjoined to the right: stop the presses,
speak slowly. The final prominence in PDE phrasal units is known as
the Nuclear Stress Rule, or the End Rule. The End Rule applies only to
unmarked, not contrastive or specially focused phrases. The promi-
nence assigned by the End Rule adds to the prominence of the stressed
syllable of the rightmost word: in fórmer stúdent the first syllable of stúdent
is prosodically stronger than the first syllable of fórmer.
Reconstructing the assignment of phrasal stress in OE is challenging
and many issues relating to phrasal prosody are under-researched and
controversial. Spelling is uninformative, and there are no early gram-
marians commenting on phrasal prosodic contours. The choice and
distribution of alliterating words in the verse remains the only testable
base of hypotheses about the prominence relations on the P-Phrase
level. However, the arrangement of alliterating elements across the OE
alliterative line is not necessarily ‘natural’, in the sense that the conven-
tions of verse composition and recitation may override and thus obscure
the prosodic patterns of speech.
One area of uncertainty is the prosodic behaviour of the inflected
verb within a verb phrase in OE. It is easy to ascertain statistically that
the OE scops treated finite verbs differently from nouns. In a verse
where the strong positions are filled by a noun and a verb, the noun will
consistently carry the alliteration.
(16) Finite verb alliteration in OE verse:
(a) Him ða Scyld gewat / to gescæphwile Beowulf 26
‘then Scyld departed / at the destined time’
Gebad wintra worn, / ær he on weg hwurfe Beowulf 264
‘lived to see of winters many / before he away turned’
(b) þenden wordum weold / wine Scyldinga Beowulf 30

‘when with words ruled / friend of Scyldings’
ne gefeah he þære fæhðe, / ac he hine feor forwræc Beowulf 109
‘not enjoyed he of that hostility / but he him far banished’
(c) Hi hyne þa ætbæron / to brimes faroðe Beowulf 28
‘they him then carried / to briny sea’s current’
The alliterative patterns in (16) show that in both NP-VP strings as in
Beowulf 26a, 264b and VP-NP strings as in Beowulf 264a, the highest met-
rical prominence is on the noun, not on the verb, even if it is phrase-final
as in Beowulf 26a, 264b. This alliterative regularity in (16a) is known as
Sievers’ Rule of Precedence: it states that ‘if an inflected verb precedes a
noun, it does not have to alliterate, that it must not alliterate if the noun
does not alliterate too, and that a non-alliterating noun can never be
followed by an alliterating finite verb’ (Sievers 1893: §§22–9). The rule
does not exclude double alliteration as in Beowulf 30a, 109a (16b). The
only circumstance in which the finite verb carries the structural allitera-
tion (16c) is when all other words in the verse are prosodically weaker,
as in Beowulf 28a: Hi hyne þa ætbæron ‘they him then carried’.
The extent to which the linear conventions of alliteration with
respect to the verb match the prosody of speech is thus hard to deter-
mine, but the fact that clause-initial finite verbs are regularly skipped in
alliteration, while this does not happen to clause-initial nouns, suggests
that verbs were indeed less prominent than nouns. Throughout the
modern Continental West Germanic languages and therefore presum-
ably in PrG, complements are stronger than their verbs, irrespective of
the linear order. This typological comparison and the consistency with
which complement-verb prosodic relations are respected in the verse –
the complement always alliterates – is a good argument for projecting
the complement-strong prosodic contour to OE. By itself, however, the
argument from verse – lack of alliteration on phrase-final finite verbs
if there are other major-class words in the same verse – remains unin-
formative. OE verse-ends and phrase-ends typically coincide, and the
last stress in the a-verse does not have to alliterate and may not alliter-
ate in the b-verse, irrespective of the morphosyntactic category of the
item filling that position.
Prominence in noun and adjective phrases and coordinate phrases is
not directly recoverable from the verse. In this area the linear rules of
alliteration may be more of a handicap than actual help: alliteration on
the first adjective or noun in an OE verse line is obligatory. However,
identifying a metrical convention that privileges linear order of allit-
erative signals for the purpose of keeping the line together, with the
prosodic realisation of the same sequences in speech is unjustified. :

First, in noun and coordinate phrases, the normative syntactic order
overlaps with the linear alliteration on the first stressed word in the
phrase. This takes instances such as lange hwile ‘a long while’ (Beowulf
16a), hond ond rond ‘hand and shield’ (Beowulf 656a) out of the evidential
basis for prosodic reconstruction. Second, projecting the metrical con-
vention onto the prosody is made suspect by the freedom with which
the poet switches components to fit the scheme in the line.
(17) NP modifier-head positioning in OE verse:
Head-final Non-head-final
Geata dryhten (Beowulf 2561b) dryhten Geata (Beowulf 2901a)
‘lord of the Geats’
Leofa Biowulf (Beowulf 2663a) Wiglaf leofa (Beowulf 2745a)
‘Dear Beowulf’ ‘Wiglaf dear’
The unstable internal order of such syntactically equivalent phrases
suggests very strongly that the principle of selection is based not on
prosodic prominence, but on the alliterative needs in the rest of the line.
The absence of double alliteration in the b-verse is clearly metrically
determined because all types of words can be positioned there.13 Some
other facts also prompt scepticism about the link between alliteration
and linguistic prominence: as noted above, phrasal right-prominence
is typologically very common. Moreover, the default contour for noun
phrases and coordinate phrases in the modern Germanic languages is
right-prominent.14 Right-hand prominence is attested also in copula-
tive combinations of the type Anglo-Sáxon, Native Canádian; they also
typically align with syntactically coordinated phrases. The density of
double alliteration in on-verses coextensive with noun + prepositional
phrase (bat under beorge ‘boat under cliff’, Beowulf 211a) and in conjoined
phrases (word ond wísdom ‘word and wisdom’, Andreas 569a) exceeds by
far the overall 47 per cent ratio of double alliteration in the on-verse,
as reported in Hutcheson (1995: 112). This asymmetrical distribution
precludes a linguistic bias towards left-prominence in such phrases,
but does not rule out equal or right-hand prominence. The most eco-
nomical account that does not require a historical shift, therefore, is that
the right-prominent prosodic contour of phrasal stress has been in the
language since Old English times.
13
See Russom (1987: 114); Hutcheson (1995: 271).
14
For German see Selkirk (1984: 225–30).
9.5 ME stress placement: the native component

The prosodic system of Middle English continues to show stability of
the GSR, aligning primary stress with the left edge of all words and
with the left edge of the root for prefixed verbs and adverbs. This is
easily demonstrated in verse: the first syllables of all disyllabic words in
Chaucer’s Wife of Bath’s Tale (WBT) 278–9 are in strong metrical posi-
tions. Confirmation comes also from the alliteration of disyllables, as in
Sir Gawain and the Green Knight (SGGK) 49–50.
(18) Continuity of the GSR in Middle English:
Thow seyst that dróppyng hóuses, and eek smoke,
And chídyng wýves máken men to flee WBT 278–9
With lórdez and ládies, as léuest him þot.
With all þe wéle of þe wórlde þay wóned þer sámen SGGK 49–50
Words derived by native suffixation also show the expected main stress
on the leftmost root syllable, as in kíngship, wísdom, wítness, hóly, blíssful.
In words in which the derivational suffix follows a disyllabic stem the
suffix may preserve a rhythmically induced degree of stress. This non-
primary suffixal stress can be used in verse with the suffix in a strong
metrical position, most often in rhyme, for example présse: wantownésse
(GP 263–4), sóbrelý: cóurtepý (GP 289–90), but also in other positions, as
in: And spéciallý from évery shíres énde (GP 15).
The prosodic contour of compounds in ME can be inferred from
synchronic scribal variation and from the deployment of compounds in
the verse line. ME variant spellings for <cuppe bord> (1375), <copard>
(1400), <copberd> (1450), <coberd> (1474) ‘cupboard’ indicate a
window of only a century between the first attestation of the word and
its reanalysis as a simplex initially stressed word. Similarly, necklace,
first attested in 1577, appears as <neklas>, <necles> in the seventeenth
century (OED), indicating rapid loss of stress on the second element of
the compound. The pair ALWAYS-CAUSEWAYS illustrates well two options:
loss of secondary stress and reinterpretation of the compound as a
monomorphemic word, as in always ending in either (formal) [-weiz] or
(casual) [-wiz], and preservation of the secondary stress as in causeway
(1440) ending in [-wei]. : Initial stress is stable in the well-known
instances of compound obscuration as in (19).
(19) Compound stress and obscure compounds:
Source ME/EModE
OE dǽes ḕae ‘day’s eye’ ME <daysy> (1440) ‘daisy’
OE hū́swī̀f ‘housewife’ EModE <hussy> (1647) ‘hussy’
ON víndàuga ‘wind-eye’ ME <windo(u)(e), windew(e)> ‘window’
The salience of the secondary stress is an important factor in the

preservation of the orthographic transparency of the etymology in
compounds such as breakfast, Christmas, gunwale [-nəl], island, brimstone
[-stən]. Of interest is also the fact that items such as whetstone, starboard,
waistcoat used to have reduced vowels in the second element, but since
their spelling highlights the morphological structure of the compound,
the secondary stress in them is restored. A good example of spelling-
induced reinstatement of secondary stress comes from the realisa-
tion of -day in the weekdays. Early reduced forms are well attested;
compare fourteenth-century spellings <Sonde>, <Sonede> for Sunday,
fifteenth- and sixteenth-century spellings <Mundy>, <Mondy> for
Monday, <Tysdy> for Tuesday. Walker (1791–1826: §223) endorses
‘Sunday, Monday, &c. as if written Sundy, Mundy &c.’ Secondary stress
on -dày [-dei] is still in competition with the reduced form [-di] in
AE, where [-dei] is definitely the preferred realisation in slow, careful
speech.
In the verse, predictably, the first root syllable of compounds is
usually aligned with a strong metrical position. The dominant pattern
of placement of compounds in alliterative verse also supports initial
main stress. :
(20) Compound stress in ME verse:
Ful thrédbare was his overeste courtepy; GP 290
A shípman was ther, wonynge fer by weste GP 388
And þe gréhòundez so grete, þat geten hem bylyue SGGK 1171
And more he is þen any mon vpon mýddelèrde, SGGK 2100
In contrast to OE, in the ME alliterative corpus secondary stresses
in compounds are occasionally allowed to carry the alliteration, with
the first element of the compound completely out of the alliterative
schema. :
(21) Salience of secondary stress in ME alliterative verse:
And hit lyfte vp þe ỳe-lýddez and loked ful brode, SGGK 446
Dòuble-félde, as hit fallez, and fele kyn fischez SGGK 890
In the iambic compositions in ME the testability of secondary stress is
dependent on the syllabic structure of the components.
(22) Compound stress in ME iambic verse:15
(a) If éven-sòng and mórwe-sòng accorde GP 830
His hérte-blòod hath bathed al his heer KnT 2006
15
The hyphenation in the cited forms is editorial.
(b) And póudre-màrchant tart and galyngale GP 381

But by hir cóte-armùres16 and by hir gere KnT 1016
Compounds in which an unstressed syllable separates the two roots
normally start in even (strong) positions in the iambic verse line, both in
native (22a) and in non-native forms (22b). In these examples the rela-
tive prominence of the two parts of the compound is not testable; both
stresses are aligned with strong metrical positions. If a buffer unstressed
syllable separating the left edges of the roots is not available, the com-
pound can be left-aligned either with a strong or with a weak metrical
position, as in (23).
(23) Compounds with monosyllabic first part in ME iambic verse:
(a) A lóve-knòtte in the gretter ende ther was GP 197
That hadde a fýr-rèed cherubynnes face GP 624
(b) Grèhóundes he hadde as swift as fowel in flight GP 190
He was shòrt-shóldred, brood, a thikke knarre GP 549
In (23a) the compounds are left-aligned with even/strong positions,
mirroring the compound stress in speech. In (23b), however, the first
stress is subordinated and appears in an odd/weak metrical position,
while the second root is in a strong position. Such ‘inversions’ are found
most commonly in the first foot, as in GP 190, but not exclusively.
Further study of the distribution of compounds with a monosyllabic
first element in both types of ME verse has the potential of revealing the
strength of the ‘compound stress’ principle in earlier English.
The testability of phrasal stress in ME verse is limited. In an iambic
line, if there is a buffer weak syllable between the stressed syllables,
and if the left edge of the phrase aligns with a strong metrical position,
as in (of) sóndry fólk . . . (GP 25), (and) máde fórward . . . (GP 33), . . . týme
and spáce (GP 35), the relative prominence of the two stresses cannot
be ascertained. Monosyllabic adjectives in noun phrases do provide
some corroboration for continuing right-prominent phrasal stress – ne
pólax, né short knýf . . . (KnT 2544), Gret swéryng is . . . (PardT 631) – but
the stress-alternating nature of the verse, the availability of optional -e
and metrical slot-fillers, as well as the flexibility of monosyllabic items
in metrical matching obscure the picture. As argued in Minkova and
Stockwell (1997a), there are no good arguments in favour of positing
dramatic changes in the prosody of phrasal stress from Old English to
Present-Day English. Even if we assume a more level phrasal stress in
16
‘A garment embroidered or painted with heraldic arms’, OFr cōte, n. and armūre
(MED). The first root is monosyllabic: <-e> in cōte is elided before <a->.
Old English than in Present-Day English, the right-hand prominence of

Old French and Anglo-Norman would have contributed to the present
contour.
9.6 ME prosodic innovations

The non-native component in the OE vocabulary was never higher
than about 3 per cent (see 9.3.1). The population was relatively
homogeneous, with OE the dominant language, even if we assume
marginal Old English–Insular Latin bilingualism for a portion of the
educated population. The prosody of Old English remained firmly
Germanic.17
The situation changed gradually after 1066 (see 1.3). The control of
the administrative, economic and sociocultural resources was trans-
ferred to speakers whose primary language was not English. A con-
siderable non-native component was introduced into the vocabulary
of Middle English. Within three hundred years after the Conquest
the portion of the non-Germanic, primarily Romance vocabulary
had grown to about 25 per cent. Many of the new words, especially
the earliest borrowings, were ‘core’ vocabulary: beauty, colour, danger,
diet, jealous, journey, liquor, mountain, river, season, story, tender. OE
words associated with government and military power were either
duplicated or replaced by Romance borrowings: army, council, empire,
mayor, navy, parliament, record, soldier, statute. Words from the spheres of
literature, art, science and medicine came into the language in large
numbers, including the words sphere, literature, art, science, medicine, lan-
guage, large, number; also figure, grammar, image, logic, music, poet, remedy,
romance, study, surgeon, tragedy. Many of these items did not have initial
stress in the source language and their entry into English raises the
question of their rate of adaptation to the native Germanic stress-
pattern and their potential effect on the system of stress-placement
in ME.
One type of evidence for initial stress in loanwords was noted in 9.3:
participation of the stressed vowel in the long vowel shift, thus council,
n. (1125), mountain (1275), silence (1225), libel (1297), season (1300), labour
(1300), science (1340), navy (1375). Another important source of infor-
mation is the placement of the new items in verse. Looking again at
Chaucer’s practice, we can illustrate the methodology of reconstruction
with the word season.
17
In neighbouring or substratum Celtic, too, primary word-stress was fixed on initial
syllables excluding proclitics (Bennett 1970: 465).
(24) Testing word-stress in Chaucer’s verse:

Bifil that in that séson on a day GP 19
After the sondry sésons of the yeer GP 347
And eek the lusty séson of that May KnT 2484
The word season appears fifteen times in Chaucer’s verse. In all but one
of the attestations the first syllable matches an even/strong metrical
position, as in (24). The single exception where the second syllable is
matched to a strong position is . . . thy déclináción: his tyme and his sesón
(FrT 1033–4). The end-stressed sesón in rhyme position allows two
interpretations. One is that the Anglicised ‘normal’ pronunciation séson
was artificially changed to fit the prominence required by the rhyme
position, as in native swerýng: thyng in (7); such stress-shifts were a
common verse convention and they tell us nothing about the actual
word prosody (see 9.3.2). Another interpretation is that an ‘unshifted’
pronunciation sesón could still be heard in the last decade of the
fourteenth century, so that Chaucer had two alternate realisations to
choose from. The likelihood of these hypotheses has to be tested on a
word-by-word basis, on the assumption that prosodic accommodation
of loanwords is lexically diffuse. Thus country (1275) is used forty-
five times in Chaucer’s pentameter verse, twenty-one of which are in
rhyme position and are realised as end-stressed. Of the twenty-four
line-internal attestations, however, there is not a single example of
end-stress on the word; they are all of the type illustrated by SumT
1710: A mérsshy cóntree cálled hóldernésse. The absence of finally stressed
variants in mid-line position is a strong indication that the initial stress
was ‘normal’ in the language of London speakers at the end of ME. A
comprehensive and statistically testable database of the stress-patterns
of ME borrowings is still missing, but as argued in Minkova (2000,
2006a), the methodology of data-gathering from rhymed verse has to
be revised to control for the convention of matching a prosodically
weak syllable to a metrically strong rhyme position in the final foot of
the verse line.
When rhyme position is kept out of the picture, the rate of assimila-
tion of the early ME loanwords to the native stem-initial prominence
is remarkably steady. This is partly due to a serendipitous overlap
between the Germanic Stress Rule and the Latin Stress Rule, accord-
ing to which stress falls on the penultimate syllable if it is heavy,
otherwise, on the antepenultimate syllable. The final syllable is invis-
ible to the stress rule and CV syllables are light, while all other sylla-
bles are heavy. (25) shows how this works for disyllabic and trisyllabic
words.
(25) The Latin Stress Rule:

Disyllabic fā́..ma ‘fame’, ér.gō ‘ergo, therefore’
Heavy penult
Trisyllabic co.mḗ.ta ‘comet’, co.lū́m.na ‘column’
Disyllabic cró.cus, ó.nyx
Light penult
Trisyllabic á.ba.cus, Lú.ci.fer
In words of more than three syllables the stress fell on the penult or
the antepenult depending on the penult’s weight: co.ri.án.der, ba.sí.li.ca.
Disyllabic words get initial stress by default: fama, ergo, crocus, hérpes,
mórtar, onyx, sphére, stúpor, etc. would require no accommodation. The
same is true of trisyllabic words with a light penult, the type exempli-
fied by abacus, Lucifer. It was only the subset of trisyllabic words with a
heavy penult like cometa, columna, cucumber, and words of more than three
syllables like coriander, mediator, memorandum, paralysis, that did not fit the
Germanic pattern. For that portion of the lexicon the position of the
stress was determined by syllable weight. The adoption of such words
provides the foundation for phonologically assigned primary stress as a
new prosodic model in PDE.
Words borrowed in ME after the Conquest could be direct transfers
from the Classical languages, or they could be entering English via
Anglo-Norman or Old French. Many early Latin borrowings lost their
inflectional markers (-a, -(t)is, -us, –um, and so on), so the source of the
loan was often obscured. AN and OFr words were stressed depending
on the weight of the final syllable, or the ultima.
(26) Stress in Old French and Anglo-Norman loans:
Final stress in disyllables: pité, barón, reál, merchánt
Heavy ultima
Alternating stress in polysyllables: dàngerús, comàndemént
Initial stress in disyllables: frére, dánce, róuge, ríme

Light ultima Penultimate stress in trisyllables: madáme, servíse, piéce
Alternating stress in polysyllables: crèatúre, pìlgrimáge
The most notable prosodic outliers in the loan vocabulary coming from
OFr and AN would have been disyllables with a heavy final syllable,
like pité, barón, reál, merchánt, and the trisyllabic words with penultimate
stress, like madáme, servíse, piéce, folýe, mirácle, viságe. Since final syllables
ending in schwa are unstressable, in originally polysyllabic words with
a heavy penultimate syllable like bataille the stress was on the penulti-
mate, as in Latin. Additionally, the loss of the final schwa in OFr and
AN, well under way in the thirteenth century (Short 2007), results in
more finally stressed disyllabic items. Throughout ME, however, these
disyllabic borrowings show a strong tendency towards leftward stress-
shifting, in conformity with the GSR, as in pity, baron, battle, merchant,
madam, service, piece, visage, juncture, human, chaplain, novice. The leftward
stress-shift disregards syllable weight; indeed in many cases the stress
shifts leftwards from a heavy to a light syllable, as in chaplain (Lat.
cappellān-us, OFr chapelain), also battle, folly, miracle, novice.
Polysyllabic words with a heavy final syllable, like dàngerús, ìnnocént,
comàndemént, cònsecratión, justìficatión, or words of more than three syl-
lables ending in schwa, like crèatúre, pìlgrimáge, vìlanýe, have their right
edge stressed depending on the weight of the ultima. The distribution
of the prominences to the left of that first window appears to follow the
principle of rhythmic alternation. This is very similar to the preserva-
tion of prominence on derivational affixes in the native vocabulary,
where the rhythmically induced prominence of native suffixes is utilised
in verse: the suffixes in sóbrelỳ, nórissỳng drónkenèsse, dóutelèes, mártyrdòm
are aligned with strong metrical positions (see 9.5). The combination
of dominant word-initial stress and the rhythmic preference for stress
alternation in borrowed words produces a comparable effect in the new
loan vocabulary. The difference between native sóbrelỳ, nórissỳng and
borrowed dàngerús, ìnnocént, or between drónkenèsse and pìlgrimáge is in
the relative strength of the stresses: in the native words the stress on
the affix is subordinate, while the rightmost stress in the loanwords is
primary, and additional stresses to the left are less prominent, at least
initially.
Once again, verse provides the basis for testing and confirmation. In
iambic verse, polysyllabic loanwords are hard to fit to a metrical frame
of alternating prominences. Ignoring morphological structure for the
moment, one can observe that the linguistic sequence /w w s/ in the
source language is realised in ME as /s w s/: àrgumént, chàritée, làxatíf,
gènerál, òpposít, òrisóun, plèntevóus, règióun. As noted above, it is possible that
in such cases, at least for words of lower frequency, it was the leftmost
syllable of the word that carried the secondary stress at first, judging from
the strong preference for placement of such words in rhyme position.
(27) Romance polysyllables in rhyme position in Chaucer:

aváunt: rèpentáunt GP 227–8 hóus: plèntevóus GP 343–4
dìgestíble: Bíble GP 437–8 whít: òpposít KnT 1893–4

wróoth was shé: chàritée GP 451–2 adóun: règióun KnT 2081–2
comàndemént: ysént KnT 2869–70 màriáge: ìn myn áge WBT 23–4
of blóod roiál: in gènerál Tr III: 1800–2 dýe: of bìgamýe WBT 85–6
On the other hand, the initial syllables of many trisyllabic loans with
original rightmost prominence can alliterate: áudience, bénefys, béneson,
élementz, équite, órisouns are attested in Langland’s Piers the Plowman
(Tamson 1898: 72–3). It is probable, therefore, that the switch from
word-initial secondary to primary stress in such trisyllabic words
started in Middle English. A more precise dating is not recoverable
from iambic verse, where both primary and secondary stresses may fill
strong positions. Since ME alliterative verse was no longer orally com-
posed and transmitted, scribes could have used eye-alliteration, so the
evidence from alliterative verse is suggestive but not compelling.
The preservation of some degree of stress on the final syllable in
polysyllabic Romance loans beyond Middle English is documented in
Dobson (1968: §§265–92). The pattern of alternating stress as in gráciòus,
submíssiòn, rèsolútiòn, éxcellènce is found in the poetry throughout the
seventeenth century. :
Danielsson (1948: 26–9, 39–54) attributes the eventual demotion of
the original primary rightmost stress in loanwords of more than three
syllables to the school pronunciation of Latin in Middle English and
Early Modern English. He uses the term ‘countertonic accentuation’ to
describe the shift of, for example, Gk melancholía > Lat., AN melancolie
(1375) to mélanchòly, similar to the native model of máidenhòod, drúnken-
nèss. The picture is complicated by the fact that along with borrowing
entire words, English ‘nativised’ some Latin and French derivational
affixes. Such affixes can attach to native roots without affecting the stress
placement, thus AN -able, ME singable (1340), believable (1382); OFr -ard,
ME wizard (1440) < wise, adj. + -ard; AN, OFr diminutive -erel, ME
pykerell (1290) < pike + -erel; OFr -age, ME bondage (1330) < OE bonda +
age, also brewage (1542), leafage (1599). Other affixes can be attached only
or mainly to borrowed bases: -acy, -ate, -ee, -erie, -ment, -ous, -ic(al). We
return to the innovative stress-patterns related to non-native suffixation
in 9.8.
9.6.1 Grammar, meaning and stress-shifting: PERFÉCT-PÉRFECT,

CANÁL-CHÁNNEL
The examples of the variable treatment of main stress in prefixed

nouns and prefixed verbs in OE (see 9.4.1) seem to be identical to PDE
upload, v. [$ploυd] (1870) vs upload, n. [p$ləυd] (1979); overhang, v.

[$oυvərhæŋ] (OE) vs overhang, n. [oυvər$(h)æŋ] (1853). The process
is known as functional stress-shifting. For the native vocabulary, the
prosodic marking of morphosyntactic category requires the presence
of a recognisable prefix. Although many details remain unexplored,
it seems safe to assume that borrowed words in ME could also fit the
pattern of prosodic differentiation of the morphosyntactic nature of
the stems. Verbs in which the prefixation is transparent behave like
the native prefixed verbs, that is, main stress on the first syllable of the
root: Perfóurme it out . . . (Tr III 417), ye nát discóvere me (MerT 1942).
Prefixed nouns and adjectives vary. Chaucer uses both initial and final
stress on proverb, a word first recorded in his works (OED): Wel may that
be a próverbe . . .(WBT 284), And therfore this provérbe is . . . (RvT 4319).
Etymologically non-transparent prefixed nouns and adjectives tend to
follow the native rule: Ben humble súbgit . . . (Tr II. 828), . . . in joye and
pérfit heele (KnT 1271). Of greater interest are, however, the loanwords
which are synchronically monomorphemic, which nevertheless are also
subject to functional stress-shifting, as in the examples in (28), with
first-recorded dates from the OED.
(28) Functional stress shifts in synchronically monomorphemic ME
loans:
Verb Noun/Adjective
ally 1297 1380
augment 1400 1430
ferment 1398 1420
frequent 1477 1531
rebel 1340 1297
record 1225 1300
torment 1290 1290
Notice that the dates for the verb tend to be earlier than for the cor-
responding noun or adjective. The direction of the shift in the loans is
almost always from end-stress for verbs to initial stress for nouns and
adjectives, confirming the observations on the strength of the GSR in
ME (see 9.6). The full history of stress-shifting in ME is an area which
has not been fully researched yet; it is an inquiry that promises to throw
light on the continuity and/or reintroduction of a prosodic pattern in
English which has been has been growing steadily since the second half
of the sixteenth century, as in ábstract-abstráct, récord-recórd, rébel-rebél,
pólice-políce. :
In addition to differentiating word-class, stress-shifting can involve
semantic differentiation, with or without word-class change. There is
nothing surprising about variant pronunciations of polysemous words

resulting in the split of the original base into two separate lexical entries.
A famous case in point is the semantic bifurcation in the pair: palace-
palate < Lat. palātum ‘roof of the mouth, one of the seven hills of Rome,
the house of Augustus situated there, any roof-like form’, where the [-s]
form is from AN and OFr; similarly the history of person-parson, arrant-
errant (see 8.3.2). Variable prosodic forms can also result in a lexical
split. One of the earliest examples of transfer of meaning, accompanied
by a stress-shift for one of the meanings is the pair Áugust (OE) - augúst
(1664). : (29) shows more instances of historical stress variability
producing semantically independent lexical entries, either because of
repeated borrowing or because of synchronic variability in ME.
(29) Stress-shifts related to meaning-shifts:
drágon, n. (c. 1220) < OFr dragon, Lat. dracōn dragóon (1622)
crític, adj. (1543); †crític ‘criticism’ (1656) critíque (1719)
chánnel (1300) < OFr chanel canál (1449) < Lat.
canāl-em
húman (1450) < AN humane (fem.) humáne (1500) < Lat.
hūmānus
úrban (1634) < Lat. urbānus urbáne (1533)18
trável, v. (1270) < OFr travaillier traváil (1250)
pétty (1372) ~ pétit (1390) < AN petit, masc.) petíte (1766) < petite,
fem.
ínvalid, n. (1642)19 < Lat. invalid-us inválid (1635)
tóilet (1540) < Fr. toilette toilétte, post-16th c. var.
of toilet
9.7 ME compound and phrasal stress

Section 9.4.2 touched on the problem of proving or disproving the
hypothesis of different prosodic contours for compounds and phrases
in OE: while left-edge prominence in compounds is solidly supported
by the evidence, right-hand prominence, or the End Rule, for phrases,
cannot be tested reliably in the verse. The alliterative compositions
in ME are somewhat more informative, primarily because of the
18
Urbáne was originally a variant of úrban; the meaning differentiation (urbane ‘refined’) is
first recorded in the seventeenth century. The same variation occurred with travel and
travail.
19
Originally, it was a special sense of the adj. invalíd. The switch to initial stress occurred
in the nineteenth century in AmE; Webster (1828) has ínvalid, a ‘pronunciation . . .
commonly heard in England also’ (OED).
relaxation of the classical OE matrix of obligatory alliteration on the

linearly first member of a noun phrase, whether it is the head or the
modifier. Already in the thirteenth century, the alliterating portions of
some lyrics show right-hand prominence, thus helle stenches, lustes stench,
godes word, godis wille, all alliterating on the head noun.20 The phrasal
End Rule is supported also by the frequency of right-hand allitera-
tive prominence in fourteenth-century verse. Some examples from Sir
Gawain and the Green Knight are shown in (30).
(30) The End Rule in fourteenth-century alliterative verse (SGGK):
So bisied him his onge blod and his brayn wylde 89
Talkkande bifore þe hy e table of trifl es ful hende 108
Þen þe first cors come with crakkyng of trumpes 116
To þe grene chapel þou chose, I charge þe, to fotte 451
Alliteration on the head noun in noun phrases (NPs) is regular with
semantically weak modifiers – good, great, many, such, other – but it can
go both ways with more distinct adjectives. This suggests that while
phrasal stress is subject to the overall preference for right-hand promi-
nence, it is also susceptible to other factors such as semantic load and
lexicalisation. :
Testing compound and phrasal stress in ME syllable-counting verse
is complicated by the optional phrase-internal unstressed syllables.
Looking at the NPs in the opening of Chaucer’s General Prologue, we
find: . . .with his sweete breeth, . . . in every holt, The tendre croppes, . . .the
yonge sonne, . . .his halve cours, And smale foweles (GP 5–9), where all of
the italicised phrases are aligned with S W S metrical positions. This is
uninformative because both compounds and phrases fit the same met-
rical frame. The sequence S W S can be satisfied in several ways: the
modifier is disyllabic, for example in every holt; the modifier is monosyl-
labic and follows a determiner (weak declension), so there is an extra
buffer syllable before the noun, for example his sweete breeth; the noun
is not initially stressed, for example . . . an hard requeste (Tr III, 148); or
the monosyllabic modifier is plural, preserving the inflectional -e, for
example smale foweles. The relative strength of the two prominences is
indeterminate. :
EModE shows a somewhat clearer distributional evidence for com-
pound and phrasal stress. The matching of these units to metrical
positions in Elizabethan iambic pentameter tends to mirror the lin-
guistic difference between the two types of structures. Compounds
with a monosyllabic first element (long-lived, first-born, tongue-tied) and
20
Examples from Brown (1932/1962: 4, 15, 27, 46).
NP phrases of comparable syllabic composition (sweet brood, old Time,

tall building) are distributed predictably in verse: the former type are
aligned overwhelmingly with S W / even positions, while the phrases
appear mostly in W S W positions. Compare the examples in (31) from
Shakespeare’s sonnets.
(31) Compound vs phrasal placement in Shakespeare’s Sonnets:
(a) Compounds in S(trong)-Weak metrical position:
And burn the long-lived phoenix in her blood Sonnet 19
With April’s first-born flowers, and all things rare Sonnet 21
To make me tongue-tied speaking of your fame Sonnet 80
(b) Phrases in W(eak)-(S(trong) metrical position:

And make the earth devour her own sweet brood; Sonnet 19
Yet, do thy worst, old Time: despite thy wrong, Sonnet 19
He of tall building, and of goodly pride Sonnet 80
In summary, ME compound and phrasal stress cannot be shown to be
‘innovative’. There is no reason reject the assumption that both the left-
prominent compound stress contour and the right-prominent phrasal
stress contour posited for Old English remained stable in ME. By the
end of ME, phrasal stress would have been reinforced by exposure
to French prosodic patterns, with dominant right-headedness on the
phrasal level. Although the claim is only weakly supported in the verse
material, not fully recoverable from the meter of the earlier periods,
it would appear that the most reasonable position is that ‘Alfredian,
Ricardian, Elizabethan, Jacobean, and even Clintonian English obey
and have always obeyed the same rules of phrase internal prominence
assignment’ (Minkova and Stockwell 1997b: 320).
9.8 Post-ME prosodic innovations

EModE is a period characterised by higher literacy rates, a sharp rise
in book production and a parallel rapid expansion of the lexicon. Over
20,000 classical loanwords first recorded between 1500 and 1700 survive
to this day. This exceeds by far the rate of borrowing in Middle English.
Two thirds of the new forms in EModE were based on already recorded
roots and affixes and about one third were straight borrowings. Latin
and French continued to be the main donors of new words. Of note is
also that while late ME loans from French outnumbered those from
Latin, the proportion of loans from Latin increased steadily and peaked
between 1575 and 1675. Parallel to the entry of ‘whole’ lexemes, there
is also a striking increase of the relative frequency of Latinate affixes in
the overall number of new affixed forms from 20 per cent in the fifteenth
century to 70 per cent in the third half of the eighteenth century. :
The large majority of the new items were scientific and technical
terms adopted or coined by English speakers who were proficient in
Latin. Such speakers would tend to preserve the Latin Stress Rule (25)
as in ablátion, compéndium, máximum, meánder, términus. The density of
these forms and the shared literate understanding of their prosody gave
rise to a new, parallel model of stress in English, which is weight sensi-
tive, and which can apply productively to new words such as Óregon
(1765, possibly Connecticut pidgin Algonquian, OED), kaínga ‘village’
(Maori, 1820), carráncha ‘carrion-hawk’ (South American Indian lan-
guage, 1839), Anímikie ‘thunderer’ (1873, Ojibwa), mazúrka (Polish,
1818), palachínka ‘pancake’ (Serbian, 1884). The process of integrating
a right-to-left weight-sensitive stress placement stretches over the
whole EModE period, but it was not until the mid-eighteenth century
that the new model was fully recognised. Lass (1999: 130) dates ‘the
shift in grammarians’ typological intuitions’ between the 1740s and the
1780s. From then on, ‘English begins to feel more like a language with a
Latinate accentual system than one with a Germanic type’ (ibid.). The
statement is justified by the limited share of native items in the lexicon
of an adult English speaker, approximately 25–30 per cent (Minkova
and Stockwell 2006: 466–7).
The recognition of a new model of stress-placement for polysyllables
never completely obscured the tenacity of the GSR. As is common in
any body of loanwords, the EModE borrowed lexicon was composed
primarily of nouns and adjectives. The relative share of nouns in the
new lexicon of EModE was consistently above 50 per cent and reached
70 per cent in 1760–4. The proportion of adjectives is also significant,
ranging from 20 per cent to 28 per cent in 1660–74. Verbs, on the other
hand, are borrowed at a lower rate, rarely up to 20 per cent and down to
8 per cent at the end of the period.21
Since the GSR applied without exception to native nouns and adjec-
tives throughout the earlier history of English, even in prefixed forms
(see 9.4.1), it is not surprising that left-edge prominence continued to be
a salient and active principle in the adaptation of borrowed nouns and
adjectives. Consolidation of the primary stress on the initial syllable of
the stem went beyond the disyllabic shifts recorded in 9.6, as in hérpes,
mórtar, sphére, stúpor, and affected trisyllabic nouns and adjectives. Words
such as ámorous, cálendar, chárity, génial, ínfantry, láxative, mércury, ópposite,
órient, órison, plénteous, région changed their Chaucerian ‘double-stressed’
21
Statistics from Nevalainen (1999: 353).
realisation in favour of a GSR-type stress. Stress-shift to the initial

syllable often proceeded in spite of the etymological heaviness of the
penultimate syllable, as in the ME loans ámorous, fórtunate, ínfantry,
ínterval, órient and many post-ME forms such as órchestra (1596), vértebra
(1615), bálcony (1618), tálisman (1638), cóquetry (1656), sýnergy (1660),
mónocle (1772), Cávendish (1839), bádminton (1845), állergy (1911).
The emerging picture is complexly layered: the prosody of native
unprefixed words follows the OE left-alignment of the word or stem
with the main stress. The non-native vocabulary displays hybrid pat-
terns, and no single model covers all realisations without multiple
exceptions, so that only strong tendencies can be defined. New words
can fall in with the native left-strong Germanic model, or they can
follow a weight-sensitive model whereby stress in non-derived words
is assigned by syllable weight. The weight of the penultimate syllable
is responsible for agénda, Torónto, while cánopy, ínfidel fit both models.
Although the considerable overlap between the patterns noted in 9.6 for
disyllabic nouns and adjectives continues, end-stressed nouns like abýss,
baróque, cabál, canál, duréss, elíte, maláise, ravíne do occur. Verbs with heavy
final syllables are generally end-stressed, for example paráde, dený, main-
táin, oblíge, protéct, but the realisations of the verbal affix -ate may vary:
cóncentrate, dévastate, éxcavate, but demonstrate, elongate, contemplate allow
penultimate stress. :
Variable stress in borrowed words is very much a feature of the
prosody of English in the last couple of centuries. MacMahon (1998:
493–517) presents detailed chronological tables of words susceptible to
stress-shifts since the end of the eighteenth century. He uses twenty-
four pronouncing dictionaries: from Johnston (1764) and Sheridan
(1781) to the 1990 Longman Pronouncing Dictionary and the 1992 OED2.
He establishes six patterns of main stress variation, set out in (32).
(32) Post-eighteenth-century patterns of variation and change:22
Description Examples
Type 1: Zero or limited variability énterprise, predícament
súbstantive ~ substán-
tive (BrE)
Type 2: Competing forms (A ~ B) décorous ~ decórous
cóntroversy ~
contróversy
Type 3: Shorter-term changes (A > B > A) shérbet, sherbét, shérbet
22
From MacMahon (1998: 493); the examples are from his tables 5.1–5.6. The tabulation
records only primary stresses; other levels of stress are conflated with absence of stress.
splenétic, splénetic,
splenétic
Type 4: Clear change, no reversal (A > B) accéss, n. > áccess, n.
balcóny > bálcony
Type 5: Multi-stage change (A > B > C . . .) cómplaisance, com-
plaisánce, compláisance
Type 6: Variation and reversal (A > B ~ A > A) panthéon, pántheon
Focusing only on the twentieth century, Bauer (1994b: 96–103) records

items which have undergone a recent shift to penultimate stress,
for example ábdomen, ácumen, ánchovy, étiquette, molýbdenum, précedence,
quándary, sécretive, sónorous, vágary. : The continuing variability of
main-stress placement and the concomitant degree of vowel reduction
for some items is well illustrated by the OED pronunciation entry for
quadruple, n., adj., adv.: Brit. /kwɒdrυpl/, /kwɒdru
pl/, /kwədru
pl/,
US /kwɑdrup(ə)l/, /kwədrup(ə)l/, /kwɑdrəp(ə)l/, /kwədrəp(ə)l/.
As is clear from this and many other entries, the extent to which
individual loanwords favour morphological vs syllable weight-based
stress can differ in British and American English. Some well-known
examples with first attestation dates from the OED are shown in (33);
some of these are simply ‘majority’ pronunciations in variation with the
alternative pronunciation.
(33) Stress differences between American and British English:

(a) GSR in AmE:
American English Date British English
ínquiry (1440) inquíry
pólice (also políce) (1450) políce
frústrate (1447) frustráte
móustache (1585) moustáche
(b) GSR in BrE:

