6.

2 The Phonological Form of

Morphemes
We have said that, in knowing a language, a speaker possesses
largely unconscious linguistic knowledge, which subsumes
semantic, syntactic and phonological knowledge. And we have
said that the phonological units or categories we have called
phonemes are part of that phonological knowledge. As we
progress in this book, we will investigate the question of what
other sorts of phonological knowledge speakers possess,
besides phonemes alone. Let us begin this investigation by
considering the internal structure of words. You will agree that
the English word cats may be broken down into two
component parts. Let us call those component parts
morphemes. Then we may say that this word consists of a
root morpheme and a plural morpheme (which, in this case, is
a suffix). Let us say that words of this sort are
morphologically complex: they consist of more than one
morpheme. Let us say that a morpheme takes the form of a
triple: a syntax, a semantics and a phonology. Take the
morpheme cat: it has a syntax (it is a noun), a semantics (it
means ‘cat’) and a phonology, which takes the form /kæt/; we
will refer to this as the phonological form of the morpheme.
The phonological form of a morpheme may, clearly, consist of
more than one phoneme. Just as phonemes are mental objects,
so the phonological form of this morpheme is a mental object:
/kæt/ is a mental representation in the mind of a speaker,
whereas the sequence [kʰæt] is a phonetic sequence.
Let us now consider the adjectives impossible, imbalanced,

113
infelicitous, intangible, indirect, insane, incorrect and
inglorious. All consist of at least a prefix morpheme and a root
morpheme (some of these words have a prefix, a root and a
suffix). Many speakers have the following pronunciations of
these words:
(6)
(a)
[ɪmpʰɒsɪbɫ]
impossible
(b)
[ɪmbælənst]
imbalanced
(c)
[iɱfəlɪsɪtəs]
infelicitous
(d)
[ɪntʰændʒɪbɫ]
intangible
(e)
[ɪndɪɹεkt]
indirect
(f)
[ɪnseɪn]
insane
(g)
[ɪŋkəɹεkt]
incorrect
(h)
[ɪŋglɔ:ɹiəs]
inglorious
It is part of the native speaker’s unconscious linguistic

114
knowledge of English that these words all have the same
prefix. That prefix is one of the morphemes of English, and,
like all morphemes in the language, has a syntax (it is a prefix),
a semantics (it has a specific meaning) and a phonology. But
what is the phonological form of that morpheme? We know
from the data that the suffix may be realized as [ɪm], [ɪɱ], [ɪn]
or [ɪŋ]. It is clear, then, that the first phoneme in the prefix is /
ɪ/ and the second phoneme is a nasal, but which nasal
phoneme? We claimed above that English has three nasal
phonemes: /m/, /n/ and /ŋ/. So the phonological form of this
prefix might be [ɪm], [ɪn] or [ɪŋ]. Let us consider [ɪŋ]. We
could say that the /ŋ/ phoneme is realized as [n] before /t/, /d/
and /s/, and as [m] before [p] and [b]. This seems to make
sense: we can say that, when the prefix is added to a root, the
place of articulation of the nasal becomes identical to that of
the first consonant in the root. Thus, it is alveolar when
followed by an alveolar consonant (such as /t/, /d/ and /s/),
labio-dental when followed by a labio-dental consonant (such
as /f/), and bilabial when followed by a bilabial consonant
(such as /p/ or /b/). This is the process of assimilation we
referred to in chapter 2, in which one segment becomes similar,
in some respect, to another when the two are adjacent. Here,
the assimilation is in place of articulation. Further evidence that
nasals in English undergo place of articulation assimilation is
not hard to come by. Consider the following data, which are
representative of the speech of many speakers of English:
(7) Nasal assimilation in English
(a)
[ʌŋkʰlɪə]
unclear
(b)

115
 [ʌŋgɒdli]
ungodly
(c)
[ʌɱfεə]
unfair
(d)
[ʌɱvælju:d]
unvalued
(e)
[ʌntʰɹu:]
untrue
(f)
[ʌndʌn]
undone
(g)
[ʌmbεəɹəbəɫ]
unbearable
(h)
[ʌmbaɪəst]
unbiased
While the [ɪŋ] solution is plausible, it faces a difficulty: we
might equally say that the phonological form of the morpheme
is [ɪn], or [ɪm], and that, in either case, the nasal assimilates to
a following consonant. On the evidence presented thus far,
there is no non-arbitrary way of choosing between the three
alternatives: each is as plausible as the others. The following
data, however, allow us to make a non-arbitrary choice:
(8)
(a)
[ɪæktɪv]
inactive

