THE SOUND OF INDO-EUROPEAN
Phonetics, Phonemics, and Morphophonemics
Edited by
Benedicte Nielsen Whitehead
Thomas Olander
Birgit Anette Olsen
and
Jens Elmegård Rasmussen
Museum Tusculanum Press
University of Copenhagen
2012
@ Museum Tusculanum Press and the author 2012
he Sound of Indo-European: Phonetics, Phonemics, and
Morphophonemics
© Museum Tusculanum Press and the authors 2012
Edited by Benedicte Nielsen Whitehead, homas Olander,
Birgit Anette Olsen & Jens Elmegård Rasmussen
Cover design by hora Fisker
Set by homas Olander
Printed in Denmark by Specialtrykkeriet
ISBN 978 87 635 3838 1
Copenhagen Studies in Indo-European, vol. 4
ISSN 1399 5308
Published with support from:
Roots of Europe – Language, Culture, and Migrations
Museum Tusculanum Press
126 Njalsgade
DK 2300 Copenhagen S
www.mtp.dk
@ Museum Tusculanum Press and the author 2012
CONTENTS
Preface
iX
Václav Blažek
Indo-European laryngeals in the light of Afroasiatic
1
Lars Brink
he etymology of Nordic kuna ‘woman’
27
Andrew Miles Byrd
Predicting Indo-European syllabification through phonotactic
analysis
33
Paul S. Cohen & Adam Hyllested
A new sound law of PIE: Initial **h3uʢ > *h2uʢ
53
Michael Frotscher
he fate of PIE final *-rʘ in Vedic and Latin
73
José Virgilio García Trabazo
Phonologische und morphologische Bemerkungen zu den
hethitischen -e/a-Verben
97
Piotr Gąsiorowski
he Germanic reflexes of PIE *-sr- in the context of Verner’s Law
117
Aaron Griffith
Non-raising before *μ in Old Irish
129
Irén Hegedűs
he RUKI-rule in Nuristani
145
Eugen Hill
Hidden sound laws in the inflectional morphology of
Proto-Indo-European
169
Anders Richardt Jørgensen
Palatalization of *sk in British Celtic
209
Götz Keydana
Evidence for non-linear phonological structure in Indo-European:
he case of fricative clusters
223
@ Museum Tusculanum Press and the author 2012
vi
Contents
Alwin Kloekhorst
he phonological interpretation of plene and non-plene spelled e in
Hittite
243
Guus Kroonen
Consonant gradation in the Germanic iterative verbs
263
Martin Joachim Kümmel
Typology and reconstruction
291
Rosemarie Lühr & Susanne Zeilfelder
Optimale Onsets im Indogermanischen
331
Paolo Milizia
On the morphophonemics of Proto-Indo-European *-s e/opresents
361
Kanehiro Nishimura
Vowel reduction and deletion in Sabellic: A synchronic and
diachronic interface
381
Georges-Jean Pinault
Remarks on PIE amphikinetic and hysterokinetic nouns
399
Charles Prescott
Germanic and the ruki dialects
425
Giancarlo Schirru
Laryngeal features of Armenian dialects
435
Vitaly Shevoroshkin
Anatolian laryngeals in Milyan
459
Zsolt Simon
PIE ‘me’ and a new Lydian sound law
485
Thomas Smitherman
On ancient Kartvelian-Indo-European lexical contacts and their
consequences for Proto-Indo-European
501
David Stifter
Lenition of s in Gaulish?
523
@ Museum Tusculanum Press and the author 2012
Contents
vii
Brent Vine
PIE mobile accent in Italic: Further evidence
545
Gordon Whittaker
Euphratic: A phonological sketch
577
Paul Widmer
Notiz zur holokinetischen Ablautklasse
607
Nicholas Zair
A new environment for laryngeal loss in Proto-Celtic
613
@ Museum Tusculanum Press and the author 2012
@ Museum Tusculanum Press and the author 2012
Typology and reconstruction
he consonants and vowels of Proto-Indo-European1
Martin Joachim Kümmel
University of Freiburg
When we reconstruct a proto-language, we produce a hypothesis about a
non-attested synchronic stage and about the changes leading from it to the
attested languages. his means that we should evaluate both the synchronic
stage and the reconstructed changes with the help of general and typological considerations.
In this paper some problems of phonological reconstruction in PIE are
discussed from a typological perspective. First, one of the most controversial topics of PIE consonantism is addressed: the reconstruction of the stop
system (i.e., the “glottalic” question). Ater an evaluation of diferent hypotheses from both synchronic and diachronic typological data, it is argued that the best solution might be to reconstruct pre-PIE implosives (i.e.,
non-explosive non-glottalic stops) and voiced explosives that changed to
voiced explosives and breathy voiced stops in PIE or at least in the central
IE languages.
In the second part, the focus lies on the reconstruction of qualitative
ablaut in vowels. In the light of typological parallels it is argued that both
later reflexes of the vowels and some morphonological rules might be easier to understand if PIE *o was the reflex of a pre-PIE long *ā in contrast to
PIE *e/a resulting from pre-PIE short *a.
1
For reasons of space, I had to concentrate on the two major topics of the oral
presentation and leave my remarks on tectals and laryngeals for future publications. I was very pleased to learn in July 2009 that Michael Weiss had made a
very similar proposal for the reconstruction of the (pre-)PIE stops (Weiss
2009).
@ Museum Tusculanum Press and the author 2012
292
1
Martin Joachim Kümmel
Introduction
Linguistic reconstruction means to make a reasonable hypothesis about a
preceding stage of the history of a language. A reasonable hypothesis is
capable of explaining the developments that led to the later, attested stages
of the languages in question. More specifically, this means that the hypothesis should set up a probable, or at least possible, prestage, and possible, or preferably probable, changes. heoretically there are two ways of
evaluating the probability of the assumed synchronic state and of the assumed diachronic changes: (1) heoretical and/or universal facts and constraints that allow us to exclude some of the alternatives, and (2) typological parallels or even trends that allow us to judge one possibility more
probable than another. To make a reconstruction probable, it should be
corroborated by both of these evaluations – to make it possible, the first
must not exclude it, and the second should provide at least one parallel.
Bearing this in mind, we shall reconsider the reconstructed sound system of Proto-Indo-European (PIE), discussing the stop system and the
vowels. As to the term PIE, it is defined as the protolanguage of all IndoEuropean languages as it can be reconstructed from the comparative evidence. he immediate result of comparative reconstruction is called late
PIE, earlier stages reached by internal reconstruction may be labelled early
PIE. Common IE refers to developments that might have taken place in a
parallel fashion in all branches of IE without necessarily being of PIE date,
as, e.g., the loss of *h1 with compensatory lengthening.
2
he reconstruction of IE consonants: Stops
From a typological perspective, the modern reconstruction of the PIE system of stops is not very problematic as far as the traditional mediae or
tenues are concerned: their opposition is typologically well attested. It is
the mediae aspiratae that are typologically problematic, and seen in relation to other distributional peculiarities this has led to a number of reinterpretations of the traditional mediae as “glottalic” sounds.
In Table 1 I give an overview of four diferent “glottalic” models compared with the neo-traditional “non-glottalic” model. In addition to replacing plain voiced stops by a glottalized series, the former partly deviate
@ Museum Tusculanum Press and the author 2012
Typology and reconstruction
293
from the latter by the assumption of aspirated instead of plain voiceless
stops and/or the assumption of non-aspirated or voiceless stops instead of
mediae aspiratae.
Table 1
T
*t
*dȺ
*d
H
*t
*dȺ/d
*t’/
Reflexes of PIE plosive types, exemplified by dentals
G
*tȹ~t
*dȺ~d
*t’
K
*t
*dʘȹ
*dʣ [ɉdʘ]
V
*tȹ
*dʘ
*t’
An
t/t·
d/dʘ
d/dʘ
To
t
t,ts<*dȺ
ts<*d
IIr
t,tȹ
dȺ/d
d
Arm
tȹ
d/dȺ
t/t’
Gr
t
tȹ
d
It
t
f/ð
d
Celt
t/tȹ
d
d
Germ
θ
d/ð
t/tȹ/ts
BS
t
d
əd
Alb
t
d
d
(T = “neo-traditional”; H = Hopper 1973/1977; G = Gamkrelidze & Ivanov 1973; K
= Andreev 1957; Kortlandt 1978, 1985; V = Normier 1977, Vennemann 1984)
Kortlandt’s “preglottalized lenis” may be interpreted as “creaky voiced” (Ladefoged
& Maddieson 1996: 53f.), “voiced laryngealized” (Maddieson 1984: 111f.), or
“voiceless implosive” (Ladefoged & Maddieson 1996: 87–90), all of which may
represent one phonological class (vgl. Clements & Osu 2002: 313); on the relics of
voicing and aspiration in Tocharian cf. Ringe 1996: 47f., 64f.; for Italic see StuartSmith 2004
Now, what is the basis for reconstructing these PIE series of stops? hree
main types of correspondences between the individual IE branches can be
observed: In one case, the main reflexes are plain voiceless or voiceless aspirated, traditionally reconstructed as plain voiceless stops (tenues). In the
second case, we mostly find voicing and/or aspiration, and therefore
voiced aspirates (mediae aspiratae) were reconstructed. Here I have to
comment on the notion of “voiced aspirates”. Phonetically, in IE languages
these sounds are neither voiced nor aspirated; they are articulated with a
diferent phonation type called “breathy voice” (meaning the same as
“murmured”, cf. Ladefoged & Maddieson 57f.) – but phonologically, they
may, but need not, count as voiced and/or aspirated. In the third case, the
reflex is normally voiced, but it may be voiceless in languages where the
tenues are distinctly aspirated, and this series was reconstructed as plain
voiced (mediae). hus we arrive at a system with one plain voiceless or
“fortis” series, but two voiced (or “lenis”) series.
@ Museum Tusculanum Press and the author 2012
294
2.1.1
Martin Joachim Kümmel
Synchronic typology
Against this kind of system a number of typological objections have been
raised:
1 According to Hopper (1973: 141) “a typologically plausible triple stop
system should have only one voiced series”. Is this true?
No. Even if most types of systems with two voiced series are quite rare, one
such type is rather frequent: about 16% of the three-series languages
counted by Maddieson (1984: 28f.) exhibit a system of voiceless stops opposed to plain voiced explosives and voiced implosive stops. In this sample, implosives were not distinguished from laryngealized voiced stops (see
Clements & Rialland 2005: 19; Hamann & Fuchs 2008: 104f.). Recent research has provided evidence that “implosives” should rather be defined as
“non-explosive” or “non-obstruent” stops – ingressive airstream being only
a secondary feature (see Stewart 1989: 231f.; Clements & Osu 2002; 2005;
Clements & Rialland 2005: 17f.) – and that “laryngealized voiced stops”
are “non-explosives” with distinctive glottalization and less voicing, something like [ɉɗʘ] (Clements & Rialland 2005: 19f.). But even if this distinction is made, the type remains rather frequent (if only in Africa).
he parallel to PIE is not perfect, however, since the type does not show
two voiced explosive series, as traditional Indo-European is assumed to do.
But changes from non-explosives to explosives seem to be rather common
(cf. Stewart 1989: 236f.; Clements & Rialland 2005: 20).
2 According to Jakobson (1958: 22f.), voiced aspirates never occur without contrasting voiceless aspirates (and an h-sound). Is this true?
Not entirely: here is at least one language that exhibits “voiced aspirates”
without possessing voiceless aspirates, namely Kelabit in Northern Borneo
(cf. Blust 1974; 2006). In fact, the voiced aspirates in this language have an
even better claim to be called “voiced aspirates” than the sounds attested in
IE languages, since the latter are breathy voiced [b̤Ⱥ], while the Kelabit aspirates are not really murmured, but rather described as beginning voiced
and ending in voiceless aspiration, [bpȹ] (Blust 1974: 50, cf. Ladefoged &
Maddieson 1996: 62f.). But still, these diferent phones might be considered as belonging to one phonological class, combining “voicing” and “aspiration”.