American English Date British English
premíer (1500) prémier
debrís (1708) débris
café (1802) cáfe
miráge (1812) mírage
garáge (1902) gárage
The examples in (32) and (33) leave no doubt as to the hybridity of
the PDE prosodic system. As is the case for ME too, however, the pro-
sodic behaviour of new words in EModE and PDE has not been fully
documented and analysed. What is still needed is a full-scale statistical
overview of the patterns of adaptation, controlled for factors such as

word-class, number of syllables, the place of the item in a derivational
set and frequency of the morphemes in non-monomorphemic items,
though Berg’s (1999) comparison between stress in British and American
English offers both valuable data and some interesting conclusions. :
Before we conclude the survey of the prosodic adaptation of loan-
words in English, we must note a further complicating factor: the role
of suffixes in stress placement. As noted in section 9.4.1, OE suffixation
was never associated with main-stress reduction; the highest level of
prominence in a suffixed word was always inherited from the stem:
bérend ‘bearer, pregnant’ - bérendnes ‘fecundity’, fǽder ‘father’ - fǽderlic
‘fatherly’. Through ME, and into PDE, the suffixes listed in (34) never
carry the primary stress in a derivative.
(34) Stress-neutral native derivational suffixes:
-dom: as in mártyr - mártyrdom
-en: as in forgíve - forgíven
-er: as in intérpret - intérpreter
-ful: as in regrét - regrétful
-hood: as in néighbour- néighbourhood
-ish: as in yéllow - yéllowish
-less: as in compássion - compássionless
-ly: as in mátron - mátronly
-ness: as in invíncible - invíncibleness
-some: as in advénture - advénturesome
-ward(s): as in héaven - héavenward(s)
-wise: as in óther - ótherwise
Persistence of primary stress in words derived with native suffixes
continued in ME. At the same time, the adoption of a large number
of foreign suffixed words triggered the introduction of suffix-specific
main-stress placement at the right edge of the word.
The behaviour of the Latinate suffixes in PDE has been the subject of
intense linguistic scrutiny, yet none of the analyses can be considered
conclusive.23 Some of the more general patterns exemplified in (35)–
(38) include suffixes which behave like the native suffixes, that is, they
are stress-neutral; suffixes that always attract stress; and suffixes that
push the stress onto a particular syllable to their left. We will ignore
secondary stress, noting that some suffixes, for example -ise, -oid, are
stress-neutral with respect to the stem but they attract secondary stress.
23
A good descriptive coverage is found in Fudge (1984); the analytical problems are
addressed in Giegerich (1999).
Also, the process of derivation with suffixes that are not stress-neutral
may involve preservation of the stress of the original base, so the initial
primary stress on móuntain, pícture, vélvet is inherited as secondary stress
in the derivatives mòuntainéer, pìcturésque, vèlvetéen.
(35) Stress-neutral borrowed suffixes:
-ist: (e)vangelist (1175), exorcist (1382), dogmatist (1541)
-ty: specialty (1330), mayoralty (1387–8), admiralty (1419)
-ise: authorise (1383), crystallise (1600), generalise (1425)
(36) Stress-attracting borrowed suffixes:
-ade: grenáde, lemonáde
-air: debonáir, corsáir
-ane: arcáne, mundáne, germáne, urbáne
-ee: payée, devotée
-een: velvetéen, cantéen
-eer, -ier: mountainéer, brigadíer
-elle: bagatélle, villanelle
-esce: effervésce, acquiésce
-esque: statuésque, grotésque, picturésque
-ese: journalése, viennése
-ette: majorétte, serviétte
-oon: tycóon, ballóon, dragóon, picaróon
(37) Penultimate stress with borrowed suffixes:
-ic: numéric, idiótic, históric, económic
-id: carótid, myópsid
-ion: rebéllion, compánion
(38) Antepenultimate stress with borrowed suffixes:
-(cra)cy: demócracy, aristócracy
-ast: icónoclast, enthúsiast
-ity/-ety: tranquílity, humílity
-ose: cómatose, béllicose
-tude: similitude, áttitude
The antepenultimate is stressed also in combining forms such as -ólogy,
-ósophy, -ógraphy, -ólatry, -ócracy, and so on. These new patterns of stress-
assignment extend to native roots under foreign suffixation as in Icelándic
(1674), weatherólogy (1823), speedómeter (1904), Chàplinésque (1921). Then
there are patterns that are not based on an actual affix, for example -eau,
which tends to be stress-attracting, but the final stress may be shifted
in chateau, tableau, plateau, trousseau, bureau, especially in British English.
Similarly, -oo attracts stress: bambóo, shampóo, tabóo, although the etymology
of the <-oo> words is quite disparate, so the fixing of the stress on the
final syllable is within English, where [u
] is always a stressed vowel.
The examples in (35)–(38) are selected to illustrate some general pat-
terns, but they are far from exhaustive. They are indeed only slightly
better than a bewildering laundry list, and there are exceptions galore
in each group: in (35) we list -ise as stress-neutral, but chastíse (1325) has
alternated with initially stressed chástise from the beginning.24 In (36)
with -ade, we find grenáde, lemonáde, but AmE also lémonàde, and both
BrE and AmE allow initial main stress in prómenàde; with -air we find
debonáir, corsáir, but AmE also córsair; with -ee we find payée, devotée, but
both employée and emplóyee, and both refugée and réfugee in AmE. Some
highly productive suffixes, for example -able, are hard to fit in the tax-
onomy because they produce variable results: -able can be stress-neutral:
colléctable, excúsable, récognisable, réplicable; it can go against the stress of
the base: irréparable, cómparable, préferable, ádmirable, demónstrable; and the
stress varies in applicable, despicable, hospitable, (in)explicable. The pro-
ductive suffix -age has two forms: unstressed [-əd/-id] in pilgrimage,
steerage, orphanage, [-eid] in verbs: engage, rampage, presage; and stressed
[ɑ(
)] as in still unassimilated barrage (1859).
The difficulty of describing stress in derived words in PDE is exac-
erbated by the tendency for the prosody of the base to be inherited in
the derivative. Stress placement in derived words can ignore the nature
of the suffix and preserve the prosody of a pre-existing and frequently
used base, thus cápital, prefér are the bases which trigger the change of
old capítalist to current cápitalist, and of old préferable to AmE preférable. :
In summary, stress-placement in PDE is a mixture of prosodic pat-
terns, some inherited from Old English, some introduced in Early
Modern English. With Old English we share left-edge prominence in
disyllabic noun and adjective bases, left-edge marking of compounds,
lack of stress on function words, head-prominence in clitic groups and,
most probably, right-hand phrasal prominence. The loan vocabulary
of Middle and Modern English complicates the picture by introduc-
ing weight-sensitive stress assignment, suffix-driven stress assignment,
tolerance for final stress in nouns, and specific rhythmic constraints
for the avoidance of stress-clashes and strings of unstressed syllables.
Many relevant details in the prosodic history of English remain under-
24
‘The stress was originally always on the first syllable: chastise, is generally so with
Shakespeare (7 times against 2), and also in later poets, as still in chastisement; but
already in Chaucer sometimes, and Gower often, on the second, as now’ (OED). The
etymology of the word is unclear – it may have final stress on the analogy of comprise,
demise, despise, devise.
researched: the prosodic behaviour of borrowings in Middle English

needs fuller documentation, and the relevance of competing factors
such as phonological composition, frequency, morphological marking
and transparency, social prestige and spelling have to be evaluated. The
relationship between innovations in verse form and prosodic innova-
tions is also of considerable linguistic and cultural interest. Other areas
that invite further inquiry are the prosodic patterns in the regional and
ethnic varieties of English, and the contact-induced changes in English
spoken as a second language.
The historical diversity of the stress-patterns in English is related
directly to the rise of new word-formation models. The loss of post-
tonic syllables goes back to OE; it is unremarkable in a language
with stress on the first root syllable. Though the loss normally affects
inflectional syllables (see 7.6), a process of ‘back-clipping’ in English is
attested early and its productivity grew in EModE. :
(39) Back-clipping of initially stressed words:
trent < tréntal (1389) ult < últimum (1750)
chat < chátter, v. (1440) vis < vísit, v. (1754)
coz < cóusin, n. (1559) bod < bódy (1788)
mob < mobile (1688) lunch < lúncheon, n. (1829)
hack < háckney, v. (1721) beaut < béauty (1866)
In (39) the clippings preserve the stressed syllable of the input, and the
output words can be both nouns and verbs. A somewhat different model
of back-clipping is illustrated in (40), where the main stress is ignored.
(40) Back-clipping of non-initially stressed words:
Oxon < Oxónian, n., adj. (1439) extra < extraórdinary (1776)
sol < solútion, n. (1588) prof < proféssor, n. (1800)
phyz < physiógnomy, n. (1687) advert < advertisement (1814)
gin < Genéva, n. (1714) math < mathemátics (1847)
ally < alabáster, n. (1720) perk < perqúisite, n. (1869)
In the set in (40) it is the left edge of the word that is preserved, in spite
of the lack of primary stress, or any stress on the remaining portion.
If the remaining portion is disyllabic, it has initial stress. The process
appears to be almost categorically restricted to nouns and adjectives.
Thus, along with the introduction of new vocabulary, we can discern
a tendency for adapting the polysyllabic words to the mono- and
disyllabic initially stressed mould of the native word-stock. Shortened
forms, for example rep(utation) (1677), rep(ublic) (1701), mob(ile) (1688),
penult(imate) (1490), were the target of complaints by some eighteenth-
century language commentators, yet such forms have continued to
enter the language and the model is very much in vogue in PDE. : Bio,
econ, lit, math, psych are in the vocabulary of any college student, the telly
has ads and sitcoms, we use our cells to download apps. There are clearly
two distinct factors at work: stress-preservation, as in zine (1965), and
left-edge alignment/first-syllable anchoring, as in app (1985). Their
interaction is an important component of the prosodic morphology of
English (see further Lappe 2007).
The loss of pre-tonic syllables is known as aphesis.25 Pre-tonic
loss of unstressed stem vowels (gýpsy < Egyptian, n. (1514), Mér(ri)
kin, n. < American (1872), lectric < electric (1955)), is an innovation
which can be associated with the adoption of the weight-based model
of stress-placement and conformity to the preferred word-length in
English. Deletion of initial unstressed prefixes did occur in OE, espe-
cially with the prefix ge-, thus fere < gefere ‘companion’ (975), mung <
gemong ‘mixture’, n. and adj. (1175), but the frequency of attested forms
increased significantly from early ME on. Not surprisingly, in view of
the native model of unstressable verbal prefixes (bespéak, forgét, with-
stánd) and the variable use of the a- prefix (down ~ adown, mid ~ amid,
mend ~ amend, rise ~ arise), the most frequent new aphetic forms are
Latinate prefixed verbs (spute < dispute, v. (1225), dite < endite, v. (1300),
stall < install, v. (1300)). There are, however, some instances of aphesis
in nouns – merlin < esmerilun (1382), †colet (1382) < acolyte, larum (1533) <
alarum, cello (1848) < violoncello, zine (1965) < magazine – where the delet-
able part is not prefixal. Here belong also the truncated forms of some
names such as Bert, Ria, Gene. The early history of these forms is not yet
fully documented and analysed, but there is some evidence that the
process peaked between 1300 and 1500 and has been gradually getting
more marginal. :
The recessiveness of pre-tonic syllable loss in English has not been
discussed in the literature. Possible directions of inquiry are (a) the
uneven historical rate of unstressed syllable loss at the right or the left
edge of the word as a diagnostic of the dominant prosodic model, and (b)
the sociolinguistic implications of the chronological fluctuation in the
productivity of aphesis in English. A full exploration of the data prom-
ises to be quite revealing about the interplay between weight-based
stress and left-edge prominence, a rivalry which has been characteristic
of English prosody for many centuries.
25
The usual definition of ‘aphesis’ refers to the pre-tonic loss of a vowel (see 2.5), but
here the term is used more broadly to cover the loss of any pre-tonic syllabic material.
10 Early English verse forms:
from Cædmon to Chaucer
This chapter offers an overview of the main verse forms in the early
history of English. After defining some terms specific to the study
of poetic meter, it addresses the evolution of alliterative verse: its
structure in Old English, the continuity of alliterative versification in
early Middle English and the reinvention of alliterative verse in the
fourteenth century. Section 3 turns to the introduction of rhyme and
syllable-counting. Section 4 focuses specifically on Chaucer’s contribu-
tion: the iambic pentameter. The principles of matching stress to meter
outlined in the chapter are applicable to post-Chaucerian verse forms
as found in the works of Skelton, Lydgate, Wyatt, Shakespeare, Milton,
Wordsworth, Byron, Shelley and Keats, to name only some of the great
poets of the previous millennium.
10.1 Preliminaries: speech prosody vs poetic meter, stress vs ictus

The term prosody, as used here is restricted to stress phenomena in
the spoken language, so ‘prosody’ is short for ‘speech prosody’ (see 9.1).
Meter describes the modes of versification, where linguistic material is
arranged in specific recursive frames; as used here, ‘meter’ is short for
‘verse or poetic meter’. Meter always draws on the prosodic structures
available in the language. The two systems rely on the same funda-
mental distinctive categories and relations, though meter obeys verse-
specific conventions such as repetition, alliteration and end-rhyme,
and not all prosodic features have to be harnessed into a specific verse
form. Nevertheless, there are important correspondences between the
structure of verse and the structure of the ambient language. This makes
meter a valuable source of information about language. The linguistic
competence of the poet provides the raw material for verse; the way in
which that material is manipulated to fit the metrical constraints of a
particular tradition reveals the poet’s, or the copyist’s intuitions about
prosody. :
323
The study of poetic meter relies on characteristics defined on the

basis of linguistic primes: segments, syllables, words, stress, weight,
rhythmic preferences and hierarchical relations. For English the syl-
lable, the carrier of linguistic stress, is the main constituent that forms
the basis for the historical study of prosody and meter. As shown in 9.1
and 9.3, word-stress can be seen as a gradient property, with absence of
stress at one end of the continuum and primary stress at the other end.
The prominence of an individual syllable is also a function of the posi-
tion of the word in a larger domain: the first syllable of léather [lεðə(ɹ)]
in the phrase leather handbag is less prominent than the first syllable in
handbag, but in the compound leather-coated [lε-] is more prominent
than [koυ-] in coated. Prosodic prominence relations are most stable at
the level of the word; in a word like leather the first syllable has to be
stronger than the second. Commonly the same prominence relationship
persists irrespective of the embedding of the stem in a derivative, for
example leatherette [$lεðərεt].
The syllable is also the basic building block of the verse units in
the various English metrical forms. In verse, syllables fill metrical
positions; metrical positions are the prime constituents of verse. The
metrical positions in the line differ in their prominence. A promi-
nent, or strong metrical position is called an ictus, here marked with
S(trong); ictic positions are usually, but not always, filled by stressed
syllables, while non-ictic W(eak) positions attract unstressed syl-
lables. S and W metrical positions are further organised into higher-
level constituents known as metrical feet. All metrical feet contain an
ictus. Within the foot, the strong position must be either initial as in
S W or trochee, S W W or dactyl, or final as in W S or iamb, W W
S or anapaest. The number of weak positions in a foot varies in the
different metrical forms. The weak/non-ictic position in a metrical
foot can be left unfilled; this is the case in the so-called ‘headless’ lines
(see 10.4). :
Above the level of the foot the unit that includes one or more feet is
usually known as a colon (pl. cola), or a hemistich, though for OE the
accepted term for a hemistich is verse (see 10.2). Hemistichs are usually
separated by a caesura, a metrical pause commonly coterminous with a
syntactic boundary. The structure uniting the cola is the metrical line:
the lines in the verse-types we will be looking at are made up of two
colas. In English alliterative compositions, the two verses make up the
alliterative long line.
The overall number of syllables per line can be the same for each
line, as is the case in Chaucer’s or Shakespeare’s verse: such verse is
known as isosyllabic. The number of feet per line in isosyllabic verse
E ARLY ENGLISH VERSE FORMS : FROM CÆDMON TO CHAUCER 325
determines whether a line is a monometer (one), dimeter (two), tri-

meter (three), tetrameter (four), pentameter (five), hexameter (six) or
heptameter (seven). Alternatively, the number of syllables per line can
be flexible – this is true of English alliterative verse and of free verse,
both of which are non-isosyllabic. Our discussion of the early forms of
English verse will stop at the level of the line, though clearly lines can
build larger verse units such as couplets, triplets, quatrains, stanzas and
cantos.
In ‘ideal’ syllable-counting verse, single metrical positions are
filled by single syllables, S-positions align with stressed syllables and
W-positions align with unstressed syllables. This important con-
straint on matching prominences in speech and verse is a valuable
historical diagnostic for stress. Further alignments of linguistic units
to metrical constituents are optional. This includes word bounda-
ries coinciding with metrical foot boundaries, phrase boundaries
coinciding with hemistichs, and clause boundaries matching line-
ends. A simple alignment between meter and prosody is shown in
Figure 10.1; prosodically stressed syllables are marked s, unstressed
syllables are w. 1
Line
Hemistich/Colon
F F F F Foot
W S W S W S W S Positional Prominence
1 2 3 4 5 6 7 8 Positions
The birds a round me hopped and played1
w s w s w s w s Prosodic prominence
Word
w s s w s w s Clitic group
w s w s Phrase
Clause
Figure 10.1 Matching meter to prosody
1
Wordsworth, Lines Written in Early Spring (1798).
In this example, the alignment on the syllabic level – syllabic promi-

nence to prosodic prominence – is perfect: all metrical S’s are filled
by stressed syllables (birds, -round, hopped, played) flanked by weakly
stressed syllables within a word (a-) or a clitic group (the, me, and). All
word boundaries coincide with the metrical foot boundaries in positions
2, 4, 6 and 8. The clause boundary appears at the end of the line. The
only mismatch is between phrase and colon: the birds around me straddles
the hemistich, so that the caesura is heard after the fifth, and not after
the fourth position. This type of mismatch is common and benign; the
verse is well-formed and the mismatch avoids monotony. On the other
hand, the iambic tetrameter line in Figure 10.1 would be metrically
much less acceptable if it started with *Around me the birds. . . because
the would be in the S position in the second foot. Moreover, if there
were no mismatches, as in the construct The birds around, they hopped and
played, where phrase and hemistich boundaries align, the line would be
rhythmically less interesting. Thus the mismatch in the actual line is
not just metrically tolerated, it contributes to the aesthetic value of the
line by introducing a small irregularity. The example also illustrates an
important analytical principle: metrical well-formedness is a gradient
property of all types of verse.
The simplicity of the correspondences in Figure 10.1 is deceptive; the
forms of English verse covered in the following sections are much more
complex, but the principles remain the same. The metricality of a verse
piece is evaluated on the basis of matches and mismatches between the
abstract metrical template and the properties of the linguistic material
used in the verse.
10.2 Alliterative verse

As is evident from the name of this type of versification, alliterative
verse relies on alliteration, defined for the purpose of Germanic and
OE meter as identity of the onsets of stressed syllables. More spe-
cifically, the recurrent alliterating linguistic units are equivalent along
three parameters: phonetic properties, onset position in the syllable and
placement in a prosodically prominent syllable.
The etymological relation between Lat. littera ‘letter’ and alliteration
can be taken too literally; the reference to identity of initial letters was
pervasive in the early definitions of the term. The orthographic aspect
of alliteration should be kept out of the linguistic analysis: <f> in folc
‘folk’ and <ph> in Pharaon ‘pharaoh’ alliterate in OE, and PDE city
slicker is an alliterating phrase, while giant gorilla is not. Although the
dictionary definitions today emphasise the phonological rather than
the visual identity in alliteration, the restriction to stressed syllables is

often ignored, so that strings such as you better believe it, below the belt, are
considered ‘alliterating’ in the broader sense of the term, satisfying two
out of the three parameters listed above: phonetic identity and onset
position. :
As will become clear in this section, the interpretation of what con-
stitutes identity in alliterative verse is not static. The requirements can
be more or less restrictive; the practice of the poets may shift depend-
ing on the modes of composition and consumption: primarily oral, as in
Germanic and OE, or mixed oral and written, as in ME.
Proto-Germanic and all of the older Germanic languages were char-
acterised by the placement of primary stress on the first root syllable
(see 3.4.3). Against this prosodic foundation the choice of left-edge
identity as a unifying feature of the verse line appears to be a natural
artistic extension of the left-prominent prosodic patterns of Germanic.
Put differently, in a language with initial word-stress, the development
of a verse form organised around alliteration corresponds to one of the
most basic principles of the interface between language and meter, the
‘fit’ between stress and ictus. Our starting point, then, is the assump-
tion that the prosodic structure of the Germanic languages, placing
highest prominence on the initial syllable of the root, is ideally suited
for alliteration. :
10.2.1 Classical OE alliterative verse

The OE alliterative tradition is part of the indigenous Germanic verse
tradition. Attempts to look for the origin of OE alliteration outside
Germanic are unconvincing; ‘cultural’ borrowing of the model is
ruled out by the alliterative inscriptions from Anglian territory before
AD 400.2 Alliteration flourished in Germanic both for the mnemonic
and aesthetic reasons that generate it in the poetic models of other
languages, and for internal linguistic reasons.
The composition of alliterative verse in Anglo-Saxon times stretches
over the entire OE period. The datable chronological end points of the
surviving material range from 737, the Northumbrian ‘Moore’ manu-
script of Cædmon’s Hymn, the verse account of the miraculous invention
of alliterative meter by an illiterate cowherd, to the 1065 poem The
Death of Edward, the last OE composition which can be described rea-
sonably as belonging to the classical metrical tradition of Anglo-Saxon
2
See Lapidge (1979: 219–20) for a summary of the arguments and a vigorous defence of
the indigenous nature of OE versification.
versification. The entire OE poetic heritage was meticulously edited

and annotated in the monumental six-volume collective edition of The
Anglo-Saxon Poetic Records (Krapp and Dobbie 1931–53).3
Germanic and OE versification is notoriously difficult to model.
Although new theories of OE meter continue to appear, no new
approach rivals the descriptive adequacy and broad scholarly acceptance
of the observations and patterns in Sievers (1893). : Although there
are no agreed principles of interpretation of all the relevant parameters
determining the metrical structures of OE alliterative verse, the obliga-
tory presence of alliteration as a structural binding factor and as a central
linguistic means of creating an artistic effect is (almost) universally rec-
ognised. As noted in 9.3.2, the vanishingly rare occurrence of verse lines
lacking alliteration, amounting to as little as 0.001 per cent in the poetic
corpus, allows us to rely on alliteration as a basic correlate of stress.
Other regularities in the metrical structure of OE verse are listed in (1).
(1) Some basic descriptive features of OE alliterative verse:
• A (long) line consists of two verses the ‘on-verse’ (‘a-verse’) and
the ‘off-verse’ (‘b-verse’).
• Each verse/half-line has two feet and at least four positions.4
• Each foot must contain an ictus (S), also known as a ‘lift’, and at
least one non-ictic position (W), also known as a ‘dip’.
• The first foot is stronger than the second.
• The verses are linked by alliteration. The on-verse may have
two alliterating strong positions. In the off-verse, only the ictic
syllable of the first foot participates in the structural alliteration.
This allows us to represent the structure of the line as in Figure 10.2,
where F stands for FOOT and the numbers at the bottom stand for
POSITIONS.
The binary representation in Figure 10.2 is an abstract template
based on the minimal line structure in terms of syllable count. Verses
shorter than four syllables are defective. Four-syllable verses like ne leof
ne lað ‘not friend nor enemy’ are not the most common type; not exceed-
ing 25 per cent in the corpus.5 The foot-internal prominence relations
have to be left unspecified; S’s (lifts) and W’s (dips) can appear in either
order, allowing a mid-verse stress-clash, where positions 2 and 3 are
filled by stressed syllables, as in beléan míhte ‘dissuade could’.
3
Accessible as public domain at <http://www.sacred-texts.com/neu/ascp/> (last
accessed 30 May 2013).
4
Verses of more than four positions are considered ‘hypermetric’.
5
The estimate is based on the statistics in Hutcheson (1995: 175–269).
(Long) Line
On-verse/a-verse Off-verse/b-verse
Feet F F F F
Positions 1 2 3 4 1 2 3 4
ne leof ne lað be- lean mih- te
‘not friend nor enemy dissuade could’ (Beowulf 511)
Figure 10.2 The structure of the Old English verse line
One of the most striking features of OE meter, making it speech-like,

is that the length of the long line is variable and can contain from eight
to as many as fifteen syllables. This is related to the uneven size of the
lifts – one or two syllables, and especially the dips – zero to three, four,
very occasionally even more syllables. :
As stated in the third bullet in (1), each foot must contain an ictus
(S). A monosyllabic foot, where the W position is empty (optionally
analysed as filled by a zero syllable), is allowed only if the adjacent foot
is minimally trisyllabic; otherwise a foot must have at least one syllable
filling a non-ictic position W.
The size of W positions varies. If a W position is filled by more than
one syllable, it is known as an ‘expanded’ dip, that is, W = [ww], [www],
and so on. Verse-final weak positions cannot be expanded: W ≠ [ww]verse.
Expanded W positions are preferred in the first foot of a verse: [wws],
[wwws], or [sww], [swww] are very common in the first foot of the
verse. Expanded dips appear in the second feet of verses only in 2–3
per cent of the verses. Note that foot-boundaries never separate a
string of weak syllables – adjacent w’s are always subsumed under a
single non-ictic position. This is a metrical convention: for the purpose
of meter, in, for example, œ̄ nne ofer ȳðe ‘alone over waves’ (Beowulf 46a),
scanned as [swww / sw], the syllables . . . ne o.fer are not distinguished
in metrical strength, though in speech the first syllable of ófer would be
realised as more prominent than the adjacent syllables to the left and
the right.
A strong position S is typically filled by a single syllable, and it
must be filled by at least one syllable. The one-to-one correspondence
between a syllable and a position may be disregarded for S-positions
under a special metrical convention known as ‘resolution’. Resolution
is a metrical equivalence: one and only one stressed heavy syllable (H)
can fill a lift, but a light stressed syllable (L) and any other unstressed
syllable (X) may jointly fill a lift to avoid an unacceptable metrical
violation, such as two unstressed syllables/an expanded dip at the right
edge of the verse. This metrical equivalence, under which, for example,
the words cyning ‘king’ and ēa ‘river’ are ‘the same’, requires onset-
maximal syllabification, that is, intervocalic singleton consonants are
syllabified as onsets of the vowel to the right (see 2.3.2). The convention
is verse-specific and suggests that the recitation style observes onset-
maximality, which may not have been true in casual speech. Thus in
the verse réceda under róderum ‘of halls under heavens’ (Beowulf 310a), the
syllables ró.de- are metrically subsumed under the S position to avoid
the unacceptable verse-final [ww] in-de.rum. One notation for resolution
is the insertion of a hyphen between the light stressed syllable and the
following syllable, so ró.de is [s-ww].
Verse-internal resolution is illustrated in (2) where slashes separate
feet, lower-case s’s and w’s match syllables, and upper case S’s and W’s
represent positions in the abstract metrical template.
(2) Verse-internal resolution in OE verse:
ond féo rum gú me na → [w s-w / s-w w] = W S / S W
| | | |
L H L L
‘and lives of men’ Beowulf 73b
hǽ leð under héo fe num → [s-w w w / s-w w] = S W / S W

| | | |
L H L L
‘heroes under heaven’ Beowulf 52a
The norm of four positions in the verses in (2), marked with the capi-
talised equivalent scansions (W S / S W for Beowulf 73b and S W / S W
for Beowulf 52a), is achieved only if the light stressed syllables in these
verses fill an ictic position together with the following syllable. When
this adjustment is made verse-initially or after an unstressed syllable,
the resolvable syllable may be heavy, as in féorum ‘lives’ and hǽleð
‘heroes’. Indeed, the examples in (2) can be taken as illustrating another
specificity of OE verse: in some frameworks resolution is considered
obligatory if the first ictic position in the verse is aligned with a light
syllable and if the word is not a finite verb – we will return to the special
status of inflected verbs below. On the other hand, subsuming a heavy
syllable under a resolved ictus is generally prohibited verse-finally:
an unstressed heavy syllable blocks resolution with a preceding light
stressed syllable, as shown in (3). :
(3) Verse-final suspension of resolution:

gúð rìnc mó. nig → [s s / s w] = S W / S W, not *[s s / s-w]
| | or *S W / S
L H
‘battle-man many’ Beowulf 838b
wið wrāð wé. rod → [w s / s w] = W S / S W, not *[w s / s-w]

| |
L H
‘against hostile troops’ Beowulf 319a
The restriction illustrated in (3) is known as Kaluza’s Law. Structurally
féorum ‘lives’ and hǽleð ‘heroes’ in (2) and mónig ‘many’, wérod ‘troops’
in (3) are identical, yet the heavy second syllable in these items counts
as filling an independent weak position only at the verse-end. This is
not to say that resolution is precluded from occurring at verse-ends;
compare with Beowulf 478a: on Gréndles grýre ‘in Grendel’s terror’, where
grýre ‘terror’ has two light syllables and the verse scans [w s / w s-w] =
W S / W S. It is hard to see how such complex metrical conventions
reflect linguistic realities. It is more likely that the general principle of
closure in metrical composition – flexibility at the left edge, and tighter
constraints at the right edge of the verse or the line – is at play in this
case.6
Creative compounding was a defining stylistic feature of OE versi-
fication. In compounds the primary stress is on the first root syllable
with secondary stress on the first syllable of the second root, thus báncòfa
‘bone-chamber’, méodosètla ‘mead-benches, lándfrùma ‘land-ruler (see
9.4.1). In the meter the secondary stress is preserved and a polysyllabic
compound can occasionally fill a whole verse.
(4) Compounds as whole verses in OE:
wílġesī̀þas ‘willing companions’ (Beowulf 23a) [s w / s w] = S W
/SW
híldewæpnum
̀ ‘with battle-weapons’ (Beowulf 39a)
lándgemỳrcu ‘shore-boundaries’ (Beowulf 209b)
In (4) the second elements -sī̀þas ‘companions’ and -wæpnum ̀ ‘with
weapons’ function as prosodically independent entities, and it is only
on the verse level that the prosodic strength relations are reflected – the
first foot is stronger than the second. When embedded within a longer
6
See Hayes (1983: 373) for the principle of closure in other verse traditions.
verse, the second elements of compounds can be within the same foot as
the first element. In such cases – for example médoærn
̀ micel ‘mead-room
big’ (Beowulf 68a), wiġes wéorðmỳnd ‘battle’s honour’ (Beowulf 65a) – the
second elements -ærn ̀ ‘room’, -mỳnd ‘-mind’ are metrically subordi-
nated on the foot level, leading to a somewhat different foot-internal
prosodic contour in performance. With this in mind, and allowing for the
syllabic expansions within the metrical foot, we can posit three types
of foot cadences, shown in (5), where parentheses indicate the optional
recurrence of unstressed syllables within a single metrical W. :
(5) Types of foot cadences in OE verse:
Falling [s w (w)] = S W monegum (mœ̄gþum) ‘from many tribes’
Rising: [w (w) s] = W S on bearm (scipes) ‘in the ship’s bosom’
Cascading: [s-w s w(w)] = S W méodosètla (oftēah) ‘meat-benches
seized’
As already noted, the arrangement of the feet in a verse is only mini-
mally linearly regulated in the sense that a syllabically short foot is
balanced by a longer foot; there are never two feet within a verse both
missing a filled weak position. Since the W position in the first foot is
most commonly the one that hosts multiple unstressed syllables, the first
foot tends to be longer, containing more syllables. In terms of metrical
prominence, the first foot in a verse is considered stronger, because nor-
mally its ictus carries the structural alliteration – this is recorded in the
fourth bullet in (1) above. The left-hand prominence can be taken one
level further: the selection of alliterating ictic positions across the long
line suggests that the a-verse is metrically stronger than the b-verse,
because only the former allows both ictic positions to be filled by alliter-
ating syllables. The second ictus in the b-verse, or the fourth ictus in the
long line, does not contribute to the structural alliteration – alliteration
is systematically avoided in that position, the fifth bullet in (1).
Trivial as it may sound, this statement needs further elaboration.
Identity of the onsets of the syllables in ictic position in at least the first
foot of each verse is a requirement for the line to be metrical. Double
alliteration is allowed only in the a-verse; the occurrence of double
alliteration can be quite high, ranging up to 69 per cent of the a-verses in
some poems, 49 per cent in Beowulf and a mean of 46 per cent across the
corpus.7 In some verse-types, notably some A-, D- and all E-type verses
(see (9) below), double alliteration is the norm.
Further alliteration conventions which refer to segmental identity in
the stressed-syllable onsets are shown in (6).
7
The statistics are from Hutcheson (1995: 271).
(6) Alliteration conventions for OE meter:

(a) Singleton onset consonants in ictic positions must be identical:
bēaga bryttan, / on bearm scipes
m&¯rne be mæste. / Þ&¯r wæs mādma fela Beowulf 35–6
‘the bestower of rings / in the bosom of the ship,
mighty by mast / there were treasures many’
(b) Vowels alliterate with each other:

īsig ond ūtfūs, / æþelinges fær Beowulf 33
‘icy and out-keen, / hero’s vessel’
Þā þæt Offan m&¯g / ǣrest onfunde
þæt se eorl nolde / yrhðo geþolian Maldon 5–6
‘when Offa’s kinsman / first understood
that the earl would not / slackness endure’
(c) sp-, st-, sc- alliterate as groups:

<sp>: <sp>:
ond on spēd wrecan / spel gerāde Beowulf 873
‘and on success create / a skilful tale’
<st>: <st>:
Þā stod on stæðe, stīðlīce clypode Maldon 25
‘stood then on the place / sternly called’
<sc>: <sc>:
Oft Scyld Scēfing / scēaþena ðrēatum Beowulf 4
‘oft Scyld Scefing / from bands of enemies’
The identity of the onsets in (6a) is straightforward. The only complica-
tion arises with the velar and palatal <g>’s, which alliterate freely with
each other only in the early verse, for example Beowulf.
(7) Alliteration on voiced velars in early OE verse:
Swā sceal ġeong guma / gōde ġewyrcean Beowulf 20
‘So shall young man / good deeds work’
In Beowulf 20 ġeong ‘young’ has the allophone [j] from PrG /γ/, while
guma ‘man’ and gode ‘good deeds’ have the [γ] allophone, which became
a separate phoneme /g/ in the tenth century (see 4.2.1, 5.1.2). Syllables
beginning with the consonant spelled <c>, occasionally <k>, alliterate
freely throughout the corpus, irrespective of the later history of the con-
sonant, so we get cýnerìċe ‘kingdom’ matching ċild ‘child’ (The Death of
Edgar 975). Initial <h->, either alone or in clusters, alliterates with itself.
Vowel alliteration is based not on the quality of the vowel peak in the
stressed syllable, but on the insertion, especially in slow declamatory
performance, of a glottal stop [ʔ-] in the onset. Indeed, calling the exam-
ples in (6b) ‘vowel’ alliteration is a misnomer – the identity is based on
a consonant characterised by the laryngeal feature of constricted glottis.
Cluster alliteration, as in (6c), points to the strong cohesiveness of
the [s-]-initial clusters when the second consonant in the cluster is a
voiceless stop. There are no cases of splitting these clusters in the OE
poetic corpus: alliteration of <sp-> with <sm->, or <st->, or just <s->
does not occur. The cohesiveness of [sp-, st-, sk-] in English is a salient
phonological property: these are the only clusters that tolerate a third
consonant in the onset: spl-, str-, skw- are well-formed onsets in English,
but there no other CCC- onset clusters. The tradition of treating [sp-,
st-, sk-] the same way as singletons in alliteration continued in ME (see
10.2.2), although it was not as rigorously observed as in OE.
There are also conditions on alliteration which have to do with the
relative prominence of syntactic constituents. As discussed in 9.4.2,
Sievers’ Rule of Precedence stipulates that ‘if an inflected verb precedes
a noun, it does not have to alliterate, that it must not alliterate if the
noun does not alliterate too, and that a non-alliterating noun can never
be followed by an alliterating finite verb’ (1893: §§22–9). (8) illustrates
some of the options; the alliteration is marked in bold.
(8) Alliteration involving finite verbs in in OE verse:
(a) hē þæs frōfre gebād Beowulf 7b
‘he of the relief experienced’
(b) þenden wordum wēold Beowulf 30a
‘when with words wielded’
(c) in worold wōcun Beowulf 60a
‘in the world were born’
(d) ne hyrde ic cymlicor Beowulf 38a
‘heard not I comelier’
(e) wēox under wolcnum Beowulf 8a
‘grew under the skies’
(f) Hī hyne þā ætb&¯ron Beowulf 28a
‘They him then carried’
The higher strength of complements in relation to their verbs is a
shared and persistent feature of Germanic prosody. : The alliterative
behaviour of the verb in OE verse reflects this situation. When the verb
is in clause-final position, its complement carries the alliteration as in
(8a), Beowulf 7b; in this case the linear Rule of Precedence and the pro-
sodic relations of complement-verb overlap, and the verb alliterates only

if the complement does too, as in (8b), Beowulf 30a. Intransitive verbs are
no exception to the reluctance of predicates to be stressed over argu-
ments, as in (8c), Beowulf 60a. When the verb is in clause-initial position,
the alliteration can skip it, as in (8d), Beowulf 38a; or the verb can co-
alliterate as in (8e), Beowulf 8a. If the finite verb is the most prominent
word in the string, and there are no nouns, adjectives or heavy adverbs
in the same verse, the verb carries the alliteration, as in (8f), Beowulf 28a.
The type represented in (8f) is known as Type A3; the most remark-
able aspect of this type is that if one posits an ictus earlier in the verse,
it would be a non-alliterating ictus. The single alliterating ictus in such
verses does not have to be a finite verb. Since the alliterating ictic syl-
lable is always preceded by at least two monosyllables, the question
of whether one of them carries higher stress is open; what unifies the
pattern is the unambiguous alliteration on the first syllable of the last
word in the a-verse. Predictably, verses with single alliteration on
the last ictic position in the verse are almost entirely restricted to the
a-verse. Such verses can either be classified as a separate type or they
can be subsumed under type A; we will follow Hutcheson (1995: 198–
203) in bundling those together with the common Type A. :
Although the principles of scansion are still debatable, and the typology
of verses based on Sievers’ original classification has been critiqued and
modified numerous times, the classical division into alphabetically named
types continues to dominate the discourse on OE meter. The Sieversian
taxonomy is shown in (9), with one addition, the three-stress type.
(9) Sievers’ Five Types (expanded):
Scansion Mnemonic Examples
Type A: [s w (w) / s w] Arnie’s army Adam ærest Glc 826a
Type B: [(w) w s / (w) in briny baths þurh þæs Andreas
w s] beornes breost 1279a
Type C: [(w) w s / s w] by cost-cutting ofer cald Andreas
cleofu 310
Type D1: [s (w) / s s w] dumb dogcatchers deorc Beowulf
dēaþscua 160a
Type E: [s s w / s (Ø)] egg-laying hen eallisig tungl MB 24.23
Type 3: [s / s w / s] fierce fighting holm heolfre Beowulf
tribe weoll 2138a
Again, parenthesised w’s indicate the possibility of expansion of the dip.
The phrases in the middle column are PDE constructs intended to imitate
the cadences of the verses and possibly help with the type identification.
The variety of verse-types may seem bewilderingly rich, and indeed it is,
but it has to be acknowledged that statistically Type A verses are by far
the best represented type, amounting to 47 per cent of the a-verses and
39 per cent of the b-verses in the corpus.8 A second observation about the
density of types is that the first three types, A, B and C together comprise
76 per cent of the a-verses and 82 per cent of the b-verses. Types D and
E always involve compounds: in Type D the compound is to the right,
and in type E the compound is initial in the foot. In spite of their relative
rarity overall, these two types represent a major characteristic property
of OE meter. Their abandonment is an important diagnostic of ‘the end’
of the Classical OE meter.9 Type 3 [‘three’], introduced by Hutcheson
(1995), is distinguished from all other types in that it has three independ-
ent lexical stresses, and the items bearing the stresses are not identifiable
as compounds. The overall number of such verses is not high: 2.7 per
cent in the a-verse and 2.6 per cent in the b-verse.
Nearly all verses are complete syntactic units. The smallest morpho-
syntactic units that occupy a verse are compounds, as shown in (4). Most
often, however, a verse is coextensive with a phrase or a clause: Hī hyne
þā ætbœ̄ ron / tō brimes faroðe ‘they him then carried / to the sea’s current’
(Beowulf 28).
Classifying a particular verse according to the typology shown in (9)
is not always straightforward, and the descriptive details are complex. In
order to fit the linguistic material under the idealised four positions in
Figure 10.2, one needs to accommodate the uneven matching of syllables
to metrical positions. One such accommodation is resolution. Another
one is the expandability of the non-final weak position in the verse.
Additionally, there is a consensus in the OE metrical scholarship that
certain unstressed syllables are ‘invisible’ to the meter; they are extra-
metrical. Extrametricality is commonly associated with the prefixes ġe-,
be- and the negative proclitic ne. Occasionally the same proclitics have
to be counted to get the obligatory fourth position, for example Fyrst
forð ġewat ‘time forth passed’ (Beowulf 210a), where ġe- has to fill a weak
metrical position. The placing of extrametrical syllables, not just clitics
but any unstressed syllables, before the first foot of the verse is known as
anacrusis. An initial S W foot in an a-verse may be preceded by at most
two extrametrical weak syllables. Anacrusis is rarer in the b-verse and
there can be at most one unstressed syllable in anacrusis there.
8
Percentages from Hutcheson (1995: 297), calculated from a database of 13,044 long
lines, representing approximately 40 per cent of the extant OE poetic texts.
9
For statistics and discussion of the loss of these types in early ME see Cable (1991:
52–65).
Assigning syllabic structure and prosodic values to OE verse, like

scanning any verse extant only in written form, involves judgements
based on the reader’s understanding of the template, the independently
reconstructed phonology, and the reader’s interpretation of what needs
to be foregrounded in the line – aesthetic and semantic preferences that
can be intentionally emphasised in performance. The latter aspect of
verse lies outside the narrower limits of the linguistic correspondences
of prosodic to metrical units, as idealised in Figure 10.1. The matching
of the template to the phonology is less problematic when the phonol-
ogy is categorical about a certain property, as are the stress contours of
the words in the examples in (10).
(10) Unambiguous stress-to-ictus and syllable-to-position
correspondences:
strēamas wundon ‘streams wound’ [s w / s w] = S W / S W Beowulf 212b
cealde strēamas ‘cold streams’ [s w / s w] = S W / S W Beowulf 1261a
In such verses a word like OE strēamas ‘streams’ will always have its
initial syllable more prominent than the second syllable containing
the plural inflection. The syllable -mas in this word must be placed in a
metrically weak position, it cannot be ignored in scansion, it cannot be
resolved because the preceding syllable is heavy: the word is unambigu-
ously disyllabic and its syllables must be placed in metrically distinct
positions.
In many instances, however, the phonological underpinning may be
ambiguous; compare the PDE options of hárass-haráss, or the variable
syllable-count in botany, natural, Indian – three or two syllables depend-
ing on the rate and style of speech. The competition of variables is
always present in language and it is an important part of the process
of language change. There can be no doubt that poets are aware of the
options in the ambient language and can take advantage of them in
order to conform more closely to the chosen metrical template. The
‘metrically relevant range of phonological representations’ of the forms
used in verse is known as paraphonology (Kiparsky 1977: 190). The
paraphonology refers to non-categorical properties used selectively in
the verse. It is the ‘phonology of opportunity’, drawing on the variability
of spoken forms, available to poets at all times. :
Reference to optional features is essential to the reconstruction of
both the metrical forms and the linguistic forms in ME verse (see 10.2.2,
10.4), where the statistical preponderance of unambiguous matches
gives us a basis for formulating hypotheses about segmental and pro-
sodic change. For the OE corpus, we can mention two paraphonological
properties that are of interest: the varying syllabicity of post-obstruent

coda sonorants, because it is valuable in reconstructing the date of com-
position of a poetic piece; and elision of final <-e>, because it plays a
central role in the later verse forms.
In OE, original monosyllables with codas in obstruent + sonorant
could be treated either as monosyllabic or as disyllabic, where the
second syllable peak was either the sonorant, which became syllabic, or
a schwa, followed by the sonorant; the development of a second sylla-
ble is labelled parasiting. The ‘parasitic’ schwas can be reflected in the
orthography, for example <wœ̄ pn ~ wœ̄ pen> ‘weapon’ < Goth. *wēpn,
ON vapn, Modern Icelandic vopn. As shown convincingly by Fulk (1992,
2007), the monosyllabic forms are most common in poems that ‘have
traditionally been regarded as likeliest to be relatively early composi-
tions’ (Fulk 2007: 306). Other items that can have monosyllabic realisa-
tions in verse are bēcn ‘monument’, hleahtor ‘laughter’, winter ‘winter’,
wundor ‘wonder’. The ‘parasitic’ schwas in such words survive to this
day, and they alternate with syllabic sonorants in fast speech.
Elision involves the deletion of an unstressed final [-ə] before another
vowel to the right; the avoidance of two unstressed vowels in hiatus is
common cross-linguistically; PDE disallows word-internal adjacency of
two vowels, of which the first one is schwa or any short vowel.10 As noted
in 7.6, the orthographic evidence for unstressed vowel elision goes back
to OE; for example, the ‘correct’, expected forms of the verbs in <wene
ic> ‘I hope’, <sæġde ic> ‘I said’ appear as <wen ic, <sæġd ic>. In the verse
final [-ə] elision appears to have been optional.
(11) OE elision optional; ambiguous scansions:
&¯ nne ofer ȳðe [s w w w / s w] or [s w w / s w] = SW/SW
Beowulf 46a
‘alone over waves’
wanode ond wyrde [s-w w w / s w] or [s-w w /s w] = S W /S W
Beowulf 1337a
‘lessened and destroyed’
r&¯sde on ðone rōfan [s w w w w / s w] or [s w w w / s w] = S W /S W
Beowulf 2690a
‘rushed at the renowned’
10
Allerton (2000) discusses the current tendency to avoid such sequences in morpho-
logically complex words and across word boundaries in BrE.
The examples in (11) illustrate that in most cases where elision can be
posited, the verse will be well-formed with or without the extra weak
syllable, so the performance/realisation of the verse is open-ended. The
choice can be based on the relative frequency of either option, so that
for Beowulf 2690a the second option, with a shorter dip in the first foot ([s
w w / s w]), would be statistically better attested. Alternatively, one can
argue that elision would be avoided in the formal declamatory style of
oral poetry, giving preference to the first, more extended-dip scansions
in (11). :
A more detailed account of OE verse was given here partly because
of the general unfamiliarity of the topic, and partly because OE verse
provides a basis for comparison with later forms of versification, the
topic of our next section.
10.2.2 Continuity and reinvention of alliterative versification in ME11

