CHAPTER IX

PHONOLOGICAL PROCESSES AND RULES

Basic Competencies
After learning chapter IX, the students are required to achieve the
following competencies. They are able to:
a. explain the different types of alternation
b. explain the steps in writing the phonological rules
c. identify the types of phonological processes
d. formulate the rule of a certain phonological process

The previous chapter was concerned with establishing the phonemic system which
underlies the phonetic inventory of a language; that is deciding what the underlying set
of contrasts is. This chapter takes a closer look at this part of the phonological
component of the grammar, starting with some discussion of the range of phenomena
we have to account for as phonologists, and moving on to a more formal explication of
the conventions of rule writing.

A. Introduction
It is impossible to find two languages having exactly the same inventory of
phonemes which are realized by the same set of allophones; no two languages have
exactly the same phonological rules regulating the deployment of their sounds. The
phonological systems of different languages are obviously different. However, it is
certainly wrong to ignore the fact that there are the similarities between languages
because they are no less impressive than the differences.
The explanation in Chapter 5 makes it feasible to use a relatively small number of
DISTINCTIVE FEATURES like [±back], [±high], [±low], [±round], [±voice] etc. to
characterize the phonological contrasts found in all the world's languages. One possible
explanation for this is the fact that human anatomy and physiology impose limits on the
range of sounds which people call produce as speakers and discriminate as hearers.
Thus, for instance, since no human is endowed with a tongue which is so long that the
tip can curl all the way back to the throat, it is safe to predict that no language has apico-
uvular consonants made with the tip of the tongue and the uvula as the articulators (the
uvula is the fleshy bit that hangs down from the centre of the soft palate at the very back
of the mouth). On the other hand, given the case with which the tip and the blade of the
tongue can be raised towards the upper front teeth and the teeth ridge, it is not
surprising to discover that all languages have either dental or alveolar sounds, if not
both. Distinctive feature theory claims that there is a universal inventory of phonological
construction materials from which various languages chose different elements which
they use in building their phonological systems. Alternatively, distinctive features can
be likened to cooking ingredients on a supermarket shelf. The selection of ingredients
that a particular language puts in its shopping basket depends on the recipe which it
wishes to concoct.
It is significant, but not unexpected, that the phonological recipes which are
available fall within the range permitted by human biology. What is intriguing is the fact
that not everything that is biologically possible is equally likely to occur. Within the
range of possible sounds, certain articulatory parameters are exploited by languages
much more commonly than others.
Furthermore, besides exhibiting similarities in the features they use in structuring their
sounds, languages also show other phonological similarities. For example, although the
phonological systems of different languages are governed by different rules, the
variation which occurs does, for the most part, fall within certain parameters. Similar
phonological processes turn up, in language after language.

B. Alternations, Processes and Rules

Much of the focus of recent phonological thinking concerns the characterization
of predictable alternations between sounds found in natural languages. We've already
seen many examples of these alternations, such as that between [p] and [pH] in English.
Under specific conditions, there is an alternation between these phones: we get one, [p],
and not the other, [pH], after [s], as in [spIt], not *[spHIt]. That is, while at the
underlying (phonemic) level there is only one element, /p/, there is an alternation in the
representation of this element on the surface (phonetic) level between [p] and [pH],
which is determined by the environment in which the phoneme occurs.
We can characterize such alternations in terms of being caused by, or being due
to, some phonological process. In this particular case, we might call the process involved
‘aspiration'. In English, a voiceless stop is aspirated when it occurs in absolute word-
initial position before a stressed vowel (i.e. not following [s]).
We can represent processes, and thus characterize the alternations that result
from them, by means of rules. Rules are formal statements which express the
relationship between units on the different levels of the phonological component. In the
case of aspiration in English, we might have a rule such as:

(1)
- cont + syll
- voice [+ spread glottis] / # ______ + stress
- del rel

The feature [spread glottis] is used to characterize glottal states, including that
for aspiration. The above rule is a formal statement of the set of phonemes affected
(voiceless stop phonemes), the change which occurs (such stops are represented by the
aspirated allophones) and the condition under which such a change takes place (after a

148
word boundary - # - and before a stressed vowel). Note that the facts of aspiration in
English are somewhat more complex than our rule suggests, in that aspiration occurs
before any stressed vowel, even when the stop is not word-initial, as in 'a[pH]art'. A
fuller account involves reference to syllable boundaries.
It is the identification of such alternations, and of the phonological processes
behind them, and the formalizing of the most appropriate rules to capture them; that is
the main thrust of much of generative phonology. These alternations are a central part of
what native speakers 'know' about their language, and the goal of the generative
enterprise is the formal representation of such knowledge.