@ Museum Tusculanum Press and the author 2012
Typology and reconstruction
295
All this means that the (neo-)traditionally reconstructed system of PIE
is improbable, but certainly not impossible, and we should be cautious not
to normalize data without a compelling reason (see Kortlandt 1985: 185,
following Dunkel 1981: 566, cf. Haider 1983: 81). But of course we may try,
so let us test the diferent models of PIE: Which of them are better attested
and therefore more probable?
In what follows, I give an overview of three-stop series systems and
their frequency in languages of the world (as available in the UPSID database, see Maddieson 1984: 28f.). Even if some caution is at place when using this database containing information of quite diferent quality, it
should be suicient for tracing more general trends. Parallels to systems
reconstructed for PIE are indicated by the abbreviations used in table 1
above.
2.1.1.1
hree-stop system types cross-linguistically
a Frequent types
• tȹ~ t ~ d (Romani, Panjabi, Shina, Kashmiri; Khotanese? (see Kümmel 2007: 289–294, 441), Northern Kurdish; Agulis/Sasun Armenian; hai; Burmese; Burushaski …)
• H' = t ~ d ~ t’ (Ossetic; South Arabian, Ethiopic; Dizi, Kefa; Tlingit;
Haida; Chontal, Tzeltal; Quileute, Puget Sound, Pomo; GununaKena …)
• V = tȹ ~ t ~ t’ (Artvin Armenian; Georgian; Haida, Navaho,
Chipewa, Tolowa, Hupa; Wichita; Quechua, Jaqaru …)
• V' = tȹ ~ d ~ t’ (Kabardian; Klamath; Kwakw’ala, Yana, Acoma …)
NB: he last 3 types are rather similar to each other, their diferences only
depending on which features of the non-ejective stops is taken to be phonologically most relevant. E.g., Ossetic, Georgian, and Kabardian have
pretty much the same system, the most adequate interpretation probably
being the last one (since there can be voicing assimilation).
• K' = t ~ d ~ ɗ (Katcha, Kadugli, Kpelle, Dan, Ogbia, Tarok, Doayo;
Kohumono; Tama, Mursi, Daju; Angas, Margi, Dangaleat; *Protohai; with ɓ only: Aizi, Bete, Gwari, Kpelle; Mumuye)
@ Museum Tusculanum Press and the author 2012
296
Martin Joachim Kümmel
b Rare types
•
•
•
•
•
•
•
•
•
•
•
•
tȹ ~ t ~ ɗ (Swahili; Vietnamese, Khmer, Khmuʔ; Karen, Phlong)
K' = t ~ d ~ dʣ (Lugbara; Kera, Lame?, Kanakuru?)
t ~ d ~ t’/ɗ (Ik; Iraqw, Hausa; Hamer)
tȹ ~ d ~ dʣ (Somali; Wapishana)
tȹ ~ t ~ dʣ (Lakkia, Lungchow; Cham)
tȹ ~ dʘ/t ~ (Korean; Tol)
tȹ ~ t ~ nd (Hakka; Nambakaengo)
t ~ d ~ nd (Sinhalese, Divehi)
tȹ ~ t ~ d̤ (Erevan Armenian; Xhosa; *Old Chinese?)
tȹ ~ d̤ ~ d (Sivas Armenian)
tȹ ~ t’ ~ ɗ (Maidu)
T = t ~ dȹ ~ d (Kelabit)
c Systems not (yet) attested
• G = tȹ/t ~ d̤/d ~ t’
• H = t ~ d̤ ~ t’
• K = t ~ d̤/dʘȹ ~ dʣ
his overview shows that some proposed “glottalic” systems are not attested at all if all phonetic specifications are taken into account: there is no
system with breathy voice and ejectives, as in the two earliest “glottalic”
proposals, nor is there a system with laryngealized lenes and lenis aspirates, as originally proposed by Kortlandt. Variants of these proposals,
however, do much better: if breathy voice was post-PIE, all “glottalic” systems correspond to system types attested by more than one known language, while “non-glottalic” PIE does not.
We have to conclude that “glottalic” reconstructions of PIE do in fact
provide a more probable synchronic system. But is this synchronic plausibility matched by a corresponding diachronic plausibility?
2.1.2
Diachronic typology
1 he traditional tenues are represented as follows: tȹ 3, t 7 = [-voice] all;
[+asp] 3, [-asp] 7
he most probable alternatives would be to reconstruct voiceless unaspirated stops or voiceless aspirated stops. In the first case, we would have to
@ Museum Tusculanum Press and the author 2012
Typology and reconstruction
297
assume secondary aspiration and (possibly secondary) fricativization in a
minority of branches – changes that are quite frequent and assured for
younger stages of IE languages from diferent branches (see Kümmel 2007:
168f.). In the second, we would have to assume secondary deaspiration in
most branches; this is certainly possible, but not very well attested (Job
1989: 128f.; Kümmel 2007: 93f.), and it should thus be judged less probable.
Summing up: a system with unaspirated voiceless stops as predecessors of
the tenues is most probable from a diachronic viewpoint. his means that
Hopper’s and Kortlandt’s models are as good as the traditional model,
while Vennemann’s and Gamkrelidze’s are not favoured.
2 he traditional aspiratae show a more complicated picture: tȹ 2, dȺ 2,
d/dȺ 1, d/ð 1, d 4 = [+voice] 8, [-voice] 2 / [+asp] 5, [-asp] 5
he majority of Indo-European languages exhibit voiced plosives. Aspiration is directly attested in two branches, and at least two additional
branches show some traces of it: in Modern Armenian we find plain voiced
stops, breathy voiced stops, voiceless lenis or fortis stops, and voiceless aspirated stops; therefore, breathy voiced stops have to be reconstructed for
previous stages of quite a number of modern dialects (though it has been
assumed that this was a post-Old-Armenian innovation from simple
voiced stops). In Italic, we find voiceless fricatives in initial position, but
voiced ones in internal position. However, internal voicing might be secondary, so voiceless fricatives might be the primary reflex, and these might
go back to voiceless aspirates, as in Greek. In Tocharian, there is evidence
for earlier aspiration, leading to a plain voiced reflex before a following
aspirate, so we might add this language to those that attest some kind of
“voiced aspirates”. In Germanic, most dialects show voiced fricatives in
internal position, but stops in initial position, corresponding to voiced or
voiceless stops in High German. Without any doubt, all these reflexes can
easily be derived from the “voiced aspirates” of the traditional model and
the identical “murmured” stops of Hopper’s model or the “allophonically
aspirated” voiced stops of Gamkrelidze’s, but what would the other models
presuppose? For 5 branches we would have to assume a secondary development to breathy voice, which is not too well attested (at least for languages that do not already exhibit an aspiration opposition, cf. Kümmel
2007: 171f.). If we start from original voiceless lenis stops, we would have to
assume unconditional voicing for most branches, and typological evidence
@ Museum Tusculanum Press and the author 2012
298
Martin Joachim Kümmel
for such a change seems to be meagre (Kümmel 2007: 47f.). Models with
voiceless aspirates would do even worse, since it is certainly less probable
for (voiceless) aspirated stops to become unconditionally voiced or breathy
voiced (for conditioned voicing cf. Kümmel 2007: 53). Summing up: the
non-glottalic model is as probable diachronically as all glottalic models
with “voiced aspirates”, i.e., K, H, G, but clearly better than the others.
3 Last, but not least, the traditional mediae show the following reflexes: t
(~t’~tȹ~ȹt~’t) 2, d 8
= [+voice] 8, [-voice] 2 / [-asp] 10 / [+glott] 1–2, [-glott] 8–9
Voiced reflexes are obviously dominant, and only two branches clearly
show voiceless fortis stops. In Armenian, the dialectal reflexes include ejectives, voiceless lenis stops and fortis stops (with a rather long duration). In
Germanic, dialectal reflexes include aspirated stops, preaspirated stops,
preglottalized stops, and africates, though plain voiceless stops are attested
in some regions of almost all sub-branches, and their distribution rather
looks like an archaism (cf. Kümmel 2007: 295). For these two subfamilies,
the traditional model requires the assumption of a “Lautverschiebung”
from voiced to voiceless stops, and evidence from loanwords seems to corroborate this (cf. Rasmussen 1987: 9–12 = 1999: 224–227). Since changes of
that kind are clearly attested in the later history of both IE and non-IE languages (cf. Kümmel 2007: 138f.), they cannot be considered problematic,
though, of course, the preservation of voiceless stops as per most “glottalic”
models would not be a problem either. Although the appearance of glottalic or laryngealized articulation in some dialects would be most easily
accounted for by the “glottalic” models, it might also be secondary. Moreover, for the majority of branches of the IE family most “glottalic” models
would presuppose an unconditional change from voiceless glottalized
stops to plain voiced ones. Such a change is not easy to support by typological parallels, and therefore rather unlikely (cf. Job 1989; Kümmel 2007:
47f., 189f.; conditioned changes caused by lenition or dissimilation are
something diferent). Since it is much more probable for preglottalized
lenis stops (= laryngealized plosives) than for ejectives, we can conclude
again: the phonetic details of Hopper’s, Gamkrelidze’s and Vennemann’s
models are not favoured by diachronic typology, Kortlandt’s are clearly
better, but the traditional model is the most probable.
@ Museum Tusculanum Press and the author 2012
Typology and reconstruction
299
Our conclusion is, therefore: the diachronic typology of systemic developments clearly favours the traditional reconstruction of the plosives as
against all “glottalic” models, but the best of the latter seems to be Kortlandt’s.
Preliminary typological conclusion
Summing up: synchronic typology favours “glottalic” models, but diachronic typology rather contradicts this. his means that typology does
not allow for a clear decision.
2.1.3
Comparative and internal arguments
here have, however, been other, more specific arguments for a glottalic
articulation of the traditional mediae, relying on particular sound laws
involving mediae or their reflexes in individual languages, as proposed in a
series of articles by Kortlandt (conveniently summarized in Kortlandt 1985,
but see also Kortlandt 1978, 1981, 1988, 1997).
2.1.3.1
Comparative evidence for “glottalic” mediae?
1 In Balto-Slavic, an indirect reflex might be provided by Winter’s lengthening before old mediae. he resulting “acute” intonation is partly
reflected by glottalization in Latvian and Žemaitian Lithuanian, which
could be taken as evidence for a preglottalized articulation.
But since the Balto-Slavic outcome of Winter’s law is in no way diferent
from the regular development of long vowels resulting from compensatory
lengthening before lost laryngeals, the “glottalic” efect need not be triggered by the stops themselves, and might be as secondary as the Danish
“stød” resulting from an earlier intonational contrast. he lengthening itself may be explained by stronger voicing of the plain voiced stops as
against the former aspirates (cf. Rasmussen 1987: 99f. = 1999: 233f.; 1992b =
1999: 63–77; Kümmel 2007: 306f.).
2 In Armenian, the dialectal reflexes of voiceless plosives include ejectives, and the various sound shits of the modern Armenian dialects,
might be explained by taking these ejectives as original, since an additional feature is needed anyway.
@ Museum Tusculanum Press and the author 2012
300
Martin Joachim Kümmel
But there is another possibility: Armenian dialectal t can be derived directly from breathy d̤, so that breathy voice might have been the additional
distinguishing feature (see Pisowicz 1976: 45–60; 73; Vaux 1998: 238–241;
Garrett 1998) and we do not need glottalization of the voiceless stops. Recent phonetic investigations suggest that the main diference between the
Armenian “voiced” and “voiceless” stops might be interpreted as slack
voice vs. stif voice (cf. G. Schirru in this volume).