The 1065 poem The Death of Edward is the last composition that can
be described reasonably as belonging to the Classical metrical tradi-
tion of Anglo-Saxon versification. Very revealing in this respect are
the statistics and the comments presented in Cable (1991: 54–5). He
notes one single metrically dubious verse (soþfæste sawle ‘soothfast soul’
(28a)) in the sixty-eight verses of The Death of Edward, while on the
other side of the 1066 chronological divide the next extant poem with
prominent alliteration, Durham, composed c. 1100, shows a very high
level of unmetricality. In Durham 38.1 per cent of the forty-two verses
fail to conform to the classical rules. Thus, while the cataclysmic effect
of the Norman Conquest with respect to changes affecting phonology
and morphosyntax can be questioned, the demarcation line in terms of
versification modes is clearer, at least within the inevitable limitations
imposed by the surviving texts.
The absence of records showing that alliterative verse continued to
be composed after the Conquest leaves us in the dark as to the familiar-
ity and popularity of this kind of versification in the next two centuries.
Some superficial formal similarities, such as the use of alliteration
and free distribution of unstressed syllables, have been seen as a link
between some late tenth-century Ælfrician prose texts and early ME
verse compositions such as The Proverbs of Alfred, The Worcester Fragments
of the Soul’s Address to the Body, The Bestiary and Lagamon’s Brut. However,
the link is tenuous, and positing continuity of an oral poetic form on the
basis of vague analogy to literary prose is unjustified, as Fulk (2004) has
11
This section uses material and arguments from Minkova (2009a, 2009b).
argued persuasively. It is much more probable, though not testable

because of the nature of oral traditions, that a body of alliterative verse
compositions did exist in the twelfth and thirteenth centuries. The
early ME compositions mentioned above are ‘hybrids’ both in structure
and genealogy: they mix rhyme, alliteration and syllable-counting in
often erratic patterns, and they combine features of the traditional oral
alliterative versification with the more literary features of rhyme and
syllable-counting which were gaining prestige in the Anglo-Norman
cultural setting. :
In the fourteenth century a significant portion of the literary activ-
ity in many northern, western and south-western areas was channelled
into the composition of alliterative verse, culminating in masterpieces
like Winner and Waster (W&W), Sir Gawain and the Green Knight (SGGK),
The Wars of Alexander (WA), Piers Plowman (PP) and The Alliterative
Morte Arthure (MA). These works are associated with a literary period
in English known as the Alliterative Revival. The Revival flourished for
about a century, starting around the 1330s and dying out in Scotland in
the fifteenth century.
If we accept the logical premise that a body of oral alliterative
verse must have existed in post-Conquest England continuously until
the fourteenth century, our description of the meter of the Revival
metrical forms can start with a comparison of the features of OE verse
with the features of ME alliterative verse. Before proceeding with
the structural similarities and differences, we will survey the ways in
which the technique of the ME versifiers matched or differed from
the technique of the Old English scops. The patterns of onset identity
observed in OE, shown in (6a–c), were still present in ME: singleton
onsets have to be identical, different vowels may alliterate with each
other, and [sp-, st-, sk-] alliterate cohesively. However, all three pat-
terns are subject to further qualifications in ME which reflect new
phonological variables in the language. Singleton onsets have to allit-
erate in ME verse, but some poets or copyists resort to dialectal bor-
rowing in alliteration, so we find /f-/ : /v-/, where /v-/ is the south
and south-west Midlands pronunciation of /f-/, and /w-/ : /v-/, also
dialectal variants of the fricative in initial position in the south (see
4.4). Another innovation is the occasional matching of <h->-initial
words with vowel-initial words, indicating the instability of onset /h-/
(see 5.1.3).
One of the most interesting aspects of the choice of alliterating words
in ME verse is the systematic avoidance of vowel alliteration. In OE
the rate of vowel alliteration is within the expected range: 15.5 per cent
of the lines in Beowulf have vowel alliteration. The ME corpus shows a
dramatic drop in vowel alliteration, ranging from 1.1 per cent to at most
5 per cent of the lines, which is significantly below the expected random
distribution of alliteration on consonants vs alliteration on vowels.12
Moreover, even in the rare cases of vowel alliteration, the poets strongly
prefer identical vowels, as illustrated in (12).
(12) ME alliteration on identical vowels:
And his arsounz al after / and his aþel skyrtes, SGGK 171
There alienes, in absence / of all men of armes MA 273
Offirs all his old gods / his honour þam thankis WA 3658
Of Appils & almands / & all manere of frutis WA 4718
The rationale for these choices is best explained with reference to the
obligatory nature of stressed-syllable onsets in OE (see the discussion
of (6b) in 10.2.1). In OE the glottal stop provided the basis of onset
identity in orthographically vowel-initial stressed syllables. In ME
the realisation of a glottal stop in the same position became optional.
Anglo-Norman does not mandate a filled onset and it is likely that this
could influence the realisation of vowel-initial stressed native words
too, especially for non-monolingual speakers. Although some residual
mixing of vowels across the line survives, loanword phonology would
be an incentive for the selection of identical vowels.
Another aspect of ME alliteration that goes hand in hand with the
change in the onset-constraint on stressed syllables is the so-called
Liaison Alliteration or Stab der Liaison, illustrated in (13).
(13) Liaison Alliteration (Stab der Liaison):
‘Þat schal I telle þe trwly,’ quoþ þat oþer þenne SGGK 2444
Takis þam with him to his tent & þam at ese makis WA 1955
Vmquile he noys as a nowte, as a nox quen he lawes WA 4871
In these examples the poet relies on resyllabification within the clitic
phrase: the clitic coda becomes an onset to the stressed vowel-initial
heads in the groups that other, at ease, an ox. That onset is then the one
that carries the alliteration, as in tell, truly: (tha)tother, take, tent: (a)tease,
noys ‘annoys’, nowte ‘bull’: (a)nox. The reality of this reassignment in
speech can be confirmed independently; for example, the phrase ‘at an
oven’ is attested in the surnames Roger atte Novene of Walyngford (1323),
Rico ate Nouene (1327) (MED entry under ‘oven’). No such examples are
found in OE verse, which suggests that the resyllabification in ME is
enabled by an ongoing change in the realisation stressed vowel-initial
words.
12
See the discussion in Minkova (2003: 239–40 and passim).
The speech-based innovative alliterative techniques discussed here

point to the survival of a strong oral component in ME alliterative
verse. This is not to say that the visual image of a letter can be excluded
completely from the alliterative techniques, especially as ‘orality’ may
be more a matter of delivery than of composition. In other words,
one should ask the sceptical question, how much of alliteration was
‘literal’, also ME ‘letterall’ (1450)? By definition, all of the alliterative
material we have inherited has survived by virtue of being written
down or copied, mostly anonymously. The only reliably identifiable
‘author’ of OE poems was Cynewulf (post-Beowulf, pre-Alfredian, late
eighth century, Mercian), who left his signature in the poems. Other
attributions (Cædmon/Bede, Wulfstan, Aldhelm) are questionable. All
of the major Anglo-Saxon poems are anonymous, and so are the ME
alliterative poems. Authorial anonymity is not synonymous with oral
composition, however, and the possibility of influence of the written
shape of words on the selection of matching onsets cannot be fully
excluded.
Already in OE, the difference between a letter and the sound it
represents would have been familiar to any well-educated cleric. :
Two aspects of the ME alliterative technique suggest that the identity
of letter-shape should also be considered in evaluating the poets’ craft.
The first one is the apparent stress-promotion of prefixes as in rewarde:
Rome; persayfede: Peter, where the allure of letter-sameness seems a better
explanation of the matching than positing initial stress on *pérceived
or *réward. The second point, admittedly negative, is the absence of
‘erroneous’ alliteration of consonants that are phonetically very close.
If, as noted above, cross-dialectal /f-/ : /v-/ and /w-/ : /v-/ could be
matched in alliteration, we would expect highly confusable spirants in
the same dialect, such as /f/ and /θ/ to be paired in alliteration, albeit
rarely.13 That such alliterations are not found in ME suggests that the
poets or copyists tolerated letter over sound identity and cross-dialectal
forms, but they did not deviate from the sound-identity norm if the
letters were not identical.
A comparison of the OE and the ME alliterative long line reveals
both similarities and differences. The basic descriptive features of the
Classical OE compositions, listed in (1), are all identifiable in ME too: a
line consists of a-verses and b-verses, each verse has a minimum of two
feet and four positions, each foot contains an ictic (S) and a non-ictic
(W) position, the first foot is stronger than the second and the verses are
13
On the confusability of /f/ -/θ/ and <th->-fronting see Wells (1982, II: 329); Tabain
(1998).
Long Line
a-verse b-verse
F F F F F Feet
P P P P P P( ) P P P P( ) Positions
Þe tulk þat þe tram- mes of tre- soun þer wrot
w S W S () w S W S
Figure 10.3 Structure of the Middle English alliterative long line in

fourteenth-century verse:
S = a strong position, W = a weak position filled by two or three syllables,
w = a weak position filled by a single syllable, ( ) = an optional verse-final
w(eak) position
linked by alliteration. Within these general properties, however, there

are important aspects in which the two traditions differ.14 Using the
same primes as Figure 10.2, Figure 10.3 shows the structure of the ME
alliterative long line.
In addition to the familiar line-constituents – a-verse, b-verse, feet
and positions – the ME material justifies separate treatment of weak
metrical positions filled by a single syllable, which we can mark as [w],
and a ‘strong’ dip, corresponding to the ‘expanded’ weak position of
OE, marked [W]. The parentheses at the end of the verse indicate that
a single unstressed syllable at the end of the verse is not part of the
template – it is ‘extrametrical’.
Now we turn to the metrical frames in ME, bearing in mind that each
generalisation covers only statistically dominant attestations. The first,
and probably most conspicuous metrical difference between Old and
ME, recognised by several generations of metrists, is that within the ME
alliterative long line the a-verse and the b-verse are rhythmically differ-
ent. The concrete manifestations of the rhythmic difference are shown
in (14) and (15). :
14
Imprecise as it is to refer to the entirety of the OE poetic records as fitting a single
monolithic metrical template, it is even more of an idealisation to subsume all ME
alliterative compositions under a single template. Some differences are ignored here
in favour of presenting a more general schema applicable to the majority of the texts
and the lines.
(14) Three alliterating S-positions in the a-verse in ME:

And meled þus much with his muthe / as e may now here SGGK 447
A lowde latir he loe / and to þe lede said WA 96
To laite þaire lord at was lost / with latis vnblythe WA 152
First, the a-verse is heavier than the b-verse because it allows up to
three strong metrical positions, and all three may, but do not have to
alliterate. Verses with three ictic positions are known as ‘extended’.
The dashed line for the rightmost branch in the a-verse in Figure 10.3
marks the optionality of the third foot. Extended verses are avoided in
the second half-line.
Second, the a-verse and the b-verse differ in the distribution of W’s,
weak positions filled by two or more unstressed syllables: two W’s are
common in the a-verse, while the percentage of b-verses with two
heavy dips is very low. :
(15) Avoidance of two strong dips in the b-verse:
And bot the lengthe of a launde / thies lordes bytwene W&W 54
Quen he was semely vp set / with septour in hand WA 198
With alle þe mete and þe mirþe / þat men couþe avyse SGGK 45
The b-verse is more tightly constrained. It requires two lifts. As in OE,
the first of these lifts carries the obligatory alliteration, but unlike the
OE schema, alliteration on the second ictus is not unusual.
(16) Second ictus alliterating in the b-verse:
In a somer seson, / whan softe was the sonne PP P.1
Gawayn gly3t on þe gome / þat godly hym gret SGGK 842
Double-felde, as hit fallez, / and fele kyn fischez SGGK 890
The stricter limits on the arrangement of syllables in the b-verse war-
rants a summary statement on the types that can be labelled ‘metrical’.
The optional verse-final weak syllables are excluded.
(17) Metrical patterns in the b-verse in decreasing frequency:
w S W S: of trésoun þer wrot SGGK 3b
S W S: flákerande with wýnges W&W 92b
W S w S: vndyr the héuene rýche MA 108b
W S S: that one the bent houes W&W 105b
The rare types are listed in (18), again in decreasing frequency. Only
the types attested above 1 per cent are illustrated.
(18) Rare or unmetrical b-verse types:
W S W S: when the púrpos was takyn MA 415b
w S w S: with géntill knyghtes MA 372b
SwS
wSS
wSwSwS
SwSw
Clearly, the composition of alliterative verse in England in the four-
teenth century is based on a system that imitates closely, but does not
replicate the design of OE verse. There are new techniques of allitera-
tion, new constraints on the length of the verses and the distribution
of syllables in the dips. In addition to the overall rhythmic asymmetry
of the a- and b-verses, the preferred verse-internal cadences are dif-
ferent. One of the most striking discontinuities is the rarity in ME of
four-syllable verses, which, though not the most frequent type in OE,
still make up about one-quarter of the verse data in OE (see 10.2.1).
Another remarkable difference is that the pattern S w S w is unmetri-
cal in Middle English, while this is a common (sub-)A-Type in OE.15
The common OE (sub-)B-Type w S w S is at most marginally met-
rical in ME. Note that in both of these types the intervals between
single strong and weak syllables are identical; these are types where
the foot structure can be labelled purely trochaic or purely iambic.
This throws into relief the avoidance of simple stress-alternating
rhythm: the presence of a ‘strong’ dip is practically obligatory, making
this a salient characteristic of this tradition of versification, both in
comparison with OE, and in comparison with the contemporary
iambic meters. :
10.3 Introduction of rhyme, syllable-counting and binary foot

structure
Grounded in the prosodic pattern of stress on the first root syllable,
alliteration as a cohesive device survived in Middle English, but at
the same time new modes of versification based on rhyme and syl-
lable counting were gaining popularity. Early instances of occasional
ornamental rhyming can be found in the OE alliterative corpus.
(19) Rhyme as ornamentation in OE alliterative verse:
fylle gef&¯gon; / fægere geþ&¯gon Beowulf 1014
‘rejoiced in the feast / partook of it with relish’
Byrhtnoð maþelode / bord hafenode Maldon 42
‘Byrhnoth spoke / raised his shield’
15
The S w S w subtype in OE amounts to 16 per cent of the data in Hutcheson (1995:
175).
Such lines are rare and it is clear that rhyming, exact and inexact,
including assonance, was not an obligatory cohesive device. One famous
exception is the so-called Rhyming Poem, which combines alliteration
and rhyme, sometimes in consecutive lines, in what has often been
described as a tour de force of metrical technique. :
(20) Combining alliteration and rhyme in OE (The Rhyming Poem):
Scrifen scrād glād / þurh gescād in brād;
wæs on lagustrēame lād, / þær me leoþu ne biglād
Hæfde ic hēanne hād, / ne was mē in healle gād 13–15
‘The appointed ship glided through the distance into the broad sea;
there was a path upon the ocean stream, where I was not without
guidance.
I had high rank; I lacked nothing in the hall’
Scealcas w&¯ron scearpe, / scyl wæs hearpe,

hlūde hlynede, / hleoþor dynede 27–8
‘The servants were active, / the harp was resonant,
loudly rang; / sound pealed’
Swā nū world wendeþ, / wyrde sendeþ 59

‘thus now the world goes, fate brings’
The West European tradition of rhyming is associated with Church
Latin, starting with some rather crude correspondences in the second
century AD. : McKie describes three stages in the acceptance and
spread of rhyme in England:
Initially, rhyme was imported from Church Latin by Anglo-Saxon ver-
sifiers steeped in the culture of the Church. . . . Yet such use was occa-
sional, and in no sense prosodic. In the second stage, under the pressure
of linguistic changes brought about by the Conquest, narrative verse
in rhymed couplets emerged. . . . In the final stage . . . the simplicity of
couplet-rhyme was overtaken by the more highly prized complexity of
stanzaic rhyme-patterning. (McKie 1997: 829)
The Latin ‘rhythmi’, variously impure, are an important source for

the adoption of rhyme in Old English. The memory-aiding nature of
rhyming also contributes to the process. A good example of how the
Latin practice of rhyming was taken over by the Anglo-Saxon clerics is
the translation of a Latin proverb.
(21) Rhyming in Latin-English translation:
Ardor frigesscit, nitor squalescit, amor abolescit, lux obtenebrescit.
Hat ācolað, / hwit āsolað,

leof ālaðaþ, / leoht āðystrað
‘Heat cools / white(ness) tarnishes
Love becomes hateful, / light darkens’
By the eleventh century end-rhyme had become an essential feature
of Medieval Latin verse. That feature is never absent from Old French
verse. English versifiers, who were probably also good Latinists, could
not escape the influence of rhyme – they made increasing use of rhyme
for both aesthetic and mnemonic purposes. A possibly humorous entry
in The Anglo-Saxon Chronicle for 1076 tells us: he [Raulf] þa lædde þæt wif
to Norðwic, þær wæs þæt brȳdealo, þæt wæs manegra manna bealo (‘He led
the wife to Norwich, there was that bridal that was the destruction
of many men’). The early rhymes were commonly feminine rhymes: a
stressed syllable followed by an unstressed syllable, and often involved
inflectional and derivational suffixes, as in the examples in (21). In
the twelfth century the practice of rhyming was firmly established in
England, its popularity enhanced by the troubadour poets patronised
by Henry II and Eleanor of Aquitaine (McKie 1997: 828). In bi- and
trilingual England the versifiers sometimes mixed languages in choos-
ing the rhyming words: ‘in tabulis / la vile de Paris / so wel me is’.16 As
the use of rhyme spread, the matching was perfected, and the identity
of endings became a way of keeping verse lines together, giving rise
to the couplet.
(22) The short rhymed couplet in early ME:
Ure feder þet in heouene is,
þet is al soð ful iwis,
weo moten to þes wordes iseon
þet to liue and to saule gode beon Paternoster (12th c.)17
By the beginning of the thirteenth century the rhyming couplet had
become a form completely native to English versifiers, producing ver-
nacular masterpieces such as the anonymous Owl and the Nightingale. At
the same time, techniques combining alliteration and rhyme are used
well into the thirteenth century, as in the lines in (23) from The Proverbs
of Alfred.18
16
The macaronic example is from the love poem ‘Dum ludis floribus’, cited in McKie
(1997: 821). Further mixed rhyming AN-English (et leal ‘and faithful’: it fele ‘feel it’,
sovent ‘often’: ysend), or English-Latin (Kyng of Blys: virginis ‘by the Virgin’), and a dis-
cussion of code-switching in medieval English poetry is found in Schendl (2001).
17
Old English Homilies, First Series (Early English Text Society OS 29), p. 55.
18
Another text which famously bridges the techniques of alliteration and rhyme is
(23) Mixing alliteration and rhyme in ME:

þe lond to leden / mid laueliche dedin 74–5
þat þe chireche habbe gryþ. And þe cheorl beo in fryþ 87–8
Histories of English versification frequently emphasise ‘the pressure of

linguistic changes brought about by the Conquest’ (McKie 1997: 829)
as the impetus behind the second stage in the evolution of rhyming.
Optimally, there is a ‘fit’ between metrical form(s) selected by an indi-
vidual artist or a group of people and the language (see 10.2). However,
there is nothing ‘suboptimal’ in OE in terms of availability of rhyming
words, the ease of counting syllables or placing beats at regular inter-
vals. Therefore the repeated references to ‘language change’ are too
general; not every type of linguistic change can be associated with the
adoption of rhyme in English. :
The relationship between the loss of inflections in ME and the accept-
ance of rhyme is dubious; numerous verse traditions, including the
oldest known tradition of rhyming in Chinese, dating back to the fifth
century BC (McKie 1997), employ rhyme. The gradual perfection of
the rhyming patterns in Latin in the first millennium proceeded without
the crutches of newly developed analyticity, and rhyming characterises
the verse tradition in a strongly synthetic language like Russian. The
very existence of occasional rhymes, including some monosyllabic or
masculine rhymes in OE, makes the dependence of rhyming on loss of
inflections a difficult hypothesis to defend. The borrowing of Romance
vocabulary is a different matter; the presence in the language of adjec-
tives and nouns with final stress and the fact that many polysyllabic
loanwords had alternating stress might indeed have facilitated rhyming
by increasing the pool of available items, but it should not be consid-
ered the determining factor. Twelfth- and thirteenth-century rhyming
compositions, for example Poema Morale, Havelok the Dane, The Owl and
the Nightingale, use vocabulary that is almost exclusively Germanic. The
primary reason why rhyme became the verse-line marker of choice in
ME was the cultural model of Latin and Anglo-Norman.
Another important innovation in ME versification was the introduc-
tion of syllable-counting, also an imported metrical feature. Verses of
equal numbers of syllables, ‘isosyllabic’ verses, are found in OE poetry,
but their recurrence was not structurally regulated; a verse could have
from a minimum of four to fourteen syllables (see 10.2.1). The first
notable compositions based on the iteration of isosyllabic lines in ME
Lawman’s Brut, where one finds ‘pure’ alliterative lines, ‘pure’ rhyming couplets and
hybrid lines such as: he is king & heo is quene; of þine kume nis na wene (14,046).
are the mid-twelfth-century Poema Morale, which does what it says –

gives moral advice in 270–400 lines, depending on the manuscript –
and the autograph 19,000-line long poem The Ormulum (late twelfth
century). The pieces are monotonously regular, with fifteen syllables in
each long line, divided by a caesura after the eighth syllable. The form
is known as septenary or septenarius, a popular form in classical Latin
versification, illustrated with the first lines of these poems in (24).
(24) The septenary in early Middle English:
Ic em nu alder þene ic wes, a wintre ent a lare Poema
Morale 1
‘I am now older than I was in years and in knowledge’
Þiss boc iss nemmnedd forrþi þatt Orrm itt Preface 1
Orrmulum wrohhte
‘this book is named Ormulum because Orm wrote it’
These compositions were the work of highly trained monastics who
must have had exposure to the Latin form, which the monk Orm, for
one, applied with considerable care and precision. The septenary line
can also be analysed as a sequence of four and three iambic feet or eight
plus seven syllables, with a special constraint at the right edge of the
long line: the fourteenth syllable is always heavy and it is invariably
followed by a weak syllable which belongs to the same word.19 The
vocabulary of both texts is almost exclusively Germanic.
Another schema based on the iteration of isosyllabic lines is the
octosyllabic line, practised with various precision from the end of
the twelfth century onwards. The earliest extended poem of lasting
literary value which is in octosyllabic rhymed couplets is The Owl and
the Nightingale, whose composition is dated early rather than late in the
period 1189–1216 (Stanley 1972: 19). The opening lines of the poem are
given in (25).
(25) The Owl and the Nightingale, lines 1–6:
ICH was in one sumere dale,
In one suþe diele hale,
Iherde ich holde grete tale
an Hule and one nitingale.
Þat plait was stif & starc & strong, 5
19
The rightmost edges of any type of verse are known to be especially rigid metrically
(see note 6). In The Ormulum the letter of the fifteenth-syllable vowel is always <-e(-)>,
most likely a kind of schwa. The metrical significance of this fact is discussed in
Minkova (1996: 102–4).
sum wile softe & lud among;

(Stanley 1972: 49)
In addition to the prevocalic elision of <-e>, as in iherde ich ‘heard I’ (3),

hule and ‘owl and’ (4), and so on, the octosyllabic frame suggests the pos-
sibility of word-internal weak-vowel syncope, as in sumere ‘summer’ (1),
though both in this case and in the case of diele ‘hidden’ either elision
or syncope will get to the right number of syllables.
Exact syllable-counting cannot be claimed for all of the verse written
outside the ME ‘Alliterative Revival’ tradition. Even the highly skilled
early poets, and the author of The Owl and the Nightingale was certainly
one of them, allow syllabic ambiguity and occasional clear violations of
the syllabic count. But the target of isosyllabicity is undeniable: even a
conservative approach to the flexibility of the final unstressed vowel in
early ME shows that the syllabic regularity in this text reaches 70 per
cent. A less restrictive approach to the paraphonology of the unstressed
vowels will undoubtedly indicate even higher regularity. : The issue
of counting syllables will be revisited after a discussion of a related
feature of ME syllable-counting poetry, namely the organisation of
syllables into metrical feet.
The distribution and the syllabic content of weak and strong positions
in OE verse was flexible, and it is only after disassociating the numeri-
cal correspondence of syllable(s) from metrical positions that we can get
to the abstraction of W S and S W metrical sequences, as in (5) above.
Verses exhibiting one-to-one correspondence of syllable to position
did exist in OE, but they were not the statistical norm. The adoption of
isosyllabic versification in ME is linked to the rise of a new structural
component of verse: the binary foot. Using the terminology of classical
Greek and Latin versification, we will refer to the unit of a S(trong)/
ictic + W(eak)/non-ictic position as a trochee (S W), and to the unit of a
weak + strong position as an iamb (W S).20 The two metrical feet have
ready counterparts in the prosody of speech: unprefixed disyllabic words
(silver, scatter, holy, seldom) and monosyllables followed by clitics (heard it,
milk and, found them) are trochaic, prefixed verbs, adverbs, and groups with
monosyllabic clitics (beseech, outdid, about, the king) are iambic. In OE verse
the occurrence of the left-strong type (S W) was more frequent than the
(W S) type.
20
Originally, the terms are applied to quantitative meter, where S stands for ‘long’ and
W for ‘short’ syllables. Tri-positional feet: S W W (dactyl) and W W S (anapest), will
not be discussed here, because regular dactylic and anapestic versification in English
is outside the scope of this chapter.
In Middle English isosyllabic verse the iamb became the dominant

metrical foot, making it the cadence of choice for English poetry for
the entire second millennium. : The regular alternation of weak
and strong prominences, each one associated with a single syllable,
is characteristic even of the earliest post-Conquest syllable-counting
compositions such as The Ormulum and The Owl and the Nightingale.
(26) Iambic footing in early ME verse:
W S / W S / W S / W S /
An énn- gell comm off héoff- ness ærd;
‘An angel came from heaven’s region
W S / W S / W S / (W)
inn á- ness wé- ress hé- we
in one man’s likeness’ Ormulum 3336–7
The lines in (26) illustrate the default matching of syllables to posi-
tions: strong (S), or even, positions are filled by stressed syllables:
énngell ‘angel’, héoffness ‘heaven’s’, áness ‘(of) one’, wéress ‘man’s’, héwe
‘likeness’ are aligned with S W positions. Lexical monosyllables – comm
‘came’, ærd ‘region’ – also occupy S positions. The unstressed syllables
of disyllabic lexical words and function words – an, off, inn – are placed
in weak (W) or odd-numbered positions. Some derivational mor-
phemes such as -dom, -ing, -like, -ness can be treated as second elements
of compounds and like them, they can fill a strong metrical position;
compare (27a) and (27b), where the former illustrates the S-matching
of -dom.
(27) Derivational suffixes in strong metrical position in early ME
iambic verse:
And ec forr þatt he wollde swa
(a) W S W S W S
Þurrh hiss þeww- dom ut lesen
‘through his thraldom release’
(b) W S W S W S W S
Off def- less þeww- dom al le þa
‘from devil’s thraldom all then’ Ormulum 3618–20

As in OE (see 10.2.1, (4)), compounds such as twifald ‘twofold’, mannkinn

‘mankind’, summwhatt ‘somewhat’ allow the placement of either the first
or the second element in a strong metrical position; compare Orm’s All
mánnkinn forr to lokenn (3283) with Off all mannkínn onn eorþe (3273). :
An important principle of matching speech prosody to meter relevant
to the establishment of a binary foot structure in English verse is that the
matching of monosyllabic lexical items to metrical positions is flexible.
Monosyllables in isolation do not really have ‘stress’; their prominence
is determined in relation to adjacent syllables. In the early isosyllabic
verse placing a prosodically weak monosyllable in S is not a violation of
the metrical template, as in lines 3409 and 3411 of The Ormulum in (28),
where and in line 3409 and þurrh in line 3411 fill strong positions.
(28) Unstressed monosyllables in strong positions in The Ormulum:
& tokenn innwarrdlike godd;
W S W S W S (W)
To lo fenn & to þann kenn
‘To praise and to thank’
All þatt teȝȝ haffdenn herrd off himm;

W S W S W S
& se ȝhenn þurrh his are
‘and seen through his grace’ 3408–11
The observation that monosyllables are flexible in syllable-counting

binary-footed meters has been around for a long time. George Gascoine
(1575: 469) noted that ‘woordes of one syllable will more easily fall to
be shorte or long as occasion requireth, or wilbe adapted to become cir-
cumflexe or of an indifferent sound’.21 The free correspondence between
monosyllabic words and metrical positions in the pentameter was formal-
ised as the Monosyllabic Rule in generative metrics. : The rule permits the
scansion of lines such as in (29) as fully metrical, though the italicised words
correspond to odd-numbered positions in the metrical template (see 10.5).
(29) The Monosyllabic Rule in syllable-counting verse (Chaucer):
Ne no man shal unto his felawe ryde
But o cours with a sharpe ygrounde spere KnT 2548–9
I wole han twelf pens, though that she be wood FrT 1576
21
From ‘Certayne Notes of Instruction’ (Gascoine 1575: 465–73).
The fashion for iambic versification in ME was another cultural import

from the Continent, but, as noted above, it could not have been adopted
with such ease if the prosodic conditions had not been favourable.
Again, the correlation between the prosodic shape of the majority of
the vocabulary items and the choice of verse form was recognised early.
Gascoine’s advice to prospective versifiers is:
I thinke it not amisse to forewarne you that you thrust as few wordes of
many sillables into your verse as may be: and hereunto I might alledge
many reasons: first the most auncient English wordes are of one sillable,
so that the more monasyllables that you use, the truer Englishman you
shall seeme, and the lesse you shall smell of the Inkehorne. (Gascoine
1575: 468)
The gradual loss of final <-e> and vowel syncope in inflectional endings
made the native vocabulary predominantly monosyllabic, allowing flex-
ibility in the prosody-to-meter matching. Increased use of prepositions
compensating for inflectional loss created new W S clitic groups: at
níght, to rést, with chéer. Prefixed verbs and adverbs supplied another set
of natural iambic structures: befóre, forgíve, perfórm, asléep. Phrasal stress
continued to be right-strong; phrases made up of stressed monosyl-
lables easily match an iambic foot: five bóoks, tall mén, full glád, God knóws.
The poets also draw from an inventory of handy ‘fillers’, semantically
dispensable monosyllabic words, for example and, now, for, full, some;
past tense did marking; and the grammatically redundant ‘pleonastic’
this, that. Thus, although individual underived disyllabic words (summer,
weapon, silver) had root-initial stress, in connected speech metrical W S
cadences were frequent and easy to construct; this permits an effortless
‘fit’ between language and the iambic meter.22
The gradual adoption of isosyllabic versification in early ME pro-
ceeds against the background of unstable realisation of unstressed
vowels. Some accommodations, such as pre-vocalic elision of <e->, are
commonly recognised in the scholarship on the history of English met-
rical forms (see (25) above, 7.6). Some further paraphonological options
whose application affects the syllabic count in a verse line are illustrated
with Chaucerian examples in (30)–(35) below.
An extension of the pre-vocalic loss of final <-e> in disyllables is the
22
The lasting appeal of the iambic form is recognised by the earliest English prosodists.
Gascoine (1575: 466) writes: ‘note you that commonly now a dayes in english rimes
(for I dare not cal them English verses) we use none other order but a foote of two
sillables, wherof the first is depressed or made short, & the second is elevate or made
long: and that sound or scanning continueth throughout the verse’.
loss of final <-e> in monosyllabic clitics followed by a vowel-initial

word. This can be termed apocope within the domain of the clitic
group. After apocope, resyllabification places the consonant of the clitic
in the onset of the following syllable. Apocope is more likely when the
lexical head starts with an unstressed syllable as in (30a).
(30) Apocope within clitic groups:
(a) Thestaat, tharray, the nombre, and eek the cause GP 716
Tespien where he myghte wedded be MerT 1257
(b) In the ende of which an ounce, and namoore CYT 1266
And to the ymage of Juppiter hem sente SNT 364
Apocope is a remedy against a string of unstressed syllables and against
onsetless syllables; it is confirmed by spellings such as <thestaat>,
<thespien> for ‘the estaat’, ‘to espien’ in the Chaucerian manuscripts, as
well as by the poet’s metrical practice even when the scribes preserved
<-e>, as in the ende (CYT 1266) and the ymage (SNT 364) in (30b).
Another phonological option is synizesis, the fusion of two sylla-
bles into one, when an unstressed [i, υ] is followed by an unstressed
vowel.
(31) Synizesis ([i, υ] + [ə]):
Nowher so bisy a man as he ther nas
And yet he semed bisier than he was GP 321–2
This Pardoner hadde heer as yelow as wex GP 675
Loss of word-internal schwa occurs when [-ə-] is adjacent to the
sonorants [r, l, m, n], as in PDE mystery, summoner, asterisk, lingering.
Syncope can occur within a word or it can be triggered by a following
vowel-initial or weak /h/-initial unstressed word.
(32) Contraction/syncope ([ə] + [r, l, m, n]):
And everemoore he hadde a sov(e)reyn prys GP 67
By wat(e)r he sente hem hoom to every lond GP 400
Which cause is of my deth for sor(o)we and thought Tr I 579
Like apocope and synizesis, word-internal syncope avoids the place-
ment of two syllables in a single metrical position.
Decisions on syllabicity in (31)–(32) rest primarily on the segmental
environment of the unstressed vowel. Other paraphonological adapta-
tions can be triggered by the inherent low prosodic prominence of a
word, or by the position of schwa in the word, without reference to what
the adjacent vowels or consonants are. Predictably, lack of stress on
function words, pronouns and auxiliaries will increase the probability
of dropping -e.
(33) Loss of <-e> in words of low prosodic prominence:

That down were sent in scornynge of nature Tr I 105
As don thise foles that hire sorwes eche Tr I 705
In speech, long strings of unstressed syllables are rhythmically undesir-
able; instability of final -e in potentially trisyllabic words with initial
stress is therefore expected, see 7.6 (20).
(34) Final schwa loss in trisyllabic words:
Lest it were wist on any mánere syde Tr I 321
And of som goodly ánswere yow purchace Tr II 1125
He somwhat is fro wépynge now withdrawe Tr IV 886
Within the native vocabulary trisyllabic words are either compounds,
or they are inflected forms of originally disyllabic roots. Therefore,
the processes of pre-vocalic elision and trisyllabic loss may affect final
vowels in which the unstressed vowel is an inflection. :
(35) Inflectional loss of schwa:
With buttokes brode, and brestes rounde and hye RT 3975
For I loued ek, though ich vnworthi were Tr IV 329
ME isosyllabic verse was composed and copied at a time when the syl-
labicity of the inflections -ed, -es, -eth still depended on the prosodic and
segmental context. Inflectional syncope is more likely in a disyllabic
stem, in elision environments, and depends also on the nature of the
stem-final consonant: liquids leading nasals, followed by fricatives and
stops.23
Clearly, statistics on the strictness of syllable-counting will differ
depending on the way the optional syllables are treated. The uncer-
tainties are both linguistic and textual. Linguistically, a decision on
what counts as a syllable hangs on the application or non-application
of paraphonological options. The presumed dialect and chronology of
the poem and the modes of its transmission are also barriers to iron-
clad conclusions: what we have in the extant texts are often copies of
poems, not authorial compositions. Random survival of variant texts,
possibly influenced by intermediate oral transmission from memory, is
another obstacle. With all of these caveats in mind, it is still reasonable
to describe the development of new metrical modes in ME as resting on
23
These results for the weak preterits are reported in Minkova (2009c: 325–31). For
earlier verse attestations of syncope in -es see Fitzgerald (2008) and Putter (2005:
292, esp. fn. 22). The evidence from syllable-counting verse is supported by scribal
evidence for inflectional syncope (see LALME vol. III, questionnaire items 56, 59, 60
and vol. IV: 108–14).
three important innovations: the extended use of rhyme as a cohesive

structural device, the adoption of syllable-counting and the increasingly
sophisticated use of binary metric feet. Skipping over some hard ques-
tions about violations of isosyllabicity, correspondence violations, espe-
cially the difference between singly filled weak positions and doubly
filled weak positions, we will move to the last section of our survey:
Chaucer’s metrical art. :
10.4 Chaucer and the invention of the iambic pentameter

The poetic work of Geoffrey Chaucer (?1342–1400) is universally
recognised as the highest achievement of Middle English isosyllabic
versification. In the Introduction to The Man of Law’s Tale the author draws
attention to his own success as a storyteller and famously comments on
his metrical skills:
. . . but nathelees certeyn
I kan right now no thrifty tale seyn
That Chaucer, thogh he kan but lewedly
On metres and on rymyng craftily,
Hath seyd hem in swich Englissh as he kan
Of olde tyme, as knoweth many a man. MLI 45–50
The obligatory modesty convention forces Chaucer to play down his
verse-craft, which in his words he kan but lewedly, ‘he knows only poorly’.
His protestations notwithstanding, his superior command of meter and
rhyme, as well as his pioneering approach to the Continental model,
have made him one of the major innovators in the history of English
verse.
In some of his early work Chaucer was using a familiar verse form. He
must have been acquainted with the octosyllabic line both from native
and from Latin, Anglo-Norman and French versification. Trochaic or
iambic tetrameter meter is widespread across the world’s verse tradi-
tions. Its binarity on both the foot level and the colon level (Figure 10.1)
makes it easily accessible and replicable – it is the preferred model
for nursery rhymes, and it can be described as the ‘unmarked’ mode
of footed syllable-counting meter. It is also the type of meter that can
easily lapse into monotony. Fragments of Chaucer’s early translation of
the Romaunt of the Rose, his poem The Book of the Duchess, dating from
1368–9, as well as The House of Fame, completed 1378–80(?), are in iambic
tetrameter. Chaucer approached the inherited template with original-
ity. His lines are generally metrically smooth, the syllable count is more
consistent than the earlier octosyllabic verse, allowing for occasional
headless line (empty W position in the first foot) and a feminine rhyme,
which adds an extrametrical ninth syllable at the right edge of the line.
While in the previous English tetrameter compositions line-ends coin-
cided with a major syntactic break, Chaucer diversified the rhythm by
introducing a large number of run-on lines, or lines with enjambment,
in which single syntactic units straddle two lines. In the following exam-
ples this is the case with ne may slepe (BD 22–3), cam to do, take good herte
(HF 602–5). :
(36) Enjambment in Chaucer’s octosyllabic verse:
And I ne may, ne nyght ne morwe,
Slepe; and thus melancolye BD 22–3
And whider thou shalt, and why I cam

To do thys, so that thou take
Good herte, and not for fere quake HF 602–5
After Chaucer turned to the pentameter as his meter of choice,
which he used in Troilus and Criseyde (1381–6) and The Canterbury
Tales (1387–1400), he must have regarded the tetrameter as an infe-
rior metrical scheme. Echoes of the tetrameter are heard in The Tale
of Sir Thopas, a parody inserted in the Canterbury Tales, described
by Chaucer himself as being in ‘a rym I lerned longe agoon’ (Pro
Thop 709). The Host’s rude interruption of the tale, dismissing it as
‘rym doggerel’ (Thop 925) and ‘drasty ryming’ (Thop 930), and his
disparaging ‘Thou dost noght elles but despendest tyme’ (Thop 931),
suggest that like his fictional mirror-image of a rhymester, the mature
Chaucer considered the previously popular form a good target of
ridicule.
There is no known body of iambic pentameter verse in English
prior to Chaucer. Troilus and Criseyde is the first long, self-contained,
finished major work in the iambic pentameter in the English language.
The individual components of the iambic pentameter were not original
with Chaucer: as discussed in 10.3, iambic verse had been composed in
English since the end of the twelfth century, but the earlier iambic feet
were used in either eight- or fourteen-syllable lines. Decasyllabic lines
were known in France, Portugal and Spain, and were a popular form in
fourteenth-century Italian verse, though the rhythmic contour of the
Continental models was not strictly iambic. Chaucer’s metrical inno-
vation, then, consisted in combining the iambic rhythm of the English
tradition with the decasyllabic line borrowed from the Continent.
Chaucer himself identifies Dante as his metrical source in The Wife of
Bath’s Tale. :
(37) Chaucer’s familiarity with decasyllabic verse:

Wel kan the wise poete of florence,
That highte Dant, speken in this sentence.
Lo, in swich maner rym is Dantes tale . . . WBT 1125–7
The two main features that define the template of Chaucer’s pentam-
eter verse line are a fixed number of syllables per line and iambic feet.
Ideally, each foot includes a weak syllable followed by a strong syllable.
The line is often split in two parts by a caesura. The caesura occurs
commonly after the second foot, but also (with diminishing frequency)
after the third foot, the fourth and the first foot, and, occasionally, even
within the third and the fourth foot. As in every other type of verse,
internal line-breaks tend to coincide with syntactic breaks, but the
correspondence is not obligatory.
Some of the nuances of scansion can be helpful to the reader of
Chaucer’s poetry. Looking at the syllabic count first, the model iambic
pentameter line has ten syllables, each one of which corresponds to a
single metrical position. Weak (W) and strong (S) positions alternate
in strict succession. There may be an optional eleventh syllable, which
is not part of the template; as in other verse-forms an unstressed syl-
lable after the last ictus can be extrametrical; the isosyllabicity of the
line is not affected by its presence or absence. The canonical iambic
pentameter matching is illustrated in (38).
(38) Canonical iambic pentameter in Chaucer:

Foot
W S W S W S W S W S Position
1 2 3 4 5 6 7 8 9 10 (11) Syllabic count
In al this world ne was ther noon hym lik GP 412
I holde hym riche al hadde he nat a sher (te) WBT 1,192
The treatment of unstressed syllables follows the patterns discussed in

10.3 (30)–(35): loss of <-e> before vowels and weak <h->, synizesis and
all the various contractions should be taken into consideration when scan-
ning the line; the crossed <e->’s in holde, riche, hadde in WBT 1192 have to
be elided. The basic syllabic count is not affected by extrametricality at
the right edge, as the boxed eleventh syllable <-te> in WBT 1192, where
the extra syllable is the second syllable of a feminine rhyme. :
Headlessness occurs at the left edge of the line when the initial
foot is missing its weak syllable. Again, this variation of the schema is
shared with other types of iambic verse, resulting in seven syllables for
the tetrameter line, or nine syllables for the pentameter. Headless lines
must start with a stressed syllable.
(39) Headless iambic pentameter lines in Chaucer: 24
Foot
1 2 3 4 5 6 7 8 9 10 Syllabic count
Ø Swere and ly en, as a wom man kan WBP 228
Ø Wom men may go sauf ly up and doun24 WBT 878
If a line is headless, it can never have more than ten syllables, and if
a tenth syllable is present, it is extrametrical. Extrametricality and
headlessness can be combined in a ten-syllable line, as in (40).
(40) Combining headlessness and extrametricality in a ten-syllable
line:
Foot
1 2 3 4 5 6 7 8 9 10 11 Syllabic count
Ø ‘Gladly’, quod she, ‘sith it may yow li (ke)’… WBP 188
The appearance of ten syllables in such cases is deceptive: only nine

of the basic metrical positions are filled, and the lines are headless; any
other scansion would result in unnatural stressing within the line, for
example gladlý, and a strange placement of an unstressed -e in the strong
tenth position.25 Besides, the previous line of the couplet is: And teche us
yonge men of your praktíke, so that *liké would destroy the rhyme. From
this follows an observation which should guide the reader in scansion:
whether the leftmost position is filled or empty, the tenth position must
always be matched to a stressed syllable – we return to this constraint
on stress alternation in the iambic pentameter below.
Over 98 per cent of the lines in Chaucer’s pentameter verse conform
to the decasyllabic norm if the scansion incorporates the adjustments
24
This is the syllabic structure of the line both in Hg and El – the two best manuscripts.
Some editions have now inserted after go, a questionable emendation.
25
The fact that the line can be scanned as fully trochaic is irrelevant; as noted above,
what matters is that it is embedded in iambic metrical context.
to the syllable count described in (30)–(35) and (39)–(40). Bad lines do

occur, but those are genuinely rare and attributing them to the poet or
to one of his later scribes remains speculative, especially if comparing
manuscripts is uninformative.
Besides isosyllabicity, the iambic template presupposes regular stress
alternation: odd positions filled by unstressed syllables, and even posi-
tions by stressed syllables. With respect to this parameter too, there
is variation in the realisation of the basic schema. Any deviation from
the schema is a mismatch; the mismatches between stress and metrical
position are known as labelling mismatches (Kiparsky 1977), or simply
stress mismatches.
Stress mismatches are not of the same severity, nor are they equally
tolerated across the verse line. Word-stress mismatch, for example
placing summer in metrical W S or placing behold in S W metrical
sequence, is more disruptive than placing find it in W S or his belt in S
W, which is more disruptive than placing brimstone in W S, and so on. A
hierarchy of stress mismatches is proposed in (49), but first we survey
the distribution of mostly word-stress mismatches across Chaucer’s
pentameter line.
The most common type of stress mismatch, which breaks the monot-
ony and produces a more interesting prosodic contour is the reversal
of the strength relations in the first foot, known as iambic reversal or
trochaic substitution. This means that an unambiguously stressed syl-
lable is placed at the left edge of the line, followed by two unstressed
syllables, resulting in what is known as a rhythmic triple.
(41) Trochaic substitution at the left edge of the iambic line:
Foot
Un- der his belt he bar ful thrif- ti- ly GP 105
s w w s w s w s w s Stress
Unless one posits a completely artificial delivery: *undér his belt . . ., the
inversion in the first foot creates a triple. Trochaic substitution in the
first foot must be considered a valid prosodic option even in the case of
French loanwords.
(42) Trochaic substitution involving French loanwords:
(a) Jústice he was ful often in assise GP 314
Compare:
(b) Now was ther thanne a jústice in that toun PhysT 121
The type of trochaic substitution shown in (41) and (42a) is most common
in the first foot; it is associated with metrical and syntactic boundaries.
Post-pausal mismatches are characteristic of all types of isosyllabic meter.
Predictably, the next highest incidence of trochaic substitution is in the
third and the fourth foot, which corresponds to the frequent division of
the Chaucerian pentameter line in colas of 2 + 3 or 3 + 2 feet.26 Pre-pausal
feet, either the last foot of the line or line-internally the last foot of the first
colon, are locations where the prosodic and the metrical prominences
are much more strictly matched. In Chaucer’s iambic pentameter the
second foot undergoes substitution rarely, in under 2 per cent of the lines
(Youmans and Li 2002: 157), and when the second foot is matched to a
trochaic word, it usually follows another trochee or two monosyllables.
(43) Second-foot trochaic substitution in the iambic line:
Of his óffryng and eek of his substaunce. GP 489
Whan myn hóusbonde is fro the world ygon WBP 47
Under that cólour hadde I many a myrthe WBP 405
Third- and fourth-foot substitutions are in the range of 3.6–3.7 per
cent (Youmans and Li 2002: 157). As the examples in (44) show, the
substitutions occur after syntactic breaks.
(44) Third- and fourth-foot trochaic substitutions:
If that I speke áfter my fantasye WBP 190
What nedeth yow, Thómas, to maken stryf? SumT 2000
The smylere with the knyf únder the cloke KnT 1999
I governed hem so wel, áfter my lawe WBP 219
Occasionally, there will be more than one reversed foot in a decasyl-
labic headless line, creating a strikingly different prosodic cadence in an
otherwise fully iambic metrical context.
(45) Multiple substitutions in the iambic line:
Bácyns, lávours, er that men hem bye,
Spóones and stóoles, and al swich housbondrye WBP 287–8
Bléssinge hálles, chámbres, kíchenes, bóures

Cítees, búrghes, cástels, hýe tóures . . . WBT 869–70
The fifth foot disallows substitution; in practice it can accommodate
only iambic sequences. In that position compounds must shift the main
26
About half of Chaucer’s pentameter lines show a syntactic break after the second foot.
Statistical details and comments on the uneven distribution of inversions across the
line are found in Tarlinskaja (1976: 279–80), Youmans and Li (2002: 156–8).
stress from the first to the second element, and derivational suffixes are
metrically promoted.
(46) Strictness of the fifth foot in the iambic pentameter:
Right so a wyf destroyeth hire housbónde WBP 377
And al was fals, but that I took witnésse WBP 382
The strong position in the rightmost foot of the line, where the rhyme
is located, is metrically demanding in that it enforces prominence
on the syllable filling that position. This is a verse convention, pos-
sibly observed in performance, but its implications for reconstructing
speech prosody are limited. In ME rhyming practice, suffixes are
allowed to fill the rhyme position: . . . sette at réste: the worthiéste (Tr II
761), and make a thyng: . . . at his writỳng (GP 325–6), drónkenèsse: witnèsse
(WBT 381–2). The metrical strictness of the last strong position is
such that it can even invert the prosodic contour of a native word by
suppressing its primary stress – . . . upon a mére: . . . and a millére (GP
541–2), answèrde: hérde (CIT 21–2) – although this is rare in Chaucer.
The convention is linguistically constrained only to the extent that
inflectional syllables are not allowed to fill the last strong position in
the line.
The uneven distribution of trochaic inversions and the very strong
tendency for the alignment of syntactic breaks with even metrical
positions, especially after the second foot, supports the claim that the
pentameter line invented by Chaucer is not just a sequence of W’s and
S’s. The prototype iambic pentameter structure is hierarchical: different
feet have different strength indices for the strong positions. :
(47) The prototypical iambic pentameter line:
[W 2S] [W 3S] // [W 2S] / [W 2S] [W 5S]
The description of Chaucer’s pentameter, or any type of non-free
verse, raises the question of the difference between meter and perfor-
mance. The template of Chaucer’s iambic pentameter, as presented
here, requires that at least four weak and exactly five strong metri-
cal positions in the line must be filled. If weak positions are filled by
unstressed syllables, and strong positions by stressed syllables, the fit
between prosody and meter is perfect. However, in PDE, as well as in
earlier English, weak and strong syllables do not alternate at absolutely
regular intervals: strings of unstressed syllables, as well as stress-clashes
between two or more stressed syllables occur commonly in speech.
This inevitable discrepancy between the rhythm of speech and the
abstract metrical schema is sometimes taken as an argument against the
metrical regularity of Chaucer’s iambic pentameter. One of the much-
debated issues in Chaucerian metrical studies is whether Chaucer’s

ten-syllable line can legitimately be described as a five-stress line. It
is indeed the case that Chaucer’s narrative verse naturally reflects the
imperfect stress alternation found in the spoken language, so that the
number of full stresses in a line realised in recitation can be four or even
three.
(48) Discrepancy between prosodic peaks and metrical S’s in Chaucer:
For of góod náme and wísdom and manére TC 1.880
Póynaunt and shárp, and rédy al his géere GP 352
Thanne wolde he spéke and críe as he were wóod GP 636
It is laughable to advocate reading these lines as mechanically mir-
roring the meter, for example ‘For óf good náme and wísdom ánd
manére’, matching pedantically the surface string of syllables to the
metrical W’s and S’s. This does not mean that the lines in (48) are not
perfect instantiations of the iambic pentameter. In verse performance,
as in speech in general, the underlying categories and their surface
realisation can differ within some well-defined limits. In this case, the
limits are set by the syllabic composition of the words and by their
grammatical function. As pointed out in 10.3, monosyllabic words can
occupy both strong and weak positions, though in practice nouns,
verbs and adjectives gravitate towards the strong positions, while
function words are placed in the weak positions. Strings of unstressed
monosyllables, also known as lapses – But as a [child of twelf month oold,
or lesse] (PrT 484); [I wole han twelf pens,] though that she be wood (FrT
1576) – are freely accommodated by the meter; such strings do not
constitute evidence against the abstract stress-alternating schema
used by Chaucer. The same logic of performative freedom applies to
adjacent full lexical monosyllables, which some metrists interpret as
‘spondaic’ stress-clashes, as in [But as a child of] twelf month oold, [or lesse]
(PP 484), [I wole han] twelf pens, [though that she be wood] (FrT 1576). In
this case too, the schema is observed, whether the prominence of one
of the items is adjusted up or down or not in performance.
Mismatches of prosodic prominence to metrical positions are dif-
ferently evaluated not just by where they occur in the line, but also in
terms of the domain of the violation, where a mismatch on the word-
level is most noticeably disruptive. A possible hierarchy of the severity
of matching violations is shown in (49), where matching the syllable
prominence of underived words to the corresponding metrical positions
is most strictly enforced, and the matching for phrasal stress is most
easily ignored.
(49) Gradient violability of stress-to-ictus matching:

Underived words (father, lyved, speken, muchel, under)
↓
Clitic groups (a lord, the best, in swich)
↓
Derived words (ending, childhood)
↓
Compounds (brimstone, offspring, col-blak)
↓
Phrases (old men, brend gold, white thing)
The hierarchy is tentative and it includes only disyllabic sequences. It
correlates with the morphological and semantic independence of the
syllables in the string. The categories are rather crude. Unstressed sylla-
bles belonging to the stem will be least likely to fill a strong position; this
type of mismatch triggers trochaic substitution. Noun and verb inflec-
tions are similarly resistant to metrical promotion, but the comparative
-est and the gerundial -yng are occasionally found in the tenth position,
for example almageste: hyeste (WBP 325–6); thyng: grucchyng (WBP 405–6).
The placement of French derivational suffixes in rhyme position (-able,
-aunce, -ioun, -ie, and so on) is much more frequent than the use of native
suffixes in rhyme. There are subtler differences in the behaviour of
different clitics, and of different derivational affixes. Determiner clitics
(a, the) are very rarely placed in S metrical position. Prepositions and
pronouns, as well as auxiliaries (be, have, wol) and conjunctions (and,
or) are more flexible. Not least, the line separating derived items from
compounds and compounds from phrases is often blurry. Nevertheless,
the hierarchy provides a basis for testing the Chaucerian verse corpus, a
task which has not been undertaken yet.
In the idealised schema in Figure 10.1 the edges of most prosodic
domains are aligned with the edges of the metrical divisions: clause
boundaries match line boundaries, and word boundaries and clitic
group boundaries match foot boundaries. Deviations from these ‘ideal’
alignments are known as bracketing mismatches. Bracketing mismatches
avoid the monotony of the line: the more bracketing mismatches there
are, the more metrically interesting a line is. As in the case of stress mis-
matches, mismatches of prosodic and metrical domain boundaries are of
different severity. Typically, bracketing mismatches are hierarchised in
the order given in (50).
(50) Hierarchy of bracketing mismatches:
(a) Clause boundary – line-end
That they were maked for purgacioun
Of uryne . . . WBP 120–1

↓
(b) Phrase boundary – colon
But, lord crist! whan that it remembreth me WBP 469
↓
(c) Clitic group and word boundary – foot
A bettre envyned man was nowher noon GP 342
The rarest of the bracketing mismatches in Chaucer’s pentameter

involves enjambment, as in (50a). Non-observance of the alignment
at the right edge of the line makes for a looser verse-form and bridges
the gap between verse and prose. : In (50b) the syntactic breaks after
the first and the third positions are quite unusual – as noted above,
Chaucer’s pentameter cola-breaks occur most commonly after the
fourth position, or after the fifth and the sixth position. In (50c) the
second foot starts inside the second word with elision of -e on <bettre>:
[-tren-], the third foot straddles <. . . -ned man> and the fourth foot
ends inside <nowher>. All three examples, in fact, show word-to-foot
misalignment, yet all three examples are within the range of metri-
cally well-formed lines. Bracketing mismatches are more common in
Chaucer’s narrative verse than in his lyric verse.
The cumulative effect of both stress-to-ictus matching and the
matching of prosodic to metrical units in the earlier syllable-counting
verse has not been explored for Chaucer or Gower. Sophisticated
mathematical modelling of constraints, their interrelation and predict-
ability, as proposed by Hayes et al. (2012) for Shakespeare and Milton,
has yet to be attempted on pre-Elizabethan verse. What is evident from
the survey in this chapter, however, is that isosyllabic verse and iambic
footing have been an option in the language since before the end of
the twelfth century. Even the modern reader of poetry is aware of the
continuity of this tradition:
. . . the iamb has been the predominant foot in metrical English poetry for
upwards of seven centuries and will remain so for the forehearable future.
The oft-anthologized pieces in The Classic Hundred: All-Time Favorite
Poems show a good deal of metrical and stanzaic variety, but in ninety-
nine of the poems the fundamental rhythm is iambic, with the remaining
lone exception a poem in free verse. (Harmon 1997: 15)
The principles of metrical template construction for isosyllabic verse
have been the same from Orm to this day, and the variability in line
length, stress matching and domain alignment that characterises indi-
vidual poets and compositions can all be addressed in terms of the
parameters discussed here. And before we take leave of this all too hasty
survey, we should reiterate the advantages of familiarity with the struc-
ture of early verse for the historical phonologist. Alliterative verse is an
essential source for formulating and testing hypotheses about segmental
inventories, their categorisation and their phonological function. All
measured verse is a primary testing ground for prosodic reconstruction:
theories of early English syllable structure, word- and compound stress,
and phrasal prominence are strengthened or invalidated by parallel, yet
independently learned and replicated metrical systems.
Bibliography
Ahn, Sang-Cheol (2003), ‘Towards an optimal account of diachronic chain

shifts: part I (Grimm’s law)’, in J. Spenader, A. Eriksson and Östen Dahl (eds),
Variation within Optimality Theory. Proceedings of the Stockholm Workshop on
‘Variation within Optimality Theory’, April 2003, at Stockholm University,
Sweden, Stockholm: Stockholm University, 7–16.
Aitken, A. J. (1962), ‘Vowel length in modern Scots’, unpublished mimeograph,
Edinburgh: Department of English Language, University of Edinburgh.
Albright, Adam (2009), ‘Feature-based generalisation as a source of gradient
acceptability’, Phonology 26, 9–41.
Alexander, James D. (1985), ‘R-methathesis in English: a diachronic account’,
Journal of English Linguistics 18.1, 33–40.
Algeo, John (ed.) (2001), The Atlas of North American English, Cambridge:
Cambridge University Press.
Allerton, D. J. (2000), ‘Articulatory inertia vs “Systemzwang”: changes in liaison
phenomena in recent British English’, English Studies 81:6, 574–81.
Anderson, John (2001), ‘A major restructuring in the English consonant system:
the de-linearization of [h] and the de-consonantization of [w] and [j]’, English
Language and Linguistics 5:2, 199–212.
Anderson, John and D. Britton (1997), ‘Double trouble: geminate versus
simplex graphs in the Ormulum’, in Jacek Fisiak (ed.), Studies in Middle English
Linguistics, Trends in Linguistics. Studies and Monographs 103, Berlin:
Mouton de Gruyter, 23–58.
Aroui, Jean-Louis and Andy Arleo (eds) (2009), Towards a Typology of Poetic
Forms, Language Faculty and Beyond, 2, Amsterdam: Benjamins.
Barber, Charles (1976), Early Modern English, London: André Deutsch.
Barber, Charles (1997), Early Modern English, Edinburgh: Edinburgh University
Press.
Bauer, Laurie (1992), ‘The Second Great Vowel Shift revisited’, English World-
Wide 13, 253–68.
Bauer, Laurie (1994a), ‘English in New Zealand’, in Robert Burchfield (ed.), The
Cambridge History of the English Language, Vol. V: English in Britain and Overseas,
Origins and Development, Cambridge: Cambridge University Press, 382–428.
Bauer, Laurie (1994b), Watching English Change, London: Longman.
367
Bauer, Laurie (2002), An Introduction to International Varieties of English, Edinburgh:

Edinburgh University Press.
Becker, Michael, Andrew Nevins and Jonathan Levine (2012), ‘Asymmetries in
generalizing alternations to and from initial syllables’, Language 88:2, 231–60.
Beddor, Patrice Speeter, Rena Arens Krakow and Louis M. Goldstein (1986),
‘Perceptual constraints and phonological change: a study of nasal vowel
height’, Phonology Yearbook 3, 197–217.
Bell, Alan, Jason M. Brenier, Michelle Gregory, Cynthia Girand and Dan
Jurafsky (2009), ‘Predictability effects on durations of content and func-
tion words in conversational English’, Journal of Memory and Language 60,
92–111.
Bennett, William H. (1970), ‘The stress patterns of Gothic’, PMLA 85:3, 463–72.
Benson, Larry D. (ed.) (1987), The Riverside Chaucer, 3rd edn, Boston: Houghton
Mifflin.
Berg, Thomas (1999), ‘Stress variation in British and American English’, World
Englishes 18:2, 123–43.
Bermúdez-Otero, Ricardo (1998), ‘Prosodic optimization: the Middle English
length adjustment’, English Language and Linguistics 2:2, 169–97.
Bermúdez-Otero, Ricardo (2005), Review of Alliteration and Sound Change in Early
English by Donka Minkova, Diachronica 22, 438–45.
Bermúdez-Otero, Ricardo (2006), ‘Phonological change in Optimality Theory’,
in Keith Brown (ed.), Encyclopedia of Language and Linguistics, vol. 9, 2nd edn,
Oxford: Elsevier, 497–505.
Bermúdez-Otero, Ricardo (2007), ‘Diachronic phonology’, in Paul de Lacy
(ed.), The Cambridge Handbook of Phonology, Cambridge: Cambridge University
Press, 497–517.
Bermúdez-Otero, Ricardo and Richard Hogg (2003), ‘The actuation problem:
phonologization, rule inversion and rule loss’, in Holt (ed.) (2003), 91–119.
Bermúdez-Otero, Ricardo and April McMahon (2006), ‘English phonology and
morphology’, in Bas Aarts and April McMahon (eds), The Handbook of English
Linguistics, Oxford: Blackwell, 382–410.
Bermúdez-Otero, Ricardo and Graeme Trousdale (2012), ‘Cycles continua: on
unidirectionality and gradualness in language change’, in Terttu Nevalainen
and Elizabeth Closs Traugott (eds), The Oxford Handbook of the History of
English, Oxford: Oxford University Press, 691–720.
Berndt, Rolf (1965), ‘The linguistic situation in England from the Norman
Conquest to the Loss of Normandy (1066–1204)’, Philologica Pragensia 8,
145–63.
Berndt, Rolf (1972), ‘The period of the final decline of French in medieval
England (fourteenth and early fifteenth centuries)’, Zeitschrift fur Anglistik und
Amerikanistik 20, 341–69.
Berndt, Rolf (1992), ‘The history of the English language and social history:
French vs. English’, in Wilhelm G. Busse (ed.), Anglistentag 1991, Düsseldorf:
Proceedings, Tübingen: Niemeyer, 276–92.
Bibire, Paul (2001), ‘North Sea language contacts in the early Middle Ages:
BIBLIOGR APHY 369
English and Norse’, in Thomas Liszka and Lorna E. M. Walker (eds), The
North Sea World in the Middle Ages, Dublin: Four Courts Press, 88–108.
Bigham, Douglas Stephen (2008), Dialect Contact and Accommodation among
Emerging Adults in a University Setting, PhD thesis, The University of Texas
at Austin.
Blevins, Juliette and Andrew Garrett (1998), ‘The origins of consonant-vowel
metathesis’, Language 74:3, 508–56.
Boersma, Paul (1998), Functional Phonology, The Hague: Netherlands Graduate
School of Linguistics (LOT).
Bondar, Vladimir A. (2006), ‘On hit-it variation in Old English’, paper delivered
at the Central European Student Conference in Linguistics (CESCL), Budapest,
May–June 2006.
Bosworth, Joseph and T. Northcote Toller (1898, 1921), An Anglo-Saxon
Dictionary (1898) and Supplement (1921), <http://www.bosworthtoller.com>
(last accessed 3 June 2013).
Bradley, Travis G. (2007), ‘Morphological derived-environment effects in ges-
tural coordination: a case study of Norwegian clusters’, Lingua 117:6, 950–85.
Branchaw, Sherrylyn (2010), Survival of the Strongest: Strong Verbs in the History of
English, PhD thesis, University of California, Los Angeles.
Bredehoft, Thomas (2005), Early English Metre, Toronto Old English Series,
Toronto, Buffalo and London: University of Toronto Press.
Brinton, Laurel and Margery Fee (2001), ‘Canadian English’, in John Algeo
(ed.), The Cambridge History of the English Language, Vol. VI: English in North
America, Cambridge: Cambridge University Press, 422–40.
Britton, Derek (2007a), ‘The dialectal origins of the language of Henry Machyn’,
in Christopher M. Cain and Geoffrey Russom (eds), Studies in the History of the
English Language III. Managing Chaos: Strategies for Identifying Change in English,
Berlin: Mouton de Gruyter, 251–67.
Britton, Derek (2007b), ‘A history of hyper-rhoticity in English’, English
Language and Linguistics 11:3, 525–36.
Browman, Catherine P. and Louis Goldstein (1990), ‘Gestural specification
using dynamically-defined articulatory structures’, Journal of Phonetics 18,
299–320.
Brown, Carleton (ed.), [1932] (1962), English Lyrics of the XIIIth Century, Oxford:
Clarendon Press.
Brunner, Karl (1963), An Outline of Middle English Grammar, trans. Grahame
Johnston, Cambridge, MA: Harvard University Press.
Burling, Robbins (1966), ‘The metrics of children’s verse: a cross-linguistic
study’, American Anthropologist 68, 1418–41.
Burton, T. L. (2010), William Barnes’s Dialect Poems: A Pronunciation Guide,
Adelaide: Chaucer Studio Press.
Bybee, Joan L. (2007), Frequency of Use and the Organization of Language, Oxford:
Oxford University Press.
Bybee, Joan and Dan I. Slobin (1982), ‘Rules and schemas in the development
and use of the English past tense’, Language 58, 265–89.
Cable, Thomas (1991), The English Alliterative Tradition, Philadelphia: University

of Pennsylvania Press.
Cable, Thomas (2009), ‘Foreign influence, native continuation, and metrical
typology in alliterative lyrics’, in Jefferson and Putter (eds) (2009), 219–34.
Campbell, Alistair (1959), Old English Grammar, Oxford: Clarendon Press.
Carter, Paul and John Local (2007), ‘F2 variation in Newcastle and Leeds
English liquid systems’, Journal of the International Phonetic Association 37:2,
173–82.
Cercignani, Fausto (1981), Shakespeare’s Works and Elizabethan Pronunciation,
Oxford: Clarendon Press.
Chambers, J. K. (2006), ‘Canadian Raising retrospect and prospect’, The
Canadian Journal of Linguistics 51:2/3, 105–18.
Ching, Marvin, K. L. (1996), ‘GreaZy/GreaSy and other /Z/-/S/ choices in
Southern pronunciation’, Journal of English Linguistics 24, 295–307.
Chomsky, Noam and Morris Halle (1968), The Sound Pattern of English, New
York: Harper and Row.
Clark, Cecily (1992) ‘Onomastics’, in Norman Blake (ed.), The Cambridge History
of the English Language, Vol. II: 1066–1476, 542–606.
Clarke, Sandra (2006), ‘Nooz or nyooz?: The complex construction of Canadian
identity’, Canadian Journal of Linguistics 51:2/3, 225–46.
Cohn, Abigail (1993), ‘Nasalization in English: phonology or phonetics’,
Phonology 10:1, 43–81.
Cohn, Abigail (2003), ‘Phonological structure and phonetic duration: the role
of the mora’, Working Papers of the Cornell Phonetics Laboratory, vol. 15, 69–100.
Collie, Sarah (2007), English Stress Preservation and Stratal Optimality Theory, PhD
dissertation, University of Edinburgh.
Collie, Sarah (2008), ‘English stress preservation: the case for “fake cyclicity”’,
English Language and Linguistics 12:3, 505–32.
Colman, Fran (1983), ‘Old English /ɑ/ ≠ /æ/ or /ɑ/ ~ /æ/?’, Folia Linguistica
Historica 4:2, 265–85.
Colman, Fran (ed.) (1992), Evidence for Old English: Material and Theoretical Bases
for Reconstruction, Edinburgh: John Donald.
Côté, Marie-Hélène (2004), ‘Syntagmatic distinctness in consonant deletion’,
Phonology 21, 1–41.
Crisma, Paola (2009) ‘Word-initial h- in Middle and Early Modern English’, in
Minkova (ed.) (2009), 130–68.
Crosswhite, Katherine (2004), ‘Vowel reduction’, in Hayes et al. (eds) (2004),
191–231.
Cruttenden, Alan (2008), Gimson’s Pronunciation of English, 7th edn, London:
Hodder Education.
Crystal, David (2006), ‘English worldwide’, in Hogg and Denison (eds) (2006),
420–40.
Crystal, David (2008), ‘Think on my Words’: Exploring Shakespeare’s Language,
Cambridge, Cambridge University Press.
Culpeper, Jonathan and Phoebe Clapham (1996), ‘The borrowing of Classical
BIBLIOGR APHY 371
and Romance words into English: a study based on the electronic Oxford
English Dictionary’, International Journal of Corpus Linguistics 1:2, 199–218.
Curzan, Anne and Kim Emmons (eds) (2004), Unfolding Conversations: Studies in
the History of the English Language II, Berlin: Mouton de Gruyter.
Cutler, Anne and Sally Butterfield (1992), ‘Rhythmic cues to speech segmenta-
tion: evidence from juncture misperception’, Journal of Memory and Language
31, 218–36.
Dance, Richard (2003), Words Derived from Old Norse in Early Middle English:
Studies in the Vocabulary of the South-West Midland Texts, Medieval and
Renaissance Texts and Studies, vol. 246, Tempe, AZ: Arizona Center for
Medieval and Renaissance Studies.
Danchev, Andrei [1975] (1976), On the Phonemic and Phonetic Values of the Short
ea and eo Diagraphs in Old English. Annuaire de l’Université de Sofia. Faculté
des Lettres. Langues et littératures romanes et germaniques. Vol. LXX, Sofia:
Naouka I Izkoustvo, 1, 37–88.
Danielsson, B. (1948), Studies on the Accentuation of Polysyllabic Latin, Greek, and
Romance Loan-Words in English with Special Reference to those Ending in -able, ate,
-ator, -ible, -ic, -ical, and -ize, Stockholm: Almqvist and Wiksell.
Daunt, Marjorie (1939), ‘Old English sound changes reconsidered in relation to
scribal tradition and practice’, Transactions of the Philological Society 38, 108–37.
Davenport, Mike and S. J. Hannahs (2010), Introducing Phonetics and Phonology,
3rd edn, London: Hodder Education.
Delattre, Pierre and Donald C. Freeman (1968), ‘A dialect study of American r’s
by x-ray motion picture’, Linguistics 44, 29–68.
Denton, Jeannette Marshall (1999), North and West Germanic Consonant Gemination:
A Philological and Phonetic Reanalysis, PhD thesis, University of Chicago.
Denton, Jeannette Marshall (2003), ‘Reconstructing the articulation of Early
Germanic *r’, Diachronica 20:1, 11–43.
Diamond, Jared and Peter Bellwood (2003), ‘Farmers and their languages: the
first expansions’, Science 300:5,619, 597–603.
Dobson, Eric John (1957), English Pronunciation 1500–1700, vol. 1, Oxford:
Clarendon Press.
Dobson, Eric John [1957] (1968), English Pronunciation 1500–1700, vol. 2, 2nd
edn, Oxford: Clarendon Press.
DOE, Dictionary of Old English: A to G, online, ed. Antonette diPaolo Healey,
Toronto: Dictionary of Old English Project, University of Toronto, <http://
www.doe.utoronto.ca/pages/pub/fasc-a-g-web.html> (last updated 22 July
2011, last accessed 3 June 2013).
Dresher, B. Elan [1978] (1985), Old English and the Theory of Phonology, PhD dis-
sertation, University of Massachusetts, Amherst. 1985 reprint, New York:
Garland.
Dresher, B. Elan (2009), The Contrastive Hierarchy in Phonology, Cambridge:
Dresher, B. Elan and Nila Friedberg (eds) (2002), Formal Approaches to Poetry:
Recent Developments in Metrics, Berlin: Mouton de Gruyter.
Duffell, Martin (2008a), A New History of English Metre, Legenda, Studies in

Linguistics 5, London: Modern Humanities Research Association and Maney
Publishing.
Duffell Martin J. (2008b), ‘Some observations on English binary metres’,
Language and Literature 17, 5–20.
Duggan, Hoyt N. (2001), ‘Some aspects of a-verse rhythms in Middle English
alliterative poetry’, in Robert F. Yeager and Charlotte Morse (eds),
Speaking Images: Essays in Honor of V. A. Kolve, Ashville, NC: Pegasus Press,
479–503.
Eckert, Penny (n.d.), ‘Vowel shifts in Northern California and the Detroit
suburbs’, <http://www.stanford.edu/~eckert/vowels.html> (last accessed
3 June 2013).
Eddington, David and Michael Taylor (2009), ‘T-glottalization in American
English’, American Speech 84:3, 298–315.
Eddington, David and Caitlin Channer (2010), ‘American English has goʔ a loʔ
of glottal stops: social diffusion and linguistic motivation’, American Speech
85:3, 338–52.
Ekwall, Eilert (1960), The Concise Oxford Dictionary of Place-names, 4th edn,
Oxford: Clarendon Press.
Ellis, Lucy and William J. Hardcastle (2002), ‘Categorical and gradient proper-
ties of assimilation in alveolar to velar sequences: evidence from EPG and
EMA data’, Journal of Phonetics 30:3, 373–96.
Erickson, Blaine (2002), ‘On the development of English r’, in Minkova and
Stockwell (eds) (2002), 183–207.
Evans, Arthur Benoni (1881), Leicester Words, Phrases, and Proverbs. Edited with
Additions and an Introduction, by Sebastian Evans. London: Published for
the English Dialect Society by Truebner and Co., 57 and 59, Ludgate Hill.
Fabb, Nigel (2010), ‘Is literary language a development of ordinary language?’,
Lingua 120:5, 1,219–32.
Fabricius, Anne (2002), ‘Weak vowels in modern RP: an acoustic study of
happY-tensing and KIT/schwa shift’, Language Variation and Change 14, 211–37.
Ferragne, Emmanuel and François Pellegrino (2010), ‘Formant frequencies of
vowels in 13 accents of the British Isles’, Journal of the International Phonetic
Association 40:1, 1–34.
Fikkert, Paula (2007), ‘Acquiring phonology’, in P. de Lacy (ed.), Handbook of
Phonological Theory, Cambridge, MA: Cambridge University Press, 537–54.
Filppula, Markku, Juhani Klemola and Heli Paulasto (2008), English and Celtic in
Contact, New York and Abingdon: Routledge.
Filppula, Markku, Juhani Klemola and Heli Pitkänen (eds) (2002), The Celtic
Roots of English. Studies in Language, vol. 37, Joensuu: Faculty of Humanities,
University of Joensuu.
Finegan, Edward (2006), ‘English in North America’, in Hogg and Denison (eds)
(2006), 384–420.
Finkenstaedt, T., E. Leisi and D. Wolff (eds) (1970), A Chronological English
Dictionary, Heidelberg: Carl Winter Universitätsverlag.
BIBLIOGR APHY 373
Fischer, David Hackett (1989), Albion’s Seed: Four British Folkways in America,
Oxford: Oxford University Press.
Fisher, John Hurt (2001), ‘British and American, continuity and divergence’, in
John Algeo (ed.), The Cambridge History of the English Language, Vol. VI: English
in North America, Cambridge: Cambridge University Press, 59–85.
Fisiak, Jacek (1968), A Short Grammar of Middle English, Warsaw: Państwowe
Wydawnictwo Naukowe.
Fitzgerald, Christina (2008), ‘“Ubbe dubbede him to knith”: the scansion of Havelok
and ME -es, -ed, and -ede’, in Susan Fitzmaurice and Donka Minkova (eds),
Studies in the History of the English Language IV: Empirical and Analytical Advances in
the Study of English Language Change, Berlin: Mouton de Gruyter, 187–205.
Flasdieck, Hermann F. (1958), ‘Die Entstehung des engl. Phonems [ʃ], zugleich
ein Beitrag zur Geschichte der Quantität’, Anglia 76, 339–410.
Flemming, Edward (1995), Auditory Representations in Phonology, PhD dissertation,
University of California, Los Angeles.
Flemming, Edward (2003), ‘The relationship between coronal place and vowel
backness’, Phonology 20, 335–373.
Flemming, Edward (2004), ‘Contrast and perceptual distinctiveness’, in Hayes
et al. (eds) (2004), 232–77.
Flemming, Edward (2009), ‘The phonetic weakness of schwa vowels’, in
Minkova (ed.) (2009), 78–99.
Flemming, Edward and S. Johnson (2007), ‘Rosa’s roses: reduced vowels in
American English’, Journal of the International Phonetic Association 37, 83–96.
Fridland, V. (2000), ‘The Southern Shift in Memphis, Tennessee’, Language
Variation and Change 11, 267–85.
Fromkin, Victoria (ed.) et al. (2000), Linguistics: An Introduction to Linguistic
Theory, Oxford: Blackwell.
Fudge, Eric, (1984), English Word-Stress, London: George Allen and Unwin.
Fulk, Robert D. (1992), A History of Old English Meter, Philadelphia: University
of Pennsylvania Press.
Fulk, Robert D. (1997), ‘Ambisyllabicity in Old English: a contrary view’, in
Irmengard Rauch and Gerald F. Carr (eds), Insights in Germanic Linguistics II:
Classic and Contemporary, Trends in Linguistics. Studies and Monographs 94,
Fulk, Robert D. (1998), ‘The role of syllable structure in Old English quantita-
tive sound changes’, NOWELE: North-Western European Language Evolution 33,
3–35.
Fulk, Robert D. (1999), ‘Evaluating the evidence for lengthening before homor-
ganic consonant clusters in the Ormulum’, in G. F. Carr and W. Harbert (eds),
Interdigitations: Essays for Irmengard Rauch, New York: P. Lang, 201–9.
Fulk, Robert D. (2002), ‘Early Middle English evidence for Old English metrics:
resolution in Poema morale’, Journal of Germanic Linguistics 14, 331–55.
Fulk, Robert D. (2004), ‘Old English poetry and the alliterative revival: on
Geoffrey Russom’s “The evolution of Middle English alliterative meter”’, in
Curzan and Emmons (eds) (2004), 305–12.
Fulk, Robert D. (2007), ‘Old English meter and oral tradition: three issues bearing
on poetic chronology’, The Journal of English and Germanic Philology 106:3,
304–24.
Fulk, Robert D. (2012), An Introduction to Middle English: Grammar: Texts,
Peterborough, ON: Broadview Press.
Garrett, Andrew and Juliette Blevins (2009), ‘Analogical morphophonology’,
in Kristin Hanson and Sharon Inkelas (eds), The Nature of the Word: Essays in
Honor of Paul Kiparsky, Cambridge, MA: MIT Press, 527–45.
Gascoine, George (1575), The Posies, Printed at London for Richard Smith,
and are to be solde at the Northweast doore of Paules Church. (Edited by
John W. Cunliffe, MA, DLit (London) Professor of English in the University
of Wisconsin, USA. Cambridge: at the University Press 1907).
Geist, Robert J. (1948), ‘Notes on “The King of Tars”’, The Journal of English and
Germanic Philology 47:2, 173–8.
Getty, Michael (2002), The Metre of Beowulf, Berlin: Mouton de Gruyter.
Gick, Bryan (1999), ‘A gesture-based account of intrusive consonants in
English’, Phonology 16:1, 29–54.
Gick, Bryan (2002), ‘The American intrusive L’, American Speech 77, 167–83.
Giegerich, Heinz (1992), English Phonology: An Introduction, Cambridge:
Giegerich, Heinz (1997), ‘The phonology of /ɔ:/ and /ɑ:/ in RP English: Henry
Sweet and after’, English Language and Linguistics 1:1, 25–79.
Giegerich, Heinz (1999), Lexical Strata in English: Morphological Causes, Phonological
Effects, Cambridge: Cambridge University Press.
Giegerich Heinz J. (2009), ‘Compounding and lexicalism’, in Rochelle Lieber
and Pavol Štekauer (eds), Handbook of Compounding, Oxford: Oxford University
Press, 178–200.
Gimson, A. C. (1970), An Introduction to the Pronunciation of English, 2nd edn,
London: Edward Arnold.
Gnanadesikan, Amalia E. (2009), The Writing Revolution: Cuneiform to the Internet,
Chichester: Wiley-Blackwell.
Goldsmith, John (1995), ‘Phonological theory’, in J. Goldsmith (ed.), The
Handbook of Phonological Theory, Oxford: Blackwell, 1–23.
Gordon, Matthew J. (2001a), ‘Small-town values and big-city vowels: a study of
the Northern cities shift in Michigan’, Publications of the American Dialect Society
(PADS) 84, Durham, NC: Duke University Press.
Gordon, Matthew J. (2001b), ‘Investigating chain shifts and mergers’, in
J. K. Chambers, P. Trudgill and N. Schilling-Estes (eds), The Handbook of
Language Variation and Change, Malden, MA: Blackwell, 244–66.
Gordon, Matthew Kelly (2004), ‘Syllable weight’, in Hayes et al. (eds) (2004),
277–312.
Gordon, Matthew Kelly (2006), Syllable Weight: Phonetics, Phonology, Typology,
New York: Routledge.
Grant, Colin (2011), An Acoustic Analysis of Anglian Smoothing (MA thesis,
Indiana University).
BIBLIOGR APHY 375
Greenberg, Joseph (1978), Universals of Human Language, Vol. II Phonology,

Stanford: Stanford University Press.
Greenfield, Stanley and Daniel Calder (1986), A New Critical History of Old English
Literature. With a Survey of the Anglo-Latin Background by Michael Lapidge, New
York and London: New York University Press.
Gussmann, Edmund (2002), Phonology: Analysis and Theory, Cambridge:
Hall, Nancy (2006), ‘Cross-linguistic patterns of vowel intrusion’, Phonology 23,
387–429.
Hall, Tracy A. and Silke Hamann (2010), ‘On the cross-linguistic avoidance of
rhotic plus high front vocoid sequences’, Lingua 120, 1,821–44.
Halmari, Helena (2011), ‘Alliteration in inaugural addresses: from George
Washington to Barack Obama’, in Roper (ed.) (2011), 45–61.
Hamann, Silke (2005), ‘The diachronic emergence of retroflex segments in three
languages’, Tijdschrift voor linguistiek te Utrecht 15, 29–48.
Hammond, Michael (1997), ‘Vowel quantity and syllabification in English’,
Language 73:1, 1–17.
Hammond, Michael (1999), The Phonology of English: A Prosodic Optimality-
Theoretic Approach, Oxford: Oxford University Press.
Hanham, Alison (ed.) (1975), The Cely Letters 1472–1488, EETS 273, London:
Oxford University Press.
Hanson, Kristin and Paul Kiparsky (1996), ‘A parametric theory of poetic
meter’, Language 72:2, 287–336.
Harmon, William (1997), ‘English versification: fifteen hundred years of conti-
nuity and change’, Studies in Philology 94:1, 1–37.
Harrington, Jonathan, Felicity Cox and Zoe Evans (1997), ‘An acoustic phonetic
study of broad, general, and cultivated Australian English vowels’, Australian
Journal of Linguistics 17:2, 155–84.
Haugen, Einar (1938), ‘Notes on “voiced t” in American English’, Dialect Notes
6:16–17, 627–34.
Hawkins, Sarah and Jonathan Midgley (2005), ‘Formant frequencies of RP
monophthongs in four age groups of speakers’, Journal of the International
Phonetic Association 35:2, 183–99.
Hay, Jennifer and Andrea Sudbury (2005), ‘How rhoticity became /r/-sandhi’,
Language 81:4, 799–824.
Hayes, Bruce (1983), ‘A grid-based theory of English meter’, Linguistic Inquiry
14, 357–93.
Hayes, Bruce (1988), ‘Metrics and phonological theory’, in Frederick Newmeyer
(ed.), Linguistics: The Cambridge Survey, Cambridge: Cambridge University
Press, 220–49.
Hayes, Bruce (1989), ‘The prosodic hierarchy in meter’, in Paul Kiparsky and
Gilbert Youmans (eds), Phonetics and Phonology, Vol. 1: Rhythm and Meter, San
Diego: Academic Press, 201–60.
Hayes, Bruce (1995), Metrical Stress Theory: Principles and Case Studies, Chicago:
University of Chicago Press.
Hayes, Bruce (2009a) ‘Syllabification in English’, <http://linguistics.ucla.