C. Alternation types
Phonological alternations come in many shapes and sizes and the processes
behind them are equally varied, as are the kinds of factor which condition them.
Consider the following sets of data from English; in what ways do the alternations
represented in (2) differ from one another?

(2)
a. [wIt] vs. [wI‚n]
[tu˘l] vs. [tu‚˘m]
b. i[n]edible, i[n] Edinburgh vs.
i[m]possible, i[m] Preston vs.
i[N]conceivable, i[N] Cardiff
c. book[s] vs. dog[z] vs. hors[Iz]
bat[s] vs. pen[z] vs. watch[Iz]
d. lea[f] vs. lea[v]es
hou[s]e vs. hou[z]es
e. electri[k] vs. electri[s]ity
medi[k]al vs. medi[s]inal

In (a), we see an alternation between purely oral vowel allophones - [I] and [u˘]
- which occur before an oral segment, and nasalized vowel allophones - [I‚] and [u‚˘]
which occur before a nasal segment. In (b) there is an alternation between different
realizations of the final nasal consonant in both the prefix 'in-' and the preposition 'in'; it
agrees in place of articulation with a following labial or velar consonant. In (c) we see
different realizations of the plural marker - orthographic '(e)s' - which may be [s], [z] or
[Iz], depending on the nature of the preceding segment. In (d) there is an alternation in
voicing for a root final fricative, voiceless in the singular, voiced in the plural. Finally, in
(e) we see alternation between a stop vs. fricative for the segment represented
orthographically by the 'c' in 'medical' and 'medicinal' and by the second 'c' in 'electric'
and 'electricity'.

149
 These sets of alternations are different from each other in a number of ways. The
type of alternation involved can vary: one or more of the allophones involved in the
alternation may be restricted to just one set of environments - like nasalized vowels in
English in (a), which only occur before nasal consonants - or the allophones may occur
'independently' elsewhere - and represent a different phoneme, as in the [m] of 'i[m]
Preston', which occurs 'in its own right' in words like 'ru[m]'. Or the factors conditioning
the alternation may vary. The alternation may occur whenever the phonetic
environment is met (as in vowel nasalization or nasal place agreement). On the other
hand, the alternation may be more restricted, and may only be found in the presence of
particular suffixes (like the plural) as in (c), or even particular lexical items, as in the [k]
vs. [s] alternation in 'electric/ity' in (e). In both these cases, the phonetic environment by
itself is not sufficient to trigger the alternation; if it were, words like 'dance' or 'rickety'
would be impossible in English - 'dance' has [s] following a voiced segment (compare
'dens'), 'rickety' has medial [k] not [s] (compare 'complicity'). Further, the alternation
may be 'optional' - or at least determined by factors other than the immediate phonetic
environment - like the variation in the final consonant of the preposition 'in', which
typically happens in faster speech styles rather than in slower ones (where the nasal may
not necessarily assimilate). The following sections deal with each of the types of
alternation in turn.

1. Phonetically conditioned alternations

Alternations like those in (a)—and (b), assuming normal speech style, given the
observation about slow speech immediately above—can be characterized as being
conditioned purely by the phonetic environment in which the phones in question occur,
with no other factors being relevant. If a vowel phone in English is followed by a nasal
consonant, the vowel is nasalized, irrespective of anything else (such as morphological
structure). Indeed, it is very difficult for English speakers to avoid nasalizing vowels in
this position, hence the designation of such alternations as 'obligatory'; there are unlikely
to be any exceptions to this process. Note, however, that this particular alternation is not
universally obligatory; in French, vowels in this position are not nasalized - [bçn] not
*[bç‚n] for bonne 'good (feminine)'.
Similarly, for (b), in English the alveolar nasal /n/ assimilates to the place of
articulation of a following labial or velar consonant, whether this is within a word or
across a word boundary. Again, this is difficult for speakers to avoid, although it is
somewhat easier than with vowel nasalization, possibly due to the influence of the
orthography. As with vowel nasalization, this assimilation is not universal; it does not,
for instance, occur in Russian - [funksj´] not *[fuNksj´] – compare English
[f√NkS´n].
Other alternations of this sort in English include aspirated vs. non-aspirated
voiceless stops discussed above, the lateral and nasal release of stops, 'flapping' in North