3 In westernmost Indo-Aryan (Sindhi and neighbouring languages), Old
Indo-Aryan unaspirated voiced stops are reflected as “implosives” (injectives) if they were preserved and not lost by internal lenition – i.e., in
initial position and when geminated by Middle Indo-Aryan cluster assimilation (cf. Turner 1924; Allen 1957). In neighbouring Panjabi, initial
voiced aspirates are reflected by plain voiceless stops, but there was no
deaspiration in voiceless aspirates. his has been taken to mean that the
“voiced aspirates” were never aspirated, but distinguished from the
“mediae” by some other feature.
However, to consider Sindhi implosives an archaism is problematic for
more than one reason (cf. Rasmussen 1999: 231f.; Kümmel 2007: 189, 304):
(1) PIE “aspirates” deaspirated by Grassmann’s law are also continued as
injectives, which means that many of the injectives actually attested cannot
go back to PIE glottalized stops, but must have acquired this feature secondarily; If so, then why not in all cases? (2) A more recent dissimilation of
aspirates took place only in Sindhi and related dialects, and the voiced
stops that evolved from this process may have triggered a special pushchain development of the older voiced stops. (3) his scenario would presuppose a transitional stage without simple voiced stops, but with injectives and breathy-voiced stops, which constitutes a rather improbable system. As to Panjabi, the “tone-depressing” efect of the new plain voiceless
stops is paralleled by the development of old h [ɦ], pointing to earlier
breathy voice rather than the opposite, and a direct development from
breathy voiced to voiceless stops is possible, even if voiceless aspirates are
preserved. he indirect evidence presented by “Lubotsky’s law” (cf.
Lubotsky 1981) depends on its acceptance, but the law is quite controversial; for a detailed critique see now Lipp (2009: I 161f.).
4 Lachmann’s law in Latin might show lengthening before preglottalized
stops, as in the case of Winter’s law.
@ Museum Tusculanum Press and the author 2012
Typology and reconstruction
301
However, it might also reflect lengthening before analogically restored
voiced stops as opposed to fricatives (see most recently Jasanof 2004). he
distribution rather favours the latter solution, since a compensatory
lengthening before a lost *ʔ would be more natural in open syllables than
solely before clusters.
5 In Germanic, dialectal reflexes include preglottalized stops (in Western
Danish and English), preaspirated stops (in Northern Scandinavian),
africates and geminated fricatives (in High German), and all of these
have been considered reflexes of Proto-Germanic preglottalized stops.
However, the “West-Jutland stød” presupposes all pre-12th century soundlaws (cf. Ringgaard 1960, not falsified by Kortlandt) and may represent an
innovation; it might also be due to a secondary development from older
preaspiration (for possible parallels cf. Kusmenko 2008: 135f.). Modern
English preglottalization might “be much more ancient than is commonly
assumed” (Kortlandt 1985: 197; cf. Kortlandt 1997), however, there is no
clear evidence that it must, and Danish influence may be considered.
Northern Scandinavian preaspiration is better explained by Saami substrate influence than by a “weakening” of preglottalization (see Rießler
2004; Kümmel 2007: 305f.; Kusmenko 2008: 129f.). he various geminations in Germanic may simply be caused by syllable-structure adjustments
and need not reflect a segmental ʔ.
2.1.3.2
Inner-PIE Evidence for “glottalic” mediae: *d ~/> *h1?
A similar kind of evidence has been based on alleged dialectal IE sound
laws. he most famous case, proposed by Kortlandt (1983) and considered
as a post-PIE process, is an alternation *d ~ *h1. To explain it, a sound law
has been formulated according to which *d was replaced by *h1 before a
following stop and/or by dissimilation in various languages. While it has
also been accepted by colleagues that do not assume a “glottalic” model
(e.g., Meiser 1998: 172f.; Meier-Brügger 2000: 218), such a law would indeed be most easy to understand if *d was [ɉd] and *h1 was [ʔ], and much
more diicult if *d had no special glottalic feature and/or if *h1 was [h].
However, we should not forget that even simple non-glottalic stops might
be replaced by glottal stops, at least in coda position (see Kümmel 2007:
107), so that *d > *ʔ need not presuppose [ɉd].
@ Museum Tusculanum Press and the author 2012
302
Martin Joachim Kümmel
But was there really a development of *d to *h1? Let us revisit the evidence:
a *d t- ‘ten’ in *h1 tóm ‘100’ > Greek ἑκατόν? Unfortunately, there is
no evidence from any other branch for an initial laryngeal (cf. especially Rau 2009: 17 n. 15). he secondary h- in Greek presupposes analogical influence from *hen- ‘one’, and then *e- might likewise derive
from that source; an original *ha-katón (formed like Vedic sa-hásram
‘1000’) may possibly have changed to he- in much the same way as
ἕτερος from ἅτερος (cf. Sihler 1995: 423 and now also Rau 2009: 17f.
n. 15). Alternatively, we might assume an anaptyxis in *dk° > *edk° (or
*ʔk° > *eʔk°) parallel to *h1k° > *eh1k° which need not presuppose identical sounds (Ø > əɹ /#_CT; *əɹ > *ə > e /_C>ØC).
b *d omt-/d t- ‘ten(th)’ > *-h1 omt-/-h1 t- in various compound
decadic numerals. What we can really observe in this case is a lengthening of the preceding vowel without any “colouring”. But this is just what
we would expect if the original cluster *d had been simplified with
compensatory lengthening, possibly by way of a PIE sound law for this
kind of clusters (see Schumacher 2005); as always, it is diicult to distinguish between *h1 and other lost consonants. Alleged cases of specifically “laryngeal” sound laws are not certain: Greek ἑβδομήκοντα
might be a readjustment of *hebdmēk° < *septh1k°, but it might also
reflect influence from pentēk°; ἐνενήκοντα is easier to explain from an
analogical *enewē° than from †eneunēk°. In Latin septuāginta, nōnāginta, the °ā- might have been generalized from quadrāginta as it certainly was in the case of quīnquā°, sexā°.
c *duʢi- > *h1uʢi- in the word for ‘twenty’: **duʢi-d ti- (> *h1uʢi-d ti-?)
(> *h1uʢih1 ti-?) > Proto-Greek *ewīknʘti → Southern Greek *ew;kosi >
Homeric ἐείκοσι, Ionic-Attic εἴκοσι. Here we face the problem that
Western Greek (ϝ)ίκατι seems to presuppose Proto-Greek *wīknʘti- (<
*uʢih1 ti-?), and we have no means of telling which variant is original
(cf. Beekes 1969: 62; Rau 2009: 18 n. 16 assumes aphaeresis in fast
counting). he adverb *uʢi (PII *uʢí, Latin -ui-, PT *wə, etc.) ‘apart’ has
also been adduced, but here there is absolutely no comparative evidence for a reconstruction *h1uʢi < **duʢi ‘in two’.
@ Museum Tusculanum Press and the author 2012
Typology and reconstruction
303
2.1.3.3 More general internal arguments
Other internal arguments have been based on the distribution of stop series in grammatical elements: it has been observed that mediae are suspiciously absent from endings and suixes (cf. Dunkel 2001: 3f., 9), while
aspiratae can occur (if not in the most basic layers) and tenues seem to be
the normal case. his points to a more marked or more secondary status of
the mediae as opposed to the tenues and aspirates obviously constituting
one class, as is shown by the root-structure constraints: two stops in one
root (if not preceded by *s) must be either both tenues or aspirates, while
mediae may co-occur with each of the other types, but not with another
media. However, all this need not tell us much about the phonological system of late PIE; it will rather be an efect of an earlier system, as it still is in
the attested IE languages (cf. Haider 1983: 85f.). For the latest stage of the
protolanguage, voicing assimilation is normally reconstructed for stops
and *s which seems to presuppose that the mediae were phonologically
voiced. But since this might be due to later parallel innovations, this argument is not decisive.
herefore, the old assumption of Pedersen’s (1951) might be the best solution: pre-PIE could have had a diferent system that shited to the traditionally reconstructed one already in PIE (cf. Miller 1977a; 1977b; Haider
1983): “**t ~ d ~ D[+voiced] > PIE *t ~ d̤ ~ d. he most probable source of
the PIE “mediae” would have been “voiced”, as was the source of the “aspirates”. If the latter were ordinary voiced stops, what might have characterized the former?
2.1.4
A “non-obstruent” solution
From the perspective of system typology, the most probable solution will
be an implosive or rather “non-explosive” stop. Such a reconstruction
might also explain why PIE stops were neutralized to “mediae” in final position (cf. Kümmel 2007: 301f.): stops were probably not released in this
position, and non-released stops were more similar to non-explosive stops
than to explosives. Already Haider (1983: 84f.) proposed to reconstruct
pre-PIE implosives as the source of the “mediae”, but with the assumption
that such sounds must be classified as glottalic (maybe one reason why his
thoughts did not receive the attention they deserved). A very similar sce-
@ Museum Tusculanum Press and the author 2012
304
Martin Joachim Kümmel
nario has now independently been argued for by M. Weiss (2009), adducing a possible diachronic parallel: in Cao Bang, a northern hai language,
an older opposition of implosives and voiced explosives was changed into
an opposition of simple voiced stops and breathy-voiced stops.
It is also interesting that a similar opposition is found in West African
Kwa systems with “lenis” implosives = non-explosive stops vs. “voiced
fortes” with possibly redundant murmur, as argued by Stewart (1989:
236f.). In the neighbouring and cognate Tano languages, a system shit
parallel to Grimm’s law seems to be attested (Stewart 1989: 237f.; 1993): *t
> *tȹ > θ; *d > *dȺ > ð; *ɗ > t, which might provide a parallel for Germanic
and Armenian.
If we thus assume original non-obstruent stops, we have to consider the
chronology of their shit to simple voiced stops that caused the original
voiced stops to develop phonemic breathy voice. Was it already PIE, or did
it only apply to dialectal IE? It should be borne in mind that there is no
evidence for breathy voice in Anatolian, Celtic, and Balto-Slavic, and,
seemingly, the languages from Messapian to Phrygian – i.e., in some rather
peripheral languages. Could this point to a central IE innovation that
spread to most, but not all, dialects?
2.1.5
Further questions and possibilities
For non-obstruent stops, it is well known that they sometimes alternate
with other non-obstruents, i.e., nasals or liquids (cf. Haider 1983: 86; Stewart 1989: 239f.; Clements & Rialland 2005: 18). herefore, sporadic alternation of PIE “mediae” with such sounds might reflect this state of afairs,
e. g., *d ~ *l in Hitt. dā- vs. Luwian lā- ‘to take’, or *d ~ *n in Luwian
tappas-/tipas-, Lithuanian debesìs ~ Hittite nepis-, Slavic *nebes-, Slavic
*domъ ~ Lithuanian nãmas. In this perspective, the famous rarity of *b
might find an explanation, as already pointed out by Haider (1983: 86): if
we assume a simple and typologically well paralleled change **ɓ > *m
(perhaps with some exceptions), we can explain the rarity of *b as well as
the striking fact that *m° is much more frequent in PIE roots (48 in LIV²)
than *n° (18) and that only *m° appears before resonants. Another, but not
necessarily contradictory, possibility would be that **ɓ became *uʢ that is
also exceptionally frequent in roots (as argued by Weiss 2009).