edu/people/hayes/251English/Readings/HayesAmbisyllabicity.pdf> (last
accessed 3 June 2013).
Hayes, Bruce (2009b) Introductory Phonology, Oxford and Malden, MA:
Wiley-Blackwell.
Hayes, Bruce and Donca Steriade (2004), ‘Introduction: the phonetic bases of
phonological markedness’, in Hayes et al. (eds) (2004), 1–34.
Hayes, Bruce, Robert Kirchner and Donca Steriade (eds) (2004), Phonetically
Based Phonology, Cambridge: Cambridge University Press.
Hayes, Bruce, Colin Wilson and Anne Shisko (2012), ‘Maxent grammars for the
metrics of Shakespeare and Milton’, Language 88:4, 691–731.
Hebda, Anna (2002), ‘Open syllable lengthening before /t/ and /k/ in the
language of Cursor Mundi – the evidence from rhyme vowels’, Studia Anglica
Posnaniensia 38, 229–37.
Heselwood, Barry (2009), ‘R vocalization, linking R and intrusive R: accounting
for final schwa in RP English’, Transactions of the Philological Society 107:1, 66–97.
Hickey, Raymond (1986), ‘Possible phonological parallels between Irish and
Irish English’, English World-Wide 7, 1–21.
Hickey, Raymond (1998), ‘The Dublin Vowel Shift and the historical perspec-
tive’, in Jacek Fisiak and Marcin Krygier (eds), Advances in English Historical
Linguistics 1996, Trends in Linguistics. Studies and Monographs 112, Berlin:
Hill, Archibald A. (1940), ‘Early loss of [r] before dentals’, PMLA 55:2, 308–59.
Hogg, Richard M. (1992a), A Grammar of Old English, Vol. I: Phonology, Oxford and
Cambridge, MA: Blackwell.
Hogg, Richard (1992b), ‘Phonology and morphology’, in Hogg (1992a), 67–168.
Hogg, Richard (1996), ‘Old English open-syllable lengthening’, Transactions of
the Philological Society 94:1, 57–72.
Hogg, Richard (2000), ‘On the (non-)existence of High Vowel Deletion’, in
Aditi Lahiri (ed.), Analogy, Levelling, Markedness, Berlin: Mouton de Gruyter,
353–77.
Hogg, Richard (2002), An Introduction to Old English, Edinburgh: Edinburgh
University Press.
Hogg, Richard (2006a), ‘English in Britain’, in Hogg and Denison (eds) (2006),
352–84.
Hogg, Richard (2006b), ‘Old English dialectology’, in Kemenade and Los (eds)
(2006), 395–416.
Hogg, Richard M. (gen. ed.) (1992–c. 2001), The Cambridge History of the English
Language in 6 volumes, Cambridge: Cambridge University Press.
Hogg, Richard and David Denison (eds) (2006), A History of the English Language,
Cambridge: Cambridge University Press.
Hogg, Richard and Robert D. Fulk (2010), A Grammar of Old English, Volume II:
Morphology, Chichester: Wiley-Blackwell.
Holt, D. Eric (ed.) (2003), Optimality Theory and Language Change, Studies in
Natural Language and Linguistic Theory 56, Dordrecht: Kluwer.
BIBLIOGR APHY 377
Honeybone, Patrick (2005), ‘Diachronic evidence in segmental phonology: the

case of obstruent laryngeal specifications’, in M. van Oostendorp and J. van de
Weijer (eds), The Internal Organization of Phonological Segments, Berlin: Mouton
de Gruyter, 317–53.
Honeybone, Patrick and Joseph Salmons (eds) (2014), The Oxford Handbook of
Historical Phonology, Oxford: Oxford University Press.
Hoover, David (1985), A New Theory of Old English Meter, New York: Peter
Lang.
Horn, Wilhelm and Martin Lehnert (1954), Laut und Leben. Englische
Lautgeschichte der neueren Zeit (1400–1950), 2 vols, Berlin: Deutscher Verlag
der Wissenschaften.
Horobin, Simon and Jeremy Smith (2002), An Introduction to Middle English,
Edinburgh: Edinburgh University Press.
House, Arthur S. (1961), ‘On vowel duration in English’, The Journal of the
Acoustical Society of America 33:9, 1,174–8.
Howell, Robert B. (1987), ‘Tracing the origin of uvular R in the Germanic lan-
guages’, Folia Linguistica Historica 7:2, 317–49.
Howell, Robert B. (1991), Old English Breaking and its Germanic Analogues,
Tübingen: Max Niemeyer.
Howell, Robert and Katerina Somers Wicka (2007), ‘A phonetic account of
Anglian Smoothing’, Folia Linguistica Historica 28:1, 187–214.
Hughes, Geoffrey (2000), A History of English Words, Oxford: Blackwell.
Hume, Elizabeth and K. Johnson (eds) (2001), The Role of Speech Perception in
Phonology, New York: Academic Press.
Hutcheson, Bellenden Rand (1995), Old English Poetic Metre, Cambridge:
D. S. Brewer.
Hyman, Larry (2002), ‘Is there a right-to-left bias in vowel harmony?’, paper pre-
sented at the 9th International Phonology Meeting, Vienna, 1 November 2002,
<http://www.linguistics.berkeley.edu/~hyman/Hyman_Vienna_VH_
paper_forma.pdf> (last accessed 3 June 2013).
Ikegami, Masa T. (1984), Rhyme and Pronunciation: Some Studies of English Rhymes
from King Alisaunder to Skelton, Tokyo: Hogaku-Kenkyu-Kai.
‘Indo-European languages’ (2009), in Encyclopædia Britannica Online, <http://
search.eb.com/eb/article-74551> (last accessed 5 July 2009).
Ingham, Richard (2012), ‘The decline of bilingual competence in French in
medieval England: evidence from the PROME database’, in Nils Langer,
Steffan Davies and Wim Vandenbussche (eds), Language and History, Linguistics
and Historiography: Interdisciplinary Approaches, Studies in Historical Linguistics
9, Frankfurt am Main: Peter Lang, 271–92.
IPA Handbook (1999), Handbook of the International Phonetic Association: A Guide to
the Use of the International Phonetic Alphabet, Cambridge: Cambridge University
Press.
Ito, Junko and Armin Mester (2001), ‘Structure preservation and stratal opacity
in German’, in L. Lombardi (ed.), Segmental Phonology in Optimality Theory,
Cambridge: Cambridge University Press, 261–95.
Iverson, Gregory K. and J. C. Salmons (1995), ‘Aspiration and laryngeal repre-

sentation in Germanic’, Phonology 12, 369–96.
Iverson, Gregory K. and J. C. Salmons (2005), ‘Filling the gap. English tense
vowel plus final /š/’, Journal of English Linguistics 33:3, 1–15.
Iverson Gregory K. and Sang-Cheol Ahn (2007), ‘English voicing in dimen-
sional theory’, Language Sciences 29, 247–69.
Jefferson, Judith and Ad Putter (eds) (2009), Approaches to the Metres of Alliterative
Verse, Leeds Texts and Monographs New Series 17, Leeds: University of
Leeds.
Jespersen, Otto (1909), A Modern English Grammar on Historical Principles. Part I,
Sounds and Spellings, Heidelberg: Carl Winter’s Universitätsbuchhandlung.
Johnson, Wyn and David Britain (2007), ‘L-vocalisation as a natural phenom-
enon: explorations in sociophonology’, Language Sciences 29, 294–315.
Johnston, Paul A., Jr. (1992), ‘English vowel shifting: one great vowel shift or two
small vowel shifts?’, Diachronica 9:2, 189–226.
Jones, Charles (1989), A History of English Phonology, London: Longman.
Jones, Daniel (1924), An English Pronouncing Dictionary (showing the pronunciation of
over 50,000 words in international phonetic transcription), New York: E. P. Dutton.
Jordan, Richard (1974), A Handbook of Middle English Grammar: Phonology, trans.
and rev. by Eugene Joseph Crook, Janua Linguarum Series Practica 218, The
Hague: Mouton.
Kastovsky, Dieter (1992), ‘Semantics and vocabulary’, in Hogg (1992a), 290–408.
Kastovsky, Dieter (ed.) (1994), Studies in Early Modern English, Berlin: Mouton
de Gruyter.
Kaun, Abigail (1995), The Typology of Rounding Harmony: An Optimality-Theoretic
Approach, PhD thesis, University of California, Los Angeles.
Kawahara, Shigeto (2007), ‘Sonorancy and geminacy’, in University of Massachusetts
Occasional Papers in Linguistics 32: Papers in Optimality III, Amherst: Book Surge
Publishing, 145–86.
Keenan, Edward (2002), ‘Explaining the creating of reflexive pronouns in
English’, in Minkova and Stockwell (eds) (2002), 325–55.
Kelly, Michael (2004), ‘Word onset patterns and lexical stress in English’,
Journal of Memory and Language 50, 231–44.
Kemenade, Ans van and Bettelou Los (eds) (2006), The Handbook of the History of
English, Oxford: Blackwell.
Kenyon, John S. and Thomas A. Knott (1949), Pronouncing Dictionary of American
English, <http://www.archive.org/details/pronouncingdicti00unse> (last
accessed 3 June 2013).
Keyser, S. J. and W. O’Neil (1985), Rule Generalization and Opacity in Language
Change, Dordrecht: Foris.
Kim, Myungsook (1993), ‘On lengthening in the open syllables of Middle
English’, Lingua 91, 261–77.
King, Anne (1992), ‘You say [æjðər] and I say [æjhwæðər]? Interpreting Old
English written data’, in Colman (ed.) (1992), 20–43.
Kiparsky, Paul (1973), ‘The role of linguistics in a theory of poetry’, Daedalus
BIBLIOGR APHY 379
102, 231–45. [Reprinted in Donald C. Freeman (ed.) (1981), Essays in Modern

Stylistics, London: Methuen, 9–23].
Kiparsky, Paul (1977), ‘The rhythmic structure of English verse’, Linguistic
Inquiry 8, 189–247.
Kiparsky, Paul and W. O’Neil (1976), ‘The phonology of Old English inflex-
ions’, Linguistic Inquiry 7, 527–57.
Kirchner, Robert (2004), ‘Consonant lenition’, in Hayes et al. (eds) (2004),
313–46.
Kirkby, John [1746] (1971), A New English Grammar, or Guide to the English Tongue,
with Notes . . ., facsimile reprint, Menston: Sclar Press.
Kirwin, William J. (2001), ‘Newfoundland English’, in John Algeo (ed.), The
Cambridge History of the English Language, Vol. VI: English in North America,
Cambridge: Cambridge University Press, 441–55.
Kitson, Peter (1997), ‘When did Middle English begin? Later than you think!’,
in Jacek Fisiak (ed.), Studies in Middle English Linguistics, Trends in Linguistics.
Studies and Monographs 103, Berlin: Mouton de Gruyter, 221–71.
Knappe, Gabrielle (1997), ‘Though it is tough: on regional differences in the
development and substitution of the Middle English voiceless velar fricative
[x] in syllable coda position’, in Heinrich Ramisch and Kenneth Wynne (eds),
Language in Time and Space: Studies in Honour of Wolfgang Viereck on the Occasion
of his 60th Birthday, Zeitschrift für Dialektologie und Linguistik, Beihefte, Heft
97, Stuttgart: Franz Steiner, 139–63.
Kökeritz, Helge (1953), Shakespeare’s Pronunciation, New Haven, CT: Yale
University Press.
Kökeritz, Helge (1959), ‘English i for Old French ü’, in J. Vendryes (eds),
Mélanges de linguistique et de philologie, Fernand Mossé in memoriam, Paris: Didier,
218–25.
Kökeritz, Helge (1961), A Guide to Chaucer’s Pronunciation, New York: Holt,
Rinehart and Winston.
Kortmann, Bernd and Benedikt Szmrecsanyi (2004), ‘Global synopsis: morpho-
logical and syntactic variation in English’, in Bernd Kortmann and Edgar
Schneider (eds), A Handbook of Varieties of English, Berlin: Mouton de Gruyter,
1,142–202.
Krapp, George Philip and Elliot van Kirk Dobbie (1931–53), The Anglo-Saxon
Poetic Records, 6 vols, New York: Columbia University Press.
Kreidler, Charles (2004), The Pronunciation of English: A Coursebook, Oxford and
Malden, MA: Blackwell Publishing.
Krygier, Marcin (1996), ‘Plural markers of the Old English nouns of relationship
in the two manuscripts of Laamon’s Brut’, in Jacek Fisiak (ed.), Middle English
Miscellany: From Vocabulary to Linguistic Variation, Poznań: Motivex, 47–69.
Kuryl̃owicz, Jerzy (1970), Die sprachlichen Grundlagen der altgermanischen Metrik,
Innsbruck: Institut für vergleichende Sprachwissenschaft der Universität
Innsbruck.
Labov, William (1991), ‘The three dialects of English’, in P. Eckert (ed.), New
Ways of Analyzing Sound Change, New York: Academic Press, 1–44.
Labov, William (1994), Principles of Linguistic Change. Vol. I, Internal Factors,

Oxford: Blackwell.
Labov, William (2008), ‘Triggering events’, in Susan Fitzmaurice and Donka
Minkova (eds), Studies in the History of the English Language IV: Empirical and
Analytical Advances in the Study of English Language Change, Berlin: Mouton de
Gruyter, 11–54.
Labov, William, Paul Cohen, Clarence Robins and John Lewis (1968), A Study
of the Nonstandard English of Black and Puerto Rican Speakers in New York City,
Cooperative Research Report no. 3288, Washington DC: United States Office
of Education.
Labov, William, M. Yaeger and R. Steiner (1972), A Quantitative Study of Sound
Change in Progress, Philadelphia: US Regional Survey.
Labov, William, Sharon Ash and Charles Boberg (2006), The Atlas of North
American English (ANAE), Berlin: Mouton de Gruyter.
Ladefoged, Peter (1982), A Course in Phonetics, 2nd edn, New York: Harcourt
Brace Jovanovich.
Ladefoged, Peter (2005), Vowels and Consonants, Oxford: Blackwell Publishing.
Ladefoged, Peter and Ian Maddieson (1996), The Sounds of the World’s Languages,
Oxford and Malden, MA: Wiley-Blackwell.
LAEME, A Linguistic Atlas of Early Middle English, 1150–1325. Compiled by Margaret
Laing. [http://www.lel.ed.ac.uk/ihd/laeme1/laeme1.html] (Edinburgh: ©
2008– The University of Edinburgh).
Lahiri, Aditi and Paula Fikkert (1999), ‘Trisyllabic shortening in English: past
and present’, English Language and Linguistics 3:2, 229–67.
Laing, Margaret (1999), ‘Confusion wrs confounded: litteral substitution sets in
early Middle English writing systems’, Neuphilologische Mitteilungen 100, 251–70.
Laing, Margaret (2009), ‘Orthographic indications of weakness in early Middle
English’, in Minkova (ed.) (2009), 237–316.
Laing, Margaret and Roger Lass (2003), ‘Tales of the 1001 nists: the phonologi-
cal implications of litteral substitution sets in some thirteenth-century South-
West Midland texts’, English Language and Linguistics 7:2, 257–78.
Laing, Margaret and Roger Lass (2006), ‘Early Middle English dialectology:
problems and prospects’, in Kemenade and Los (eds) (2006), 417–51.
Laing, Margaret and Roger Lass (2009), ‘Shape-shifting, sound-change and the
genesis of prodigal writing systems’, English Language and Linguistics 13:1, 1–31.
LALME, A Linguistic Atlas of Late Middle English (1986), ed. Angus McIntosh,
Michael L. Samuels and Michael Benskin, with the assistance of Margaret
Laing and Keith Williamson, Aberdeen: Aberdeen University Press.
Lamb, Sydney and E. Douglas Mitchell (eds) (1991), Sprung from Some Common
Source: Investigations into the Prehistory of Ancient Languages, Stanford: Stanford
University Press.
Lapidge, Michael (1979), ‘Aldhelm’s Latin poetry and Old English verse’,
Comparative Literature 31, 209–31.
Lapidge, Michael (1986), ‘The Anglo-Latin background’, in Greenfield and
Calder (1986), 5–37.
BIBLIOGR APHY 381
Lappe, Sabine (2007), English Prosodic Morphology, Dordrecht: Springer.

Lass, Roger (1976), English Phonology and Phonological Theory, Cambridge:
Lass, Roger (1980), On Explaining Language Change, Cambridge: Cambridge
University Press.
Lass, Roger (1985), ‘Minkova noch einmal. MEOSL and the resolved foot’, Folia
Linguistica Historica 6, 245–66.
Lass, Roger (1987), The Shape of English: Structure and History, London: J. M. Dent
and Sons.
Lass, Roger (1992a), ‘Phonology and morphology’, in Norman Blake (ed.), The
Cambridge History of the English Language, Vol. II: 1066–1476, 23–155.
Lass, Roger (1992b), ‘Front rounded vowels in Old English’, in Colman (ed.)
(1992), 88–116.
Lass, Roger (1992c), ‘What, if anything, was the Great Vowel Shift?, in Matti
Rissanen, Ossi Ihalainen, Terttu Nevalainen and Irma Taavitsainen (eds),
History of Englishes: New Methods and Interpretations in Historical Linguistics,
Lass, Roger (1994), Old English: A Historical Linguistic Companion, Cambridge:
Lass, Roger (1999), ‘Phonology and morphology’, in Roger Lass (ed.), The
Cambridge History of the English Language, Vol. III: 1476–1776, Cambridge:
Cambridge University Press, 56–186.
Lass, Roger (2000), ‘Language periodization and the concept “middle”’, in Irma
Taavitsainen, Terttu Nevalainen, Päivi Pahta and Matti Rissanen (eds),
Placing Middle English in Context, Berlin and New York: Mouton de Gruyter,
7–42.
Lass, Roger (2009), ‘On schwa: synchronic prelude and historical fugue’, in
Minkova (ed.) (2009), 47–77.
Lass, Roger (2014), ‘Interpreting alphabetic orthographies: early Middle English
spelling’, in Patrick Honeybone and Joseph Salmons (eds), The Oxford
Handbook of Historical Phonology, Oxford: Oxford University Press.
Lass, Roger (ed.) (1969), Approaches to English Historical Linguistics, New York:
Holt, Rinehart and Winston.
Lass, Roger and John Anderson (1975), Old English Phonology, Cambridge:
Lass, Roger and Susan Wright (1985), ‘The South African chain shift’, in Roger
Eaton, Olga Fischer, Willem Koopman and Frederike van der Leek (eds),
Papers from the 4th International Conference on English Historical Linguistics,
Amsterdam: John Benjamins, 137–62.
Lass, Roger and Margaret Laing (2005), ‘Are front rounded vowels retained in
West Midlands Middle English?’, in Nikolaus Ritt and Herbert Schendl (eds),
Rethinking Middle English: Linguistic and Literary Approaches, Frankfurt am Main:
Peter Lang, 280–90.
Lauttamus, Timo (1981), ‘A note on the development of Old English [x] to
Middle English [f]’, Neuphilologische Mitteilungen 82, 1–5.
Lavoie, Lisa (2009), ‘Testing consonant weakness phonetically’, in Minkova

(ed.) (2009), 29–46.
Lawson, Eleanor, Jane Stuart-Smith and James Scobbie (2008), ‘Articulatory
insights into language variation and change: preliminary findings from an
ultrasound study of derhoticization in Scottish English’, in K. Gorman (ed.),
University of Pennsylvania Working Papers in Linguistics 14:2: Selected Papers from
NWAV 36, 102–10.
Leach, Alfred (1880), The Letter H Past, Present, and Future, London: Griffith and
Farran.
Lehiste, Ilse (1970), Suprasegmentals, Cambridge, MA: MIT Press.
Lehmann, Winfred P. [1956] (1971), The Development of Germanic Verse Form, New
York: Gordian Press.
Leith, Dick (1983), A Social History of English, London: Routledge and Kegan Paul.
Lewis, M. Paul, Gary F. Simons and Charles D. Fennig (eds) (2013) Ethnologue:
Languages of the World, 17th edn, Dallas: SIL International. Online version
available at <http://www.ethnologue.com> (last accessed 16 May 2013).
Liberman, Anatoly (1965a), ‘K istorii dolgoty v sredneangliiskom’ [On the
history of length in Middle English], Filologicheskie nauki 1, 125–35.
Liberman, Anatoly (1965b), ‘K voprosu o sredneangliiskom udlinenii v otkry-
tom sloge’ [On the question of Middle English lengthening in open syllable],
Vestnik Leningradskogo Universiteta 8, 129–37.
Liberman, Anatoly (1992a), ‘Vowel lengthening before resonant + another
consonant and svarabhakti in Germanic’, in Irmengard Rauch, Gerald F. Carr
and Robert L. Kyes (eds), On Germanic Linguistics: Issues and Methods, Berlin:
Liberman, Anatoly (1992b), ‘A bird’s-eye view of open syllable lengthening in
English and in other Germanic languages’, NOWELE: North-Western European
Language Evolution 20, 67–87.
Liberman, Anatoly (1995), ‘The beginning and end of the (great) vowel shifts
in Late Germanic’, in Irmengard Rauch and Gerald F. Carr (eds), Insights in
Germanic Linguistics: Methodology in Transition, Trends in Linguistics. Studies
and Monographs 83, Berlin: Mouton de Gruyter, 203–31.
Liebl, Christian (2006), ‘The A and O of a medieval English sound change:
prolegomena to a study of the origins and early geographical diffusion of
/ɑ:/ > /o:/’, in Nikolaus Ritt, Herbert Schendl, Christiane Dalton-Puffer
and Dieter Kastovsky (eds), Medieval English and its Heritage: Structure, Meaning
and Mechanisms of Change, Studies in Medieval Language and Literature 16,
Frankfurt am Main: Peter Lang, 19–35.
Luick, Karl (1898), ‘Beiträge zur englischen Grammatik III. Die
Quantitätsveränderungen im Laufe der englischen Sprachentwicklung’,
Anglia 20, 335–62.
Luick, Karl [1914, 1921–40] (1964), Historische Grammatik der englischen Sprache,
2 vols, Leipzig: Tauchnitz.
Lutz, Angelika (1991), Phonotaktisch gesteuerte Konsonantenveränderungen in der
Geschichte des Englischen, Tübingen: Niemeyer.
BIBLIOGR APHY 383
Lutz, Angelika (1994), ‘Vocalization of “post-vocalic r” – an Early Modern

English sound change?’, in Kastovsky (ed.) (1994), 167–85.
Lutz, Angelika (2002), ‘When did English begin?’, in Teresa Fanego, Belén
Méndez-Naya and Elena Seoane (eds), Sounds, Words, Texts and Change:
Selected Papers from 11 ICEHL, Santiago de Compostela, 7–11 September 2000,
Amsterdam: John Benjamins, 145–71.
Lutz, Angelika (2004), ‘The first push: a prelude to the Great Vowel Shift’, Anglia
122:2, 209–24.
Maclagan, Margaret and Jennifer Hay (2007), ‘Getting fed up with our feet: con-
trast maintenance and the New Zealand English “short” front vowel shift’,
Language Variation and Change 19, 1–25.
MacMahon, Michael K. C. (1998), ‘Phonology’, in Suzanne Romaine (ed.),
The Cambridge History of the English Language, Vol. IV: 1776–1997, Cambridge:
Maddieson, Ian (1984), Patterns of Sound, Cambridge: Cambridge University
Press.
Maddieson, Ian (1985), ‘Phonetic cues to syllabification’, in Victoria Fromkin
(ed.), Phonetic Linguistics. Essays in Honor of Peter Ladefoged, Orlando: Academic
Press, 203–21.
Maddieson, Ian (1993), ‘Splitting the mora’, UCLA Working Papers in Phonetics
83, 9–19.
Magnuson, Thomas J. (2007), ‘The story of /r/ in two vocal tracts’, Proceedings
of the 16th International Congress of the Phonetic Sciences, 6–10 August 2007,
Saarbrucken, Germany, 1,193–6.
Malone, Kemp (1930), ‘When did Middle English begin?’, in James T. Hatfield,
Werner Leopold and A. J. Friedrich Zieglschmid (eds), Curme Volume of
Linguistic Studies, Language Monograph No. 7, Philadelphia: Linguistic
Society of America, 110–17.
Marchand, Hans (1969), The Categories and Types of Present-Day English Word-
Formation: A Synchronic-Diachronic Approach, Munich: Beck.
Mayr, Robert (2010), ‘What exactly is a front rounded vowel? An acoustic
and articulatory investigation of the NURSE vowel in South Wales English’,
Journal of the International Phonetic Association 40:1, 93–112.
McCully, Christopher B. (1997), ‘Stress, survival and change: Old to Middle
English’, in Jacek Fisiak (ed.), Studies in Middle English Linguistics, Trends
in Linguistics. Studies and Monographs 103, Berlin: Mouton de Gruyter,
301–35.
McDavid, Raven I., Jr. (1952), ‘Some social differences in pronunciation’,
Language Learning 4:3–4, 102–16.
McKie, Michael (1997), ‘The origins and early development of rhyme in
English verse’, The Modern Language Review 92:4, 817–31.
McMahon, April (1994), Understanding Language Change, Cambridge: Cambridge
University Press.
McMahon, April M. S. (2000), Lexical Phonology and the History of English,
Cambridge: Cambridge University Press.
McMahon, April (2002), An Introduction to English Phonology, Oxford: Oxford

University Press.
McMahon, April (2006), ‘Restructuring Renaissance English’, in Linda
Mugglestone (ed.), The Oxford History of English, Oxford: Oxford University
Press, 147–77.
McMahon, April (2009), ‘Perspectives on weakness from English /r/’, in
Minkova (ed.) (2009), 99–115.
Menken, H. L. (1910), ‘The two Englishes’, Baltimore Evening Sun, 15 September.
Milroy, James (1983), ‘On the sociolinguistic history of /h/-dropping in
English’, in Michael Davenport and Erik Hansen (eds), Current Topics in
English Historical Linguistics, Odense: Odense University Press, 37–54.
Minkoff, Marco [1967] (1972), English Historical Grammar, 3rd edn, Sofia: Naouka
I Izkoustvo.
Minkova, Donka (1982), ‘The environment for open-syllable lengthening in
Middle English’, Folia Linguistica Historica 3, 29–58.
Minkova, Donka (1985), ‘Of rhyme and reason: some foot-governed quantity
changes in English’, in Roger Eaton, Olga Fischer, Willem Koopman and
Frederike van der Leek (eds), Papers from the 4th International Conference on
English Historical Linguistics, Amsterdam: John Benjamins, 163–78.
Minkova, Donka (1991), The History of Final Vowels in English, Berlin and New
York: Mouton de Gruyter.
Minkova, Donka (1992), ‘Verse structure in the Middle English “Genesis and
Exodus”’, Journal of English and Germanic Philology 91:2, 157–78.
Minkova, Donka (1993), ‘On leap-frogging in historical phonology’, in Papers
from the Tenth International Conference on Historical Linguistics Amsterdam, ed.
Jaap van Marle, Amsterdam Studies in the Theory and History of Linguistic
Science, Series IV, Current Issues in Linguistic Theory, vol. 107, Amsterdam:
John Benjamins, 211–29.
Minkova, Donka (1996), ‘Non-primary stress in early Middle English accentual-
syllabic verse’, in Chris McCully and J. J. Anderson (eds), English Historical
Metrics, Cambridge: Cambridge University Press, 95–120.
Minkova, Donka (1997), ‘The credibility of Pseudo-Alfred: prosodic insights in
post-Conquest mongrel meter’, Modern Philology 94:4, 427–54.
Minkova, Donka (2000), ‘Middle English prosodic innovations and their test-
ability in verse’, in Irma Taavitsainen, Terttu Nevalainen, Päivi Pahta and
Matti Rissanen (eds), Placing Middle English in Context, Berlin and New York:
Minkova, Donka (2003), Alliteration and Sound Change in Early English, Cambridge
Studies in Linguistics 101, Cambridge: Cambridge University Press.
Minkova, Donka (2004), ‘Philology, linguistics, and the history of [hw] ~[w]’,
in Curzan and Emmons (eds) (2004), 6–46.
Minkova, Donka (2006a) ‘Old and Middle English prosody’, in Kemenade and
Los (eds) (2006), 95–125.
Minkova, Donka (2006b), ‘Phonetic naturalness vs. orthography in the forma-
tion of a standardized consonantal inventory of English’, in Ursula Schäfer
BIBLIOGR APHY 385
(ed.), The Beginnings of Standardization. Language and Culture in Fourteenth

Century England, Frankfurt am Main and New York: Peter Lang, 147–75.
Minkova, Donka (2007), ‘The forms of verse’, in Peter Brown (ed.), A Companion
to Medieval English Literature and Culture c. 1350–1500, Oxford: Blackwell
Publishing, 176–96.
Minkova, Donka (2008a), ‘Prefixation and stress in Old English’, Word Structure
1:1, 21–52.
Minkova, D. (2008b), Review of Early English Metre by Thomas Bredehoft,
Toronto, Buffalo, London: University of Toronto Press, Speculum 83,
673–5. Full text at <http://www.english.ucla.edu/faculty/minkova/
BredehoftSpeculumlong.pdf> (last accessed 3 June 2013).
Minkova, Donka (2009a) ‘Diagnostics of metricality in Middle English allitera-
tive verse’, in Jefferson and Putter (eds) (2009), 77–115.
Minkova, Donka (2009b), ‘On the meter of Middle English alliterative verse’, in
Aroui and Arleo (eds) (2009), 207–25.
Minkova, Donka (2009c) ‘Inflectional syncope vs. epanthesis. Notes on the
history of weak preterites in English’, in Minkova (ed.) (2009), 316–46.
Minkova, Donka (2011a), ‘Syllable weight and the weak-verb paradigms in
Old English’, in David Denison, Ricardo Bermúdez-Otero, Chris McCully
and Emma Moore (eds), Analysing Older English, Cambridge: Cambridge
University Press, 194–213.
Minkova, Donka (2011b) ‘Phonemically contrastive fricatives in Old English?’,
English Language and Linguistics 15:1, 31–60.
Minkova, Donka (2012a), ‘Prosody’, in Alexander Bergs and Laurel Brinton
(eds), Historical Linguistics of English, Handbücher zur Sprach- und
Kommunikationswissenschaft, Berlin: Mouton de Gruyter, 113–29.
Minkova, Donka (2012b), ‘Functional stress-shifting in Old English’, handout
from ICEHL 17, Zürich, August 2012, <https://www.researchgate.net/
profile/Donka_Minkova/publications/> (last accessed 18 June 2013).
Minkova, Donka (2013), ‘Reconstructing stress in Old and Middle English’, in
Manfred Krug and Julia Schlüter (eds), Research Methods in Language Variation
and Change, Cambridge: Cambridge University Press, 260–78.
Minkova, Donka (2014), ‘Establishing phonemic contrast in written sources (with
reference to Old English)’, in Patrick Honeybone and Joseph Salmons (eds),
The Oxford Handbook of Historical Phonology, Oxford: Oxford University Press.
Minkova, Donka (ed.) (2009), Phonological Weakness in English: From Old to Present-
Day English, Basingstoke: Palgrave Macmillan.
Minkova, Donka and Robert Stockwell (1992), ‘Homorganic clusters as moric
busters in the history of English: the case of -ld, -nd, -mb’, in Matti Rissanen,
Ossi Ihalainen, Terttu Nevalainen and Irma Taavitsainen (eds), History
of Englishes: New Methods and Interpretations in Historical Linguistics, Berlin:
Minkova, Donka and Robert Stockwell (1994), Review of Norman Blake (ed.),
The Cambridge History of the English Language, Vol. II: 1066–1476, Journal of
Linguistics 30:2, 528–66.
Minkova, Donka and Robert Stockwell (1997a), ‘Chaucerian phonemics:

evidence and interpretation’, in Raymond Hickey and Stanisław Puppel
(eds), Language History and Linguistic Modelling, Trends in Linguistics. Studies
and Monographs 101, Berlin: Mouton de Gruyter, 29–58.
Minkova, Donka and Robert Stockwell (1997b), ‘Against the emergence of the
nuclear stress rule in Middle English’, in Jacek Fisiak (ed.), Studies in Middle
English Linguistics, Trends in Linguistics. Studies and Monographs 103,
Minkova, Donka and Robert Stockwell (1998), ‘The origins of long-short allo-
morphy in English’, in Jacek Fisiak and Marcin Krygier (eds), Advances in
English Historical Linguistics, Berlin: Mouton de Gruyter, 211–40.
Minkova, Donka and Robert Stockwell (eds) (2002), Studies in the History of the
English Language: A Millennial Perspective, Berlin: Mouton de Gruyter.
Minkova, Donka and Robert Stockwell (2003a), ‘English vowel shifts and
“optimal” diphthongs: is there a logical link?’, in Holt (ed.) (2003), 169–90.
Minkova, Donka and Robert Stockwell (2003b), ‘Emendation and the Chaucerian
metrical template’, in Donka Minkova and Theresa Tinkle (eds), Chaucer and
the Challenges of Medievalism, Frankfurt am Main: Peter Lang, 129–40.
Minkova, Donka and Robert Stockwell (2005), ‘Clash avoidance in morpho-
logically derived words in Middle English. (Why [-hυd] but [-dəm])?’, in
Nikolaus Ritt and Herbert Schendl (eds), Rethinking Middle English: Linguistic
and Literary Approaches, Frankfurt am Main: Peter Lang, 263–80.
Minkova, Donka and Robert Stockwell (2006), ‘English words’, in Bas Aarts and
April McMahon (eds), The Handbook of English Linguistics, Oxford: Blackwell,
461–83.
Minkova, Donka and Robert Stockwell (2009), English Words: History and
Structure, 2nd edn, Cambridge: Cambridge University Press.
Minkova, Donka and Kie Ross Zuraw (forthcoming), ‘Testing ambisyllabicity in
early English’, in Merja Kytö and Päivi Pahta (eds), The Cambridge Handbook of
English Historical Linguistics, Cambridge: Cambridge University Press.
Moen, Per (1979), ‘The laxing of the high-back vowel: historical development
and recent trends in Minnesota speech’, American Studies in Scandinavia 11,
67–79.
Montgomery, Michael (2001), ‘British and Irish antecedents’, in John Algeo
(ed.), The Cambridge History of the English Language, Vol. VI: English in North
America, Cambridge: Cambridge University Press, 86–153.
Montgomery, Michael and Connie Eble (2004), ‘Historical perspectives on the
pen/pin merger in Southern American English’, in Curzan and Emmons (eds)
(2004), 415–34.
Moore, Samuel and Albert H. Marckwardt (1951), Historical Outlines of English
Sounds and Inflections, Ann Arbor, MI: George Wahr.
Morén, Bruce (2003), ‘Weight typology: an Optimality Theoretic approach’,
The Linguistic Review 20, 281–304.
Moreton, Elliott and Erik Thomas (2007), ‘Origins of Canadian Raising in
voiceless-coda effects: a case study in phonologization’, Papers in Laboratory
BIBLIOGR APHY 387
Phonology 9, ed. Jennifer Cole and José Ignacio Hualde, Berlin: Mouton de
Gruyter, 36–74.
Morris, Richard (ed.) (1863), The Pricke of Conscience (Stimulus Conscientiae), A
Northumbrian Poem by Richard Rolle de Hampole, Berlin: A. Asher.
Mossé, Fernand and James Walker (trans.) [1952] (1968), Handbook of Middle
English, Baltimore: Johns Hopkins University Press.
Mugglestone, Lynda (2003), Talking Proper: The Rise and Fall of the English Accent
as a Social Symbol, New York: Oxford University Press.
Mulcaster, Richard [1582] (1925), THE FIRST PART OF THE ELEMENTARIE
WHICH ENTREATETH CHEFELIE OF THE right writing of our English tung, set
furth by RICHARD MULCASTER. Imprinted at London by Thomas Vautroullier
dwelling in the blak-friers by Lud-gate. 1925 reprint, Oxford: Clarendon Press.
Murray, Robert W. (2000), ‘Syllable cut prosody in early Middle English’,
Language 76, 617–54.
Muthmann, Gustav (1999), Reverse English Dictionary Based on Phonological and
Morphological Principles, Berlin: Mouton de Gruyter.
Nagy, Naomi and Patricia Irwin (2010), ‘Boston (r): Neighbo(r)s nea(r) and
fa(r)’, Language Variation and Change 22, 241–78.
NED, New English Dictionary on Historical Principles (1884–1928), Oxford:
Clarendon Press.
Nevalainen, Terttu (1999), ‘Lexis and semantics’, in Roger Lass (ed.), The
Cambridge History of the English Language, Vol. III: 1476–1776, Cambridge:
Nevalainen, Terttu (2006), An Introduction to Early Modern English, Edinburgh:
Edinburgh University Press.
Nevalainen, Terttu and Ingrid Tieken-Boon van Ostade (2006), ‘Standardisation’,
in Hogg and Denison (eds) (2006), 271–311.
Nevanlinna, Saara (2007), ‘The loss of final plosive consonant in Late Middle
English rhyme words’, in Hans Sauer and Renate Bauer (eds), Beowulf and
Beyond, Frankfurt am Main: Peter Lang, 271–92.
Oakden, J. P. [1930, 1935] (1968) Alliterative Poetry in Middle English, Hamden,
CT: Archon Books.
Ogden, Richard (2009), Introduction to English Phonetics, Edinburgh: Edinburgh
University Press.
Ogura, Mieko (1987), Historical English Phonology: A Lexical Perspective, Tokyo:
Kenkyusha.
Oppenheimer, Stephen (2006), The Origins of the British: A Genetic Detective Story,
London: Constable and Robinson.
Ostler, Nicholas (2005), Empires of the Word: A Language History of the World, New
York: HarperCollins.
Page, B. Richard (2006), ‘The diachrony and synchrony of vowel quantity in
English and Dutch’, Diachronica 23:1, 61–104.
Page, Richard (2012) ‘Regular vowel lengthening versus sporadic vowel length-
ening in Middle English’, paper presented at GLAC 18, Bloomington, IN,
April 2012.
Pająk, Bożena (2009), ‘Contextual constraints on geminates: the case of Polish’,

Proceedings of BLS 35, Berkeley: University of California.
Palsgrave, John (1530), Lesclarcissement de la langue francoyse, London: J. Haukyns.
Parkes, M. B. (1983), ‘On the presumed date and possible origin of the manu-
script of the Orrmulum’, in E. G. Stanley and D. Gray (eds), Five Hundred
Years of Words and Sounds: A Festschrift for Eric Dobson, Cambridge: Cambridge
Petrova, Olga (2000), ‘Grimm’s Law in Optimality Theory’, in Karlene
Jones-Bley, Martin E. Huld and Angela Della Volpe (eds), Proceedings of the
Eleventh Annual UCLA Indo-European Conference, Los Angeles, 4–5 June 1999,
Washington DC: Institute for the Study of Man, 45–67.
Petrova, Olga (2004), ‘Perceptual contrast in Verner’s Law’, in Curzan and
Emmons (eds) (2004), 371–411.
Phillips, Betty (1981a) ‘The phonetic basis of a Late Old English sound change’,
in Phonologica 1980, ed. W. U. Dressler, Oskar Pfeiffer and John Rennison,
Innsbruck: Institut für Sprachwissenschaft, 337–41.
Phillips, Betty (1981b), ‘Lexical diffusion and Southern tune, duke, news’, American
Speech 56:1, 72–8.
Phillips, Betty (1983), ‘Constraints on syllables and quantitative changes in early
English’, Linguistics 21, 879–95.
Phillips, Betty (2002), ‘Lexical diffusion and competing analyses of sound
change’, in Minkova and Stockwell (eds) (2002), 231–43.
Phillips, Betty (2006), Word Frequency and Lexical Diffusion, Basingstoke: Palgrave
Macmillan.
Polka, Linda and Megha Sundara (2012), ‘Word segmentation in monolin-
gual infants acquiring Canadian English and Canadian French: native
language, cross-dialect, and cross-language comparisons’, Infancy 17:2,
198–232.
Pope, Mildred [1934] (1961), From Latin to Modern French with Especial Consideration
of Anglo-Norman, Manchester: Manchester University Press.
Proctor, Michael (2010), Gestural Characterization of a Phonological Class: The
Liquids, PhD thesis, Yale University.
Pullum, Geoffrey and William Ladusaw (1986), Phonetic Symbol Guide, Chicago:
University of Chicago Press.
Purnell, Thomas C. and Malcah Yaeger-Dror (2010), Accommodation to the Locally
Dominant Norm, special issue of American Speech 85:2.
Putter, Ad (2005), ‘Weak E and the metre of Richard Spalding’s Alliterative
Katherine Hymn’, Notes and Queries 52, 288–92.
Putter, Ad, Judith Jefferson and Donka Minkova (2014), ‘Dialect, rhyme, and
emendation in Sir Tristrem’, Journal of English and Germanic Philology 113:1.
Pyles, Thomas (1943), ‘The pronunciation of Latin learned loan words and
foreign words in Old English’, PMLA 58:4, 891–910.
Quirk, Randolph and C. L. Wrenn (1957), An Old English Grammar, 2nd edn, New
York: Holt, Rinehart and Winston.
Restle, David and Dietmar Zaefferer (eds) (2002), Sounds and Systems: Studies in
BIBLIOGR APHY 389
Structure and Change: A Festschrift for Theo Vennemann, Trends in Linguistics.