116
 (b)
[ɪnɐpɹətɪv]
inoperative
(c)
[ɪnεfəbɫ]
ineffable
(d)
[ɪnədvaɪzəbɫ]
inadvisable
(e)
[ɪnɔ:dɪbɫ]
inaudible
(f)
[ɪneɪliənəbɫ]
inalienable
In each case, there is no consonant at the beginning of the
root to which the nasal could assimilate: each root begins with
a vowel. From the fact that vowel-initial roots are realized with
the [ɪn] form, we can therefore conclude that the phonology of
the prefix takes the form [ɪn], and that the nasal does not
change its place of articulation if the root-initial segment is a
vowel or an alveolar consonant. Note that this is generally true
of alveolar nasals in English, as the following data, involving
the prefix seen in (7), suggest:
(9)
(a)
[ʌneɪdəd]
unaided
(b)
[ʌnətʰɹæktɪv]
unattractive

117
 (c)
[ʌnivεntfəɫ]
uneventful
(d)
[ʌnɔ:θədɒks]
unorthodox
We might, of course, have said that the morpheme in
question has four different phonological forms: /ɪm/, /ɪɱ/, /ɪn/
and /ɪŋ/, and that words such as impossible, infelicitous,
indirect and incorrect are each stored mentally with the
appropriate prefix. There are two problems with this approach.
Firstly, there is no independent evidence that there is an /ɱ/
phoneme in English ([ɱ] never functions contrastively with
any other nasal). Secondly, even if there were no [ɪɱ] forms,
we would be committed to claiming, under the ‘several
phonological forms’ approach, that it is mere coincidence that
the /ɪm/ form is attached only to roots beginning with a bilabial
consonant, the /ɪŋ/ form only to roots beginning with a velar
consonant, and the /ɪn/ form only to roots beginning with
alveolar consonants and vowels. But that is surely an
implausible claim. So, for this sort of case at least, the idea that
we should postulate more than one phonological form for a
morpheme is deeply unattractive and implausible. Given the
data we have seen thus far, it appears much more plausible to
say that any given morpheme has a single phonological form.
And if that is the case, then it is the phonologist’s task to
hypothesize as to what that form might be. In doing so, she or
he will be guided by evidence and argumentation: the facts of
the matter, since they are mental in nature, and thus not
directly observable, will not be available for immediate

118
inspection via the five senses.
In adopting this ‘one phonological form per morpheme’
approach, we are allowing that, while any given morpheme has
only one phonological form, that phonological form may
‘correspond’ in some sense to a variety of different phonetic
forms. In the case we have just looked at, the prefix has the
phonological form [ɪn], but that in turn corresponds to four
different phonetic forms: [ɪm], [ɪɱ], [ɪn] and [ɪŋ]. Phonologists
refer to such phonetic forms as alternants: we say that there is
an alternation between the four forms. Which alternant of a
given morpheme occurs in a given word is entirely predictable;
there is a generalization which captures that predictability, and
we are able to express it in the form of a phonological rule, just
as we did with [ɾ] and [l] in Korean. In our English case, the
rule in question concerns nasals in general; it might be put
informally as:
(10) The rule of nasal assimilation in English
If the phonological form of a prefix ends in a nasal
then that nasal will assimilate to the place of
articulation of a following consonant.
We could have formulated this generalization as a formalized
rule, or as some kind of constraint on the phonological form of
morphemes in English. We will not go into the types of
formalism required to express such generalizations, or inquire
whether they are best expressed as rules or as constraints. The
most important point is that native speakers appear to be in
possession of generalizations of this sort, and that these appear
to constitute a part of their largely unconscious phonological
knowledge.
The data we have just considered also exemplify an
important phenomenon: that of phonemic overlapping. On

119
the basis of the data in (5), we postulated the following nasal
stop phonemes, with the realizations shown:
(11)

However, we have allowed that the /n/ phoneme may also be

realized as [m] if it precedes a bilabial consonant, or [ŋ] if it
precedes a velar consonant. This means that a given
occurrence of [m], for instance, may be either a realization of
/m/, as in map, or a realization of /n/, as in improbable. That
is, the /m/ and /n/ phonemes overlap in their realizations. We
may depict this as follows:
(12)

The question arises: how can the speaker of English tell

whether a given [m] is a realization of /m/ or of /n/? The
answer is that the phonological context allows the speaker to
tell: an [m] which does not precede a bilabial consonant will be
a realization of the /m/ phoneme. The phonemic contrast
between /m/ and /n/ is said to be neutralized before a bilabial
consonant. Neutralization is the suspension of phonemic
contrasts in one or more specifiable contexts.

6.3 English Vowel Phonemes

120