@ Museum Tusculanum Press and the author 2012
Typology and reconstruction
305
In this context, it might be interesting to consider the possibly most
certain case of PIE *b, namely the present stem *píb-o/e- ‘to drink’. Since it
belongs to the root *peh3-, it should reflect < *pi-ph3-o/e-, and is therefore
taken to show a voicing efect of *h3 on a preceding stop (cf. Mayrhofer
1986: 143f.). A possible parallel is provided by the possessive “Hofmann
suix” *-Hen(H)-, e.g., in *h2ap- ‘water’ → **h2ap-h3on- > *h2abon- ‘river’
(Hamp 1972; 1979: 169f.; McCone 1994; but cf. Willi 2004), but in this case,
the only independent argument for the identification of the laryngeal as
*h3 would be derivation from the root *h3en(h1)- in Latin onus etc. (cf.
Dunkel 2001: 12). If we assume that *b was originally “non-explosive”, we
might not have to deal with voicing here, but with an unreleased stop in
the coda before a following obstruent. Before stops and *s, later voicing
assimilation removed the non-explosive stops, but before a laryngeal, the
unreleased stop may have been identified with **ɓ and regularly yielded *b
which remained until the laryngeal was lost. he restriction to *h3 might
be accidental: these formations might have been the only old cases of *TH
clusters followed by a vowel in the whole paradigm, so that no analogical
restitution of the stop was possible. Indeed, other clearly old formations
with *VTHV are diicult to find. Maybe we should not automatically reconstruct *h3 in cases of “voicing”.
If we want to speculate about pre-PIE implosives and their cognates, we
might look for resonants as well as for stops. E.g., there is at least one good
case of a PIE “media” corresponding to a Uralic nasal, viz., PFU *jäŋe/jäŋi
‘ice’ (Rédei 1988: 93; Sammallahti 1988: 543) ≈ PIE *jeg-i/o- < PIU **jæɠ -?
Since one major diference between the Uralic and PIE sound systems is
the presence of four phonemic nasals in the former in contrast to a much
larger number of stops in the latter, we might suspect a Proto-Uralic
change from implosives to nasals (or vice-versa?). Another possible instance of this change might be PIE *dek- ‘to perceive’ ~ PFU *näke-/näki‘to see’ (Rédei 1988: 302; Sammallahti 1988: 546) < PIU **ɗæk-.
2.2 Conclusion
he development of the PIE stop system may thus be reconstructed as in
Table 2 (for the dorsal series cf. Kümmel 2007: 310–327).
@ Museum Tusculanum Press and the author 2012
306
Martin Joachim Kümmel
Table 2
Early and late PIE stops
labial
voiceless
*p
voiced > breathy *b>b̤Ⱥ
implosive > voiced *ɓ>b
3
coronal
*t
*d>d̤Ⱥ
*ɗ>d
“palatal” = velar
*k
*DZ>DZɻȺ
*ɠ>DZ
labiovelar
*kɀ
*DZɀ-DZɻɀȺ
*ɠɀ>DZɀ
“velar” = uvular
(*q?)
(*Dz>Dz̤Ⱥ?)
(*ȫ>Dz?)
he reconstruction of PIE vowels
3.1 Vowel quality: height
Here, no substantial typological problems seem to appear: generally, a triangular three-level system is reconstructed, but with the assumption that
the diference between the low and mid vowels was secondary, low *a
normally being an allophone of *e, we get an older quadrangular two-level
system in which *e and *o were the lowest vowels. herefore, they would
rather have been something like /æ/ and /Ǣ/, since all languages have low
vowels. he overall development of the IE vowels agrees with the hypothesis that *e, *o were rather low: they seem to be quite consistently diferent
from high *i, *u in most ancient languages. hus, the systems given in Table 3 seem probable.
Table 3
Late
i
e
Prehistorical IE vowel systems
PIE
u
o
a
<
Early
i
ɛ
PIE
u
Ǥ
[a]
Pre
i
-PIE
u
æ-a
Ǣ
<
3.2 Vowel quantity
here is no doubt that length was a phonological feature of the (late) PIE
vowels. But with the acceptance of the laryngeal theory, most of the previously reconstructed long vowels disappeared, so that vowel length became
largely restricted to instances of the lengthened grade (“Dehnstufe”) of IE
ablaut and, possibly, to lengthening in monosyllables. hus, long *ī and *ū
could only occur in monosyllables, and the occurrence of the other long
@ Museum Tusculanum Press and the author 2012
Typology and reconstruction
307
vowels was heavily restricted, since lengthened grade only occurred in the
nominative of certain athematic nouns (mostly instead of a segmental
marker, be it *-s or *-h2), maybe in some “vrʘddhi” formations, and in the
strong stems of some ablauting patterns of the “acrostatic” type. Elsewhere
I have argued against the concept of “Narten roots” with a basic underlying
long *ē (see Kümmel 1998), and these types are oten judged rather recent
innovations (e.g., the “Narten” present is oten presented as a secondary
development from earlier reduplicated formations, see Rix apud Harðarson 1993: 29 n. 12; Ringe 2006: 19f.). In any case, the lengthening of the
vowels could have been quite recent and might have arisen in the latest
phase of the protolanguage, so late that we may easily reconstruct a morphonological system of PIE that had no phonological quantity at all.
3.3 he status of apophonic *o
While there are good reasons to consider the alternation *e ~ *Ø to be the
result of pre-PIE syncope of the unaccented predecessor of *e, the alternation *e ~*o is much less easy to explain. Traditionally, the *o-grade was
interpreted as a weaker grade than the *e-grade, reflecting some kind of
unaccented variant. A phonetic explanation has been given by Rasmussen
(2003: 354f.): the lower pitch of the unaccented vowel might have made it
“darker”. Indeed, back vowels have a lower pitch, especially in the second
formant. But the inherent pitch level of segments is not usually confused
with prosodic pitch (tone), and parallels for a development of an unaccented front vowel to a back vowel are not easily found: younger developments in IE languages typically show centralization and/or raising of unaccented vowels, but no backing. So there are doubts whether this scenario is
plausible.
However, even the basic assumption that *o was typically unaccented is
problematic: while there are a number of certain cases where an unaccented *o-grade seems to alternate with accented e-grade, this is contradicted by some quite clear examples of accented *o-grade: to mention only
the least controversial, root nouns like *dóm- ‘house’ or *pód- ‘foot’, the
strong stem of the perfect, thematic action nouns, or the type reflected in
the Indo-Iranian passive aorist and similar o-grade verbs (whatever their
origin may be, cf. Jasanof 2003: 64f., 144f.; Kümmel 2004; Villanueva
Svensson 2006) cannot easily be dismissed as being secondary. he ac-
@ Museum Tusculanum Press and the author 2012
308
Martin Joachim Kümmel
cented *o-grade even seems to be the stronger ablaut grade opposed to *egrade in other forms. In addition, the fact that *o grades seem to be rather
typical for strong stems in general and appear only very rarely in weak
stems does not agree well with the hypothesis that the *o-grade was some
kind of variant of the zero grade: the phonological environment in strong
and weak stems is not generally diferent. Aside from these distributional
arguments, there is even some evidence for a greater phonetic strength of
PIE *o as such which shall be considered now.
3.3.1
Evidence for a greater strength of *o
he reflexes of the vowel usually reconstructed as *o, functioning as an
ablaut grade alternating with *e and zero, display some peculiarities that
are not easily understood from a simple short *o. In some IE languages, its
development seems to presuppose a vowel that was “stronger” than the
other non-high vowels *e and *a.
3.3.1.1
Evidence from attested IE languages
1 Anatolian: Accented *ó became long ā in Luwian, while *é remained
short before an obstruent (with consonantal gemination, Melchert
1994: 263f.), cf., e.g., *pód- > pāt- ‘foot’ vs. *médu- > maddu ‘wine’,
*mélit- > mallit- ‘honey’. In Hittite, *ó was lengthened unconditionally,
while *é yielded short ă or short ĭ before some clusters, and even when
it was preserved, it seems to have remained phonologically short (see
now Kloekhorst in this volume): e.g., *kónk- > kānk- ‘to hang, to weigh’,
*mórg- > mārk- ‘to divide’, *spónd- > ispānt- ‘to libate’, *pód- > pāt- ‘foot’
vs. *léng- > link- ‘to swear’, *kérs- > kars- ‘to cut’, *éndo > anta ‘in(to)’,
*sés- > ses- ‘to sleep’. Also PIE *a seems to have remained short, cf., e.g.,
Hittite *h2ánt- > hant- ‘forehead; in front’.
he only phonetic explanation for this phenomenon so far has been given
by Kloekhorst (2008a: 65, 98 and esp. 2008b): *ó must have counted as a
long vowel already in PA (at least in first and final syllables), but the Luwian gemination only took place ater short vowels. If old *o was a long
vowel already in PA, it could have caused lenition of a following single fortis consonant (Kloekhorst 2008a: 98). Such a lenition rule would explain
two unrelated, hitherto problematical phenomena, namely the lenited con-
@ Museum Tusculanum Press and the author 2012
Typology and reconstruction
309
sonant in the 3rd singular of many hi-verbs and the lenis consonant in
some instances of old *kɀ, cf. Hittite nāhi ‘fears’ from nahh- < *nóh2-; sākuwa ‘eyes’ < *sókɀā. he alternative explanations for these developments
are problematic: Melchert (1994: 61f.) claimed unconditional lenition of
*kɀ, but this is not only diicult to understand phonetically; it is also based
on problematic evidence (on allegedly “Luwian” taru- see Oettinger 1979:
225; Kloekhorst 2008a: 843). For the hi-verbs, Melchert (1994: 81f.) must
assume analogical spread from the prototype ispār- ~ isparr- ‘to spread, to
trample’ < *spór(h1)- ~ *spérh1-. While some kind of analogical redistribution is probable, the alleged prototype is impossible: ispār- ‘spread’ has to
be distinguished from isparr- ‘trample’ and should be reconstructed as
*spór- (see Kloekhorst 2008a: 406f.); thus it did not originally contain a
geminate and cannot have provided the model for lenition in this verb
class. here are some cases of a preserved fortis ater *ó, but none of them
really disproves the rule: sākki ‘knows’ certainly contains a cluster *kH, and
tākki ‘is similar’ might be similarly explained (a direct connection with
*dek- ‘to accept’ is not imperative, as Kloekhorst has rightly observed), or it
is analogical to sākki (for the morphologically conditioned distribution of
lenition in hi-verbs, cf. Melchert 1994: 81f.); sakkar ‘faeces’ might be analogical from the oblique cases, and āppa(n) ‘ater, behind’< *ópV(n) (cf.
Lycian epñ) might be derived from an unaccented or ablauting variant. If
nakku- ‘ritual substitute’ goes back to a *noku- and is not originally Luwian
(cf. Rieken 1999: 202f.; Kloekhorst 2008: 594), the fortis might have been
generalized from the original weak stem or forms with non-syllabic *w.
Kloekhorst (2008b: 132) argues that the *o resulting from *h3e did not
cause lenition. Unfortunately, his Hittite example hāppar ‘business’ is not
probative (we might consider an original meaning ‘contract’ from *h2ep‘to fit’), and Luwian harran(i)- would be a much better example if we could
be sure that it is the equivalent of Hittite hāran- ‘eagle’ (cf. Kloekhorst
2008a: 301f.).
2 Tocharian: Here *o remained a strong vowel, but *e eventually merged
with *i and became a rather weak vowel (cf. Ringe 1996: 90f., 125f.).
3 Finally, according to Brugmann’s Law, as is well known, apophonic *o
yielded *ā in PII open syllables, in contrast to all other IE short vowels
that remained short. A phonetic explanation for this phenomenon has
never been given.
@ Museum Tusculanum Press and the author 2012
310
Martin Joachim Kümmel
hus we have evidence from three branches of IE for a somewhat stronger
status of *o in contrast to *e and its tendency to be longer than any other
short vowel.