150
American, Northern Irish and Australian English, clear vs. dark /l/ and intrusive 'r' in
non-rhotic Englishes.

2. Phonetically and morphologically conditioned alternations

The alternations in (c) are also clearly motivated by the phonetic environment;
the form of the plural is dependent on the nature of the final segment of the noun stem.
If the noun ends in a sibilant, i.e. [s], [z], [S], [Z], [tS], or [dZ], the plural takes the form
[Iz]. If the final segment is a voiceless non-sibilant, the plural is a voiceless alveolar
fricative [s]. If the final segment is a voiced non-sibilant, the fricative is voiced [z].
However, unlike the alternations in (a) and (b) discussed above, the alternations
in (c) do not necessarily occur whenever the phonetic environment alone is met. If they
did, forms like [fEns] 'fence' or [beIs] 'base' would be impossible, since they involve
sequences of a voiced segment followed by a voiceless alveolar fricative. So the phonetic
environment cannot be the only relevant conditioning factor; something else must be
taken into account as well. The 'something else' in this instance is clearly the internal
complexity of the words, in that the plural marker's' has been added. The word can be
seen to consist of two separable units, known as morphemes - e.g. 'fen + s' consists of the
stem 'fen' plus the plural marker '-s'. Words like 'fens' are said to be morphologically
complex. The final fricative only agrees in voice with the preceding segment if it
represents the plural marker, i.e. if there is a morpheme boundary between the two
segments. Thus voicing agreement will occur in 'fens' (fen + s, where '+' indicates a
morpheme boundary) and in 'bays' (bay + s), giving [fEnz] and [beIz]. On the other
hand, 'fence' and 'base' are both morphologically simple forms: they have no internal
morphological boundaries, and thus no voicing agreement takes place.
Like the alternations previously discussed, this type of alternation is obligatory
and automatic; it occurs whenever both the phonetic and morphological conditions are
met. Speakers never say things like *dog[Iz] or *book[z], and the alternations will occur
even with completely new words; if we were to launch some product called a 'plotch',
the plural would have to be 'plo[tSIz]', and not *'plo[[tSz]' or *'plo[tSs]'. When an
alternation behaves in this predictable, automatic manner, applying freely to new forms,
it is known as productive.

3. Phonetically, morphologically and lexically conditioned alternations

Consider now the alternations in (d) and (e). Here there is clearly some phonetic
conditioning: fricatives are voiced between vowels (voicing assimilation) in (d), and a
velar stop [k] is fronted and fricativized to an alveolar fricative [s] before a high front
(that is palatal) vowel segment in (e). The latter is also a kind of assimilation, though
somewhat more complex, involving both manner and place of articulation—the term for
this particular process is velar softening.
There is also clearly some morphological conditioning too in that, for instance,
[bEIsIs] 'basis' and [kIt] 'kit' are both well formed (they don't become *[bEIzIz] and
*[sIt] respectively, even though their phonetic environments are the same as those

151
involved in the alternations above). But even stating that there must be a morpheme
boundary after the final fricative in cases like 'leaf' or after the final stop in cases like
'electric' is insufficient, since we don't get these alternations with, for example, 'chie[fs]'
(not *'chie[vz]') or with 'li[k]ing' (not *'li[s]ing').
In these cases we must, thus, also specify the particular (set of) lexical items the
alternation is relevant for: only some of the fricative final nouns in English show voicing
assimilation and only some [k]-final stems exhibit velar softening. Furthermore, unlike
the alternations in the previous two sections, alternations involving lexical conditioning
are not typically productive (or are at best intermittently so); a new product called a
'plee[f]' would have the plural 'plee[fs]' rather than 'plee[vz]'.
Other alternations of this type in English include the so-called vowel shift or
trisyllabic shortening pairs like 'rept[aI]l' /'rept[I]lian', 'obs[i˘]n' /'obs[E]nity',
'ins[eI]n'/'ins[a˘]nity'. Such alternations are often the 'fossilized' remains of
alternations/processes which were once productive at an earlier point in the history of
the language, but have since died out. The pairs given immediately above are due to a
series of changes during the history of English, including the late Middle English 'Great
Vowel Shift', hence one of the names given to the alternation.