Studies and Monographs 141, Berlin: Mouton de Gruyter.
Reszkiewicz, Alfred (1973), A Diachronic Grammar of Old English, Warsaw:
Państwowe Wydawnictwo Naukowe.
Riera, María, Joaquín Romero and Ben Parrell (2009), ‘Schwa in American
English V+/r/ sequences’, in Vigário et al. (eds) (2009), 15–35.
Ringe, Don (2006), A Linguistic History of English. Volume I: From Proto-Indo-
European to Proto-Germanic, Oxford: Oxford University Press.
Ritt, Nikolaus (1994), Quantity Adjustment: Vowel Lengthening and Shortening in
Early Middle English, Cambridge: Cambridge University Press.
Ritt, Nikolaus (1997), Early Modern English vowel shortenings in
monosyllables before dentals: a morphologically conditioned sound
change?’, in Raymond Hickey and Stanislav Puppel (eds), Language
History and Linguistic Modelling, Berlin and New York: Mouton de Gruyter,
65–72.
Ritt, Nikolaus (2007), ‘It seemed like a good idea at the time: why there may not
have been any early Modern English shortening before dentals after all’, in
Ute Smit, Stefan Dollinger, Julia Huettner, Gunther Kaltenboeck and Ursula
Lutzky (eds), Tracing English through Time: Explorations in Language Variation: In
Honour of Herbert Schendl on the Occasion of his 65th Birthday, Vienna: Braumüller,
315–31.
Roberts, Jane (2006), A Guide to Scripts Used in English Writings up to 1500, London
and Boston Spa: British Library Publishing.
Robinson, Orrin W. (1992), Old English and its Closest Relatives: A Survey of the
Earliest Germanic Languages, Stanford: Stanford University Press.
Roper, Jonathan (ed.) (2011), Alliteration in Culture, Houndmills: Palgrave
Macmillan.
Ross, John Robert (1972), ‘A reanalysis of English word stress’, in M. Brame
(ed.), Contributions to Generative Phonology, Austin: University of Texas Press,
229–323.
Rothwell, W. (2001), ‘English and French in England after 1362’, English Studies
82:6, 539–59.
Rusch, Willard (1994), ‘When rhymes go bad: recontextualizing Chaucer’s
rhymes with the mid front long vowels’, American Journal of Germanic
Linguistics and Literatures 6:1, 1–50.
Russom, Geoffrey (1987), Old English Meter and Linguistic Theory, Cambridge:
Russom, Geoffrey (1998), Beowulf and Old Germanic Metre, Cambridge Studies
in Anglo-Saxon England 23, Cambridge: Cambridge University Press.
Russom, Geoffrey (2004), ‘The evolution of Middle English alliterative meter’,
in Curzan and Emmons (eds) (2004), 279–304.
Ryan, Kevin (2011a), Gradient Weight in Phonology, PhD thesis, University of
California, Los Angeles.
Ryan, Kevin M. (2011b), ‘Gradient syllable weight and weight universals in
quantitative metrics’, Phonology 28:3, 413–54.
Saintsbury, George [1906] (1923), A History of English Prosody from the Twelfth
Century to the Present Day, vol. I, 2nd edn, New York: Macmillan.
Salmons, Joseph (1992), Accentual Change and Language Contact: A Comparative
Survey and a Case Study of Northern Europe, Oxford and New York: Routledge.
Samuels, Michael (1963), ‘Some applications of Middle English dialectology’,
English Studies 44, 81–94.
Sandved, Arthur O. (1985), Introduction to Chaucerian English, Cambridge:
D. S. Brewer.
Schaefer, Ursula (ed.) (2006), The Beginnings of Standardization: Language and
Culture in Fourteenth-century England, Bern: Peter Lang.
Schendl, Herbert (2001), ‘Code-switching in medieval English poetry’, in
Dieter Kastovsky and Arthur Mettinger (eds), Language Contact in the History
of English, Frankfurt am Main: Peter Lang, 305–37.
Schendl, Herbert (2002), ‘Mixed language texts as data and evidence in English
historical linguistics’, in Minkova and Stockwell (eds) (2002), 51–79.
Schlüter, Julia (2005), Rhythmic Grammar: The Influence of Rhythm on Grammatical
Variation and Change in English, Topics in English Linguistics 46, Berlin and
New York: Mouton de Gruyter.
Schlüter, Julia (2009), ‘Consonant or “vowel”? A diachronic study of initial <h>
from early Middle English to nineteenth-century English’, in Minkova (ed.)
(2009), 168–99.
Schreier, Daniel (2005), Consonant Change in English Worldwide: Synchrony Meets
Diachrony, New York: Palgrave Macmillan.
Schrijver, Peter (2009), ‘Celtic influence on Old English: phonological and pho-
netic evidence’, English Language and Linguistics 13:2, 193–211.
Schrijver, Peter (2011), ‘From Proto-Celtic to Old Irish: relative chronology’,
handout for University of California, Los Angeles (UCLA) mini-course,
11–15 April 2011.
Scobbie, James and Alan Wrench (2003), ‘An articulatory investigation of word
final /l/ and /l/-sandhi in three dialects of English’, in Proceedings of the XVth
International Conference of Phonetic Sciences, 1871–4.
Scragg, Donald (1970), ‘Initial H in Old English’, Anglia 88, 165–96.
Scragg, Donald G. (1974), A History of English Spelling, Manchester: Manchester
University Press.
Selkirk, Elisabeth O. (1984), Phonology and Syntax: The Relation between Sound and
Structure, Cambridge, MA: MIT Press.
Sharf, Donald J. (1962), ‘Duration of post-stress intervocalic stops and preceding
vowels’, Language and Speech 5:1, 26–30.
Short, Ian (2007), Manual of Anglo-Norman, Occasional Publications Series no. 7,
London: Anglo-Norman Text Society.
Sievers, Eduard (1893), Altgermanische Metrik, Halle: Niemeyer.
Simon, Ellen (2008), ‘Phonological transfer of voicing and devoicing rules:
evidence from L1 Dutch and L2 English conversational speech’, Language
Sciences 32:2,010, 63–86.
Smith, Jeremy (1994), ‘The Great Vowel Shift in the North of England, and
BIBLIOGR APHY 391
some forms in Chaucer’s Reeve’s Tale’, Neuphilologische Mitteilungen: Bulletin of

the Modern Language Society, 95:433–7.
Smith, Jeremy (1996), An Historical Study of English: Function, Form and Change,
London: Routledge.
Smith, Jeremy J. (2007), Sound Change and the History of English, Oxford: Oxford
University Press.
Smith, Jeremy J. (2009), Old English: A Linguistic Introduction, Cambridge:
Sóskuthy, Márton (2010), Analogy in the emergence of intrusive-r in English.
Richard M. Hogg Prize winning essay, <http://www.isle-linguistics.org/
prize/> (last accessed 3 June 2013).
Stanley, Eric (ed.) (1972), The Owl and the Nightingale, Manchester: Manchester
University Press.
Stanley, Eric G. (1988), ‘Rhymes in English Medieval verse: from Old English
to Middle English’, in Medieval English Studies Presented to George Kane, ed.
Edward Donald Kennedy, Ronald Waldron and Joseph S. Wittig, Cambridge:
Brewer, 115–38.
Stenbrenden, Gjertrud (2003), ‘On the interpretation of early evidence for ME
vowel-change’, in Barry Blake, Kate Burridge and Jo Taylor (eds), Historical
Linguistics 2001: Selected Papers from the 15th ICHL, Melbourne, 13–16 August 2001,
Amsterdam and Philadelphia: Benjamins, 403–16.
Stenbrenden, G. F. (2010), The Chronology and Regional Spread of Long-Vowel
Changes in English, c. 1150–1500, PhD dissertation, University of Oslo.
Stenroos, Merja (2005), ‘Spelling conventions and rounded front vowels in the
poems of William Herebert’, in Nikolaus Ritt and Herbert Schendl (eds),
Rethinking Middle English: Linguistic and Literary Approaches, Frankfurt am Main:
Steriade, Donca (1991), ‘Moras and other slots’, Proceedings of the Formal
Linguistics Society of the Midamerica 1, 254–80.
Steriade, Donca (2001), ‘Directional asymmetries in place assimilation: a per-
ceptual account’, in E. Hume and K. Johnson (eds), The Role of Speech Perception
in Phonology, New York: Academic Press, 219–50.
Stockwell, Robert (1961), ‘The Middle English “long close” and “long open”
mid vowels’, Texas Studies in Literature and Language 2, 259–68. [Reprinted in
Lass (ed.) (1969), 154–63].
Stockwell, Robert (1985), ‘Assessment of alternative explanations of the Middle
English phenomenon of high vowel lowering when lengthened in the open
syllable’, in Roger Eaton, Olga Fischer, Willem Koopman and Frederike van
der Leek (eds), Papers from the 4th International Conference on English Historical
Linguistics, Amsterdam: John Benjamins, 303–18.
Stockwell, Robert (2002a), ‘Retraction and rounding in Old English Breaking’,
in Restle and Zaefferer (eds) (2002), 121–37.
Stockwell, Robert (2002b), ‘How much shifting actually occurred in the histori-
cal English vowel shift?’, in Minkova and Stockwell (eds) (2002), 267–81.
Stockwell, Robert (2006), ‘The status of Late Middle English <ei> spellings
as early evidence of the English Vowel Shift’, in Schaefer (ed.) (2006),

175–80.
Stockwell, Robert P. and C. Westbrook Barritt (1955), ‘The Old English short
di-graphs: some considerations’, Language 31:3, 372–89.
Stockwell, Robert P. and C. Westbrook Barritt (1961), ‘Scribal practice:
some assumptions’, Language 37:1, 75–82. [Reprinted in Lass (ed.) (1969),
133–42].
Stockwell, Robert and Donka Minkova (1988), ‘The English Vowel Shift: prob-
lems of coherence and explanation’, in Dieter Kastovsky and Gero Bauer
(eds), Luick Revisited, Tübingen: Günter Narr, 355–94.
Stockwell, Robert and Donka Minkova (1997a), ‘On drifts and shifts’, Studia
Anglica Posnaniensia 31, 283–305.
Stockwell, Robert and Donka Minkova (1997b) ‘Prosody’, in Robert Bjork and
John Niles (eds), A Beowulf Handbook, Lincoln, NE: University of Nebraska
Press, 55–85.
Stockwell, Robert and Donka Minkova (1997c), ‘Old English Metrics and
the phonology of resolution’, Germanic Studies in Honor of Anatoly Liberman,
NOWELE: North-Western European Language Evolution 31/32, 389–406.
Stockwell, Robert and Donka Minkova (1999), ‘Explanations of sound change:
contradictions between dialect data and theories of chain shifting’, Leeds
Studies in English 30, 83–103.
Stockwell, Robert and Donka Minkova (2001), ‘The partial-contact origins of
English pentameter verse: the Anglicization of an Italian model’, in Dieter
Kastovsky and Arthur Mettinger (eds), Language Contact in the History of
English, Frankfurt am Main: Peter Lang, 337–62.
Stockwell, Robert and Donka Minkova (2002), ‘Interpreting the Old and Middle
English close vowels’, Language Sciences 24:3–4, 447–58.
Strandberg, Otto (1919), The Rime-Vowels of Cursor Mundi: A Phonological and
Etymological Investigation, Uppsala: Almqvist and Wiskells.
Sweet, Henry (1877) Handbook of Phonetics, vol. 2, Oxford: Clarendon Press.
Sweet, Henry (1888), History of English Sounds, Transactions of the Philological
Society for 1873–4, Oxford: Clarendon Press.
Sweet, Henry (1890), A Primer of Spoken English, Oxford: Clarendon Press.
Sweet, Henry (1922), An Anglo-Saxon Reader in Verse and Prose, 9th edn, Oxford:
Clarendon Press.
Szemerényi, Oswald J. L. [1970] (1999), Introduction to Indo-European Linguistics,
Oxford: Oxford University Press.
Taavitsainen, Irma (2000), ‘Scientific language and spelling standardisation
1375–1550’, in Laura Wright (ed.), The Development of Standard English 1300–
1800, Cambridge: Cambridge University Press, 131–54.
Tabain, M. (1998), ‘Non-sibilant fricatives in English: spectral information
above 10 kHz’, Phonetica 55, 107–30.
Tagliamonte, Sali and Rosalind Temple (2005), ‘New perspectives on an ol’
variable: (t, d) in British English’, Language Variation and Change 17:3, 281–302.
Tamson, George J. (1898), Word-Stress in English. A Short Treatise on the Accentuation
BIBLIOGR APHY 393
of Words in Middle English as Compared with the Stress in Old and Modern English,
Studien zur englischen Philologie III, Halle an der Saale: Dr Martin Sändig.
Tarlinskaja, Marina (1976), English Verse: Theory and History, The Hague:
Mouton.
Tarlinskaja, Marina (1987), Shakespeare’s Verse: Iambic Pentameter and the Poet’s
Idiosyncrasies, New York: Peter Lang.
Tarlinskaja, Marina (forthcoming), Shakespeare among Others: English Drama
1541–1642.
Terajima, Michiko (1985), The Trajectory Constraint and ‘Irregular’ Rhymes in
Middle English, Tokyo: Shinozaki Shorin.
Thomas, Eric (2001), An Acoustic Analysis of Vowel Variation in New World English,
Publication of the American Dialect Society no. 85, Durham, NC: Duke
University Press.
Throop, Priscilla (2008), Ælfric’s Grammar and Glossary: The Complete English
Translation, Medieval MS Box 363: Charlotte, VT.
Tieken-Boon van Ostade, I. (2009), An Introduction to Late Modern English,
Edinburgh: Edinburgh University Press.
Tottie, Gunnel (2002), An Introduction to American English, Oxford: Blackwell.
Tristram, Hildegard (2002), ‘Attrition of inflections in English and Welsh’, in
Markku Filppula, Juhani Klemola and Heli Pitkänen (eds), The Celtic Roots of
English, Joensuu: University of Joensuu, Faculty of Humanities, 111–49.
Trudgill, Peter (1983), On Dialect: Social and Geographical Perspectives, Oxford:
Blackwell.
Trudgill, Peter (1986), Dialects in Contact, Oxford: Blackwell.
Trudgill, Peter (1999a), The Dialects of England, 2nd edn, Oxford and Malden,
MA: Wiley-Blackwell.
Trudgill, Peter (1999b), ‘A window on the past: “colonial lag” and New Zealand
evidence for the phonology of nineteenth-century English’, American Speech
74:3, 227–39.
Trudgill, Peter (ed.) (1984), Language in the British Isles, Cambridge: Cambridge
University Press.
Trudgill, Peter, Elizabeth Gordon and Gillian Lewis (1998), ‘New-dialect
formation and Southern Hemisphere English: the New Zealand short front
vowels’, Journal of Sociolinguistics 2:1, 35–51.
Trudgill, Peter and Elizabeth Gordon (2006), ‘Predicting the past. Dialect
archaeology and Australian English rhoticity’, English World-Wide 27:3,
235–46.
Uffmann, Christian (2007), ‘Intrusive [r] and optimal epenthetic consonants’,
Language Sciences 29:2–3, 334–59.
Vachek, Joseph (1976), Selected Writings in English and General Linguistics, Prague:
Academia and The Hague: Mouton.
Vennemann, Theo (1988), Preference Laws for Syllable Structure and the Explanation
of Sound Change, Berlin: Mouton de Gruyter.
Vennemann, Theo (2010), ‘The source of the Ing rune and of the futhark’,
Sprachwissenschaft 35, 1–14.
Vigário, Marina Cláudia, Sónia Frota and Maria João Freitas (eds) (2009),
Phonetics and Phonology: Interactions and Interrelations, Current Issues in
Linguistic Theory 306, Amsterdam: John Benjamins.
Wakelin, Martyn (1972), English Dialects: An Introduction, London: Athlone Press.
Wakelin, Martyn (1988), The Archaeology of English, London: B. T. Batsford.
Wales, Katie (2006), Northern English: A Social and Cultural History, Cambridge:
Walker, John (1791–1826), Critical Pronouncing Dictionary. A critical pronounc-
ing dictionary and expositor of the English language 艐: to which are prefixed,
principles of English pronunciation艐with observations etymological, critical, and
grammatical艐with directions to foreigners, for acquiring a knowledge of the use of this
dictionary, London and Leipzig: Ernest Fleischer.
Watkins, Calvert (1985), The American Heritage Dictionary of Indo-European Roots,
Boston: Houghton Mifflin.
Watkins, Calvert (1992), ‘Indo-European and the Indo-Europeans’, in The
American Heritage Dictionary of the English Language, 3rd edn, Boston: Houghton
Mifflin, 2,081–9.
Watts, Richard J. (2003), ‘Was the Great Vowel Shift really “great”? A reap-
praisal of research work on an elusive linguistic phenomenon’, in Cornelia
Tschichoid (ed.), English Core Linguistics, Essays in Honour of D. J. Allerton, Bern:
Watts, Richard J. (2011), Language Myths and the History of English, Oxford: Oxford
University Press.
Wells, John (1982), Accents of English, vols 1–3, Cambridge: Cambridge University
Press.
Wells, John (1984), ‘English accents in England’, in Trudgill (ed.) (1984), 55–69.
Wełna, Jerzy (1978), A Diachronic Grammar of English. Part One: Phonology, Warsaw:
Pań stwowe Wydawnictwo Naukowe.
Wełna, Jerzy (1987), A Critical Survey of a Historical Phonology of English Vowels
(with special reference to Karl Luick’s ‘Historische Grammatik der englis-
chen Sprache’), Warsaw: Wydawnictwa Uniwersytetu Warszawskiego.
Wełna, Jerzy (2004), ‘Middle English ē-raising: a prelude to the Great Vowel
Shift’, Studia Anglica Posnaniensia 40, 75–83.
Wełna, Jerzy (2007), ‘The loss of [ei]: [ai] opposition in Middle English’, Studia
Anglica Posnaniensia 43, 3–16.
Wetzel, Claus-Dieter (1981), Die Worttrennung am Zeilenende in altenglischen
Handschriften, Frankfurt am Main: Peter Lang.
Wheatley, Henry B. (ed.) (1867), Manipulus Vocabulorum: A Rhyming Dictionary of
the English Language by Peter Levins (1570). Edited, with an alphabetical index
by Henry B. Wheatley, London: Published for the Early English Text Society
by Trübner and Co., 60, Paternoster Row.
White, David (2004), ‘Why we should not believe in short diphthongs’, in
Curzan and Emmons (eds) (2004), 57–84.
Wiggins, Alison (2003), ‘Guy of Warwick in Warwick?: Reconsidering the dialect
evidence’, English Studies 84:3, 219–30.
BIBLIOGR APHY 395
Williams, Fionnuala Carson (2011), ‘Alliteration in English-Language versions

of current widespread European idioms and proverbs’, in Roper (ed.) (2011),
24–44.
Windross, Michael (1994), ‘Loss of post-vocalic r: were the orthoepists really
tone-deaf?’, in Kastovsky (ed.) (1994), 429–48.
Wogan-Browne, Jocelyn (ed.), with Carolyn Collette, Maryanne Kowaleski,
Linne Mooney, Ad Putter and David Trotter (2009), Language and Culture
in Medieval Britain: The French of England, c. 1100–c. 1500, Woodbridge and
Rochester, NY: York Medieval Press.
Wolfram, Walt (1976), ‘Toward a description of a-prefixing in Appalachian
English’, American Speech 51:1/2, 45–56.
Wolfram, Walt and Natalie Schilling-Estes [1998] (2006), American English:
Dialects and Variation, 2nd edn, Cambridge and Oxford: Basil Blackwell.
Wright, Joseph and Elizabeth M. Wright (1928), An Elementary Middle English
Grammar, 2nd edn, London: Oxford University Press.
Wyld, Henry Cecil [1927] (1949), A Short History of English, London: John
Murray.
Wyld, Henry Cecil (1953), A History of Modern Colloquial English, Oxford: Basil
Blackwell.
Youmans, Gilbert (1989), ‘Milton’s meter’, in Paul Kiparsky and Gilbert
Youmans (eds), Phonetics and Phonology, Vol. 1: Rhythm and Meter, San Diego:
Academic Press, 341–79.
Youmans, Gilbert (1996), ‘Reconsidering Chaucer’s prosody’, in C. B. McCully
and J. J. Anderson (eds), English Historical Metrics, Cambridge: Cambridge
Youmans, Gilbert and Xingzhong Li (2002), ‘Chaucer: folk poet or littérateur?’,
in Minkova and Stockwell (eds) (2002), Berlin: Mouton de Gruyter, 153–75.
Zec, Draga (1995), ‘Sonority constraints on syllable structure’, Phonology 12:1,
85–129.
Zuraw, Kie (2009), ‘Treatments of weakness in phonological theory’, in
Minkova (ed.) (2009), 9–29.
Word Index
The word index includes items whose pronunciation is discussed in the hard copy portion
of the book. It does not include words discussed in the Online Companion. It excludes high
frequency words which occur in the main text: function words, pronouns, core vocabulary
items. Colons in the entries indicate a rhyming pair, other pairs are linked with – or ~.
Accents are preserved only when relevant to the context.
a-, 182 Aike-Oake 200, 203, 240 aplomb, 135

ábacus, 308 -air, 319–20 apple, 72–3, 78, 104, 120,
abbot, 78, 289 -al, 47 179, 199–200, 298
abide, 203 alderman, 125 appreciate, 144
ability-habilitation, 107 aleph, 256 April, 160
-able, 310, 320 Alexander, 66 apse, 134
able-ability, 255 Alexander-Saunders, 241 -ar, 47
able-habilitate, 105 all, 129 Arab, 108, 160, 285
abnormality, 147 állergy-allérgic, 13 arbour, 200
abound-abundance, 218 allow, 32 arbour-herbage, 107
about, 191 ally, 311 -ard, 310
above, 219 almesse, 140 are : pair, 224
abridge, 141 alóha, 101, 115 area-aerial, 129
abróad, 288 alone, 32 ark, 183
absolute, 268 altar, 289 arm, 117, 124, 178
ábstract-abstráct, 311 alum, 136 aroma-aromatic, 255
abysm, 58 alumnus, 15, 136 arrant, 277–8
account-accountancy, 216 always-causeways, 284, 303 arrant-errant, 312
acorn, 223 ambi-, 187 arrow, 126
acre, 217, 223 amidst, 49, 150 arse, 122
-acy, 310 amnesia, 142 art, 169
adder, 79, 125, 212 among(st), 49, 150, 192 ash, 44, 154
áddict-addíct, 288 ámorous, 315–16 ask~ax, 48
-ade, 319–20 -an, 49 assault, 208
adze, 135 ancient, 140, 150, 255 assembly, 12–13
afford, 125 ancient-pansy, 255 assume, 268
again, 8 -ancy, 15 -ast, 319
against, 49, 150 and- (OE), 233 asterisk, 354
-age, 310, 320 -ane, 319 asthma, 57, 231
age, 263 angel, 182, 255, 269 at ease, 25, 147, 341
aged, 232 animal, 12–13 ataxy, 187
aggravate, 32 ante-, 15 -ate, 310, 316
agnostic, 67 antenna, 15 attain, 27
ahoy, 269 antique, 256 attribute, 32
aik/oak, 201, 250 ape, 73, 78 augment, 311
396
WORD INDE X 397
Áugust-augúst, 312 beetle, 195 blood, 160, 190, 246, 272

aunt, 169, 241, 268–9 beetle-bitter, 78 blood-bleed, 162
aunt-haunt, 267 behave, 29, 102 blossom, 49, 288
Aussie, 226 behest, 150 blow, 175, 266
autogony, 290 behind, 166, 168 blowth, 215
autumn(al), 136, 140 beige, 50, 142 blue, 191
await : white, 110 being, 191 blush, 194, 197
awe, 83 belief-beliefs-believe, 74, 89, board, 124, 165, 278
axe, 134 96, 156, 159 boat, 155, 188, 190, 202,
beloved, 232 250, 271
baccara(t), 290 bench, 85 boil, 191, 268, 270
Bach, 100 bench-bank, 86 boil : bile, 270
back, 198–200 bend, 169 boing, 270
bad, 180 bereft~bereaved, 213 boink, 269–70
bairn, 68, 117 Berkeley-Berkshire-Berkley, bold, 167
bake, 250, 266 276 bolt, 169, 244
balanced, 34 berry, 179 bomb, 135
bald, 268–9 besiege, 255 bomb-bombable, 135
bale-baleful, 294 besom, 223 bone, 26, 45, 244, 331
ban, 155 best, 149 bone-bonfire, 215, 255
ban-bang, 138 bestial, 144 book, 81, 190, 225, 271, 272,
band, 169 Beth, 240 273
Bangor, 138 bethink, 91 book-buck, 246
bank, 7 betide, 203 booth-book, 271
banner, 32 betoken-behave, 27 born, 155
baptist, 217 betwixt, 49 borne, 244
bar, 240 beyond, 28 borough, 48
barber, 200 bid, 72–3 borrow, 12, 67, 87, 126, 160
bark, 200 bidder, 147 boss, 243
barn, 121 bidding, 293 botany, 337
barrier-reef, 32 bin, 155 both, 190
barrow, 84 bin-bean, 225 both~baith, 203
base, 191 bind, 166, 250, 253, 258, 265 bóther-bóthersome, 287
basilisk, 217 bingo, 138 bottom, 31, 34, 40, 69, 192
bass-bass, 20 birch, 120 bough, 48, 113, 114, 140
bath, 28, 81, 179, 188, 198–9, birch-birk, 86 bought, 206, 207, 268, 269
238, 242, 255 bird, 35 boulder, 169
bath-bathe-baths, 74, 89, 90 bishop, 182 boulevard, 290
bathe, 28, 81, 93, 97 bit, 78 bound for, 168
battle, 309 bite, 45, 78 bow, 205, 207
be, 191 bite-bit, 214, 225 bow down, 205, 258
be-, 182, 219, 296, 336 bitter, 24, 72 bower, 279
beam me, 32 black, 172 bowl, 268
bear, 63, 68 black-Blake, 224, 226 boy, 98, 190, 191, 209, 268,
bear-fear, 271 bladder, 125 270
beast, 213 blame, 263 brambles, 48, 120, 129, 212
beat-beet, 190 bland, 241 brass-brazen, 97
beauty, 38, 208, 306 blank, 241 brave-braved, 226
beaver, 120, 217 blast, 213 bread, 41, 45, 271
bed, 80, 237, 238 blend, 169 break, 250, 262, 264, 266
bedchamber, 13 bless, 93 breakfast, 304
beef, 97 blessed, 232 breast, 213
beer, 206 blind, 165, 168 breastplate, 13
beership, 182 bliss, 226 breath-breadth, 227
beet, 190 blitz, 134 breathe, 93, 97
breeches~burches, 119 call, 7, 68, 269 chew, 197, 268

brethren, 161 calm, 241, 269 chief, 92
brew, 205, 207, 268 cambric, 167 child, 41, 81, 85, 166–7, 172,
brew-new, 267 Cambridge, 167 179, 220, 254, 333
bridge, 77, 81, 86, 129, 141, camouflage, 142 child-children, 151, 165
194, 197 camp, 198 childhood, 296
Bridgestowe-Bristol, 128 canál-chánnel, 310 children, 125, 166
bridle, 205–6, 209, 257–8 candēla, 5 chimney, 30
brief, 92, 97, 202 candle, 58 chin, 85
brief-brevity, 218 cánker, 289 choice, 38, 50, 58, 151, 208,
bright, 113, 206 capitalise, 25, 147 268, 269, 270
brimstone, 304 caprice, 253–4, 256 choose, 82, 201, 204
bring-brought, 163 car, 240, 242 chorus, 57, 63
brisk-brusque, 194 carafe, 97 Christ, 213
broad-braid, 203 car-core, 241 Christmas, 304
broom, 273 cardiac, 63 chthonic, 57
broth, 244 care, 82, 85, 88 church, 8, 85, 194, 196
brother, 66, 69, 91, 120, 161 cast, 7 church-kirk, 88
brought, 207 -caster vs -chester, 86 churn, 194
bud-bird, 225 castle, 49 chutzpah, 100
buf(f)et, 194 cat, 24, 172 cigar, 30
bugle, 193 catch-chase, 86 Cindi~Cindy, 187
build, 167 cattle, 12 circle, 277
bulb, 246 cattle-chattel, 86 círcle, 289
bulge, 246 caught, 33, 50 Citi, 187
bulk, 246 cause, 208 city, 82
bull, 7 cave-cavity, 218 clatter, 68
bullet, 245 ceiling, 34 claw, 191
bulletin, 245 celestial, 144 clean, 202
bullion, 245 cell, 82 clean-cleanse, 271
bungalow, 245 cerebellum, 15 cleave-cleft, 96
buoy, 269 certain, 277 cleft~cleaved, 213
burden, 125, 137, 145 chair, 28 clergy, 124
-burgh, 84, 103, 113 chalice, 159 clerk, 82, 277
burn, 119 chalk, 268, 269 clerk-clerk, 276
-bury, 196 chamber-lamp, 255 climb, 43, 135, 166–7
bury, 195–6, 235, 237–8 champ, 241 climber, 135
bush, 50, 194, 237, 245 chance, 169 climbing, 135
bushel, 245 change, 241 cloak, 263
busy, 196, 236 channel canál, 312 close-close, 88
but, 219, 246 chaplain, 309 cloth, 35, 237, 240, 242
but-about, 219 charisma-charismatic, 285 cloth-clothe, 89
butcher, 245 chárity, 315 cloth-clothes-cloths, 96
butter, 246 Charles, 12 clothe, 94
buttock, 246 chary, 85, 86, 88 clough, 207
buy, 72, 220 chaste-chastity, 218 clue, 268
buy-bought, 163 chattel, 12 clutch, 194, 245
by, prep, 205 chav, 227, 240 coal, 202
cheap-Chapman, 215 coast, 213
cabbage, 290 check-in-chicken, 238 coat-coating, 226
cage, 141 cheese, 81 coffin, 92
Cajun, 30 chef, 98 coffin-coven, 50
Calder-called her, 104 cherub, 34 cognition, 63
cálendar, 315 chess, 29, 49 coif, 97
calf, 241, 269 chest-kist, 86 coin, 270
WORD INDE X 399
cold, 167, 264 cress, 119 steel, 261

cold-chill, 86 Cristal, 187 dealer, 155
colour, 245 critic-critíque, 312 dear, 206
colourless, 13 critique, 253–4, 256 dear-dearth, 214
colours, 124 crook, 272 dear-darling, 276, 278
column, 136, 308 crooked, 232 death, 271
comb, 135, 167 cross, 243 debt, 238
combi, 187 cross~corse, 119-20 December, 167, 217
come, 69, 82, 189, 221 crown, 49, 258, 265, 289 decision, 82
comet, 5, 58, 217, 308 crutch, 194 deduce-deductive, 218
comma, 34, 151 cucumber, 217, 308 deed, 174
commute, 32 cuddle, 27 deem, 12, 201, 213
compare-comparison, 218 cudgel, 194, 197 deep, 214
complain, 190 cul(minate)-hill, 62 deep-Dep(t)ford, 215
compound, 140 culver, 92 deer, 12, 275, 279
con, 67 cumin, 137 degree, 58, 218
condemn, 136 cunning, 67 delay, 263
cone-conical, 216, 218, 227 cupboard, 303 delight-delicious, 218
connateness, 32 curd, 35 dell-dale, 224
cook, 272 cure, 37, 206 deprive-deprivation, 255
cool, 129 curriculum, 15 depth, 214
coolth, 215 curry, 245 derision, 143
coomb, 167 curse, 124, 169 derv, 227
cord, 169 cursed, 232 desert : part, 276
cordial, 63 cursor, 277 design, 265
cordial-hearty, 74 cushion, 246 despair, 65
corps, 244 cut, 65, 189 dessert, 66, 93, 226
cortège, 142 cut-cute, 190 destroy, 208
cot, 33, 50 cut-cutting, 226 devil-devilish, 295
cot-caught, 240, 243 cute, 38 dew, 191, 205, 209, 268
cot-coat, 225 diagnosis, 67
cough, 105, 112, 114, 207, dais, 263 dice, 253
242 daisy, 220, 303 digestif, 97, 98
could-cud, 246 dale, 224 digestive, 97
council, 306 Dalwood, 224 dill, 221
country, 307 Dalziel, 83 din, 173
couple, 245 damn, 136 dinghy, 139
courage, 141 damn-damnation, 134, 136 dinner-time, 13
court-courtesy, 218 dance, 22, 237–8, 241–2, diphthongal, 138
cousin, 45, 245 255, 268–9 disciple, 87
cove, 92 danger, 241, 255, 269 discipline, 87
cover, 189 darling, 276, 278 discus, 88
covet, 245 Darlington, 122 dish-disk-discus, 74, 88
cow, 250, 258, 265 daughter, 57, 113–14, 161, dissatisfied, 32, 77
cowl, 289 207 disservice, 32
coy, 270 dawn, 84 dissolve, 66, 93
crabbed, 232 -day, 304 ditch, 82, 85
(cra)cy, 319 day, 177–8, 81, 190–1, dive, 194
cradle, 217 205–6, 263, 266, 268 divine-divinity, 216, 218,
craft, 81 day-dawn, 74 227
cramp, 102 dead, 238, 240, 271 division, 141
creature, 144 dead-bead, 234, 271 dizzy, 173, 182
creature~critter, 144 deaf, 271, 273 dizzy-busy, 184, 192, 195
creep(s)-crept, 39–40, 43, dead-leaf, 271–3 -dj, 144
212–13 deal, 213 do, 155
do~dae, 203 each, 86, 130 -est, 49, 138, 163

do-due, 267 ear, 183, 279 estate~state, 48
dog, 243 ear (of corn), 146 -ette, 319
doldrum(s), 167 earl, 124, 165, 278 euphoria, 63, 68
dollar, 237, 242–3 earth, 165, 278 even(ing)-eve, 49
-dom, 138, 215, 295–6, 318, east, 170, 213 ever, 120, 219
351 eat, 191, 262 every, 49
Don-Dawn, 240 -eau, 319 evil, 157, 221
doohickey, 101, 115 edge, 86, 101, 141 example, 66
doom, 157, 254 -ee, 310, 319–20 exclusive vs excluzive, 94
doom-deem, 162 -een, 319 excusable, 12
door, 63, 172, 181, 227 -eenth, 215 excuse, 12
double, 245 -eer, -ier, 319 exemplar, 12
doug, 207 effect-feckless, 149 exemplary, 12
dough, 81, 103, 114, 263, Egypt, 149 expound, 140
266, 268 eight, 206–7, 263 eye, 220
dough-duff, 114 elbow, 49
dove, 189, 245 elder, 125, 163, 167 fa, 127
down, 30 eldest, 163 fa[r], 127
down-Dunton, 215 electrizity for electricity, 94 façade, 256
drag, 84 Elene, 105 face, 34, 38, 94, 151, 248, 250
drágon-dragóon, 312 eleven, 49, 129 fáham, 101, 115
drat, 243 elf-elfs-elven, 74, 96 fair, 279
drau, 191 -elle, 319 fajita, 59, 115
draught, 113 ellipse, 134 fall, 95, 131
draw, 83 Elysian, 142 false, 169, 269
draw-drag, 88 ember, 167 family, 49
dray, 84 Emma, 128, 132 fan, 30, 92, 156
dream, 202 -en, 49, 318 fang-fan, 75
dreich, 100 -ence, -ency, 15 fangless, 138
dress, 34, 151 end, 169 fan-van, 24
drink, 183 ends, 227 far, 118
drink-drench, 86 endure, 63, 195 fared, 189
drive, 92 eng, agma, angma, 136 farm, 140, 278
drive-drift, 214 -e(ng), 192, 236 farthing, 276
droid, 270 English~Inglis(s), 192 farthingale, 137
drop, 243 enough, 48, 113, 114, 207, -fast, 295
drought, 47, 206, 207, 258 233 fast, 198
drown, 208 enthuse, 268 fast, 91
drunk, 155 entire, 58 fasten, 49, 91
dry, 144, 258 -er, 138, 163, 318 fast-vast, 91
dryad, 63 -er/-re, 18 fat, 225
dryness, 205, 257 ere, 226 father, 66, 69, 124, 145, 198
due, 38, 143, 268 -erel, 310 father-fatherly, 295, 318
duke, 144, 193, 197, 208, 268 -erie, 310 faucet, 129
dull-bull, 234 errant, 277 fault, 131, 208
dumb, 135, 140, 167 error, 276 favour-favourable, 216
dune, 143 -es, 49 fear, 271
dung, 194 escalade~scale, 48 fear-bear, 234
dust, 213 -esce, 319 feast, 213
duty, 38 -ese, 319 feather, 7
Dvořák, 66 especial~special, 48 feature, 144
dwarf, 96, 114–15 espionage, 142 feck, 149
dwarfs/dwarves, 96 espouse~spouse, 48 feed-fed, 212, 216
dwell, 163 -esque, 319 feel, 129
WORD INDE X 401
feet, 191, 260 fool, 250 gae-go, 200

fell, 166, 168 foot, 34, 45, 54, 63, 68, 151, gaffe, 98
fellow, 126, 293 173, 197–8, 216, 237, gala, 256
ferment, 311 245, 248, 272 game-gammon, 184, 220,
ferry-very, 91 foot-feet, 151, 157, 161 224, 226
ferther, 140 foot-podium, 54, 67 gammon-game, 49
fertile, 68, 224 foot-strut, 246 gannet, 223
fetter, 147 for-, 219, 296 gape, 7
fiddle, 92, 125 force, 244 garage, 141
field, 166–8, 172 foretaste, 13 garlic, 220, 236
fiend, 161, 191, 202, 250 forge, 244 gather, 83, 124
fiend-fiends, 159 formal, 290 gaunt, 241
fierce, 122 forsook, 272 ge- (OE), 182, 219, 232, 289
fifteen, 236 fort, 244 gendarme, 142
fifth, 215 fortune, 144 ġenoh (OE), 232
fight, 41, 49 forum, 63, 68 genre, 142
fill, 155 fought, 207 georgette, 142
filth, 170, 214 foul, 214 get, 8
fine-vine, 89 foul-filth, 163 get úp, 147
fir, 277 foul-Fulford, 215 ghost, 213
fire, 155, 188, 191, 275, 279, foul-defile, 162 ghost, 81
282 foundry, 167 gigolo, 142
first, 120, 169, 193 four, 177 gigue, 142
fish, 86–8, 95, 139, 236–7 fowl, 190, 207–8, 258, 265 gild, 167
fishes, 86 fox, 95, 163 giraffes~gira[v]es, 98
fish-piscatorial, 87 fox-vixen, 163–4 gird, 63
fist, 213 fray, 49 girdle, 119, 120, 277
fit-feet, 225 Frederick, 119 girl, 124
five, 169 freemason, 13 give, 8, 83, 156
five-fifth, 96, 214, 255 frequent, 311 give-gift, 96
flame, 269 frequent-frequency, 216 glad, 180, 199, 237
flank, 241 friar, 253 glasnost, 68
flea, 188 Friedman, 119 glass-glaze, 89, 97
flee, 188 Friedrich, 119 glazing, 91
fleece, 34, 37–8, 151, 248, friend, 22, 161 gloom, 273
250 frog, 79 gloomth, 215
flesh, 75 frontier, 144 gloss, 160
fling, 192 froth, 244 glove, 189, 272, 273
flood, 237, 245–6, 272 fruit, 208 glowing, 28
floor, 279 fudge, 246 glutton, 245
floor-poor, 274, 280 -ful, 91, 215, 318 gnat, 81
flow, 268 Fulham, 245 gnaw, 49, 205
flow, 65 full, 129, 192, 237, 245 God, 22, 154, 183
flung, 45 full-fill, 151, 157, 161 gnawing, 140
fly, 205, 257 full-length, 32 go, 156
fodder, 125 full-to fill, 162 go~gae, 203
fog, 243 fumble, 246 goat, 34, 37–8, 83, 102, 151,
foison, 209 fume, 45, 195 162, 248, 250
foist, 269 fungation, 138 goat-Gatwick, 215
-fold, 91 fur, 151 God, 22, 44, 65, 155, 172,
fold, 167 furry, 32 242–3, 295
folk, 22, 131, 189, 244 fury, 32, 38 God-egad, 236
folly, 309 fury-bury, 184, 192, 195 goddess, 164
food, 246, 273 furze, 194 golden, 80
food-feed, 162 fuss-fuse, 225 gold-gild, 162
golf, 97 gun, 246 heavens, 181

golfe, 244 gunwale, 304 hecatomb, 135
gonna-going to, 299 gymnasium, 142 heed, 266
good, 83, 154, 190–1, 215 gypsy, 49 heel, 202
goose, 34, 37–8, 97, 151, height, 47
161, 169, 173, 198, habítual, 102 heir, 99, 108, 208
250 hallúcinate, 102 helix, 107
goose-gosling, 97 Havána, 102 help, 140
gooseberry, 246 haddock, 223 hemato-, 107
gorse, 119, 120, 121, 124 hag(g), 84 hemi-, 107
gosling, 94 hagio-, 107 hence, 95
gospel, 154, 215, 242 hair, 153 hept(a)-, 107
gotcha, 143 (h)aitches, 99 herald, 106
Goth, 83 half, 129, 140, 179, 238 herb, 99, 108
grace, 263 halfpenny, 131, 241, 255 Hercules, 105
grade-gradual, 218 hall, 156, 198, 269 here-there, 234
gradual, 218 hállow, 288 heritage, 99, 108
Graffham, 220 halo, 107 hermit, 106
graft, 93, 150 hand, 169, 198 heron, 106
grand-grant, 241 handiwork, 104 hetero-, 107
grass, 226 hangar, 139 hew, 207
grass~garston, 119 hap, 7 hex(a)-, 107
grass-graze, 89, 90, 96–7, happen, 237, 238–9 hiccup, 226
184, 220, 224 happy, 91, 102, 237 hide, 193–4, 213, 266
grasshopper, 79 hárass-haráss, 284, 337 high, 104, 106, 113, 163
grave, 93 harbinger, 137 hill, 29, 54, 68, 71, 73,
gravel-stone, 13 hard, 103 84, 103, 113–14,
grease, n. (c. 1340 < OFr hardy, 106 196
graisse)-(variable) hare, 279 hill-culminate, 54, 67
grease, 97 harmony, 107 hip, 221
greasy, 94 harrow, 84 historic, 99
great, 264 harry, 84, 200, 278 hither, 125
green, 41 haste, 106, 264 hoard, 124, 165, 278
greenth, 215 hat, 30, 199 hog, 79
greet-greeted, 216 hate, 199 hoi polloi, 269
grey, 177–8, 205–6, 209, haunt, 241, 268, 269 hoist, 150
263, 268 have-behave, 224 hold, 106, 167
grief, 36 have : grave, 224 hole, 105
grief-grieve, 89, 96 haven, 7 holiday, 242
grieve, 36 haw, 84 holiday, 83
grin~girn, 119 hay, 84, 178 holy-holiday, 216
grind, 168 hazel, 199, 217 home, 175
groom, 273 hellénic, 102 home~hame, 203
gross, 94 helló, 102 honest, 108
ground, 103, 168, 172 head, 45, 190, 217, 271 honey, 99, 189
ground, 81 heal, 213 honorarium, 108
grove, 83 heal-health, 214, 255 honour, 99
grow, 83 heal-heel, 261 flour, 22
growth, 215 hear, 153 hónour-honourée, 13
guarantee, 12 hearken, 124, 278 -hood, 115, 138, 215, 295–6,
guarantee-guardian, 86 heart, 63, 117, 278 318
guardian, 12 heathen, 137 hood, 102
guess, 7 heather, 102 hoof, 273
guest, 65, 68 heave-heft, 96, 214 hoof-hoofs/hooves, 96
guild, 167 heaven, 7, 106, 182, 223 hook, 237, 245
WORD INDE X 403
hoop, 273 iamb, 135 Jew, 197