3.3.1.2 Evidence from PIE or Common IE
In PIE or Common IE itself, the peculiar development known as the “Saussure Efect” might be relevant. Normally this efect is described as the loss
of a laryngeal when adjacent to a tautosyllabic resonant next to an apophonic *o (cf. Saussure 1905: 511 n. 2; Beekes 1969: 74f., 238f.; Rasmussen
1989: 178f.; 1992a: 350f.; Melchert 1994: 49f.; Nussbaum 1997; de Lamberterie 2004): *H > Ø /$_Ro; /oR_$. But what exactly was lost here (cf.
Nussbaum 1997: 186)? We have to distinguish between loss in anlaut and
loss in inlaut.
a Initial position
he evidence for the loss of a consonantal laryngeal in Hittite is rather
doubtful:
• Hittite warsa- ‘fog, mist’, better <*h1uʢorso-: two doubtful attestations
of aers° in Greek do not outweigh all other evidence pointing to
*ewers° < *h1uʢers- (cf. Kloekhorst 2008a: 972) and the fact that assimilation of *awe° clearly is not regular in Greek, cf. ἀείρω, ἄεσα,
ἄεται, ἀέξω, ἀείδω (the only other parallel might be ἔεδνα :
ἀνάεδνος).
• Likewise, Hittite wawarkima- ‘fixing point of the door-axle’ might
derive from *h1uʢerg- ‘to enclose’, to be separated from *h2uʢerg- ‘to
turn’, Hittite hurki- ‘wheel’, cf. Kloekhorst 2008a: 992 contra LIV² s.
v.).
• Hittite watarnahh- ‘to command, to instruct’ need not be derived
from *h2uʢed(H)-, but might as well reflect *uʢedȹ- ‘to lead’.
• Hittite wast(a)- ‘miss the mark, sin’ has been reconstructed as
*h2uʢomst- and compared to Gk. *awátā-, Homeric ἄτη, but here semantics are much better than the sound correspondences (cf.
Kloekhorst 2008a: 986 with ref.).
he remaining evidence consists of the missing “prothetic vowel” in Greek
words like λοιγός, μοιχός etc. Unfortunately, most etymologies are but pos-
@ Museum Tusculanum Press and the author 2012
Typology and reconstruction
311
sible and not as compelling as we could wish. In any case, it is a vowel that
is missing, not the laryngeal itself. hus the loss might have happened at a
stage when secondary anaptyctic vowels had already developed but not yet
merged with full vowels.
b Internally, i.e. in the sequence *oRH.C:
Loss of a consonantal laryngeal seems to be attested in Hittite kalmara‘ray’, kalmi- ‘piece of firewood’, if from *kolh2m° related to * olh2m/* ¼h2m- ‘reed, straw’, but these cases have been doubted as well, since their
semantics is not clear.
hus, the most certain cases of the “Saussure Efect” show the nonappearance of the vowel that normally appears instead of a laryngeal between consonants following an *o. Since such anaptyctic vowels might be
influenced by the overall quantity of the adjacent syllables, we might assume that the vowel was shorter ater “strong” *o than it was ater “weaker”
vowels, and therefore it was syncopated or never developed into a full
vowel. If loss of the consonant itself may be ascertained, a longer quantity
of *o might also explain the development: clusters might be more strongly
reduced ater a longer nucleus, so that *āRHC became *āRC while *aRHC
remained – because of a diferent syllabification *āR.HC vs. *aRH.C? hus,
these developments might support the assumption of *o being longer than
*e.
3.3.2
Typological parallels for “strong o”?
Could this greater strength of *o vs. *e be explained by natural tendencies
of vowel systems?
3.3.2.1
Looking for a diachronic asymmetry
At first sight, some cases of an *o developing “stronger” reflexes than an *e
seem to exist. E.g., in Proto-Saami, etymological *o and *e developed alike
when followed by a non-high vowel (i.e., a or ä > PS *ā/ē). But when followed by a high vowel (i.e., *i > PS *ə), only *o yielded a long vowel and
developed into a high-mid diphthong *uo (merging with *a in this position), while *e merged with *i and became a short vowel *Ǻ (Sammallahti
1998a: 43f.; 1998b: 55f.): *o > *Ǥɷ > *oa = *e > *ɛɷ > *ea /_a, but *o > *ō > *uo
@ Museum Tusculanum Press and the author 2012
312
Martin Joachim Kümmel
vs. *e > *Ǻ /_Ǻ, e.g., PFS *sormi > *sōrmǺ > PS *suormǺ ‘finger’ vs. PFS *keri >
PS *kǺrǺ ‘bark’ – an asymmetry “the reason for which is not clear” (Sammallahti 1998a: 44; 1998b: 56). However, there might be a rather simple
explanation: In the original sequence *e…i, the similarity of the vowels
might have led to a total assimilation *i…i (or perhaps, more exactly, *Ǻ…Ǻ)
that removed *e from this environment altogether, before later developments could apply (cf. the very early i-umlaut of *e in Germanic). In a sequence *o…i the similarity was much less pronounced so that no total assimilation took place, and *o remained a non-high vowel. hus, the Saami
case does not really provide evidence of an asymmetry between preserved
e and o. A similar development can be observed in the Ob-Ugric languages
where old *e…i was assimilated to *i…i, so that *e became a high vowel
and could eventually yield a short (or reduced) vowel, while in all other
cases both *e and *o are reflected as long (or non-reduced) vowels (cf.
Sammallahti 1988: 500f., 503f.). So this case is no parallel either, and I have
not been able to find one anywhere else.
3.3.2.2
Looking for a parallel synchronic asymmetry
In the absence of a diachronic parallel, we might try to look for a synchronic asymmetry between non-high back and front vowels. Here, some
rather well-known IE languages provide interesting facts: in Farsi, i.e.
modern Persian in Iran, the vowels are phonologically distinguished by
quality alone (Ternes 1999: 168f.). But in some environments, low front /a/
[æ-a] may be shorter than low back /ɑ/ [Ǣ-Ǥ]: e.g., zadan /zadan/ [zæɑdæn]
‘to kill’ vs. dādan /dɑdan/ [dʘǢɚɑdæn] ‘to give’; likewise, /e/, /o/ may be
shorter than /i/, /u/ (Lazard 1989: 265). he historical reason for this peculiarity is clear (and still reflected in the writing system): the vowels that are
more tense and tend to be longer and more stable, continue the originally
long vowels of classical Modern Persian (where quantity was phonologically relevant, as it still is in the more conservative varieties of Eastern Iran
and Afghanistan), and the others continue originally short vowels: /ɑ, i, u/
< ā, ī, ū vs. but /a, e, o/ < a, i, u (cf. Windfuhr 1997: 128f.).
his kind of qualitative diferentiation of short and long a-sounds is
rather common. We may distinguish two main types of back-front
diferentiation of ă : ā: (a) front – back; (b) back – front.
@ Museum Tusculanum Press and the author 2012
Typology and reconstruction
313
a Front ă : back ā
1 a = [a-ɑ] vs. ā > [Ǣə] > [Ǥə] > [oə]: Tocharian a : o; Pashto a : o; Albanian a : o; Welsh a : o/aw; Modern Irish a : Ǥ; Proto-Germanic a : ō;
Common Nordic a : Ǥɷ; Modern Swedish [a] : [Ǣə]; Southern Middle
English a : Ǥɷ; Northern and Eastern Frisian, all Low German and
most High German dialects a : Ǥɷ/ō; Lithuanian a : o; Cassubian a : Ǥɷ;
Hungarian (northern dialect) a : Ǥɷ; Canaanite (Hebrew and Phoenician) and Western Aramaic a : ō
2 a > [æ] > [ɛ] vs. ā = [aə-ɑə]: European Romani e : a; Kurdish e : a;
Ethiopic æ : a
3 Both tendencies combined: Modern Danish [ɛ,æ~a] : [Ǥ]; Early
Modern English æ : ō
b Back ă : front ā
1 ă > [Ǣ] > [Ǥ] vs. ā = [aə-ɑə]: Khowar o : a; Bengali Ǥ : a; Common
Slavic o : a; Hungarian (most dialects and standard) Ǥ : ā
2 a = [a-ɑ] vs. ā > [æə] > [ɛə]: Bolognese, Eastern Engadinian, Old
French a : ɛɷ; Western Frisian and Dutch dialects a : ǣ
3.3.3
A quantitative model
hus, the best parallels for “stronger o” are provided by cases in which a
non-high back vowel reflects an originally long ā in contrast to an originally short a reflected as a non-high front vowel.
a PIE *e < **/a/
here are good independent reasons to suspect that conventional PIE *e
reflects an original /a/. As far as we know, it had three basic allophones,
two non-front vowels in the neighbourhood of certain back fricatives (i.e.,
*h2 and *h3) and a front vowel elsewhere. his is similar to the allophony
that we find in Semitic systems with just one non-high short vowel: e.g.,
Arabic /a/ and /ā/ are very oten fronted in many varieties, except when
adjacent to uvulars, pharyngeals and/or pharyngealized consonants where
they may be retracted (cf. Kaye & Rosenhouse 1997: 278). In Ethiopic,
Common Semitic *a became /æ/ [æ-Ǡ], but before syllable-final “laryngeals” (glottals and pharyngeals/pharyngealized glottals) it became a back
vowel [ɑ] and merged with originally long ā (for Ge’ez, see Gragg 1997:
@ Museum Tusculanum Press and the author 2012
314
Martin Joachim Kümmel
245); in Tigrinya this extends to the position behind "laryngeals" (Kogan
1997: 428). Also for older Modern Danish, a similar allophony is decribed
for the old short a: if it remains short, we find back [ɑ] adjacent to /r/ [ȑ],
central [a] before labials and velars, and fronted [a-æ] in all other positions
(cf. Haberland 1994: 319): e.g., kratte /kȹratƽ/ [ɑkȹȑʘɑdʘƽ] ‘to scratch’, tappe
/tȹapƽ/ [ɑtɫȹabʘƽ] ‘to tap, to draw (beer)’, takke /tȹakƽ/ [ɑtɫȹaDZɽƽ] ‘to thank’,
hatte /hatƽ/ [ɑhædʘƽ] ‘hats’. For the present situation a simpler description
with only two allophones seems to be more adequate (see Basbøll 2005:
70–71, 261–262; Grønnum 2005: 287): back centralized [ɑɻ] adjacent to /r/
and before labials and velars and front [a̝] elsewhere. If the vowel is lengthened, the result is back [ɑə] when adjacent to /r/, but in all other positions
it is strongly fronted and raised to [ɛə] (Haberland 1994: 319), e.g., rase
/raəsə/ [ɑȑɑəsə] ‘to be furious’, mase /maəsə/ [ɑmɛəsə] ‘to squeeze’, tabe
/taəpə/ [ɑtɫȹɛəbʘə] ‘to loose’.
b PIE *o [Ǥ] < [Ǥə] < /**ā/
While PIE *e might thus be explained from **/a/, traditional PIE *o might
reflect an originally long /**ā/ that had acquired a back articulation independent of its environment – it was not coloured by adjacent “laryngeals”,
since long vowels tend to be less easily influenced. hus it can be compared
to old long ā in Danish which became a low back rounded vowel [Ǥə] lowered to [Ǣ] when shortened, and merging with old short o, the only
diference being that in monosyllables, old short o remained short while
old long ā could remain long, cf., e.g., *troð > tråd /tȑǤð/ ‘step’ vs. *trāð >
tråd /tȑǤə’ð/ ‘thread’ (with suprasegmental glottalization licensed by vowel
length). Many cases of parallel developments have already been mentioned
above.