4. Non-phonological alternations: Suppletion

Consider finally alternations like 'mouse' vs. 'mice', or 'go' vs. 'went'. Are these
the same kind of alternations as those we have looked at in the preceding sections? They
might at first glance seem to be like the last set described in Section 3, in that while there
is morphological conditioning (plural and past tense, respectively) we must also refer to
specific lexical items, since the alternations do not generalize over all similar forms, or
extend to new ones (the plural of 'grouse' isn't 'grice', the past tense of 'hoe' isn't 'hent' or
some such). Importantly, however, there is one crucial type of conditioning which is
absent here: there is no phonetic conditioning of any obvious sort which might help
predict the alternations involved. That is, there are no general phonological processes
involved in getting from 'mouse' to 'mice' or from 'go' to 'went'. These forms must be
learnt by the speaker on a one-off basis, as exceptions to a rule (hence children acquiring
English often produce 'regularized' forms like 'mouses' and 'goed').
The introduction into a set of alternations (a paradigm) of a form that is not
obviously related, as in the instances here, is known as suppletion, and is not part of our
phonological knowledge (since it has no phonological basis). It thus need not be dealt
with by the phonological component.
Still, it might be thought that alternations like 'mouse/mice' are more like those in
Section 3 than the clearly unrelated 'go/went' type, in that there is some obvious relation
between the forms: only the vowel is different, rather like 'inane/inanity' (and
furthermore, like the trisyllabic shortening pairs. the 'mouse/mice' alternations are the
fossilized remains of an earlier process, Old English i-mutation). There is at least one
important difference, however: for 'inane/inanity' it is the addition of two extra syllables
to the stem which triggers the alternation (hence the term 'trisyllabic shortening', since
the alternating vowel is now the first of three syllables). For 'mouse/mice', on the other

152
hand, there is no phonetic or phonological change to trigger the alternation. It is solely
dependent on being a plural form of one of a small set of English nouns.

D. Formal rules and rule writing

In the previous two sections of this chapter, and indeed throughout this book, we
have been concerned with looking at the kinds of things that speech sounds do in
language, the changes they undergo and the processes that occur. In a certain respect
this is only half the picture since, beyond simply observing what goes on, the
phonologist wants both to characterize or represent these processes and to try to
understand how they work. The rest of this chapter will focus on representing these
processes. However, this will be only one sort of representation, and a fairly basic sort of
representation besides. In the following chapters we will see why the representations
here are not the entire story and why they need to be improved on.
At this point you might wonder why, if the representations we're about to
examine are inadequate, do we bother with these and not go straight on to other ways of
representing phonological processes that may capture greater generalizations. There are
two reasons for this. First of all, the kinds of rules and rule formulation we'll deal with in
this chapter pre-date the fuller representations we'll see in Chapter 9 and some of the
more general concerns we look at in Chapters 10 and II. Understanding the formalisms
presented here enables you to start reading some of the older papers on phonology that
would be inaccessible if you understood only where phonology currently stands.
Second, dealing first with more 'basic' sorts of representation helps us see where modern
phonology has come from and why richer representations are needed.

1. Formal rules
In Chapter 7 we looked at the fundamentals of rule formulation, that a rule in
phonology consists of some phonological element (A)—typically a segment or a
feature—which undergoes some change (B) in a particular environment:

(3) A
B/ X - Y

The rule in (3) represents the state of affairs in which A becomes B between X and Y. We
could take as a concrete example the flapping rule of American English, according to
which a /t/ is pronounced as a flap [R] when it occurs between two vowels, V, provided
that the second vowel is not stressed. So, A = /t/, B = [R], X = V, Y = V, as shown in the
following rule:

(4) /t/--
[r] / V ___V

[- stress]

This rule applies to words like writer, later, transmitter and bitter.

153
 We also understand, as in (4), that the bits of phonology represented by A, B, X,
Y are either segments, or features associated with segments. That is, they are either
complete feature matrices or individual features. As a further example, we might have a
rule like that in (5):

(5) /t/ ---
[/] / V ___#

This rule would capture the process found in many varieties of English by which
a /t/ becomes a glottal stop after a vowel at the end of a word, e.g. in words like 'cat' and
'hit': /kQt/ and /hIt/ which surface as [kHQ/] and [hI/]. That is, phoneme /t/ becomes
allophone [/] when preceded by a vowel and followed by the end of a word.
More often than not rules are written in terms of the relevant features, not whole
feature matrices represented by segments. The rule for glottalisation we have just seen
can also be recast in (6) using the feature [constricted glottis], where the' +' value
indicates glottal closure.

(6) Glottalisation:

- cont - ant
+ ant - cor / [+ syll] _____ #
+ cor + const glottis

As we saw in Chapter 6, using features, rather than segments, allows us to

capture greater generalizations. Using features in rules expresses these generalizations.
In this case using the features [ - continuant] and [- voice] allows the rule to express a
process affecting the entire class of voiceless stops of English, where a segment-based
attempt would require several rules, one for each stop.