hop, 237, 242 iamb-iambic, 99, 132, 134, jewel, 197
hope, v. 165 136 jig-juice-jest, 142
hope-hop, 78 -ian, 84 Job, 227
horæ, 108 -ic, 218, 288, 319 join, 268, 270
horary, 108 -ic(al), 310 jojoba, 59, 115
horitculture, 63 ice cream, 149 joke, 29
horizon-horizontal, 255 ice-eyes, 281 jolly, 237, 242
horology, 108 ichthi(o), 57 Joni~Jony, 187
horrible, 106 -id, 319 joy, 268, 270
horse, 48, 244 idea, 58, 128 joyful, 13
horse-hoarse, 244 idea-ideal, 129 judge, 12, 13, 194
hospital, 99 idiom, 58 Judi~Judy, 187
hospitality, 68 idiot~eejit, 144 juice, 141
host, 68, 99, 213 ídiot-idiótic, 294 June, 268
hostler, 99 idle, 217 just, 194
hot, 102, 271 if, 83
hot-heat, 162 -ig, 88, 164 karakul, 290
hound, 48, 166, 168, 254 ignore, 67 Karl, 12
hound-hundred, 151 -il, 160, 164 Karl-Charles, 86
hour, 108, 140 ilk, 129 keep, 190
house, 47, 114, 121, 155, 179, improv, 227 Kelli~Kelly, 187
188, 190–1 -in, 160 ken, 67
householder, 7 -incel, 164 kenning, 67
house-house, 89 inch, 160 kerchiefs~kerchieves, 96
house-husband, 215 (in)dent-tooth, 55 kettle, 34, 40, 69
hous-ing, 91 Indian, 337 kettle-chettel, 86
hover, 189 Indian~Injun, 144 key, 183
how, 104 Infiniti, 187 kibbutz, 134
hue, 268 -ing, 136, 138, 164, 233, 351 kid, 37, 236
huge, 28, 197 inherit, 108 kind, 193
hull, 221 inhibit, 102 kindred, 168
humane-humanity, 218 injun, 143 king, 15–16, 75, 85, 160, 181,
human-humáne, 312 ink, 192 227, 236–7, 330, 348
humble, 99, 106, 141 invalid-inválid, 312 kingdom, 138, 333
humorous, 99 -ion, 319 kingling, 138
humour, 99, 106, 140 ion-cation, 256 kin-king, 138
hundred, 166 -ir, 160, 163 kirk, 8, 85
hung, 30 Irene, 108 kirk-church, 74
Hungárian, 102 -isc, 164 kiss, 78
Hungary, 138 -ise, 318–19, 320 kissed, 193
hunger, 21, 104 -ish, 138, 318 kit, 34, 37, 151, 237, 372
hungry, 138 island, 304 kitchen, 289
hunt, 237, 245 -ist, 319 kitten, 31
hurdle, 194 isthmus, 57 kitty-kiddy, 147
husband(ry), 7, 13, 47, 215, -ity, 218 knave, 133, 199
245 -ity/-ety, 319 knealt, 129
hussar, 226 knee, 82, 204
hussy, 303 Jácob, 289 knee-knelt, 255
hydrangea, 34 jam, 30 kneeled, 132
hydro-, 107 -jan, 72 knew, 133
hymnal, 136 Japan-Japanése, 285 Knievel, 133
hyper-, 107 jargon, 141 knife, 7, 132
hypo-, 107 jazz, 28, 240 knight, 49, 132
hysto-, 107 jealousy, 141 kiss, 22
knobs, 133 -let, 138 machine, 253, 290

knotty-naughty, 240 letter, 24 macro-, 256
know, 63, 108, 175, 207, 209 letters, 124 macron, 256
know-ignorant, 67 Levi, 256 mad, 180
know-no, 20 libel, 290, 306 mæcca, 77
lid, 36 magazine, 253, 256
la, 256 life, 189 Magéllan-Magéllanic, 285
labour, 306 life-live, 91 magnesium, 142
lacuna, 15 -like, 351 maharaja, 99, 101–2
ladder, 25, 41, 125, 147, limb, 135 maiden, 81, 83, 268
202 lineage, 12 maiden-maid, 49
lager, 256 lineation, 12 mail, 264
lair, 279 line-linear, 218 main, 263
lama, 256 -ling, 138 mája, 115
lamb, 135, 156, 167, 198 linger, 192 malt, 243
lamb’s wool, 135 lingering, 354 Málta-Maltése, 294
lambskin, 135 lip, 128 man, 30, 155, 198
lamp, 241, 269 lisp, 133 mane-man, 78
lance, 241 listless, 162 mango-mangle, 129
land, 169, 237, 239 literature, 144 man-men, 151, 157, 161
language, 34 lithe, 94 mannerly, 13
larst for last, 124 live, 79 many, 83
larynx, 135 loaf-loaves, 74, 89 Marchand, 277
late-latter, 224 loan, 7 margin~margent, 48
laugh, 72, 80, 105, 112–14, loath, 190 mark, 179
178 loathe, 94 marrow, 103
laughter, 113, 338 loch, 29, 100 marrow, 81
launch-lance, 241 loin, 269–70 mártyr, 289
laundry, 167, 241, 269 loiter, 209, 269 marv, 227
law, 103, 205, 269 long-length, 163 mash potato, 149
law, 81, 83 look, 165 mason-Masonic, 255
lay, 268 look-luck, 246 mat, 240
lead, v. 36, 78, 201, 213 lord, 243, 244 match, 28
lead(Pb)-lead, v. 271 lord-lard, 244 mater, 256
lead-led, 78, 214 lore, 66 mat-mate, 225
leaf, 273 lost, 240, 242 matter-madder, 147
leaf-leave, 50 lost-post, 234 mature, 143, 144
leafy-Levy, 24 lot, 33–5, 44, 105, 108, 115, maxi, 187
learn, 117 132, 151, 243 may, must, 219
learned, 232 loud, 271 mead, 221–2, 261, 271
lease [-z]-lease [s], 94 louse, 161 meal, 155, 175, 202, 204, 250
least, 213, 262 love, 165, 172, 182, 237, 245 mean, 202
leave-left, 96 loveable, 13 mean-meant, 213
lectern, 124 low, 30, 220 meant, 254
ledge, 30 lubricate, 38 measure, 28, 142, 143
leechdom, 216 Lúcifer, 289 meat, 36, 190, 202
leek-garlic, 212, 219 luck, 189 medal, 238
left, 238 luck-lucky-luckily, 211 meed, 121
leg, 7, 238 lung, 192 meet, 47, 78, 154, 162, 190,
legal-legalise, 216 lush, 45, 50 202, 213
legal-legality, 255 lust-list(less), 162 meet-meat-mate, 262
leisure, 10, 50 lute, 268 meet-met, 78, 214, 216,
lemon, 32 -ly, 49, 139, 318 225
-less, 138, 215, 264, 295, 318 lyric, 278 melange, 141
lesser, 226 lyric – lyre, 190 ménage, 142
WORD INDE X 405
menage-manage, 142 moth, 244 North-nurse, 244

-ment, 310 mother, 124, 125, 161, 273 nose, 91, 94, 182, 202, 221–2,
Menzies, 83 moth-mother, 234, 273 224, 228, 250, 255,
merchant, 277 motive, 92 261, 263
merchant-marchand, 200 mountain, 290, 306 nose-nostril, 44
merry, 196, 238 mourn, 124, 278 nostalgia, 34
merry-mirth, 192, 195 mouse, 161, 250, 258, 260 nostril, 224
mesh, 45 mouth, 38, 94, 151, 162, 169, nothingism, 138
mesh-mask, 87 191, 250, 259 nought, 44
messenger, 137 mouth-Monmouth, 227 novelty, 243
mice, 82, 161, 174, 201, 250, mouth-mouth, 89, 97 novice, 309
253, 258, 282 mouth-Portsmouth, 212 now, 191, 201
micro-, 256 Mrs, 94 numb, 30
midday, 71 much, 130, 194 nuncle, 146
midden, 137 muckle-much, 86 nurse, 34–5, 198, 276, 383
middle, 71 mud, 248
might, 206, 207 murder, 69, 125 oaf, 97
mighty, 105 muscle, 31 oak, 170, 189, 202, 250, 263,
mild, 166, 168 mush, 246 266
mild-milth, 215 music, 268 Oake-Aike, 200, 203, 240
milk, 82 must, 236 oar, 279
milk-milch, 86 mutter, 34, 39, 69 oaths, 91
mill, 196 mutual, 143 obey, 208
mine, 201 mystery, 354 occasion, 143
miracle, 12, 13, 309 myth, 98 -oid, 270, 318
mirror, 125, 278 oik, 270
mirth, 196 nah, 44 oil, 270
mis-, 183, 296 naif~naïve 97 oink, 269, 270
mission, 143 namby-pamby, 187 old, 163, 167
missive, 71 name, 182, 223 olive, 290
Missouri, 226 nap, 105, 108 -on, 49
misspell, 71 narrow, 126 -ong, 192
mist, 213 nation-national, 216 only, 139
mnemonic, 58, 347 natural, 337 -oon, 319
mock, 243 nature, 144 ooze, 109
moist, 269 nausea, 142 open, 223
mojo, 115 navel, 223 opera, 66
mold, 167 navy, 306 opus, 66
monk, 4, 164, 189, 225 necklace, 303 -or/-our, 18
monk-minchen, 164 need, 213 ordure, 144
month, 189, 221, 273 needle, 125 órgan, 289
mood, 273 need-needed, 216 órison, 315
mood-stood-blood, 234, negotiate, 144 -ose, 319
271, 273 nephew 181, 227 ostler-hostel, 107
moon, 175, 188, 250, 260, nerd, 25 -ous, 310
271, 273 -ness, 138, 264, 318, 351 out, 39, 189
moon-Monday, 184, 212, new, 191, 205, 207, 209, 268 over, 91–2
215, 273 news, 38 overhang-overhang, 311
moor, 206 newt, 146 overture, 144
moot, 12, 162 nigh, 206, 207 owe, 220
more~mair, 203 night, 103, 206 own, 175
morning, 243 nightingale, 103, 137
mornings, 122 ninth, 215 pa, 44
mortar, 289 nook, 272 pace, 32
mórtar, 315 noon, 273 pace-setter, 77
pageant, 150 piteous, 144 prolongate, 139

pahit, 102 pizzaz, 240 prolongation, 139
pal-pale, 190 place, 263 prolonging, 139
palace-palate, 312 plaid, 45 proof-prove, 89, 96
palm, 34–5, 151 plaintiff, 98 prosper, 65
palmer, 131 plank, 138 psychology, 58
pandit, 245 play, 266 psychopaths, 98
pang, 45 please-pleasure, 255 pterodactyl, 58
pansy, 241 plenary-full, 74 pull, 131, 237, 245
Parisian, 144 plenteous, 144 pulmonary, 246
park, 148 plough, 84 pulp, 246
parsley, 200 plunge, 141 pulpit, 246
pass, 32 ply, 49 pulse, 246
passenger, 137 pneo-, 58 punish, 194, 246
passim-pass him, 104 pneu-, 134 punish(en), 245
passion, 143 pneumonia, 57 Punjab, 245
pat-hat, 27 pocket, 243 pure, 38
pater, 256 pod(ium)-foot, 62 pursue, 143
paw, 44, 129, 132 podiatry, 63 puss-pus, 246
peace n. (1160)-appease, 97 pod-pawed, 240 put-putt, 246
peace-appease, 89 point, 191, 209, 270 putt-put, 50, 236
pear, 279 point : pint, 270 pyramid, 278
peasant, 140, 150 poison, 191, 269, 270
pedal, 63 police, 253, 255 qua, 256
pence, 95 pólice-políce, 311 quack, 240
pennant, 150 pontiff, 98 quaff, 240
Pennines, 71 poor, 190, 279 quarantine, 256
people, 129 poor, pore, 279 quart, 240
perfect, 276 porridge, 147 quarter, 237, 239–40
perfect-parfait, 200 port, 237, 242, 244 quartz, 237, 239–40
perféct-pérfect, 310 posh, 45 quash-quack, 236
person-parson, 274, 312 possess, 66, 93, 226 queen, 202
pet-Pete, 190 post-, 15 question, 47
petty-petíte, 312 pouff, 98
pharaoh, 22, 111, 326 pound, 48, 140, 168 race, 82
phase, 94 preposition with, 94 race-raise, 91
pheasant, 150 présent-presént, 284–5 radish, 223
philharmónic, 107 prestige, 141 radius, 15
Phillipp, 111 presume, 38 rag, 30
phone-symphonic, 255 pretty, 121 rage, 141
photographs~photogra[v] pretty~purty, 48, 119 rain, 81
es, 98 prey, 208 Ralph, 241, 255
phthisic, 58 price, 38, 151, 248, 253, 259, rang-ran, 136
phthisis, 57, 134 266, 281 rapture, 143
physics, 57 priest, 170, 190, 213 rather, 81
picture, 290 prime-primary, 216 rathe-rather, 224
piece, 202 prime-primitive, 216 ray, 30
pig, 79 prize, 282 realm, 191
pill, 128 procedure, 143 rear, 66
pillows, 126 profile, 255, 256 rébel-rebél, 311
pin, 27 profound-profundity, 216, récord-recórd, 311
pin–pine, 190 218 red, 271
pin-pen, 192, 237 prognosis, 63, 67 redeem-redemption, 218
pin-ping-pink, 138 prohibition, 99 reduce, 144
pitcher, 30 prolong, 139 reel-real, 20
WORD INDE X 407
refer, 63 rush, 194, 237, 245 sea-see, 261

reif-reave, 96 rust, 213 season, 306
rein(deer), 208 seat, 130, 188
relief-relieve, 96 sack, 24 sedge, 81
rend, 169 sacrifice, 92 see, v. 188, 190, 220
rescind, 87 sad, 199, 237, 238, 239 see, v. 104, 156
resign-resignation, 255 safe, 131 seed, 174
resign, 66 safe-keeping, 13 seek, 202
resume-resumptive, 218 safe-save, 24 seek-beseech, 86
resume, 268 saga, 84 seek-sought, 163
rev-revved, 226 saguáro-saguarésque, 285 seen, 202
rev, 227 said, 45 see-saw, 129
rhomb, 135, 136 salad, 34 see-sea, 250
rib, 30 sale-sell, 157, 162 seethe, 66, 93, 97
rice, 78 salt, 243, 269 ségment-segmentátion, 13
rick-rich, 86 salve-valve, 241 seldom, 167
ridden, 289 same, 45 sell, 72–3
ride, 155, 282 sand, 169 sell-sold, 163
ridge, 26, 28, 141, 147, 196, sandwich, 34 send, 169
198 saner, 32 serene-serenity, 255
righteous, 144 sans, 241 Sereniti, 187
ring, 45, 69, 105, 108, 128, sap, 24, 78 serf, 92
132, 231, 291 Saturday, 49 serf-serfs, 74
ringette, 138 sauce, 269 serf-serve, 96
ringlet, 138 Saul : saw, 129 sergeant, 276–7
rioha, 115 saunter, 241 serif, 97, 278
rioja, 59 save, 250 serve, 276
rise, 66 saw, 44, 84 set, 155, 172
rivalless, 32, 77 say-saw-saga, 74 setting, 147
river, 6, 306, 330 scab, 65, 87 sever, 238
riverside, 13 scabrous, 65 sewer, 143
road-rood, 190 scale, 87, 263 shade, 75
rob–robe, 190 scan, 140 shade-shadow, 44, 224
Robin~Robyn, 187 scarves/scarfs, 96 shard, 277
rode, 214 scene-seen, 20 sharp, 75
roid, 270 sceptic, skeptic, 87 shave, 87
roof, 273 sceptre, 87 she wolf, 129
roof-roofs-rooves, 96, 216 schism, 87 shear, 154
rook, 272 schizo-, 87 shed, 87
room-rum, 246 schmeer, 148 sheep, 204, 260
roost, 213 science, 306 sheep-shepherd, 212, 215
root, 273 scissors, 124, 226 shelf-shelves, 89
Rosa’s, 34 score, 7 shell, 87
roses, 34 scot, 87 shelving, 91
rot, 243 screed-shed, 214 Shepperton, 217
rotten, 237, 242 sċrift, 87 sherd, 277
rough, 113, 115, 207 script, 87 sheriff, 220, 236
rough-though, 20 scripture, 87 shield, 167
route, 256 scrod, 87 ship, 28, 76, 81, 87, 132, 215
row, 220 scrub, 87 shire, 87
royal, 268 scuffle, 87 shire-sheriff, 215
rude, 268 sculpture, 143 shirt, 8, 87
rue, 197, 220 sea, 188–9, 191, 250 shirt-skirt, 74, 88
rule, 268 seal, 78, 129, 178 shlep, 148
run-rung, 138 seal-zeal, 91 shlock, 148
shoot, 177 sleight, 47 stance, 65, 241

shootings, 217 slew, 207 stand, 65
short, 41 slide-slid, 214, 225 Stanford, 215
shot, 87 slough, 207 star, 71, 200
shoulder, 168 slow-sloth, 214 starboard, 276, 304
shove, 87 smo(r)ther, 273 stark, 200
shred, 87, 271 smooth, 93 start, 244
shrift, 214 smurf, 98 starve : deserve, 276
shrift-script, 74, 88 sock, 30 static, 65
shrive-shrift, 96, 214 sodden, 66 statistics, 65
shrub, 87 soft, 237, 240, 242 status, 65
shtetl, 148 Soho, 115 stave, 222
shtick, 148 soil, 269 stave-staff, 44
shuffle, 87 sold, 167 stead, 11, 65, 259, 272
shut, 194 solder, 208 steal, 202, 224
shuttle, 194 soldier, 131, 143, 208 steal-stealth, 184, 214
sib(b), 80 solemnity, 136 stealth, 254
side-sidth, 215 sole-soul, 22 steer>starboard, 276
sieve-sift, 96 sólid-solídify, 13 stelan, 224
sift, 172 sólid-solídity, 294 stepchild, 238
sigh, 220 -some, 215, 318 stepfather, 217
sight, 114 some, 91, 189 -ster, 215
sign, 253, 258 son, 188–9 sthenic, 57, 134
silence, 290, 306 song, 140 stink-stench, 86
silly, 91 soon, 175, 197, 273 stomach, 245
silvery, 49 soot, 273 stone(s), 175, 181, 189,
sin, 77, 138, 188, 192, 195, soothe, 94 190–1, 215, 226, 250
197 sought, 49, 103, 113, 206, stone-stane, 203
sing, 39, 49–50, 136 207, 268, 269 stood, 45, 271, 273
singer-linger, 134 soul, 268 stool, 65
single, 138 sound, 48, 140, 166 stoop, 273
sink, 138 south-southern, 216, 271 stop, 243
sink-zinc, 91 sow, 258 stow, 191, 205, 263, 266
sin-sing-sink, 138 Spain-Spaniard, 218 strange, 250, 255
sip, 30 spare : are, 224 straw, 269
sir, 35 speak(s), 85, 121 stream, 156, 159, 176, 201,
sister, 7 speech, 85, 121 204, 337
sit, 72 speed, 65 street-Stratford, 173, 174
sit-zit, 89 spend, 169 stretch, 72, 82, 163
skewer, 38 sphére, 315 Stretham, 174
skill, 7 spider, 125 strode, 214
skim, 194 spindle, 48 strong, 192
skin, 7 spinach, 34 strongish, 138
skipper, 87 spit-spitting, 226 strong-strength, 163
Skirlaugh, 87 spittle, 130 strop-strap, 243
skirt, 8, 87 spoof, 98 strut, 33–44, 151
skirt-shirt, 74, 88 spoon, 273 stúpor, 308, 315
skulk, 131 sport, 244 sty, 265
skull, 246 square, 37 sub-, 15
sky, 7 squirrel, 278 subbie, 32
slander, 241 staff, 222 subbrachial, 32
slangy, 138 staff-stave, 224 succumb, 135
slaughter, 207, 268–9 stag, 83 such, 30, 194
sleep, 41, 174, 175, 260–1 stain-glass, 149 sucker, 30
sleep-slept, 214 stalk, 269 sue, 38, 197
WORD INDE X 409
suffer, 32 terror, 278 tomb, 135

sugar, 143, 268 -th, 213 tombic, 136
suit, 38, 143, 268 that, 199 tongue, 136, 189
sulphur, 92 thaw, 65, 207, 209, 268, 269 took-tuck, 246
summon, 32 theatre, 58 tooth, 54, 68, 161
summoner, 354 theft, 163, 170 toothache, 91
sunder, 166 there, 206 tooth-dental, 54, 67
sundry, 194 thick, 28 torment, 311
sung-sun, 136 thief, 163, 191, 201–2 tother, 146
super-, 15 thief-theft, 47, 214 toucan, 290
supper, 245, 246 thigh, 114 tough, 114
sure, 30, 143, 268 thimble, 130 town, 30, 160, 191
surf, 97 thin, 30 town-clerk, 13
surgeon, 277 thine, 81 township, 160
sw-, 139 thing, 49, 50 toy, 50, 209
swagger, 240 thinghood, 138 transact, 66
swain, 208 thing-thin, 75, 136 transcend, 32
swathe, 93 think, 189 trap, 34, 151
sway, 28 think-thought, 163 travel-traváil, 312
sweat, 271 third, 79, 119 tread, 237
sweat-treat, 272 thirst, 119, 120 treasure, 141
sweet, 155, 188 this, 28, 30 tree, 63, 176, 191, 250
Swindon, 220 thistle, 49 tri(ple)-three, 62
swine, 95 thistle-this, 91 triad, 63
swine-Swinburne, 217 thither, 125 triple, 63
sword, 139 though, 105, 113, 115, 208 triplex, 68
symposium, 142 thought, 33–5, 75, 136, 151, troth, 177, 227
syrup, 278 207 trouble, 245
though-tough, 99, 112–13 trough, 207
Tadmarton, 217 thread, 271 truce, 95
taiga, 290 threat, 271 trust, 194
tail, 264 three, 54, 63, 68, 81, 94, truth, 12
tale, 157, 221, 223, 254, 263, 139 try, 144
271 three-triple, 54, 67 tsar, 148
tale-tell, 162 threshed, 119 -tude, 144, 319
talk, 49, 241, 269 thrice, 95 Tuesday, 38, 143, 268, 304
talkative, 13 thrive-thrift, 96 tuft, 150
tambourine, 253, 290 throat, 224 tug, 112
tango, 138 throttle, 224 tumble, 167
tango-tangle, 129 through, 105, 113, 115 tumour, 144
tar, 200 thrush, 194 tumult, 144
tarpaulin, 137 thud, 194 tune, 38, 143
taught, 206 thunder, 48, 168 Tunesia, 144
taxi, 187 thyroid, 26, 63, 68, 94 turf, 124
teach-taught, 163 tide, 189 turfs/turves, 96
teethe, 94 tilde, 34 turmoil, 269
teething, 91 timber, 167 turve, 96
tell, 157 time, 189 tutor, 144
tell-told, 163 tmesis, 134 twang, 240
ten, 271 toad, 202 twice, 95
tend, 49 today, 219 twine, 121
tender, 169 toe, 191 two, 139
tenor, 238 toes~taes, 203 -ty, 319
tenth, 215 together, 125 tyrant, 140, 150
terrace, 278 toilet-toilétte, 312 tyre~tire, 187
un-, 183 vision, 12 whet-wet, 76

uncouth, 235 vixen, 95 which, 130
(un)couth-kithe, 162 vodka, 98 which-witch, 99, 105, 109,
under, 69, 188, 189 voice, 209 111
unison, 94 whiff, 109
university-varsity, 276 wa-, 239 while, 140
unnamed, 32 wāfer, 92 whilst, 49, 150
unneeded, 77 waft, 237, 240 whine, 22, 28, 50, 102
upload-upload, 311 wag, 240 whine-wine, 105, 109
upon, 30 waistcoat, 304 whip, 109
Upton, 220 waiter, 41, 147 whisk, 109
urban-urbáne, 312 wake, 72, 77 whistle, 28, 102
us, 219 wake-watch, 86 white, 21, 64, 76, 105,
usage, 94 walk, 49, 241 108–9, 115, 215, 230,
use, 195, 197 wan, 240 347
usual, 141 wand, 237, 239 white-wit, 21
-ward(s), 318 whither, 125
vaccine, 226 warden, 12 Whitsunday, 215
vagrant, 98 -wards, 124 whole, 189
vahana, 98 warm-to warm, 159 whole~hail, 203
vain-fain, 92 warn, 200 whole-heal, 162
vain-vanity, 218 warranty, 12 whoop~hoop, 109
valet, 98 warranty-warden, 86 whoops, 273
válue-váluate, 287 warsh for wash, 124 why, 205
van, 30 wash-wax, 236 wick, 221, 236
vane, 95, 182 wasp, 240 -wick vs. -wich, 86
varmint, 277 watch, 72, 82 wide-width, 212, 255
vase, 98 water, 120, 147, 199, 240 wife, 96–7, 189, 226, 250,
vast-fast, 50 wavering, 92 253, 257–8, 265
vat, 95 way, 178, 206, 209 wife-wives, 89
vault, 131 weal-wealth, 214 win, 30
veal, 92 weapon, 34, 40, 69, 182, 297, wince, 192
vehement, 99 338, 353 window(s), 126, 208,
vehicle, 99 weather, 124 303
veil, 268 weave-weft, 96, 214 wind-wind, 169
verdure, 143 wed-red, 134 wine-whine, 22
verity, 12, 13 week, 172, 221, 236 wing, 7, 192
vermin, 277 weigh, 266 winter, 338
vermin~varmint, 48 weight, 268 wisdom, 215, 236, 254
vermouth, 98 welcome, 34 -wise, 318
verse, 58 well, 190 wise, 93, 215
vertigo, 256 wench, 130 wizard, 13
victor, 85 wend, 169 wiþ-, 183
vile-file, 92 were : bear, 224 woe, 191, 264
vile-vilify, 218 west, 28 wolf, 69, 245
village, 141 wet, 271 wolf-wolves, 90
vinegar, 289 whack, 240 woman, 47
vine-vinegar, 218 whale, 28, 64, 76, 102, 109, woman-women, 161
violence, 12 223 womb, 135, 167
violent, 12 whale-wail, 76, 109 wonder, 12, 189, 338
Virgil, 22 wharf, 30, 50, 109 woo, 246
Virgilius, 58 wheat, 109, 140 wood, 221, 236, 245
virgin, 124 whereon, 110 wool, 71
visage, 12 whether, 81 work-wrought, 163
visceral, 87 whetstone, 304 worse, 189
WORD INDE X 411
wrangle, 240 -y, 15 yon, 83

wrath, 134, 240 yard, 8 yonge, 140
wreak-wretch, 86 yarn, 83 you two, 83
wreathe, 94 yawn, 83 young, 83, 103
wreck-reck, 134 yea, 83, 262, 264 youngness, 138
wrenches, 134 year, 103, 175
wretch, 133 yearn, 83, 165 zeal, 93
wrinkle, 134 yeast, 83 zephyr, 93, 98
writ : wonder, 22 yeehaw, 101, 115 zeroth, 215
write-right, 99, 132 yell, 103 zeugma, 93, 143
wrong, 7, 49 yellow, 28, 145, 156, 178–9 Zeus, 143
wrought, 105, 119–20, 133, yes, 8, 30, 83 Zhivago, 30
140, 163, 225 yet, 81 zip, 30
yield, 154 zloty, 98
Xavier, 139 yogurt, 124 zombie, 98
X-ray, 139 yoke, 83 zone, 93, 98
Names Index
Page numbers in the Names index refer to the main hardcopy volume. Entries in bold refer
to the sections in the Online Companion where the sources are cited or discussed. The
Companion is available at: www.euppublishing.com/page/ETOTELAdvanced/Minkova
Ahn, Sang-Cheol, 3.4.1 Boersma, Paul, 3.4.1

Albright, Adam, 5.6 Bondar, Vladimir A., 5.1.3
Alexander, James D., 5.2.2 Bosworth, Joseph and T. Northcote Toller,
Algeo, John, 1.6 (Further Reading) 5.2.2
Allerton, D. J., 338 Bradley, Travis G., 5.2.3
Anderson, John, 5.6 (Further Reading) Branchaw, Sherrylyn, 7.5.1.1
Anderson, John and D. Britton, 167 Bredehoft, Thomas, 10.2.1
Aroui, Jean-Louis and Andy Arleo, 10.1, Brinton, Laurel and Margery Fee, 111
10.4 Britton, Derek, 125–6, 5.2.3
Browman, Catherine P. and Louis Goldstein,
Barber, Charles, 15, 224, 243, 262, 1.6 5.2.2
(Further Reading) Brown, Carleton, 313
Bauer, Laurie, 111, 259, 317, 1.6 (Further Brunner, Karl, 7.3.1
Reading) Burling, Robbins, 10.4
Becker, Michael, Andrew Nevins and Burton, T. L., 243–4, 247, 5.3.2, Chapter 8
Jonathan Levine, 98, 4.4 (Further Reading)
Beddor, Patrice Speeter, Rena Arens Krakow Bybee, Joan and Dan I. Slobin, 6.3
and Louis M. Goldstein, 175, 7.3.1 Bybee, Joan L., 6.3
Bell, Alan, Jason M. Brenier, Michelle
Gregory, Cynthia Girand and Dan Cable, Thomas, 336, 339, 10.2.2, 10.3
Jurafsky, 7.5.1.3 Campbell, Alistair, 58, 157, 181, 1.6
Bennett, William H., 306 (Further Reading), 3.4.5, 4.1.2, 5.2.2,
Benson, Larry D., 292, 10.4 6.2, 6.5.3, 7.5.1.1, 9.3.1
Berg, Thomas, 318, 9.8 Carter, Paul and John Local, 2.3.1
Bermúdez-Otero, Ricardo, 223–4, 2.5 Chambers, J. K., 281–3
(Further Reading), 4.1.1, 5.3.2, Ching, Marvin K. L., 94
7.5.1.2, 7.5.2.1 Chomsky, Noam and Morris Halle, 196
Bermúdez-Otero, Ricardo and April Clark, Cecily, 7.5.1.3
McMahon, 7.5.1.1 Clarke, Sandra, 8.2.2.3
Bermúdez-Otero, Ricardo and Richard Cohn, Abigail, 169, 6.4.1
Hogg, 121 Collie, Sarah, 9.8
Bermúdez-Otero, Ricardo and Graeme Colman, Fran, 155, 6.2
Trousdale, 5.3.2 Côté, Marie-Hélène, 150
Berndt, Rolf, 1.6 (Further Reading) Crisma, Paola, 5.6 (Further Reading)
Bibire, Paul, 6, 6.2, 7.3.1 Crosswhite, Katherine, 181
Bigham, Douglas Stephen, Appendix 8.3 Cruttenden, Alan, 34, 36, 77, 116, 134,
Blevins, Juliette and Andrew Garrett, 120 141, 146–7, 199, 237, 258, 2.2.1, 2.5
412
NAMES INDE X 413
(Further Reading), 4.1.2, 5.1.1, 5.1.3, Flemming, Edward and S. Johnson, 2.2
5.6 (Assignments), 6.4, 6.5.4, 7.5 Fridland, V., Appendix 8.3
Crystal, David, 8.2.2, Appendix 8.1 Fromkin, Victoria, 4.4
Culpeper, Jonathan and Phoebe Clapham, 9.8 Fudge, Eric, 318
Cutler, Anne and Sally Butterfield, 9.1 Fulk, Robert D., 42, 167, 224, 338–9, 1.6
(Further Reading), 5.2.2, 6.4, 7.5.1.1,
Dance, Richard, 6, 173, 177, 208, 7.6 7.5.1.2, 9.4.1
Danchev, Andrei, 6.5.3
Danielsson, B., 310 Garrett, Andrew and Juliette Blevins, 5.3.2
Daunt, Marjorie, 5.2.2, 6.5.3 Gascoine, George, 352–3, 10.2.1, 10.3
Davenport, Mike and S.J. Hannahs, 116 Getty, Michael, 170, 10.2.1
Delattre, Pierre and Donald C. Freeman, 117 Gick, Bryan, 129, 5.2.5
Denton, Jeanette Marshall, 118, 120, 3.4.5, Giegerich, Heinz, 28, 31, 43, 239, 298, 318,
3.4.5 (Further Reading), 5.2.2, 5.6 2.1, 2.2.1, 2.5 (Further Reading),
(Further Reading) 5.1.1, 5.2.4, 5.3.2, 5.6 (Further
Diamond, Jared and Peter Bellwood, 3.4.5 Reading), 7.5.1, 9.8 (Further Reading)
(Further Reading) Gimson, A. C., 199, 7.5.1
Dobson, Eric John, 114, 123, 126, 130–1, Gnanadesikan, Amalia E., 154
137, 200, 207, 272, 276, 278, 310, 2.5, Goldsmith, John, 179–80
5.2.4, 5.4, 7.3.1, 7.5.2.1, 8.1, 8.2.2.2, Gordon, Matthew J., Appendix 8.3
Appendix 8.1, Chapter 8 (Further Gordon, Matthew Kelly, 171, 4.1.2, 7.5,
Reading) 7.5.2
Dresher, B. Elan, 3.4.5 (Further Reading), Grant, Colin, 5.1.2
7.3.1 Greenberg, Joseph, 5.1.2
Dresher, B. Elan and Nila Friedberg, 10.1 Greenfield, Stanley and Daniel Calder, 10.3
Duffell, Martin, 10.3, 10.4 Gussmann, Edmund, 5.3.2
Duggan, Hoyt N., 10.2.2
Hall, Nancy, 5.2.2
Eckert, Penny, Appendix 8.3 Hall, Tracy and Silke Hamann, 128
Eddington, David and Caitlin Channer, 147 Halmari, Helena, 10.2
Eddington, David and Michael Taylor, 147 Hamann, Silke, 123, 5.2.3
Ekwall, Eilert, 195, 209, 224, 7.3.1 Hammond, Michael, 2.2.2, 4.1.2, 6.4.1,
Ellis, Lucy and William J. Hardcastle, 5.3.2 6.4.1, 7.5
Erickson, Blaine, 5.2.2 Hanham, Alison, 5.2.4
Hanson, Kristin and Paul Kiparsky, 10.2
Fabb, Nigel, 10.2 Harmon, William, 365, 10.3
Fabricius, Anne, 8.2.1 Harrington, Jonathan, Felicity Cox and Zoe
Ferragne, Emmanuel and François Evans, 7.6
Pellegrino, 7.6, 8.2.1 Haugen, Einar, 148, 5.5.2
Fikkert, Paula, 5.3.2 Hawkins, Sarah and Jonathan Midgley, 238,
Filppula, Markku, Juhani Klemola and Heli 8.2.1
Paulasto, 1.2 Hay, Jennifer and Andrea Sudbury, 126,
Finegan, Edward, 283, 1.6 (Further 5.2.3, 5.6 (Further Reading)
Reading) Hayes, Bruce, 42, 97, 286, 294, 298, 300, 331,
Finkenstaedt, T. E. Leisi and D. Wolff, 12 2.5 (Further Reading), 5.1.1, 5.2.5,
Fischer, David Hackett, 1.4 7.5.1, 8.2.1, 9.8 (Further Reading),
Fisher, John Hurt, 8.3.2 10.1, 10.3, 10.4
Fisiak, Jacek, 7.2.1, 7.6, Chapter 8 (Further Hayes, Bruce and Donca Steriade, 79, 2.5
Reading) Hayes, Bruce, Colin Wilson and Anne
Fitzgerald, Christina, 355 Shisko, 365
Fitzmaurice, Susan and Donka Minkova, Hebda, Anna, 7.5.2.1
Appendix 8.3 Heselwood, Barry, 118, 127, 5.2.2, 5.2.4
Flasdieck, Hermann F., 4.3 Hickey, Raymond, 1.2, Appendix 8.3
Flemming, Edward, 90, 197, 264, 2.2, 7.6, Hill, Archibald A., 121–4, 5.2.3, 5.6
8.2.2.2 (Further Reading)
Hogg, Richard, 8, 19, 79, 119, 121, 130, King, Anne, 6.2
154–7, 159, 173, 176, 181, 184, 198, Kiparsky, Paul, 10.1, 10.3
222, 225, 1.6 (Further Reading), 1.6 Kiparsky, Paul and W. O’Neil, 3.4.5
(Further Reading), 3.4.5, 4.2.1, 5.1.2, (Further Reading)
5.2.2, 5.2.2, 5.2.5, 6.2, 6.3, 6.5.1, 6.5.3, Kirchner, Robert, 79
6.5.4, 7.3.1, 7.5.2.1 Kirkby, John, 246
Hogg, Richard and Robert D. Fulk, 1.6 Kirwin, William J., 95
(Further Reading) Kitson, Peter, 9
Honeybone, Patrick, 2.1.1 (Further Knappe, Gabrielle, 5.1.4
Reading) Kökeritz, Helge, 239, 254, 8.2.2.2, Appendix
Hoover, David, 9.4.1 8.1, Chapter 8 (Further Reading)
Horn, Wilhelm and Martin Lehnert, 5.2.4 Kortmann, Bernd and Benedikt
Horobin, Simon and Jeremy Smith, 1.6 Szmrecsanyi, 233
(Further Reading), 7.2, 7.4 Krapp, George Philip and Elliot van Kirk
House, Arthur S., 240 Dobbie, 328
Howell, Robert B., 117, 120, 5.1.2, 5.2.2, Kreidler, Charles, 28, 2.1, 2.5 (Further
5.2.5 Reading), 5.1.1, 8.2.1
Howell, Robert and Katerina Somers Wicka, Krygier, Marcin, 161
5.1.2, 5.2.2, 5.2.2 Kury˜lowicz, Jerzy, 9.4.1
Hughes, Geoffrey, 1.6 (Further Reading)
Hutcheson, Bellenden Rand, 291, 295, 302, Labov, William, 35, 237, Appendix 8.3
328, 332, 335–6, 345, 6.5.4, 10.2.1 Labov, William, Paul Cohen, Clarence
Hyman, Larry, 164 Robins and John Lewis, 150
Labov, William, Sharon Ash and Charles
Ikegami, Masa T., 122, 192, 7.4, 7.5.2, 7.5.2.1 Boberg, 242, 1.6 (Further Reading),
Ingham, Richard, 11 2.2.1, 2.5 (Further Reading)
Ito, Junko and Armin Mester, 5.1.2 Labov, William, M. Yaeger and R. Steiner,
Iverson, Gregory K. and J. C. Salmons, 3.4.1, Appendix 8.3
4.3 Ladefoged, Peter, 37, 101, 141, 2.1.1, 2.1.2,
Iverson Gregory K. and Sang-Cheol Ahn, 2.2.1, 5.1.4
2.1.1 Ladefoged, Peter and Ian Maddieson, 25, 77,
79, 101, 116, 118, 7.3.1
Jefferson, Judith and Ad Putter, 10.2.2 Lahiri, Aditi and Paula Fikkert, 224, 7.5.1.1,
Jespersen, Otto, 125, 131, 134–5, 147, 239, 7.5.1.2
250–1, 253, 273, 8.1, 8.2.1, Chapter 8 Laing, Margaret, 7.2
(Further Reading) Laing, Margaret and Roger Lass, 184, 5.1.4,
Johnson, Wyn and David Britain, 5.2.5 6.4.1, 7.2, 7.2.1, 7.2.2
Johnston, Paul A., Jr., 253, 316 Lamb, Sydney and E. Douglas Mitchell,
Jones, Charles, 169 3.4.5 (Further Reading)
Jones, Daniel, 23, 2.1.2, 9.8 Lapidge, Michael, 327, 10.2.1
Jordan, Richard, 122, 133, 136, 194, 197–8, Lappe, Sabine, 322, 9.8
200, 222, 254, 260, 264, 1.6 (Further Lass, Roger, 11, 123, 126, 131, 133, 137–8,
Reading), 7.3.1, 7.4 144, 157, 170–1, 173, 176–7, 179, 185,
217, 224, 226, 239, 242, 245, 248, 256,
Kastovsky, Dieter, 6.3, Chapter 8 (Further 258, 270, 277–8, 282, 287, 315, 1.3, 1.6
Reading) (Further Reading), 2.2, 2.3.1, 3.4.5,
Kawahara,Shigeto, 4.1.2 3.4.5 (Further Reading), 5.2.2, 6.2,
Kaun, Abigail, 197, 7.3.1 6.3, 6.4, 6.4.1, 6.5.4, 7.2.1, , 7.3.1,
Keenan, Edward, 7.5.1.3 7.5.2.1, 8.1, 8.2.2, 8.2.2.2, 8.2.2.3,
Kelly, Michael, 287 Appendix 8.3, Chapter 8 (Further
Kenyon, John S. and Thomas A. Knott, 23, Reading), 9.3.2
8.2.1 Lass, Roger and John Anderson, 117, 5.2.2, 7.4
Keyser, S. J. and W. O’Neil, 3.4.5 (Further Lass, Roger and Margaret Laing, 194, 7.2.2
Reading) Lauttamus, Timo, 5.6 (Further Reading)
Kim, Myungsook, 223–4 Lavoie, Lisa, 169
NAMES INDE X 415
Lawson, Eleanor, Jane Stuart-Smith and Mossé, Fernand and James Walker, 1.6
James Scobbie, 126, 5.2.3 (Further Reading), 7.3.1
Leach, Alfred, 108, 111 Mugglestone, Lynda, 107–8, 111, 126, 5.1.3,
Lehiste, Ilse, 211 5.6 (Further Reading)
Lehmann, Winfred P., 10.2, 10.2.2 Mulcaster, Richard, 226, 235
Leith, Dick, 262 Murray, Robert W., 224
Lewis, M. Paul, Gary F. Simons and Charles Muthmann, Gustav, 5.1.4, 5.2.3
D. Fennig, 56
Liberman, Anatoly, 165, 224, 6.4, 6.5.3, Nagy, Naomi and PatriciaIrwin, 126, 5.2.3,
7.5.2.1 5.2.4
Liebl, Christian, 202 Nevalainen, Terttu, 15, 236, 315, 1.4, 1.6
Luick, Karl, 138, 165, 170–1, 207, 216, (Further Reading), 9.8
222, 229, 251, 253, 5.4, 6.4, 6.4.1, 8.1, Nevalainen, Terttu andIngridTieken-
Chapter 8 (Further Reading) BoonvanOstade, 19, 236
Lutz, Angelika, 120, 131, 1.3, 5.2.4, 5.6 Nevanlinna, Saara, 5.6
(Further Reading), 8.2.2.1
Oakden, J. P., 10.2.2
Maclagan, Margaret and Jennifer Hay, Ogden, Richard, 26
Appendix 8.3 Ogura, Mieko, 255, 259, 272, 7.3.2, 8.2.2,
MacMahon, Michael K. C., 248, 316 8.2.2.2, 8.3.1, Chapter 8 (Further
Maddieson, Ian, 118, 222, 3.4.1, 6.4.1 Reading)
Magnuson, Thomas J., 116 Oppenheimer, Stephen, 3, 3.4.5 (Further
Malone, Kemp, 9, 182–3 Reading)
Marchand, Hans, 9.8 Ostler, Nicholas, 11
Mayr, Robert, 198, 7.3.1
McCully, Christopher B., 9.4.2 Page, Richard, 224
McDavid, Raven I., Jr., 8.3.1 Paja˛k, Bożena, 4.1.2
McKie, Michael, 346–8, 10.3 Parkes, M. B., 7.2.1
McMahon, April, 28, 31, 36, 37, 116, 280, Passy, Paul, 2.1.2
2.1, 2.2.1, 2.5 (Further Reading), Petrova, Olga, 3.4.1, 15
5.6 (Further Reading), Chapter 8 Phillips, Betty, 51, 169, 171, 272, 5.4.1, 7.3.1,
(Further Reading) 8.2.2.3, 8.3.1, 9.8
Menken, H. L., 2 Polka, Linda and Megha Sundara, 9.1
Milroy, James, 5.6 (Further Reading) Pope, Mildred, 106, 122, 131, 141, 4.1.2,
Minkoff, Marco, 130, 167, 5.6, 6.5.3, 8.2.2 5.2.5, 8.2.1
Minkova, Donka, 85, 105, 109, 111, 1323, Proctor, Michael, 117, 5.2.5
178, 183, 221, 223, 232, 284, 307, 339, Purnell, Thomas C. and Malcah Yaeger-
341, 349, 355, 1.6 (Further Reading), Dror, 1.2
4.1.1, 4.4, 5.1.2, 5.1.4, 5.1.4, 5.3.2, Putter, Ad, 355
5.5.1, 5.6 (Further Reading), 6.5.4, Putter, Ad, Judith Jefferson and Donka
7.5.2.1, 7.6, 9.3.2, 9.4.1, 9.4.2, 9.6.1, Minkova, 7.4, 8.2.2.2
9.8, 10.2, 10.2.1, 10.2.2, 10.3 Pyles, Thomas, 105
Minkova, Donka and Kie Ross Zuraw, 42,
4.2.1, 5.5.2, 6.4 Quirk, Randolph and C. L. Wrenn, 5.2.2
Minkova, Donka and Robert Stockwell, 61,
167, 218, 253, 255, 259, 262–3, 296, Reszkiewicz, Alfred, 157, 181, 5.2.2
305, 314–15, 378, 390, 1.6 (Further Riera, María, Joaquín Romero and Ben
Reading), 5.3.2, 6.4.1, 7.5.1.1, 8.2.2, Parrell, 5.2.2
Appendix 8.3, 9.4.2, 10.4 Ringe, Don, 180, 212, 3.4.1, 3.4.5 (Further
Moen, Per, 8.3.1 Reading)
Montgomery, Michael, 111, 148 Ritt, Nikolaus, 224, 272, 6.4, 7.5.1.1, 7.5.2.1,
Montgomery, Michael and Connie Eble, 7.3.1 8.3.1
Morén, Bruce, 6.4.1 Roberts, Jane, 7.2
Moreton, Elliot and Erik Thomas, 282 Robinson, Orrin W., 3.4.5 (Further
Morris, Richard, 249 Reading), 8.2.2.2
Rusch, Willard, 7.3.2 Terajima, Michiko, 8.2.2.2