3.3.4
New light on some problems
If we assume such an original system, some peculiarities of PIE ablaut find
a rather plausible explanation:
a he archaic “acrostatic” ablaut pattern *ó ~ é
his pattern might simply reflect shortening of unaccented **ā to **a, thus
**d_m- ~ dām- > **d_m- ~ dam- > *dóm- ~ dem- parallel to **dáuʢ- ~
dauʢ- > *dáuʢ- ~ diuʢ- > *déuʢ- ~ diuʢ-. Since we find accented *é in the weak
@ Museum Tusculanum Press and the author 2012
Typology and reconstruction
315
stem, we might assume that there had been a secondary accent shit here:
the unaccented full vowel *a in the first syllable might have attracted the
accent (cf. Rasmussen 1978: 69 = 1999: 11 and passim on *e from unaccented *ē). he same pattern underlies endings with the same alternation,
but without a zero grade, as in the 1st person plural *-mo(s) ~ *-me(s),
maybe from **-m_(s) ~ **-ma(s) (cf. Tichy 2004: 89; 2006: 88).
b Invariable *-e in the vocative of thematic nouns
his phenomenon becomes clear at once if we accept -e < ** -a as the unaccented variant of *-o- < **-ā-: it is obvious that the vocative was the only
form of thematic stems that could never be accented on the thematic
vowel. he same would apply to the thematic vowel in denominative verbs
when it was followed by the accented suix *-ó/é-, and indeed, the traditional reconstruction is *°e-ó/é-. But Tucker (1988) has shown that the
whole type of Vedic denominatives in °a-yá- might be an innovation (in
fact, not a single finite form of such a verb is attested in the RV), and thus
this reconstruction is less reliable than previously thought. his means that
now the type in °á-ya- looks rather like the only one inherited.
c he “ablaut” of the thematic vowel in verbal inflection
Since *o appears before voiced consonants (or resonants?), and *e before
voiceless consonants (or obstruents?), this alternation might be explained
as an originally quantitative alternation **ā ~ a depending on the character
of the following segment. hus, the distribution established by Rasmussen
(1989: 139f.) receives a better phonetic motivation. An originally accentbased distribution, as proposed by Tichy (2000: 53f.; 2006: 52f.), would be
compatible with our hypothesis as well, but it does not account very well
for the actual distribution of the thematic vowels in inflection.
Now, what was the original form of the suix: was **ā shortened before
*s and plosives, or was **a lengthened before resonants? he distribution
of the thematic vowel in nominal inflection speaks for an originally long
vowel that was not influenced by the following consonant under certain
conditions – presumably when accented. his means that we would have
to explain the *e-variant from a shortening of the unaccented vowel before
the least sonorant consonants. A shortening of unaccented **ā has already
been established above: could we assume an exception to this rule, i.e.,
preservation before resonants (at least in the position immediately ater
@ Museum Tusculanum Press and the author 2012
316
Martin Joachim Kümmel
the accent)? In any case, the pattern of unaccented forms was analogically
generalized in the verb in contrast to nominal inflection.
d he diference between “substantival” *kɀé- and “adjectival” *kɀóhis diference might reflect earlier **kɀá- and a derived stem **kɀ_- <
**kɀa-a-. Here I cannot follow Tichy (2000: 52f.; 2006: 51f.) in assuming
that *kɀe- was the unaccented variant of *kɀó-, even if this would fit my
theory: the alleged complementary distribution (indefinite vs. interrogative = unaccented vs. accented) in Old Avestan is based on just 2 attestations of cahiiā in the indefinite function. hese might rather represent a
residual distribution of the original substantival pronoun whose interrogative function had already largely been taken over by *ká-. And elsewhere in
IE, the distribution of these stems in IE does not fit this assumption: everywhere both stems are used in both functions, the diference being rather
one of animate vs. inanimate (Armenian, Albanian, Slavic). But in these
languages, as everywhere else, *kɀe- is clearly linked with *kɀi- in opposition to *kɀo-, and *kɀi- was not confined to inanimate reference in the oldest IE languages, so the original diference between *kɀi-/kɀe- vs. *kɀocannot have been based on animacy (pace Rix 1976: 187). Instead, the
stems *kɀi- and kɀo- clearly show another functional diference in Italic,
namely that of substantival vs. adjectival (cf. Beekes 1995: 203f.). In fact,
the Avestan attestations of ca- may all be substantival and thus fit the Italic
rule. he redistribution according to animacy might be a parallel innovation of other languages: adjectival *kɀo- replaced the old substantival animate forms but not the inanimate, just as, e.g., in Old Icelandic adjectival
*hwarja- > huerr replaced animate *hwaz but not inanimate *hwat > huat,
and in Middle Iranian, where the reflexes of Old Iranian ka- tend to mean
‘who?’ in contrast to the reflexes of ci- meaning ‘what?’ (cf. Sogdian kē vs.
ču, Middle Persian kē vs. čē) – clearly a secondary redistribution of the Old
Iranian stems which had no diference in animacy.
e he so-called *kɀetuʢóres-rule
his rule would be accounted for in a phonetically plausible way as an accent shit from a shorter to a longer vowel: **ɑkɀatwaəras > **kɀaɑtwaəras
or later **ɑkɀɛtwǤ·rɛs > *kɀɛɑtwǤ·rɛs. Unfortunately, the rule is not really
assured, as shown by Rasmussen (2001): E.g., the reconstruction *kɀetuʢóres
itself is doubtful, since Greek points to an accented first syllable, and Vedic
@ Museum Tusculanum Press and the author 2012
Typology and reconstruction
317
catv_ras might be due to a synchronic rule requiring that stress on the
ending in the weak stem presupposes stress on the last syllable before the
ending in the strong stem. A better case for the rule might be the PIE perfect in which the accented zero-grade ending of the 3rd plural points to an
accent shit that could be analogical to the singular. But the peculiar pattern presupposed by this explanation need not reflect a shit in the singular; it might also reflect the blocking of an accent shit that occurred in the
plural (see below 3.5.5 c).
f Divergent reflexes of “coloured” *o
Since most alleged diferences of *o from older *e adjacent to *h3 vs. apophonic *o involve quantity, they would be much easier to understand if
their distinction was originally quantitative. But because of the relatively
low frequency of *h3, the diference itself is rather diicult to prove.
3.3.5
New or unsolved questions
But of course, some new questions arise, and others remain.
a What about the ablaut pattern *ó ~ Ø?
Is it really a younger blending of *ó ~ e with *é ~ Ø, as has oten been assumed since Schindler (1972)? Or could it reflect an older alternation of
**ā : **a that became even more distinct by the deletion of the short variant? Maybe we might assume a shortening of *ā in certain environments,
occurring early enough to feed the later loss of unaccented *a, thus **_ ~ ā
> **_ ~ a > *_ ~ Ø > *ó ~ Ø. he distribution of *ó ~ e as against *ó ~ Ø
might be compatible with this possibility, since *Ø-grade is typical for
roots ending in two consonants, whereas weak *e-grade is typical before a
single root-final obstruent, as Schindler has shown. herefore, an Osthoflike shortening before original clusters of resonant + consonant might be
responsible for the zero reflex of original unaccented **ā in these cases:
**w_d- ~ **wād- > **w_d- ~ **wad- > **w_d- ~ **wid- > *wód- ~
*wid-. Maybe the new ablaut pattern was extended to single root-final
resonant, which could account for stems like *dȹur- and *dm-, though relic
forms like *déms show that this was not the general rule.
@ Museum Tusculanum Press and the author 2012
318
Martin Joachim Kümmel
b What about *o ~ *e ~ *Ø?
A threefold variation occurs in the ending of the genitive and ablative singular. he traditional explanation is that the *e-variant was the accented
allomorph, and the two others conditioned allophones of the unaccented
ending (cf. Meier-Brügger 2000: 183f.; Tichy 2000: 66). From our new perspective, it might have been just the other way round: *-ós could have been
the primary accented, and *-es the usual unaccented allomorph. But then
we would have serious problems explaining the zero-grade allomorph
(perhaps by an early shortening of unaccented **ā under unclear conditions?). According to a more recent interpretation, *-es is considered a secondary innovation of some languages (Beekes 1985: 176f.). Indeed, -es in
Latin competed with -os for centuries and has been explained as a Latin
innovation (Szemerényi 1969: 978; Wachter 1987: 493f.; Schrijver 2005:
581–586), and we cannot exclude a common innovation of Germanic, Baltic and Slavic – even if it seems less easy to motivate in these languages
than in Latin. hus, *-es need not be PIE, and we might be let with a
rather unusual alternation *-ós ~ *-s that also calls for a special explanation. A third possibility has been proposed by Jay Jasanof (p. c.): he assumes original athematic *-és ~ -s and thematic *-ós that was partly taken
over by athematic nouns, as *-ont(i) was oten taken over by athematic
verbs. Unfortunately, the distribution is quite diferent in these two cases,
so that the parallel is not perfect. And, last but not least, *-ós and *-és
might have been diferent case forms, e.g., of the ablative and genitive or
vice versa (cf. Rasmussen 1992a: 337 n. 9; 1999: 635–638). In this case, we
might even assume an original double variation *-ós ~ *-es vs. *-és ~ *-s
that could have helped to trigger syncretism of these case forms. At present, the question must be let open.
c What about unaccented *o?
When we look at the distribution of unaccented o-grades in ablauting formations, we see that their occurrence is typically restricted to the strong
stem if the first syllable contains an accented vowel (normally *é). Should
we therefore conclude that *ā was not shortened posttonically in medial
syllables, perhaps before voiced consonants only, as we have already assumed for the o-variant of the thematic vowel? And does this mean that
the *o-grade was only later analogically introduced in stems like *népot-,
*h2áusos-?
@ Museum Tusculanum Press and the author 2012
Typology and reconstruction
319
here might be another, very simple reason for the distribution: for
genitives like *déms, we have already posited an accent retraction to the
first syllable containing a full vowel. he very same retraction might explain unaccented o-grades in strong stems: we just have to assume that the
o-grade was originally accented, but that the accent retraction rule afected
all disyllabic word forms – or, to be more precise, all word forms having
two full vowels. hus a pattern é-o- would be expected in most, if not all,
strong stem forms of athematic nouns and active verb stems, as all these
forms originally had potentially nonsyllabic endings without a full vowel.
Since I follow Rasmussen (1978: 84f. = 1999: 30f.) and Tichy (2000: 36f.)
in considering the vowel of the nom. plural ending *-es a secondary prop
vowel, even this form might have counted as disyllabic at the relevant time,
cf. *dȹégȹōm; *gȹésōr, *gȹésorm; *dȹédȹoh1m, *dȹédȹoh1t; *kɀétuʢores;
*dóh3tōr, *dóh3torm < *dag_ms; *gas_rs, *gas_rm; *dad_hm, *dad_ht;
kɀatuʢ_rss; *daɣt_rs,*daɣt_rm, etc. etc. Even then, it is possible that a following **ā was later shortened before voiceless consonants or obstruents,
so that forms like *népotss, *népotms; *génh1os, *h1áusōs would still show
analogical ablaut.
In trisyllabic forms with syllabic endings, the retraction need not have
taken place, and thus we get a natural explanation for the reconstructed
pattern of the perfect singular, cf. *memónh2a, *memónth2a, *memóne <
**mam_nxa, **mam_ntxa, **mam_na.
If we assume an original accentuation of the second syllable in the
strong stem, the “holodynamic” accent type becomes much more natural,
since we can derive it from the same accentuation pattern as most other
types, showing stress on the last full vowel of the word form, e.g. *dȹégȹōm
~ *dȹgȹm-ái < *dag_m-m ~ *dgm-ái. he same applies to the “acrostatic”
type of *dóh3-tor/tr- that difered from the “holodynamic” type by the
presence of an e-grade vowel in the first syllable of the weak stem, e.g.,
*gȹéu-tor- ~ *gȹéu-trʘ-s < *gau-t_r-m ~ *gau-tr-ás. What remains unexplained is the diferent weak stem of these two types (together with the
endingless “holodynamic” locative), but of course, this problem is an old
one. A solution might be similar to the one considered above for the alternation *o ~ *Ø, but I cannot ofer a detailed account yet.