(7) a. /p/ [p/]/ _____ #

b. /t/ [t/]/ _____ #
c. /k/ [k/]/ _____ #

2. Rule writing

Parentheses notation
In addition to these basic rules, there are also notational devices and conventions
used to express more complex relationships and operations. One of these conventions
involves parentheses - ( ) - which are used to enclose optional elements in rules. The rule

154
in (8) shows that A becomes B either between X and Z or between XY and Z. The
optional element is Y, which mayor may not be present.

(8) A B/ X(Y)____Z

Although this is written as a single rule, it in fact encodes two separate but related rules,
namely A B/ X____Z and A B/ XY____Z

To illustrate the application of parentheses notation let us look at 'l-velarisation'

in English. We saw that most varieties of English have a clear-l [l] and a dark or
velarized-l[lÚ]. So words like 'leaf' have a clear-l and words like 'fell' and 'bulk' have a
velarized-l. The distributional facts are actually more complex than this. These two
words - 'fell' and 'bulk' - show l-velarisation occurring either at the end of a word, or
before a consonant at the end of a word. That is, there is an optional consonant which
may intervene between the /l/ and #:

(9) /l/ [lÚ] / ______(C) #

The parentheses here indicate that there may or may not be a consonant between the
lateral and the end of the word.

Braces
Another notational device used in linear rule writing is brace notation, also
known as curly brackets: { }. Brace notation represents an either/or relationship between
two environments. In other words, the same process occurs in two partially different
environments and the rule captures the fact that it is the same process, despite the
difference in environment.

(10) A B/ X ____ Y
Z

The rule in (10) shows that A becomes B either between X and Y or between Z and Y. In
other words, A B/ X ___Y or A B/ Z ___ Y. Note that in (10) parentheses
have not been used. Therefore either X or Z must be present; both cannot be absent.
Recalling the rule in (.5) glottalising final-t, we also find that /t/ [/]/ ___ C, as
in 'petrol' [pHE/®´l]. Since this /t/ isn't at the end of the word we appear to have an
either/or environment: either before the end of a word or before another consonant.

(11) /t/ [/] / _____ #

C

155
Here we see that /t/ surfaces as glottal stop [/] either before another consonant or before
the end of the word.

Both parentheses and braces can appear in the same rule, allowing overlapping
environments to be captured in terms of a single rule. Take for example the rules in (12).

(12) A B / X ______ Y
A B / XZ ____ Y
A B / X ______ #
A B / XZ_____ #

These rules can be collapsed into a single rule, as in (13).

(13) A B / X(Z) _____ Y

#

The use of devices like parentheses and braces increases the power of the model and
allows us the capacity to formulate rules of greater complexity. This rule captures the
generalization that there is some process which changes A to B and that this process
occurs in a number of different environments. The advantage of this Over the list of
rules in (12) is this: by expressing this change as a single rule we are presumably saying
something important about the relationship between A and B that is not captured by a
list. In the list there is no reason that each of the four rules should involve A B; in
the single rule each of the four statements must involve A B.

Superscripts and subscripts

Superscript and subscript numbers associated with variables let us express
minimum and maximum numbers of segments relevant to a given environment. When
C0 occurs in a rule, the interpretation is “zero or more consonants.” Any integer can
appear as a subscript on a segment, and simply means that number of segments or
more: e.g., C1 means one or more consonants, C2, two or more consonants, and so on. A
subscript on a segment imposes a lower bound on the number of those segments needed
for the rule to operate. To indicate an upper bound, superscripts are used. Thus, C10
means zero or one consonant; C , between zero and two consonants; C42, between two
and four consonants, and so on. The upper bound restriction can be equivalently
expressed through the parenthesis notation.

C10 = (C), C20 = (C) (C), C21 = C (C)

If the segment contains a superscript but no subscript, the interpretation is

“exactly that number of segments.” Hence, C1 means exactly one consonant and is
equivalent to C; C2, exactly two consonants and is equivalent to CC; and so on.

156
E. Phonological Processes

When morphemes are combined to form words, the segments of neighboring

morphemes become juxtaposed and sometimes undergo change. Consider the
morphologically related forms electric, electrical, electricity, and fanatic, fanatical,
fanaticism. Here the final k of electric and fanatic becomes s before a morpheme beginning
with i. Changes also occur in environments other than those in which two morphemes
come together—for example, word initial and word final positions, or the relation of a
segment vis-a-vis a stressed vowel. All such changes will be called phonological processes.
In an effort to avoid a bewildering array of disparate types, we will organize
phonological processes into four categories: assimilation, where segments become more
alike; syllable structure, where there is alteration in the distribution of consonants and
vowels; weakening and strengthening, where segments are modified according to their
position in the word; and neutralization, where segments merge in a particular
environment. Examples of these processes are drawn from synchronic descriptions as
well as from historical change.