Ross, John, 285 Thomas, Eric, 35, 180, 199, 237–9, 241–4,
Rothwell, W., 11 248, 259, 270, 2.5 (Further Reading),
Russom, Geoffrey, 302, 6.5.4, 9.4.1, 10.2.1, 7.3.1, 8.2.1
10.2.2 Throop, Priscilla, 154, 10.2.2
Ryan, Kevin, 171, 287, 6.4.1, 7.5, 7.5.2 Tieken-Boon van Ostade, I., 1.6 (Further
Reading)
Saintsbury, George, 10.2 Tottie, Gunnel, 7.5.1.1
Salmons, Joseph, 70, 2.5 (Further Reading) Tristram, Hildegard, 3
3.4.3 Trudgill, Peter, 19, 95, 126, 138, 147, 247,
Samuels, Michael, 184, 7.2 5.2.4
Sandved, Arthur O., 7.3.2 Trudgill, Peter, Elizabeth Gordon and
Schaefer, Ursula, 1.6 (Further Reading) Gillian Lewis, 237–8
Schendl, Herbert, 185, 347
Schlüter, Julia, 233, 5.1.1, 5.1.3, 5.6 (Further Uffmann, Christian, 5.2.4, 5.2.5
Reading)
Schrijver, Peter, 150, 4.1.2, 6.5.3 Vachek, Joseph, 270, 5.3.2
Scobbie, James and Alan Wrench, 5.2.5 Vennemann, Theo, 123, 1.6, 2.5 (Further
Scragg, Donald G., 82, 111, 202, 5.1.2, Reading)
7.3.2
Selkirk, Elisabeth O., 302 Wakelin, Martyn, 3, 63, 96, 116, 8.2.2.2
Sharf, Donald J., 7.5 Wales, Katie, 147, 244, 248, 1.4, 8.2.2.2,
Short, Ian, 309 Appendix 8.2
Sievers, Eduard, 301, 328 Walker, John, 23, 99, 107–8, 135, 143–4, 240,
Simon, Ellen, 90 244, 246, 255–6, 262, 267, 304, 5.4.1,
Smith, Jeremy, 19, 179, 186, 203, 226, 249, 8.2.1, 8.2.2.1, 8.3.2
253, 5.2.2, 6.5.3, 7.2, 8.2.2.1, 8.2.2.2 Watkins, Calvert, 56, 69, 3.4.5 (Further
Sóskuthy, Márton, 128–9, 131 Reading), 3.4.5 (Assignments)
Stanley, Eric, 349, 350, 10.3 Watts, Richard J., 251,
Stenbrenden, Gjertrud, 202, 249, 251, 253–4, Wells, John, 63, 127, 129, 137, 198, 237, 244,
257, 7.3.2, 8.2.2.2, 8.3.1, Chapter 8 246, 259, 278, 342, 5.2.3, 8.2.2.2, 8.2.2.3
(Further Reading) Wełna, Jerzy, 176, 258, 7.4, 7.5.1.1, 8.2.2.1,
Stenroos, Merja, 7.3.1 8.2.2.2
Steriade, Donca, 123, 4.1.2 Wetzel, Claus-Dieter, 7.4
Stockwell, Robert, 118, 180, 258, 263, 265, Wheatley, Henry B., 6.4
5.2.2, 7.3.2, 7.5.2.1, Appendix 8.3, White, David, 176
Chapter 8 (Further Reading) Williams, Fionnuala Carson, 10.2
Stockwell, Robert and Donka Minkova, Windross, Michael, 120, 126, 5.2.3, 8.3.2
253, 262–3, 6.5.1, 8.2.2, Chapter 8 Wogan-Browne, Jocelyn (ed.) et al., 1.6
(Further Reading), 10.2.1, 10.4 (Further Reading)
Stockwell, Robert P. and C. Westbrook Wolfram , Walt and Natalie Schilling-Estes,
Barritt, 6.5.3 35, 138, 233
Strandberg, Otto, 7.5.2.1 Wolfram, Walt, 233
Stuart-Smith, Jane, 5.6 (Assignments) Wolfram, Walt and Natalie Schilling-Estes,
Sweet, Henry, 9, 108, 5.1.3, 10.2.1 150, 8.2.1
Szemerényi, Oswald J. L., 3.4.5 (Further Wright, Joseph and Elizabeth M. Wright,
Reading) 7.5.2.1
Wyld, Henry Cecil, 95–6, 122–3, 130, 239,
Taavitsainen, Irma, 7.2, 111 254, 5.5.2, 8.2.2.2
Tabain, M., 101, 342, 3.4.1
Tagliamonte, Sali and Rosalind Temple, Youmans, Gilbert, 9.7, 10.1, 10.2.1
149–50 Youmans, Gilbert and Xingzhong Li, 36, 10.4
Tamson, George J., 310, 9.5
Tarlinskaja, Marina, 361, 5.4.1, 9.6, 10.3, Zec, Draga, 6.4.1
10.4 Zuraw, Kie, 2.5
Subject Index
Ælfric, 19, 67, 86, 154, 179, 184, 187, 228, Andreas, 76, 87, 295–6, 302, 335
285, 339 Anglo-Saxon Chronicle, 5
Lives of Saints, 228 Anglo-Norman, 10–12, 53, 58, 82, 86, 107,
131, 178, 185, 190–1, 200, 213, 218, 267,
AAVE, 17, 150, 233, 237, 240 270, 306, 308, 340–1, 348, 356
accidental gap, 79 Anglo-Saxon, 2, 5, 8–10, 20, 67, 76, 82, 86,
affix,inflectional, 91 106, 327, 328, 339, 342, 346–7
affix <-y>, 231 aphe(re)sis, 49, 322
affricates, 29–30, 74–5, 77, 90, 142, 144, 197, apocope, 49, 228, 230–2, 354
268 a-prefixing, 233
affrication, 74, 84–6, 88, 99, 144–5 articulation, 27
dating, 85 assimilation, 15, 47, 71, 75, 78, 85, 122–4,
Afrikaans, 59–60, 100, 118 128, 135, 137, 143–5, 157, 198, 259,
Aidan, Bishop, 4 266, 289
Akzentumsprung, 177 Atlas of North American English (ANAE), 35
Aldhelm, 342 athematic noun, 161
Alfred, King of Wessex, 6, 10 Augustine, 4
alignment, 326 AusE, 17, 129, 147, 237
alliteration, 13, 22, 76, 94, 291–2, 295, 297, <-aun>, 241, 268
299, 300–4, 310, 313, 323, 326–8, 332–5, Australia, 17, 127, 149
339–48
and stress, 291 back-clipping, 321
liaison alliteration, 341 Barnes, William, 243–4
OE <hw>, 76 Battle of Maldon, 76, 86, 333, 345
alliterative long line, 324, 342–3 Bede, 2–4, 342
Alliterative Revival, 340, 350 Beowulf, 76, 94, 145, 183, 291, 295, 297–8,
allograph, 21, 90 300–2, 329–40, 342, 345
allomorph(s), 52, 88 bilabial, 28, 62
allophone, 21, 25, 29, 50, 66, 83, 85, 102–3, bilingualism, 4, 11, 17, 58, 115, 306
116, 118, 135–6, 138, 141, 166, 173, 177, binary foot, 345, 350, 352
180, 199, 205, 238–9, 258, 264, 282, borrowed, 7, 13, 15, 20, 22, 41, 53–4, 58, 63,
333 65, 67, 86–8, 92, 94, 106–7, 135, 141,
alveolar, 28 178, 191, 216, 218, 235, 243, 255–6, 287,
ambisyllabicity, 41 289, 292, 308–16, 319, 357
American English borrowings, 8, 12, 15, 50, 105, 148, 200, 241,
colonial, 17 246, 252, 255, 265, 284, 310, 312, 314,
Americanism, 17 327, 348
anacrusis, 336 Breaking, 117–19, 129, 275
analogy, 49 Britain, Roman, 3–4
anapaest, 324 Bullokar, William, 123
anaptyxis, 48, 120–1, 129, 132–3 Byron, Lord, 137, 240, 323
417
Cædmon, 323, 342 coronal, 29, 38, 85, 116–18, 122–3, 134,
Caedmon’s Hymn, 24, 327 142, 165, 169, 264, 268, 274–5
caesura, 324, 326, 349, 358 dorsal, 29, 85, 117, 131
Canada, 17, 127, 149, 237, 281, 283 fortis, 27
Canadian English, 17, 111, 238, 259, 281–3 geminate, 31–2, 71, 73, 77–9, 80, 83, 93,
Canadian Raising, 281–3 112, 118, 132, 214, 225
Caribbean English, 18, 237, 239, 270 glottal, 27
Caxton, William, 14–15, 82 heterorganic, 122
<-cc>, 226 lenis, 27
Celtic, 1, 3, 4, 8, 53, 57–8, 80, 306 nasal, 27, 29, 30–1, 48, 136, 152, 169, 192,
Cely Letters, 125, 143, 239 255, 355
chain shift, 35, 64–5, 203, 251–2, 256, 260, oral, 29
265 singleton, 31, 41, 49, 72–3, 75, 77, 80, 87,
Chancery Standard, 187 334
Chaucer, Geoffrey, 14–15, 95, 144, 168, 200, voiced, 21–2, 25–8, 31, 36, 44, 48, 50,
202, 204, 230, 239, 249, 250, 254, 259, 52, 62–7, 74–5, 79, 81–4, 89–95, 97–8,
292, 303, 306–7, 309, 311, 313, 315, 320, 101–5, 112–13, 124, 134–8, 141–2, 165,
323–4, 352–4, 356–65 169, 177, 205, 238, 240, 258, 272, 282–3,
Canterbury Tales, 269, 292–3, 303–5, 307, 333
309–11, 313, 352, 354, 357–8, 360–3, voiceless, 21–2, 27–9, 31, 36, 40, 49, 50,
365 52, 62–5, 72, 74, 76, 80, 82–3, 85, 89–90,
Troilus and Criseyde, 230, 357 92–3, 95, 101–3, 105, 112, 114, 138, 142,
Chaucerian English, 204, 250 146, 148, 177, 205–6, 241, 244, 258–9,
Chaucerian rhyme, 259, 269, 292 272, 281–2, 334
Christ A, B, C, 295 consonants of OE, 71, 74, 76
Christianity, 4, 20, 58 continuant, 30, 116–17, 140, 274–5
clippings, 187, 270, 321 Cornwall, 3, 95
clitic (group), 51, 139, 145–6, 298–300, 320, coronal stop deletion, 149
325–6, 353–543, 364 countertonic accentuation, 310
cliticisation, 299 creolised vernacular
cluster, 28, 32, 40, 42–3, 48–50, 57, 76, 86–7, creolised, 7
102, 109–10, 112, 119–21, 123–4, 131–6, Cursor Mundi, 222, 260
139, 149, 165–72, 213, 220, 232, 244, Cynewulf, 342
251, 272–3, 287, 334
coda clusters, 213 /d/, 28–9, 40–1, 47–8, 52, 54, 66, 89, 102, 125,
bi-segmental, 108 142, 145, 147, 163, 168, 180, 231, 272
medial cluster, 42 dactyl, 324, 350
non-native, 57 Danelaw, 6
simplification, 49 Danish, 5, 7, 21, 59–61, 243
Cockney, 17, 63, 95, 247, 259, 266, 270 Death of Edgar, 333
codification, 14, 19, 21, 48–9, 107, 126, 148, Death of Edward, 327, 339
179, 184, 235, 257, 261, 267, 269, degemination, 80–1, 93
280–1 dental, 28, 54, 67
cognate, 54–5, 69, 87–8, 101, 163 dental-final verbs, 78
colon. See hemistich [-d(ə).r], 124
Colonial English, 16 derivational affix, 84
combining form, 57–8, 319, 149 [d3], 74, 77, 79, 81, 84, 86, 122, 141
complementary distribution, 90–1, 93, 97, diacritic, 152, 154, 179, 189–90, 202, 208,
155 222, 226, 286
compound, 91, 140, 160, 161, 164, 166, 174, digraph, 22, 43, 45, 82, 86, 88, 111, 124, 156,
183, 215, 219–20, 290, 297–8, 303–5, 159, 176, 178–80, 187, 190–1, 202, 271
313, 324, 331, 336, 366 digraph controversy, 179
compound stress, 297, 312, 314 spelling, 117
consonant dip, 328–30, 335, 339, 343–5
approximant, 27, 29–31, 37, 77, 79, 105 expanded dip, 329
SUBJECT INDE X 419
diphthong, 35, 37–8, 45, 47, 50, 58, 83, 103, 255, 263, 267–9, 276, 280, 284, 306, 308,
159, 170, 175–8, 201, 204–5, 208, 241, 310, 314, 347, 356, 360, 364
244, 255–7, 259, 266, 268–70, 282, 287 Old French, 22
centring, 206, 279 frequency, 51
falling, 177, 263 fricative, 30, 75, 102; see also spirant
rising, 177 tautosyllabic voiceless fricative, 240
dissimilation, 47–8 fricative voicing, 66, 74, 83, 88, 91, 95–6, 216
height dissimilation, 259 Frisian, 21, 59–60, 73, 118
Domesday Book, 122, 209 function word, 71, 83, 94, 211, 225, 230, 299,
Donne, John, 1, 224 320, 351, 354, 363
doublets, 83, 87, 96, 104, 214 functional stress-shifting, 311
drag-chain theory, 253
Dryden, John, 224, 261–2 /γ/, 75, 77, 79, 81, 83–4, 86, 102–3, 105
Durham, 8, 85, 111, 339 gap, 79, 253, 272
Dutch, 59–60 gemination, 93
Pennsylvania Dutch, 59–60 new in OE, 78
West Germanic Consonant Gemination,
-ed, 89, 229, 231–2 71–3, 76–9
edh, 20, 82, 90, 94 German
Elene, 105, 295–6 High, 21, 60–1, 73
elision, 229–30, 338–9, 350, 353, 355, 365 Old High, 61
Elizabeth, 137 Germanic, 1–5, 7, 10, 12–13, 17, 47, 54–5,
End Rule see Nuclear Stress Rule 57–63, 65–73, 75, 79, 82, 84–7, 103, 106,
end-stressed nouns, 316 113, 116, 118–19, 129–30, 139, 145–6,
enjambment, 357, 365 153–5, 159, 162–5, 174–6, 178, 180–1,
epenthesis, 48, 119–20, 130, 137, 206, 208, 212, 220, 223, 225, 227–8, 259, 288–9,
228, 263, 275 291, 294, 301–2, 306, 308, 315–16,
[εr] > [ar], 275–6 326–8, 334, 348–9
-(e)st, 231 diaspora, 3
eth, 20, 229, 231, 355 East Germanic, 59
etymology, 12, 23, 45–6, 69, 75, 83, 86–7, Germanic invasion, 3
103, 106–7, 109–10, 116, 124, 128, 149, North Germanic, 59–60
152, 154, 156, 167, 175, 178, 188–9, Old Germanic, 66
194–5, 214, 227, 241, 244, 253, 257, 269, Proto-Germanic, 8, 36, 57, 59–62, 65–6,
279, 316, 326 68–73, 77–8, 82–7, 104, 112, 121, 129,
etymological spelling, 45, 153 156–9, 161–3, 169, 173, 177, 199, 212,
unetymological, 49 301, 327, 333
extrametrical, 336, 357 West Germanic, 60–1, 63, 72, 78, 84,
175–6
-ə, 34, 69, 80, 93–4, 126, 206, 223, 229, Germanic Stress Rule (GSR), 70, 288–9,
230–2, 277–8, 280, 338, 354 294, 303, 307, 309, 311, 315–16
əm-, 139 in AmE and BrE, 317
-gg-, 72, 79
[f], 241 Gil, Alexander, 16
Faroese, 60 glide vocalisation, 207–8
final <-e>, 189, 338, 353–4 glides, 25, 30–1, 37, 113, 119, 168, 177, 190,
flap, 147 259, 279
Flemish, 59, 60 global English, 252
‘focused’ variety, 179, 184 globalisation, 142
foreignness, 142–3, 231, 270 glottal, see stop, glottal
fortition, 50, 75, 103 glottalisation, 146–7
French, 7, 10–13, 15, 17, 23, 48, 53, 55, 57–8, grapheme, 21
86–8, 92, 95, 97, 99, 106–7, 131, 136, Greek, 15, 54–8, 60, 62–3, 65, 67, 87, 93,
141–3, 146, 148, 152, 158, 185–7, 193–5, 100, 107, 111, 115, 134, 148, 154, 269,
197, 200, 202–3, 208, 218, 243, 245, 253, 289, 350
Grimm’s Law, 61–2, 66–8; see also First irregularity, 246

Germanic Consonant Shift isosyllabic, 324, 348–53, 355–6, 361, 365
Guy of Warwick, 264 Italian, 23, 55, 154, 245, 357
Italian loans, 226
hapax legomena, 298 Italic, 57–8, 63
Havelok the Dane, 149, 348
h-dropping, 100–1, 107–8, 111 Jamestown Colony, 16
orthographic, 104
headless line, 324, 357–8, 361 Kaluza’s Law, 331
hemistich, 324–5 Keats, John, 137, 323
heptarchy, 5 Kent, 2, 4–5, 95, 203
hiatus, 93, 146, 229, 233, 338 Knight’s Tale, 230
avoidance of, 231
High Vowel Deletion (HVD), 73 <l-> insertion, 130
Holofernes, 131 labelling mismatches see stress mismatches
homophone, 22, 104, 129, 134, 167, 179, labiodental, 28
241, 279 labiovelar, 28
homophony, 64, 111, 149, 169, 244, 250, 262 LAEME, 15, 92, 108–9, 132, 135, 179, 185,
homorganic cluster lengthening, 124, 135, 189, 192, 234–4, 290
157, 165–6, 168–9, 171 LALME, 15, 122, 185, 234, 254, 355
<hw> - [], 21–2, 28, 50, 64, 74, 76, 102, language contact, 8, 47, 80, 106
109–12 laryngeal see glottal
hybrid Latin, 1–2, 4–5, 7–8, 10–12, 15–16, 20, 22–3,
prosodic patterns 316–17 35, 48, 53–5, 57–8, 60, 62–3, 65–7, 80,
verse 340, 348 86–7, 92–3, 104–5, 107, 111, 131,
word-formation, 13 139–41, 146, 148, 153, 160, 167, 185–7,
hypermetric, 328 200, 218, 235, 248–9, 255, 276, 284, 289,
294, 306, 308–10, 314–15, 346–50, 356
iamb, 324, 350 Latin borrowing
iambic pentameter, 14, 313, 323, 356–9, Latin see loanwords
361–3 Latin Stress Rule, 307–8, 315
iambic reversal see trochaic substitution Latino/Chicano English, 17
-lC-loss, 130 lengthening before /-lC/, 241
ictic, 324, 328, 330, 332–3, 335, 342, 344, 350 lenition, 50, 83, 103–4, 106, 113–14, 149, 205
ictus (ictic), 295, 323–4, 327–30, 332, 335, Levins, Peter, 293
337, 344, 358, 364–5 lexical frequency, 124, 136, 253, 272
I-Mutation/I-Umlaut, 73, 85, 153, 156–64, lexicon, 44, 61, 71, 79, 89, 133, 157, 182, 210,
172–3, 176, 193 212, 217, 231, 255, 287, 308, 314–15
Indian English, 18 lift, 328–30, 344
Indo-European, 2, 54–9, 61–2, 67–9, 74, 79, ‘linking’ /r/, 127
180 liquid, 30
inflectional suffixes, 182, 295 loans, 7, 12, 48, 58, 65, 84, 87, 92, 96–7, 107,
International Phonetic Alphabet (IPA), 25, 139, 141, 193, 194–5, 218, 231, 263,
28, 30, 33–6, 38, 41, 46, 69, 82, 103, 152, 289–90, 308, 310–11, 314, 316
154, 158, 222, 247, 251, 258, 276, 286 early loans, 217
vowel chart, 33 loanword adaptation, 13
vowel symbols, 36 loanwords, 4–5, 12–13, 15, 32, 66, 70, 92–3,
interrogative, 109 98, 100, 107, 111, 135, 137, 141, 148,
intrusive r, 127 159, 195, 197, 200, 202, 209, 216, 218,
inverse spelling, 135, 167 226, 241, 243, 245, 252, 255, 270, 276,
-ir, -er, -ur, 277 280, 289–90, 293, 306–7, 309–11,
Ireland, 4–5, 16, 127, 243, 246, 280–1 314–15, 317, 348, 360
Irish, 4, 16, 20, 111, 125, 131, 147, 240, Classical, 15
245–6, 259, 277, 283 Greek, 57
Irish English, 4 Romance see Romance vocabulary
SUBJECT INDE X 421
London, 9, 16, 19, 63, 116, 122, 134, 146, obstruent, 30–1, 61, 64, 67, 90, 101–2, 135,
198, 202, 234, 238, 254, 259, 267, 270, 142–3, 145, 169, 241, 272
275, 307 octosyllabic, 349–50, 356–7
loss of [-n], 146 OE Breaking, 116–18, 129
Lothian, 3 OFr <-an>, 241
Love’s Labour’s Lost, 131 [oi], 191, 209, 269–70
Lowland Scots, 116–17, 185, 190, 275 Old Norse, 5–8, 10, 21, 53, 60, 65, 72, 83–7,
103, 109, 122, 167–8, 172, 177–8, 192,
Machyn, 125 203, 207–8, 223, 228, 239, 242, 262–3,
macron, 33, 45, 155, 156, 174, 180, 256 267, 303, 338
major-class words, 94 onomastic data, 219
markedness, 64, 197, 198 onset
Maximal Onset Principle, 41, 287 Onset Maximalism see Maximal Onset
MB, 297, 335 Principle
merger, which-witch, 111 onset-maximal syllabification, 330
metathesis, 48, 86, 119–20, 130, 275 open-syllable lengthening, 199, 221–4, 226,
metrical feet, 324, 350 242, 251, 261, 263
metrical positions, 285, 295, 305, 309, 313, Orm, 80, 154, 167, 189, 225, 290, 349, 352,
324–5, 336, 343–4, 350, 352, 359, 362–3 365
metricality, 326 Ormulum, 215, 225, 230, 349, 351–2
Milton, 232, 323, 365 orthoepist, 23, 126, 146, 234–5, 239, 267
Milton, John, 239 orthographic evidence for word-stress, 289
minimal pairs, 77–8, 80, 89, 92–3, 136, 179, orthographic geminates, 226
246, 296 orthographic standard, 18
mismatches, 326 Owl and the Nightingale, 230, 347–51
bracketing, 364–5
Mondegreen, 145 ɔi, 35, 38, 45, 50, 58, 151, 268, 270
monophthong, 37–8, 157–8, 201, 254, 263,
267 pagan, 4–5
monophthongisation, 153, 176, 241, 259 palatal assimilation, 142
Monosyllabic Rule, 352 palatalisation, 74, 81, 83–8, 99, 141–5
mora, 43–4, 170–1, 180, 285 eModE, 143
moraic palato-alveolar, 28
bi, 43, 45, 168, 170–1, 177, 208–9, 214, 286 Paradise Lost, 232
mono, 43, 45 paraphonology, 337, 350, 353–5
morphemes, 49, 51–2, 83, 89, 318 Paston Letters, 143, 264
morphophonemic, 52, 89, 160, 163 Pearl, 95
Mulcaster, 226, 235 pentameter, 293, 307, 325, 352, 357–62, 365
periodisation, 9
nativisation, 13, 92, 142 peripherality, 36, 151
New Zealand, 17, 100–11, 127, 148, 198 Persian, 55
New Zealand English, 17 Peterborough Chronicle, 185, 230
-ŋg, 41, 49–50, 75, 136–9, 165, 192 phonaesthemes, 149
[ŋ], 29, 41, 50, 74–5, 99, 136, 138–9 phoneme, 21, 24–5, 28, 31, 40, 45, 50, 83,
non-ictic, 324, 328–9, 342 104, 138, 149, 180, 199, 200, 236, 246,
non-isosyllabic, 325 270, 333
Norman phonemic merger, 50–1
Conquest, 9–10, 12–14, 82, 106, 121, 130, phonemic split, 50, 91
182, 184, 185, 190, 215, 306, 308, 339, phonological phrase, 298, 300
340, 346, 348, 351 phonotactics, 99, 134, 140, 237, 271
Norman French see Anglo-Norman phrasal stress, 312, 314
Northumbria, 4–5 Phoenix, 297
-ns, 241 Piers Plowman, 132, 134, 340, 344
Nuclear Stress Rule see End Rule plosive. See stop
NZE, 126, 129, 147, 237–8, 259, 266 Poema Morale, 130, 230, 348–9
poetic meter, 284–5, 323–4 sandhi, 126, 128, 132

post-ME borrowings, 255 Sanskrit, 55–6, 245
pre-consonantal shortening, 168, 212–15, Scandinavia, 5, 10, 59
217, 225, 276 Scandinavian, 5–8, 10, 60, 67, 87, 208,
pre-fricative lengthening, 240 263
prestige, 235, 269; see also stigmatisation schwa, 34–5, 37, 50, 69, 93, 120, 124, 127–9,
Prick of Conscience, 248 132, 139, 176, 180, 182–3, 223, 228–31,
pro(s)thesis, 48 233, 248, 275–8, 280, 309, 338, 349,
pronominal word, 94 354–5; see also -@
pronunciation norm, 153 in word-final position, 230–1
prosody, 4, 12, 14–15, 18, 39, 44, 51–3, 55, loss in trisyllabic words, 355
58, 61, 66, 69–70, 106, 112, 130, 145, Scotland, 5, 16, 83, 100, 110–11, 127, 149,
152, 160, 164, 170–1, 181, 211, 216–20, 246, 267, 280–1, 340
223, 230, 232–3, 236, 270, 284–5, 288–9, Scots, 16, 86, 116, 131, 147, 194, 203, 240,
292–6, 298, 300–3, 306–8, 311–12, 314, 243, 259, 274–5, 277, 282–3
317–18, 320–4, 326–7, 331–2, 337–8, semivowel see glide
345–6, 353–5, 360–6 septenary, 349
weight, 285 Shakespeare, William, 22, 50, 161, 224, 239,
Proto-Indo-European (PIE), 56, 61–5, 67–9, 262, 277, 314, 320, 323–4, 365
129 Sheridan, Thomas, 144, 267, 316
accent, 65 shibboleth, 100, 108
Proverbs of Alfred, 339, 347 Shift
push-chain theory, 253 First Germanic Consonant Shift, 61–6; see
also Grimm’s Law
Received Pronunciation (RP), 18–19, 36, Great Vowel Shift, 201, 248, 250–1, 256–7,
116, 118, 129, 134, 146, 153, 198–9, 206, 260
238–9, 248, 274 Long Vowel Shift, 38, 124, 175, 203, 207,
reduplication, 149 226, 241–2, 252–3, 255–6, 257–8, 262,
Renaissance, 144, 217–18, 284 266–7, 269, 271, 280, 290, 306
Greek, 141 Northern Cities, 35, 203, 237–8, 248
Latin, 107 Sievers’ Five Types, 334–5
resolution, 77, 329 Type A3, 335
resources, 8, 14, 45, 306 Sievers’ Rule of Precedence, 301; see also
for the study of Middle English, 14 335
r-ful. See rhoticity Type A3, 334
rhotacism, 66, 121 silent -e, 226
rhotic silent words, 230
non-rhotic. See r-less similarity
rhoticity, 99, 115–17, 121, 125, 128, 274 avoidance, 136
hyper-rhoticity, 124, 126 simplification, 40, 49, 80, 99, 109, 122–4,
rhyme, 22, 95, 249 132–6, 139, 141, 148–9, 167, 173, 197
feminine rhyme, 357 singleton, 40, 42, 75, 77, 80, 93, 104, 118, 143,
Rhyming Poem, 346 170, 214, 330, 340
rhythm, 285 Singletons, 74, 78
r-less see non-rhotic Sir Gawain and the Green Knight, 143, 303–4,
Roman, 3–4, 20–1, 36, 106, 115, 146, 153, 248 313, 340–1, 344
Romance vocabulary, 12, 84, 97, 106–7, 141, Sir William Jones, 55
148, 208, 216, 218, 241, 306, 309–10, 348 [sk], 86–7
Romans, 1, 3, 55 slashed o, 158
rounding, 239 sociolinguistic, 47, 52, 111–12, 253, 262,
/-rðər, -rð.r/ > /rd.r/, 125 267, 322
rune, 20, 59, 76, 116, 154 sonority, 31, 39, 40, 50, 95, 106, 119, 121, 123,
128, 131, 133–5, 169, 181, 232, 287
SAE, 40, 237 falling, 40, 135, 213
St Augustine, 106 rising, 40
SUBJECT INDE X 423
sonorant, 30–1, 34, 48, 50, 69, 75, 83–4, 91, length, 44
101, 103, 113, 121, 129, 132, 169, 268, loss of coda, 275
338, 354 nucleus, 39
sonority scale, 31 onset, 32, 38–42, 48, 72–3, 76, 85, 91,
South Africa, 18, 60, 100, 127 101–2, 104–6, 108–9, 112–13, 115, 118,
Spanish, 17–18, 55, 58, 115, 248 120–1, 124–6, 128, 133–6, 139, 143,
Spenser, Edmund, 232 145–6, 148–9, 165, 222, 231, 239–40,
spirants see fricative 246, 268, 287, 297, 326–7, 330, 333–4,
spoken standard, 18 340–1, 354
spread glottis, 31, 112 open, 44
-ss, 226 peak, 37–9, 43, 113, 135, 222, 338
SSBE, 16–18, 20, 28, 34–7, 40, 44, 106, 126, rhyme, 39
128, 137, 139, 146, 151–2, 173, 184, 225, stressed, 34, 40, 44–5, 71, 85, 102, 126,
231, 237–8, 240, 242, 248, 252, 265, 270, 138, 145–6, 152, 165, 170, 210, 212, 216,
278, 282 220, 227, 278, 289, 295, 305, 324–8, 330,
SSE, 38, 100, 129, 173, 245, 280 341, 351, 360, 362
Standard English, 18–19, 174, 184, 257 structure, 39, 286
standard written, 14 superheavy, 170, 287
standardisation, 52, 107, 140, 195, 236, 267 vowel-initial, 29, 146
stigmatisation, 242, 244 weight, 43–4, 210
stigmatised, 106–8, 114, 131, 146 syllable structure, 1, 39, 41, 43, 120, 123,
stop, 30; see also plosive 211–12, 219, 284, 287, 366
aspirated, 27, 40, 62–4, 67 syllable(s)
glottal, 24–5, 27, 29, 62, 74, 76, 99, 102–5, unstressed, 9, 34, 45, 66, 70–1, 85, 93, 102,
112, 145–6, 205–6, 258, 334, 341 106, 115, 121, 124–5, 136–7, 143, 181–3,
palatal, 28 210, 212, 218–20, 227–9, 231–2, 285–6,
unreleased, 149 289–91, 295, 313, 320, 324–5, 330, 332,
velar, 28 336, 339, 344, 351, 354–5, 358, 360, 362
stress, 1, 13, 15, 22, 34, 39–40, 42, 44, 49, syncope, 49, 80, 182, 223, 228, 231–2, 350,
51, 58, 65–6, 69–70, 91, 93, 95, 107, 353–5
115, 142, 145, 170, 180–1, 183, 210–12, synizesis, 354, 358
214–15, 218–19, 220, 222, 224, 227, 232,
249, 255, 281, 283–7, 288–313, 315–25, ʃ, 28, 75, 132, 142–3, 225
327–8, 331, 335, 337, 345, 348, 352–5, ʃ-3 contrast, 142
359–60, 362–6 ʃl-, 53, 149
Germanic, 5 ʃm-, 53, 149
non-primary, 286, 295 ʃm- ʃl- ʃt-, 60, 148
root-initial, 58, 353
secondary and tertiary, 286 tap, 25, 99, 145, 147–8
stress mismatches, 360 Tautosyllabic /-r/, 244
stress-attracting suffixes, 319 tautosyllabicity, 75, 118, 166, 169, 177, 200,
stress-neutral suffixes, 318–19 211, 238, 241, 244, 264, 278
stress-shift, 294, 307, 316 Tennyson, Alfred Lord, 137
suprasegmental, 284 tetrameter, 325–6, 356–7, 359
sw-, 139 The Alliterative Morte Arthure, 340, 341
syllabic sonorant, 39, 222, 287 The Pickwick Papers, 96
syllable, 39 The Sowdone of Babylon, 137–8
closed, 44 The Wars of Alexander, 340–1, 344
coda, 39–44, 70, 84–5, 92, 112–13, 117–18, thorn, 20, 82, 90
120–1, 123–8, 130–1, 134–40, 146, 149, [tj] ~ [tʃ], 144
165–7, 170, 176, 178, 192, 198, 205–6, Tolkien, J.R.R., 96
208, 211, 213–14, 216, 220, 223, 229, Townley Plays, 264
231–2, 237–40, 244, 272, 275, 277–82, trisyllabic shortening, 216–18, 227
287, 338, 341 trisyllabic words, 212, 229–30, 289, 307–8,
definition, 39 310, 355
trochaic substitution, 360–1, 364 neutralisation, 9

trochee, 324, 350, 361 quality, 32
tʃ, 74, 77, 81–2, 85–6, 88, 122, 144, 226 reduction, 9, 165, 180–2, 212, 219–20, 228,
spelling, 86 286, 289, 317
ts-, 42, 53, 148–9 round, 33
Tudor, 15 tense, 35–6, 38, 151, 166, 179, 209, 378
typology, 44, 47, 196, 264, 282, 294, 336 unround, 33
vowel alliteration, 334
[-uγ-], 208 vowel harmony, 157
[ui], 187, 190–1, 193, 195, 269–70 Vowel shortening in unstressed syllables,
Unstressed syllables, 9, 34, 41, 45, 48, 219
66, 70–1, 80, etc.; see also syllable(s), vulgar, 108, 114, 126, 242, 244, 269, 278, 299
unstressed
Unstressed vowels, 49, 51, 180, 232; see also w-, 109, 111–12, 115, 134, 140, 240, 245,
schwa, vowel reduction 299
Wales, 3, 16, 147, 198, 244, 248
Vanburgh, 243 Walker, John, 23, 99, 107–8, 135, 143–4, 240,
variable stress, 316 244, 246, 255–6, 262, 267, 293, 304
-ve, 226 Webster, Noah, 17
vernacular, 10–12, 14, 185, 187 Welsh, 4, 16, 58, 111, 233, 246, 259
vernaculars, 55 Welsh English, 4
Vernay Letters, 126 West Germanic (Consonant) Gemination,
Verner’s Law, 66, 70, 116 71–3, 76–9
in English, 66 Whitby, Synod of, 4
verse, 22 Winchester, 8, 179, 184
Old English, 5, 41, 73, 77, 86, 105, 109, Winner and Waster, 340, 344
230, 291, 295, 297–8, 300–2, 328, 330, word-minimality, 71
332–4, 337, 339–41, 345, 350 Wordsworth, William, 137, 323, 325
versification, 13–14, 22, 283, 285, 290, 323, written standard, 14, 18–19, 185, 187, 267
326–8, 339–40, 345, 348–50, 353, 356 wug test, 97
Vices and Virtues, 260 Wulfila, 60
vocal tract, 25–7, 29–32, 50, 101, 105, 116 Wulfstan, 19, 342
vocalisation, 73, 114, 126–7, 131, 165, 177–8, Wycliffite texts, 187
205–6, 208, 277, 280 wynn, 20, 76, 190
voicing, 26, 47, 91, 147
vowel y, 37–8, 152, 154–5, 173, 187, 193–8, 236,
backness, 32, 36 238, 245, 268
checked, 44 Yiddish, 59–60, 148–9
free, 44 yod-dropping, 197
height, 32, 34, 36 yogh, 82
lax, 35–6, 44, 70, 151, 171, 209, 219, 255,
272 [z] ~ [], 142
length, 35, 212 //, 28, 41, 50, 53, 99, 141–3

A Historical Phonology of English

Uploaded by

Copyright:

Available Formats

A Historical Phonology of English

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A Historical Phonology of English

Uploaded by

Copyright:

Available Formats

A Historical

EDINBURGH TEXTBOOKS ON THE ENGLISH LANGUAGE – ADVANCED

TITLES IN THE SERIES INCLUDE :

Corpus Linguistics and the Description of English

A Historical Phonology of English

A Historical Morphology of English

Grammaticalization and the History of English

A Historical Syntax of English

English Historical Sociolinguistics

A Historical Semantics of English

Visit the Edinburgh Textbooks in the English Language website at www.

Edinburgh University Press Ltd

Typeset in 10.5/12 Janson by

ISBN 978 0 7486 3467 5 (hardback)

The right of Donka Minkova

3.2 The Indo-European family of languages 55

5.2.4 Post-vocalic /r/-loss 125

7.2 Notes on ME spelling (vowels): <TAXI>-<ATAXY>,

8 Vowel quality and quantity in Early Modern English and later

8.3.2 Vowels in relation to /r/: PERSON-PARSON, TEAR,

AAVE African American Vernacular English

A Companion to this book is available at http://www.euppublishing.

This book is about the evolution of the phonological structure of

transplants. In writing, the earliest record of the noun English is found in

1.1 Periods in the history of English

1.2 Old English (450–1066)

transfer of phonological and prosodic features, though the latter are

Phonologically, the early loanwords were commonly assimilated to

usually referred to them as Danes, although Norwegian presence is also

local population, thus putting down permanent roots in England. Most

Scandinavian and late OE/early ME frequently shared the Proto-

1.3 Middle English (1066–1476)

occupation of Britain by speakers of Norman French were far-reaching.

was never very high; it is estimated that it ranged, roughly, between

(2) Replacements of Old English words with loanwords from AN/

As we will see in Chapters 5 and 9, the loanwords from AN/OFr often

syllable-counting and end-rhyme, properties characteristic of contem-

seventy-five years of lexicographical work. It is an indispensable

1.4 Early Modern English (1476–1776)

Among the social and demographic factors influencing the develop-

1.5 English after 1776

widely used native languages: only Mandarin Chinese, Hindi and

since we are limiting our scope to phonology, we can talk of ‘standard

1.6 The evidence for early pronunciation

the Carolingian script in ME. Throughout the history of the language

are the physically testable principles of production and perception, both

contemporary phonological changes. The discrepancies between spell-

: Suggested further reading on Companion website.

2.1 The consonants of PDE

Unlike /v/, the glottal stop is not a phoneme in English; it is an allo-

1. languages have more consonants than vowels;

upper hard palate

The characterisation of consonants is traditionally done with reference

reference to laryngeal or glottal specifications. Another related prop-

2.1.2 Place of articulation

Another useful descriptive parameter refers to the active articulators in

2.1.3 Manner of articulation

Labial Labio- Dental Alveolar Palatal Velar Glottal

AFFRICATES tʃ d

FRICATIVES () f v θ ð s z ʃ (x) h

Upper mid ei (FACE) / (NURSE) oυ (GOAT)

Lower mid ε (DRESS) (STRUT) ɔ (THOUGHT)

The upper mid central symbols // (GA) or /