By this accent shit rule, we can easily predict the accentuation of nearly
all PIE word forms with two full vowels. Aside from younger accent shits
(e.g., contrastive accent) and productive suix accentuation, there is just
@ Museum Tusculanum Press and the author 2012
320
Martin Joachim Kümmel
one class of notable exceptions to this rule, namely nominal thematic
stems with a full grade and suix accent. Perhaps their accentuation was
analogical to the trisyllabic forms of the paradigm. However, o-grade thematic formations in particular typically show a pretonic unaccented ograde that cannot be explained by the set of rules for athematic formations.
Maybe we should reckon with rather late formations or accent shits here –
or even something more complicated (cf. Rasmussen 1992a).
3.4 Conclusion
To conclude, a more detailed reconstruction of the original PIE vowel system might look like in Table 4. If late PIE also had phonemic *a, *ā, ī, and
ū, these were secondary. Of course, the above scenario does not necessarily
exclude the possibility that PIE *e and *o also had some other sources
diferent from **a and **ā.
he vowel system of late PIE
Table 4
short
front
/i/
long
central back
front
central
back
high
/u/
(/ī/ [iə]?)
(/ū/ [uə]?)
mid
/o/ [Ǥ·] /ē/ [ɛə]
/ō/ [Ǥə·]
/e/ [ɛ-æ]
low
/a?/ [a]
[Ǣ]
(/ā/ [aə]?)
References
Allen, W. Sidney. 1957. Some phonological characteristics of Rajasthani.
Bulletin of the School of Oriental and African Studies 20, 5–11.
Andreev, Nikolaj D. 1957. Periodizacija istorii indoevropejskogo prajazyka.
Voprosy jazykoznanija 1957(2), 3–18.
Basbøll, Hans. 2005. he phonology of Danish. Oxford & New York (NY):
Oxford University Press.
Beekes, Robert S. P. 1969. he development of the Proto-Indo-European laryngeals in Greek. he Hague & Paris: Mouton.
Beekes, Robert S. P. 1985. he origins of the Indo-European nominal inflection (IBS 46). Innsbruck: Institut für Sprachwissenschat der Universität
Innsbruck.
@ Museum Tusculanum Press and the author 2012
Typology and reconstruction
321
Beekes, Robert S. P. 1995. Comparative Indo-European linguistics: An introduction. Amsterdam & Philadelphia (PA): John Benjamins.
Blust, Robert A. 1974. A double counter-universal in Kelabit. Papers in Linguistics 7(3/4), 309–324.
Blust, Robert A. 2006. he origin of the Kelabit voiced aspirates: a historical hypothesis revisited. Oceanic Linguistics 45(2), 311–338.
Clements, George N. & Sylvester Osu. 2002. Explosives, implosives, and
nonexplosives: Some linguistic efects of air pressure diferences in
stops. In C. Gussenhoven & N. Warner (eds.), Laboratory Phonology 7,
299–350. Berlin: Mouton de Gruyter.
Clements, George N. & Sylvester Osu. 2005. Nasal harmony in Ikwere, a
language with no phonemic nasal consonants. Journal of African Languages and Linguistics 26, 165–200.
Clements, George N. & Annie Rialland. 2005. Africa as a phonological
area. ed268.univ-paris3.fr/lpp/publications/2005_Clements_Africa_as.
pdf (29 September 2009).
Dunkel, George E. 1981. Typology versus reconstruction. In Yoël L. Arbeitman et al. (eds.), Bono homini donum: Essays in historical linguistics in
memory of J. Alexander Kerns (Current Issues in Linguistic heory 16),
559–569. Amsterdam: John Benjamins.
Dunkel, George E. 2001. he sound systems of Proto-Indo-European. In
Martin E. Huld et al. (eds.), Proceedings of the Twelth Annual UCLA
Indo-European conference, Los Angeles, May 26–28, 2000, 1–14 (JIES
Monograph Series 40). Washington (DC): Institute for the Study of
Man.
Gamkrelidze, T’amaz V. & Vjačeslav Vs. Ivanov 1973. Sprachtypologie und
die Rekonstruktion der gemeinindogermanischen Verschlüsse. Vorläufiger Bericht. Phonetica 27, 150–156.
Garrett, Andrew. 1998. Adjarian’s law, the glottalic theory, and the position
of Armenian. In Benjamin K. Bergen et al. (eds.), Special session on
Indo-European subgrouping and internal relations. Proceedings of the
24th annual Meeting of the Berkeley Linguistics Society, 12–23. Berkeley
(CA): Soc.
Gragg, Gene. 1997. Geʽez (Ethiopic). In Robert Hetzron (ed.), he Semitic
languages, 242–262. London: Routledge.
Grønnum, Nina. 2005. Fonetik og fonologi. Almen og dansk. 3rd edition.
København: Akademisk Forlag.
@ Museum Tusculanum Press and the author 2012
322
Martin Joachim Kümmel
Haberland, Hartmut. 1994. Danish. In Ekkehard König & Johan van der
Auwera (ed.), he Germanic languages, 313–348. London: Routledge.
Haider, Hubert. 1983. Der Fehlschluß der Typologie. In Wolfgang Meid
(ed.), Philologie und Sprachwissenschat: Akten der 10. Österreichischen
Linguisten-Tagung Innsbruck, 23.–26. Oktober 1982 (IBS 43), 79–92.
Innsbruck: Institut für Sprachwissenschat der Universität Innsbruck.
(English translation: he fallacy of typology – remarks on the PIE-stop
system. Lingua 65 (1985), 1–27.)
Hamann, Silke & Susanne Fuchs. 2008. How do retroflex stops evolve?
Evidence from typology and an articulatory study. ZAS Papers in Linguistics 49, 97–131.
Hamp, Eric P. 1972. Palaic ḫa-a-ap-na-aš ‘river’. Münchener Studien zur
Sprachwissenschat 30, 35–37.
Hamp, Eric P. 1979. Celtica Indogermanica. Zeitschrit für Celtische
Philologie 37, 167–173.
Harðarson, Jón Axel. 1993. Studien zum urindogermanischen Wurzelaorist
und dessen Vertretung im Indoiranischen und Griechischen (IBS 74).
Innsbruck: Institut für Sprachwissenschat der Universität Innsbruck.
Hopper, Paul J. 1973. Glottalized and murmured occlusives in IndoEuropean. Glossa 7, 141–166.
Hopper, Paul J. 1977. he typology of the Proto-Indo-European segmental
inventory. Journal of Indo-European Studies 5, 41–53.
Jakobson, Roman. 1958. Typological studies and their contribution to historical comparatives linguistics. In Eva Sivertsen (ed.), Proceedings of
the Eighth International Congress of Linguists, Oslo 1958, 17–35. Oslo:
Oslo University Press.
Jasanof, Jay H. 2003. Hittite and the Indo-European verb. New York (NY):
Oxford University Press.
Jasanof, Jay H. 2004. Plus ça change …: Lachmann’s law in Latin. In John
H. W. Penney (ed.), Indo-European perspectives: Studies in honour of
Anna Morpurgo Davies, 405–416. Oxford: Oxford University Press.
@ Museum Tusculanum Press and the author 2012
Typology and reconstruction
323
Job, Michael. 1989. Sound change typology and the “Ejective Model”. In
heo Vennemann (ed.), he new sound of Indo-European: Essays in
phonological reconstruction. Proceedings of a workshop held during the
Seventieth International Conference on Historical Linguistics held Sept.
9–13, 1985 at the University of Pavia, 123–136 (Trends in Linguistics:
Studies and Monographs 41). Berlin & New York (NY): Mouton de
Gruyter.
Kaye, Alan S. & Rosenhouse, Judith. 1997. Arabic dialects and Maltese. In
Robert Hetzron (ed.), he Semitic languages, 263–311. London: Routledge.
Kloekhorst, Alwin. 2008. Etymological dictionary of the Hittite inherited
lexicon (Leiden Indo-European Etymological Dictionary Series 5). Leiden & Boston (MA): Brill.
Kloekhorst, Alwin. 2008b. Čop’s Law in Luwian Revisited. Die Sprache
46(2), 2006[2008], 131–136.
Kogan, Leonid E. 1997. Tigrinya. In Robert Hetzron (ed.), he Semitic languages, 424–445. London: Routledge.
Kortlandt, Frederik H. H. 1978. Proto-Indo-European obstruents. Indogermanische Forschungen 83, 107–118.
Kortlandt, Frederik H. H. 1981. Glottalic consonants in Sindhi and ProtoIndo-European. Indo-Iranian Journal 23, 15–19.
Kortlandt, Frederik H. H. 1983. Greek numerals and PIE glottalic consonants. Münchener Studien zur Sprachwissenschat 42, 97–104.
Kortlandt, Frederik H. H. 1985. Proto-Indo-European glottalic stops: the
comparative evidence. Folia Linguistica Historica 6(2), 183–201.
Kortlandt, Frederik H. H. 1988. Vestjysk stød, Icelandic preaspiration, and
Proto-Indo-European glottalic stops. In Mohammad Ali Jazayery &
Werner Winter (eds.), Languages and cultures: Studies in honor of Edgar
C. Polomé (Trends in Linguistics: Studies and Monographs 36), 353–356.
Berlin: Mouton de Gruyter.
Kortlandt, Frederik H. H. 1997. How old is the English glottal stop? NorthWestern European Language Evolution 31–32, 175–179.
Kümmel, Martin Joachim. 1998. Wurzelpräsens neben Wurzelaorist im
Indogermanischen. Historische Sprachforschung 111, 191–208.
@ Museum Tusculanum Press and the author 2012
324
Martin Joachim Kümmel
Kümmel, Martin Joachim. 2004. Zur o-Stufe in der idg. Verbalstammbildung. In James E. Clackson & Birgit Anette Olsen (eds.), Indo-European word formation: Proceedings of the conference held at the University
of Copenhagen, October 20th–22nd 2000 (Copenhagen Studies in IndoEuropean 2), 139–158. Kopenhagen: Museum Tusculanum Press.
Kümmel, Martin Joachim. 2007. Konsonantenwandel. Bausteine zu einer
Typologie des Lautwandels und ihre Konsequenzen für die vergleichende
Rekonstruktion. Wiesbaden: Reichert.
Kusmenko, Jurij. 2008. Der samische Einfluss auf die skandinavischen Sprachen. Ein Beitrag zur skandinavischen Sprachgeschichte (Berliner Beiträge zur Skandinavistik 10). Berlin: Nordeuropa-Institut der Humboldt-Universität.
Ladefoged, Peter & Ian Maddieson. 1996. he sounds of the world’s languages. Oxford: Blackwell.
de Lamberterie, Charles. 2004. Sella, subsellium, meretrix: sonantes-voyelles et ‘efet Saussure’ en grec ancien. In John H. W. Penney (ed.), IndoEuropean perspectives: Studies in honour of Anna Morpurgo Davies, 236–
253. Oxford: Oxford University Press.
Lazard, Gilbert. 1989. Le persan. In Rüdiger Schmitt (ed.), Compendium
linguarum Iranicarum, 263–293. Wiesbaden: Reichert.
Lipp, Reiner. 2009. Die indogermanischen und einzelsprachlichen Palatale
im Indoiranischen. Band I: Neurekonstruktion, Nuristan-Sprachen,
Genese der indoarischen Retroflexe, Indoarisch von Mitanni. Band II:
horn-Problem, indoiranische Laryngalvokalisation. Heidelberg: Winter.
LIV² = Helmut Rix et al. (eds.). 2001. Lexikon der indogermanischen
Verben. 2., erw. und verb. Aufl. Wiesbaden: Reichert.