1. Assimilation
In assimilatory processes a segment takes on features from a neighboring segment.
A consonant may pick up features from a vowel, a vowel may take on features of a
consonant, one consonant may influence another, or one vowel may have an effect on
another.

a. Consonant assimilates vowel features

Features of a vowel may be extended onto a consonant as secondary modifica-
tions. Palatalization and labialization are common processes of this type. In
palatalization, the tongue position of a front vowel is superimposed on an adjacent
consonant; in labialization, the lip position of a rounded vowel induces a secondary
articulation onto the consonant.
In English, the alternations exemplified in electric, electricity, and analogous,
analogy reflect a historical palatalization followed by a shift in place of articulation.

b. Vowel assimilates consonant features

Features from a consonant may be superimposed on a vowel. In this kind of
assimilation, the modification of the vowel is usually allophonic. It is quite common for
vowels to be phonetically nasalized when adjacent to a nasal consonant, a process which
occurs in English: see [si˘], seen [si‚˘n], cat [kHQt], can't [kHQ‚nt].

c. Consonant assimilates consonant features

One of the most widespread phenomena is for consonant clusters to agree in
voicing. This process can be seen in English where the endings for the plural, third

157
person singular, and past tense agree in voicing with a preceding consonant. Thus, one
finds s and t after voiceless consonants, z and d after voiced ones.

/k√ps/ cups /k√bz/ cubs

/pQts/ pats /pQdz/ pads
/bQkt/ backed /bQgd/ bagged
/®eIst/ raced /®eIzd/ raised

It is common for a nasal consonant to become homorganic with a following

consonant—that is, the nasal adopts the same place of articulation. In English, the
negative prefix in- becomes homorganic to a following stop: e.g., inadvisable, but
impossible, imbalance, intolerance, indecisive, incoherent, the last with N for some speakers.
.
d. Vowel assimilates vowel features
The vowel of one syllable may become more like the vowel of some other syllable.
Here we can distinguish vowel harmony from umlauting.
Vowel harmony is a case in which vowels agree in certain features. In Turkish, the
high vowels of a suffix agree in back ness- and rounding with the stem vowel. In
modern German, umlauting is no longer restricted to a following high front vowel,
although it is believed that all the umlauting environments did arise in this way.
In English, irregular plurals such as foot, feet and mouse, mice are vestiges of an
umlauting process which was fairly extensive in Old English.

2. Syllable Structure Processes

Syllable structure processes affect the relative distribution of consonants and
vowels within the word. Consonants or vowels may be deleted or inserted. Two
segments may coalesce into a single segment. A segment may change major class
features, such as a vowel becoming a glide. Two segments may interchange. Any of
these processes could cause an alteration in the original syllable structure.
We shall consider CV syllable structure-a syllable containing a single consonant
and vowel-to be basic. Any process which takes a more complex syllabic structure and
reduces it to the CV pattern leads to a preferred syllable structure. The effect of such
processes is to break up clusters of consonants or sequences of vowels. For example, a
cluster of two consonants could be simplified in one of three ways: one of the
consonants could be deleted, a vowel could be inserted between the two consonants, or
the two consonants could coalesce into a single segment.

a. Consonant deletion
In some r-less dialects of English, word final r is dropped before a consonant or in
phrase final position, but not before a vowel: e.g. father came, I saw father, father arrived.
The distribution of the indefinite article also conforms to preferred syllable structure: an
apple, a banana.

158
b. Vowel deletion
Certain English morphemes terminating in a vowel drop the vowel before a
suffix beginning with a vowel: Mexico, Mexican (derived from Mexico+an); cello, cellist
(cello+ist).

c. Consonant insertion (epenthesis)

In some dialects of English, r is inserted whenever a word ending in a schwa is
followed by a word beginning with a vowel: e.g., the idea came. the idea-r--is good.

d. Vowel insertion (epenthesis)

Latin words cannot end in a consonant-liquid cluster. In such cases, the vowel e
is inserted to break up this cluster. Note the nominative forms where there is no overt
ending for the nominative case.
In English, schwa is inserted between final consonant-sonorant clusters: e.g.,
central, center [sE‚nt´®]; cyclic, cycle [saIk´l]; spasmic, spasm [spQz´m]. That the schwa
is inserted, rather than deleted, in a medial syllable can be seen by comparing wintry
with watery; the latter exhibits schwa in the medial syllable as well.

