60 MASQUERADING ALLOMORPHS

the stem to which it is suffixed. Ignoring the subtleties of the situation, the essence of the generalisation that
we need to capture is this: allomorphs of the plural agree in voicing with the final sound of the stem to
which they are attached.
ASSIMILATION is the name given to this kind of ‘agreement’ which has the effect of making one sound
become more like another sound in its neighbourhood. So, voiceless final consonants in the stem require the
voiceless fricative allomorph /s/ while voiced final consonants go with the voiced fricative /z/. However,
where following this procedure would result in two sibilants being right next to each other, the vowel /I/ (or /
e/ in some dialects) is inserted between the last consonant and the suffix (as in lasses /læsiz/ or /læsez/).
This phenomenon of vowel insertion is also called SCHWA EPENTHESIS (schwa being the name of the
vowel /e/).
Assimilation is normally the reason for the phonological conditioning of allomorphs. A morpheme may
have more than one mask. It may masquerade as one of several allomorphs when the sounds that represent
it are modified so that they become more like some other sound (s) in the environment where it appears. An
allomorph wearing a suitable phonological mask is chosen to suit each set of phonological circumstances.
The phonological modification, as mentioned at the beginning of this chapter, can affect bases as well as
affixes. So far we have seen a suffix with several phonologically conditioned allomorphs. Now we will look
at bases that change when an affix is attached.
In some nouns which end in a labial or dental voiceless fricative, i.e. /f/ or / / (spelled as -th and -f), the
final consonant gets voiced to /v/ and // respectively when the plural suffix /-z/ is present. The voicing is
much more common with the nouns ending in /f/ than those ending in / /.

[5.2]
Plural Genitive
wife wives wife’s
wolf wolves wolf’s
calf calves calf’s
thief thieves thief’s
leaf leaves leaf’s
life lives life’s
knife knives knife’s
sheaf sheaves sheaf’s

In [5.2] the final sound of the noun stem ending in /f/ shows two contrasting behaviours when /-z/ is
suffixed. If the /-z/ represents the plural of the noun as in wives (< wife (singular)) the final /-f/ changes to /
v/. But where the /-s/ represents the genitive morpheme as in the wife’s job, no such change takes place
(*the wive’s job) (cf. Swadesh and Voeglin 1939).
This is typical. The phonological conditioning of allomorphs seldom operates in a manner that is totally
predictable on the basis of phonological information alone. It is frequently the case that given precisely the
same phonological situation, we can observe differing patterns of phonological behaviour depending on the
morphological context in which an allomorph appears. We saw this when we contrasted the voicing of the
final /f/ in the plural with the non-application of the voicing rule in the genitive where no change takes place
although in both cases as /s/ follows a voiceless fricative.
Let us now briefly turn to nouns ending in the voiceless dental fricative / / (spelled -th). The spelling
always remains the same, and the pronunciation does not normally change either. It is / / when next to a
 ENGLISH WORDS 61

consonant (as in lengths, births etc.). Usually it is also / / after vowels (e.g. moths, cloths etc.). But in a few
cases, when the final / / is preceded by a vowel, it optionally undergoes voicing. So the words below have
two possible plurals:

[5.3]
Voiceless / / Voiced //
baths /ba: s/ /ba:z/
youth /ju: s/ /ju:z/
wreath /ri: s/ /ri:z/

There is a change in progress in the way nouns ending in /f/ form their plural. However, the direction of
change is towards eliminating the allomorph of the root with the sound /v/ so that we only have the /f/ form.
Hooves and rooves are being supplanted by hoofs and roofs as the plurals of hoof and roof. There are many
nouns ending in /f/ whose last sound never changes in the plural, viz. chiefs, /t i:fs/ (*/t i:vz /) laughs, /la:fs/
(*/la:vz), belief /bIli:fs/ (*/bIli:vz/). The ongoing elimination of the (/f/)alternation (in hooves etc.) is a way
of making a quirky part of the system fall into line.
To sum up, the behaviour of /f/-final nouns illustrates two properties of morphophonemic rules:

(i) Morphophonemic rules tend to be exception ridden. They rarely apply to all forms with the
appropriate phonological properties. Much of the morphophonemic alternation reflects incomplete
phonological changes in the history of a language. In some respects a language is like an army
marching in one direction when the order is given to change direction. Some soldiers fail to hear (or
choose to ignore) the order to change direction. Thus, we have seen that some /f/s are phonetically
realised as [v] in the plural and others are not.
In this respect morphophonemic rules differ from purely allophonic rules found in phonology.
ALLOPHONIC RULES (which have the job of specifying the phonetic realisation of phonemes)
apply automatically and blindly wherever the requisite phonetic environment that triggers the
alternation is present. For instance, any vowel that appears before a voiced consonant is slightly
lengthened. Compare the /e/ in bet [bet] with the slightly longer [e·] of bed [be·d]. (The raised dot
marks this slight lengthening.) The lengthening takes place regardless of which morphemes are
represented. It only fails to happen if the phonological circumstances are not right.
(ii) Morphophonemic rules are triggered by the presence of certain morphemes. They are tied rules
which are only permitted to operate if certain morphological information sets them off. Thus, for
example, in the data in [5.2] above if suffix /-z/ represents the plural morpheme, it assimilates the
voicing of the last sound of the stem, but if it represents the genitive, it does not.

5.2.1
Selecting underlying representations
I argued earlier that by setting up a single underlying representation from which different allomorphs of a
morpheme are derived by phonological rules we are able to capture generalisations about the phonological
behaviour of morphemes. The distribution of the different allomorphs of a morpheme is normally not
arbitrary. We saw, for example, that voice assimilation is the principle behind the different allomorph guises
62 MASQUERADING ALLOMORPHS

adopted by the /-z/ plural suffix. The idea of generality is vital. It is the principle that led us to posit /-z/ as
the underlying representation from which the allomorphs /-z/, /Iz/ and /-s/ are derived.
Often the same phonological factor that determines the morphophonemic alternations of one morpheme
is also relevant in alternations of several other morphemes. Therefore what is needed is a solution that
enables us to make a general statement of phonological conditioning which is valid for several morphemes.
For instance, if you look at the suffix of the third person singular present tense forms of verbs, you will see
that it shows exactly the same alternations as the /-z/ plural suffix of nouns. If we made separate statements
for these two morphemes we would be missing this generalisation.

[5.4]
Phonologically conditioned allomorphs of the third person singular
a. Underlying /-z/ is realised as /-s/ if a stem ends in a voiceless consonant which is
not a strident coronal. This means that /-s/ is added after /p t k f /.
Underlying representation Surface representation
skips /skIp-z/ /skips/
lights /laIt-z/ /laIts/
likes /laIk-z/ /laIks/
laughs /la:f-z/ /la:fs/
baths /ba: -z/ /ba: s/
b. Underlying /-z/ is realised as /-z/ if a stem ends in a voiced sound which is not a strident coronal. This
means that /-z/ is added after the consonantal segments /b d g v m n l r w/ and after any vowel, e.g.:

Underlying representation Surface

representation
roams /reum-z/ /reUmz/
calls /k:l-z/ /k:lz/
skids /skId-z/ /skIdz/
brags /bræg-z/ /brægz/
sees /si:-z/ /si:z/
shows / eU-Z/ / eUz/
c. Underlying /-z/ is realised as /-Iz/ if a stem ends in a consonant which is both strident
and coronal.
This means that /-Iz/ is added after the consonantal segments, /s z t d/, with /I/ being
inserted to separate coronal sibilants. (See p. 85 above.)
Underlying representation Surface
representation
misses /mIs-z/ /mIs-Iz/ /mIsIz/
wishes /wI -z/ /wI -IZ/ /wi iz/
launches /l:nt -z/ /l:nt -Iz/ /l:nt Iz/
touches /t t -z/ /t t -Iz/ /t t IZ/
cringes /krInd-z/ /krInd-Iz/ /krIndIz/

The tables in [5.1] and [5.4] are virtually identical. Merely listing separately the allomorphs of the third
person singular present tense and those of the plural would fail to show that there is something very