Lubotsky, Alexander M. 1981. Gr. pḗgnumi : Skt. pajrá- and loss of laryngeals before mediae in Indo-Iranian. Münchener Studien zur Sprachwissenschat 40, 133–138.
Maddieson, Ian. 1984. Patterns of sounds. Cambridge: Cambridge University Press.
Mayrhofer, Manfred. 1986. Indogermanische Grammatik, Halbband I/2:
Lautlehre (Segmentale Phonologie des Indogermanischen). Heidelberg:
Winter.
McCone, Kim R. 1992 [1994]. OIr. aub ‘river’ and amnair ‘maternal uncle’.
Münchener Studien zur Sprachwissenschat 53, 101–111.
@ Museum Tusculanum Press and the author 2012
Typology and reconstruction
325
Meier-Brügger, Michael. 2000. Indogermanische Sprachwissenschat. 7. völlig neubarbeitete Auflage der früheren Darstellung von Hans Krahe.
Unter Mitarbeit von Matthias Fritz und Manfred Mayrhofer. Berlin &
New York (NY): de Gruyter.
Meiser, Gerhard. 1998. Historische Laut- und Formenlehre der lateinischen
Sprache. Darmstadt: Wissenschatliche Buchgesellschat.
Melchert, H. Craig. 1994. Anatolian historical phonology (Leiden Studies in
Indo-European 3). Amsterdam & Atlanta (GA): Rodopi.
Miller, D. Gary. 1977a. Some theoretical and typological implications of an
Indo-European root structure constraint. Journal of Indo-European
Studies 5, 31–130.
Miller, D. Gary. 1977b. Bartholomae’s Law and an IE root constraint. In
Paul J. Hopper (ed.), Studies in descriptive and historical linguistics: Festschrit for W. P. Lehmann (Current Issues in Linguistic heory 4), 365–
392. Amsterdam: John Benjamins.
Normier, Rudolf. 1977. Idg. Konsonantismus, germ. “Lautverschiebung”
und Vernersches Gesetz. Zeitschrit für vergleichende Sprachforschung
91, 171–218.
Nussbaum, Alan Jefrey. 1997. he “Saussure Efect” in Latin and Italic. In
Alexander Lubotsky (ed.), Sound law and analogy: Papers in honor of
Robert S. P. Beekes on the occasion of his 60th birthday, 180–203. Amsterdam & Atlanta (GA): Rodopi.
Oettinger, Norbert. 1979. Die Stammbildung des hethitischen Verbums
(Erlanger Beiträge zur Sprach- und Kulturwissenschat 64). Nürnberg:
Hans Carl.
Pedersen, Holger. 1951. Die gemeinindoeuropäischen und die vorindoeuropäischen Verschlußlaute. Copenhagen: Munksgaard.
Pisowicz, Andrzej. 1976. Le développement du consonantisme arménien.
Wrocław etc.: Zakład Narodowy im. Ossolińskich.
Rasmussen, Jens Elmegård. 1978. Zur Morphophonemik des Urindogermanischen. In Bojan Čop et al. (eds.), Collectanea Indoeuropaea I,
Ljubljana, 59–143 [= 1999: 1, 1–66].
Rasmussen, Jens Elmegård. 1987. On the status of the aspirated tenues and
the Indo-European phonation series. Acta linguistica Hafnensia 20, 81–
109 [= 1999: 1, 216–243].
@ Museum Tusculanum Press and the author 2012
326
Martin Joachim Kümmel
Rasmussen, Jens Elmegård. 1989. Studien zur Morphophonemik der indogermanischen Grundsprache (IBS 55). Innsbruck: Institut für Sprachwissenschat der Universität Innsbruck.
Rasmussen, Jens Elmegård. 1992a. One type of o-grade: A consonantal root
infix? In Robert Beekes et al. (eds.), Rekonstruktion und relative
Chronologie. Akten der VIII. Fachtagung der Indogermanischen Gesellschat, Leiden, 31. August – 4. September 1987, 335–357 (IBS, Vorträge
und kleinere Schriten 65). Innsbruck: Institut für Sprachwissenschat
der Universität Innsbruck
Rasmussen, Jens Elmegård. 1992b. Winter’s Law of Balto-Slavic lengthening – an unnatural fact? Copenhagen Working Papers in Linguistics 2,
63–77 [= 1999: 2, 527–540].
Rasmussen, Jens Elmegård. 1999. Selected papers on Indo-European linguistics. With a section on comparative Eskimo linguistics (Copenhagen
Studies in Indo-European 1). 2 vols. Copenhagen: Museum Tusculanum Press.
Rasmussen, Jens Elmegård. 2001. Against the assumption of an IE
“*kɀetuʢóres Rule”. In Martin E. Huld et al. (eds.), Proceedings of the
Twelth Annual UCLA Indo-European conference, Los Angeles, May 26–
28, 2000 (JIES Monograph Series 40), 15–28. Washington (DC): Institute for the Study of Man.
Rasmussen, Jens Elmegård. 2003. An integrated view on ablaut and accent
in Indo-European. In Brigitte L. M. Bauer & Georges-Jean Pinault
(eds.), Language in time and space: A festschrit for Werner Winter on
the occasion of his 80th birthday (Trends in Linguistics. Studies and
Monographs 144, i), 351–358. Berlin: Mouton de Gruyter.
Rasmussen, Jens Elmegård. 2007. A reflex of *H1 in Hieroglyphic Luvian?
In Karlene Jones-Bley et al. (eds.), Proceedings of the 18th UCLA IndoEuropean Conference, Los Angeles, November 3–4, 2006 (selected papers)
(JIES Monograph Series 53), 161–167. Washington (DC): Institute for
the Study of Man.
Rau, Jeremy. 2009. Indo-European nominal morphology: he decads and the
Caland system (IBS 132). Innsbruck: Institut für Sprachen und Literaturen der Universität Innsbruck.
Rédei, Károly. 1988. Uralisches etymologisches Wörterbuch. 3 vols. Wiesbaden: Harrassowitz.
@ Museum Tusculanum Press and the author 2012
Typology and reconstruction
327
Rieken, Elisabeth. 1999. Untersuchungen zur nominalen Stammbildung des
Hethitischen (Studien zu den Boğazköy-Texten 44). Wiesbaden: Harrassowitz.
Rießler, Michael. 2004. On the origin of preaspiration in North Germanic.
In Karlene Jones-Bley et al. (eds.), Proceedings of the Fiteenth Annual
UCLA Indo-European Conference, Los Angeles, November 7–8, 2003
(JIES Monograph Series 49), 168–185. Washington (DC): Institute for
the Study of Man.
Ringe, Donald A., Jr. 1996. On the chronology of sound changes in Tocharian. Vol. 1: From Proto-Indo-European to Proto-Tocharian. New Haven
(Connecticut): American Oriental Society.
Ringgaard, Kristian. 1960. Vestjysk stød. Aarhus: Universitetsforlaget.
Rix, Helmut. 1976. Historische Grammatik des Griechischen. Laut- und Formenlehre. Darmstadt: Wissenschatliche Buchgesellschat.
Sammallahti, Pekka. 1988. Historical phonology of the Uralic languages
with special reference to Samoyed, Ugric, and Permic. In Denis Sinor
(ed.), he Uralic languages: Description, history, and foreign influences,
478–554. Leiden: Brill.
Sammallahti, Pekka. 1998a. he Saami languages: An introduction. Karasjok: Davvi Girji.
Sammallahti, Pekka. 1998b. Saamic. In Daniel Abondolo (ed.), he Uralic
languages, 43–95. London: Routledge.
de Saussure, Ferdinand. 1905. D’ὠμήλυσις à Τριπτόλεμος. Remarques étymologiques. In Adolf Bauer (ed.), Mélanges Nicole: recueil de mémoires
de philologie classique et d'archéologie oferts à Jules Nicole, à l'occasion
du XXXe anniversaire de son professorat, 503–513. Genève: Kündig.
Schindler, Jochem. 1972. L’apophonie des noms-racines indo-européens.
Bulletin de la Société Linguistique de Paris 67, 31–38.
Schrijver, Peter. 2005. Substrateinflüsse und historische Lautlehre: Latein
und Sabellisch. In Gerhard Meiser & Olav Hackstein (eds.), Sprachkontakt und Sprachwandel. Akten der XI. Fachtagung der Indogermanischen
Gesellschat, Halle an der Saale, 17.–23. September 2000, 573–589. Wiesbaden: Reichert.
@ Museum Tusculanum Press and the author 2012
328
Martin Joachim Kümmel
Schumacher, Stefan. 2005. “Langvokalische Perfekta” in indogermanischen
Einzelsprachen und ihr grundsprachlicher Hintergrund. In Gerhard
Meiser & Olav Hackstein (eds.), Sprachkontakt und Sprachwandel.
Akten der XI. Fachtagung der Indogermanischen Gesellschat, Halle an
der Saale, 17.–23. September 2000, 591–626. Wiesbaden: Reichert.
Sihler, Andrew Littleton. 1995. New comparative grammar of Greek and
Latin. New York (NY) & Oxford: Oxford University Press.
Stewart, John M. 1989. Kwa. In John Bendor-Samuel (ed.), he NigerCongo languages: A classification and description of Africa’s largest language family, 216–245. Lanham: University Press of America.
Stewart, John M. 1993. he second Tano consonant shit and its likeness to
Grimm’s Law. Journal of West African Languages 23, 3–39.
Stuart-Smith, Jane. 2004. Phonetics and philology: Sound shange in Italic.
Oxford: Oxford University Press.
Szemerényi, Oswald. 1969. Etyma latina II (7–18). In Studi linguistici in
onore di Vittore Pisani, vol. 2, 963–994. Brescia: Editrice Paideia.
Ternes, Elmar. 1999. Einführung in die Phonologie. 2nd ed. Darmstadt:
Wissenschatliche Buchgesellschat.
Tichy, Eva. 2004. Indogermanistisches Grundwissen für Studierende sprachwissenschatlicher Disziplinen. 2., überarbeitete Aufl. Bremen: Hempen.
Tichy, Eva. 2006. A survey of Proto-Indo-European. Translated by James E.
Cathey in collaboration with the author. Bremen: Hempen.
Tucker, Elizabeth. 1988. Some innovations in the system of denominative
verbs in early Indic. Transactions of the Philological Society 86, 93–114.
Turner, Ralph Lilly. 1924. he Sindhi recursives or voiced stops preceded
by glottal closure. Bulletin of the School of Oriental and African Studies
3, 301–315.
Vaux, Bert. 1998. he phonology of Armenian. Oxford: Clarendon.
Vennemann, heo. 1984. Hochgermanisch und Niedergermanisch: Die
Verzweigungstheorie der germanisch-deutschen Lautverschiebung.
Beiträge zur Geschichte der deutschen Sprache und Literatur (Tübingen)
106, 1–45.
Villanueva Svensson, Miguel. 2006. Traces of *o-grade middle root aorists
in Baltic and Slavic. Historische Sprachforschung 119, 295–317.
Weiss, Michael. 2009. he Cao Bang heory. http://ling.cornell.edu/docs/
Cao_Bang_heory.pptx (29 September 2009).
@ Museum Tusculanum Press and the author 2012
Typology and reconstruction
329
Willi, Andreas. 2004. Flowing riches: Greek ἄφενος and Indo-European
streams. In John H. W. Penney (ed.), Indo-European perspectives: Studies in honour of Anna Morpurgo Davies, 323–37. Oxford: Oxford University Press.
Windfuhr, Gernot L. 1990. Persian. In Bernard Comrie (ed.), he major
languages of South Asia, the Middle East, and Africa, 108–131. London:
Routledge.
@ Museum Tusculanum Press and the author 2012