e. Consonant coalescence
Two contiguous consonants are replaced by a single one which shares features of
the two original ones. Hence, coalescence involves a kind of assimilation. In Korean,
whenever a non-continuant and h are contiguous they are replaced by an aspirated non-
continuant. Other frequent examples of consonant coalescence include: a consonant plus
glottal stop coalescing to a glottalized consonant; a consonant plus y to a palatalized
consonant; a consonant plus w to a labialized consonant; and a stop plus fricative to an
affricate.
In English, morpheme final t. d. s. and z and a following y are replaced by palato-
alveolar fricatives. This is particularly evident before the suffix-ion: e.g.. relate. relation;
erade. evasion; regress, regression; confuse, confusion.
For the English prefix in-, the n assimilates to a following liquid: e.g. irresponsible,
illegal. In the spoken language these geminate clusters have become degeminated.
Since coalescence involves both assimilation and reduction, many of these
examples could be described as the joint action of these two processes. Consider the
change in which a vowel plus nasal consonant becomes a nasalized vowel. First there is
assimilation-the vowel becomes nasalized before the nasal consonant-and the nasal
consonant is then deleted. Historically, the nasalized vowels of French did evolve in this
way. However, there is not sufficient evidence to suggest that all types of coalescence
should be treated as assimilation followed by deletion.

159
f. Metathesis
Metathesis refers to the reversal of a sequence of elements, often segments, in a
word. Modem English 'bird', 'first' and 'third' have historically earlier forms 'brid', 'frist'
and 'thridde', respectively. In each of these cases the sequence of [r] and [i] has reversed
(though in non-rhotic varieties a further change has resulted in the loss of [r] after
metathesis). This can be represented abstractly by assigning an index (number) to the
segments involved and showing a reversal of the index numbers of two of them.
In Hanunoo, a sequence of glottal stop plus consonant becomes consonant plus
glottal stop. This metathesis is likely to be the prelude to a subsequent coalescence
whereby the consonant plus glottal stop becomes a glottalized consonant, a unit
segment, thereby simplifying the syllable structure.

3. Weakening and Strengthening

Not all changes in syllabic structure necessarily lead to a simpler syllable
structure. The syllable structure would become more complex, for example, if a vowel in
an original CVCV configuration were to be deleted so that two consonants came
together. Such deletions are often caused by segments occupying a weak position in the
syllabic. In the following processes the important factor is the weakening, and any
changes in syllable structure are incidental.

a. Syncope and Apocope

In syncope a vowel near a stressed vowel is deleted. This phenomenon occurred

in the development from Latin to French. In words with antepenultimate stress-where
the stress is on the third syllabic from the end of the word-the penultimate vowel, or the
vowel between the stressed and final vowels, was dropped.

In English, when the stressed syllable is followed by two unstressed ones, the
vowel immediately following the stressed syllable is often dropped in colloquial speech,
particularly if it is followed by a single sonorant consonant: e.g., chocolate, choc'late;
happening, happ’ning; every, ev’ry, nursery, nurs’ry.

Apocope is the loss of a final unstressed vowel, most often a reduced or schwa
like vowel. In colloquial French final schwa is usually dropped, whereas it would not
necessarily be in more formal styles.

Formal Colloquial
French Frech
------------------------------------------------------------------------------
egliz´ egliz church
ruZ´ ruZ red
tabl´ tabl table
fij´ fij girl

160
b. Vowel reduction
Vowel reduction involves the weakening of unstressed vowels to schwa. English
displays morphological alternations between a stressed full vowel and an unstressed
reduced (schwa) vowel.
e¤Ib´l able ´bi¤l´ti ability

kQ@n´d´ Canada k´ne@IdIj´n Canadian

fo@Ut´g®Qf photograph f´tÅ@g®´fi photography

su˘p´” super s´pi@˘®ij´” superior

c. Diphthongization
Stressed vowels and tense vowels are the strong ones. Whereas weak vowels
may undergo syncope, apocope, or reduction, strong vowels frequently diphthongize. In
Romance, Latin e and o became diphthongs in certain environments. Italian
diphthongization took place when the vowel was stressed and in an open syllable: e
became je, and o became wo. The glide which developed has the same backness and
rounding as the following vowel.
In English, the tense vowels i, e, u, and o phonetically are often diphthongs: [ij],
[ej], [uw], [ow]. Here too the glides have the same backness and rounding as the vowel.

4. Neutralization
Neutralization is a process whereby phonological distinctions are reduced in a
particular environment. Hence, segments which contrast in one environment have the
same representation in the environment of neutralization.

a. Consonant neutralization
Neutralization of word final obstruents takes place in German. In initial and
intervocalic positions, voiced and voiceless obstruents are in contrast. Only voiceless
ones are found in word final position, so in this environment there is neutralization
between pairs of voiced and voiceless obstruents.

b. Vowel neutralization
Russian has a five-vowel system for its stressed vowels. When these vowels
appear in unstressed position there is neutralization: Both i and e appear as i, both a and
o appear as a, and u remains u. Thus, in unstressed position, Russian goes from a five-
vowel to a three-vowel system. (Russian also neutralizes voicing of obstruents in word
final position.) . In French, all nasalized vowels are low. Different oral vowels may have
the same nasalized partner. Hence, there is neutralization in tongue height for the
nasalized vowels.

161
 There appears to be an interrelationship between neutralization and assimilation,
or between neutralization and weak position. Where obstruent clusters agree in voicing,
contrasts in voicing are neutralized. By the same token, if a nasal consonant becomes
homorganic to a following consonant, then nasal consonants of different places of
articulation can no longer contrast in those environments, and one could view this
assimilation as a type of neutralization. In Russian, vowel neutralization affects the
unstressed vowels, which are weaker than stressed ones. In English, unstressed vowels
reduce. This process is also neutralization, since different unstressed vowels all merge to
schwa. Nasalized vowels are perceptually more obscure, hence weaker than oral vowels.
Nasalized vowels are neutralized in French. It may be that all neutralizations could be
subsumed under either assimilation or weakening.

Summary
In this chapter we have considered the different types of phonological
alternations and processes found in languages. We have also examined how these
alternations and processes may be expressed in terms of formal notation as rules. These
rules provide a way of linking the underlying phonemic level with the surface phonetic
level. In the next chapter we examine the nature of the phonological structures on which
such rules operate.
Most phonological processes can be explained as articulatory or perceptual
phenomena. Assimilation has a natural explanation in coarticuIation. During the
formation of a sound, the articulatory organs may be anticipating the articulation for
another sound, and consequently the first sound will be modified in the direction of the
second, or the articulation of the first may be carried over into that of the second.
Coarticulation effects are readily observed when consonants become palatalized or
labialized before palatal (front) or labial (rounded) vowels, or vowels are nasalized in
the vicinity of a nasal consonant, or the place of articulation for a consonant induces a
similar place of articulation on a preceding nasal. Other kinds of assimilation may be
related to inherent constraints on the articulatory mechanism. In languages which have
voicing contrasts for obstruents, invariably in clusters, the distinctions are neutralized
and all obstruents must agree in voicing. This type of assimilation appears to be a
consequence of inherent difficulties in adjusting the glottis for different voicing states for
sequences of segments of the same type.
Other phonological processes can be explained through perception. Segments
which are maximally differentiated, which are perceptually more opposed to one
another, are more stable than those which are less differentiated. Stressed vowels are
perceptually stronger than unstressed ones. The former frequently diphthongize, a
process which makes them even more perceptible. Unstressed vowels have less
perceptual distance among themselves, and may therefore be neutralized, a process
leading to fewer vowels, but with greater perceptual distance between adjacent ones. Or
unstressed vowels may reduce to schwa, thus being maximally opposed to tense vowels,
or they may drop out together. There are also interrelation between articulation and
perception. The optimal articulatory contrast is that between a closed vocal tract and an

162
opened one, in other words, between consonants and vowels. Processes for preferred
syllable structure lead to this optimal alternation.

Further reading
At the core of early generative phonology, focusing on rules and representations, is
Chomsky and Halle (1968) which is, however, rather daunting. More accessible and
recent works on generative phonology include Spencer (1996), Kenstowicz (1994), Carr
(1993), Durand (1990).

Exercises

1. What processes are involved in the relationship between:

a. [g®Qnd] grand and [g®QmpA] grandpa
b. [post] post and [posm´n] postman
2. Using distinctive feature specifications, explain why palatalization is an example of
assimilation. What features seem to be particularly crucial in accounting for this
assimilation?
3. Label the phonological processes which account for the types of changes indicated
below. Be sure to give the specific subcategory of a process where appropriate.
a. s becomes z when followed by a voiced sound
b. k is added when preceded by a nasal and followed by s
c. ps becomes sp at the end of the word
d. t becomes s when followed by s
e. b becomes p when preceded by a voiceless sound

4. The word-initial sequences of /pl-/, /bl-/, /kl-/, /gl-/, /sl/, and /fl/ are permissible in
English, but other sequences involving /l/ are not. Can you produce a description of
the required features a consonant must have in order to precede /l/?

163