Linguistics

Introduction to Phonetics and Phonology

Generative Phonology: Introduction

Principal Investigator Prof. Pramod Pandey

Centre for Linguistics, SLL&CS, Jawaharlal Nehru University,
New Delhi-110067
pkspandey@[Link]
011-26704226; 011-26741258, -9810979446
Paper Coordinator Prof. Pramod Pandey

M odule ID & Name Lings_P2_M _30: Generative Phonology: Introduction

Content Writer Pramod Pandey

Content Reviewer Prof Hemalatha Nagarajan

Lings_P2_M31: Generative Phonology

Objectives:
• To introduce students to the model of the theory of Generative Phonology
presented in Chomsky & Halle’s 1968 book, Sound Pattern of English (also
known as SPE).

Contents

1 Introduction
2 Origin of generative phonology as a critique of American Structuralist Phonology
3 Premises of the theory of generative phonology
4 Main tenets of the theory of generative phonology
5 Illustration of the tenets with phonological data.
6 The model of generative phonology
7 Analysis of data and testing hypotheses
8 Summary

1
1 Introduction

In the present and the following modules, we introduce ourselves to the theory of

Generative Phonology. We discuss the origin of the theory as a criticism of American

Structural Phonology and go on to present its premises and tenets. We will look at the

model of the theory and how it helps achieve the goal of the theory as part of the larger

goal of generative grammar. The introduction to the theory will be followed by data

analysis and summary.

2 Origin of generative phonology

The goal of phonological theory came to be part of the goal of linguistic theory. Instead

of describing a language, phonological theory aimed at explaining the nature of

phonological knowledge as part of linguistic knowledge. Towards this end, the notion of

the phoneme had come to be questioned as the best way of representing speech sounds.

Morris Halle had worked on the notion of distinctive features in collaboration with the

Prague School linguist Roman Jakobson and the the Swedish linguist Gunnar Fant.

Jakobson, along with Trubetzkoy, had proposed the notion of distinctive features, as

seen in the previous modules. Morris Halle attempted to show with very convincing

arguments that the representation of segmental sounds in terms of distinctive features

could help make phonological generalizations more plausible. They could also help

capture the notion of a natural class, found to be typically involved in phonological

processes. He also showed how the use of distinctive features could make grammars

simpler and more elegant.

2
Around the beginning of generative grammar, Chomsky had strongly criticized the

goals and methods of analysis of American linguistics. He had shown how the basic

principle of Biuniqueness could not be adhered to in an attempt at explaining the

phonological knowledge of a speaker. It was also shown, famously by Halle through the

analysis of Russian data, that the case of phonological processes which are found to be

at work both within and across morphemes militate against the notion of the phoneme.

They also militate against the notion of a strict division of levels of linguistic analysis- a

tenet of American Structural linguistics. (See the discussion of these issues in an earlier

module on ‘Problems in Phonemic Analysis’). There was need to replace the notion of

the phoneme and allophone with notions of underlying representation and surface

phonetic representation. (Although this attempt has been successful, practicing

phonologists both within generative linguistics and outside it continue to use the notion

of the phoneme and the allophone as part of basic linguistic descriptions.). In what

follows, we take a look at the various aspects of structural phonology that was found to

be inadequate in explaining phonological knowledge and helped the growth of

generative phonology.

3
[Link]

[Link]

3 The Main Premises of Generative Phonology

The model of the theory of generative phonology presented in SPE was formally

introduced in Morris Halle’s 1962 paper, ‘Phonology in Generative Grammar’ practiced

for well over two decades, with its first. The main premises of the theory are the

following: (see Kenstowicz

[Link]

(1)

a. Phonology is part of a generative grammar and phonological knowledge is

part of linguistic knowledge in that phonological structure reflects the
linguistic competence of the individual native speaker;
b. The competence of the individual native speaker lies in the ability to
compute the phonetic representation on the basis of which to interpret the
syntactic structure of a sentence in a language;

4
 c. Phonological competence as part of an individual speaker’s linguistic
competence can be investigated following the methods of normal science.

These premises of the theory of generative phonology have proved very fertile in

influencing the growth of knowledge of sound structure in diverse fields such as

language acquisition, language change, language loss and language cognition. Let us

briefly explain the premises.

Generative grammar professedly differed from American structural linguistic theory in

its goal. Whereas Structural linguistics had the description of language as its goal,

generative grammar considered explanation of linguistic knowledge as its goal. What

this meant was that linguistic analysis must account for the way a language is known in

the mind/brain of the speaker-hearer, and not simply describe a language. In order to

achieve the goal of accounting for linguistic knowledge, a theory of language had to be

based on universal principles, like a theory of science. Generative grammar was thus

proposed as an instance of a theory that accounted for the knowledge of a language as

an instance of universal grammar. The theory had three main components in its first

proposal in Chomsky’s Syntactic Structures- a lexicon, a syntactic component and a

phonological component. Phonology was assumed to depend on the syntactic structure

generated by the syntactic component. The syntactic and the phonological components

were generative in the sense that the syntactic and the phonological components

generated an infinite number of sentences with their syntactic and phonetic forms by

means of finite sets of syntactic and phonological units and rules. All the idiosyncratic

properties and exceptions were stored in the lexicon. Thus with respect to phonology,

words such as cat and dog were stored in the lexicon because the sounds in them could

5
not be predicted. However, the plural forms of those words- cats /kæts/ and /dɔɡz/- are

not stored, rather they are generated in the phonological component. A phonological

rule applies to give the phoneme /s/ in one word and the phoneme /z/ in the other. The

fact that other words, such as socks and bags also have the regular difference in the

pronunciation of the plural morpheme /sɔks/ and /bæɡz/ supports the claim. What is

interesting is that non-sense forms such as *zock and *bleig will also have the same

pronunciations as they end in /k/ and /ɡ/ respectively. You will recall that the plural

morpheme has the allomorphs /s/ and /z/ following voiceless and voiced sounds

respectively, excepting /s z ʃ ʒ ʧ ʤ/.

The syntactic structure and the phonological structure are inter-related in the following

sense. Given a sentence and its syntactic structure, phonological rules apply to yield its

phonetic form. And given the phonetic form of a sentence the phonological units and

rules of the phonological component phonological rules work out the syntactic form of a

sentence. The two components are integrated. (As the theory of generative grammar

developed the model was revised to meet with the growth of the theory. See the

modules on syntax in Paper 3 and Paper 6).

The theory has grown because of its emphasis on refutation and modification, the

two main characteristic features of normal science, as argued by Karl Popper. The

property of a theory to be refutable requires it to be presented in an objective way so

that it can be proved to be wrong. (For example, the sentence There are many students

in this class is not refutable, as the word many has subjective meaning and can have

multiple interpretations. However, the sentence There are 25 students in this class is

refutable, as the number of students being 25 can be verified to be true or false). When

6
a scientific theory is refuted it needs to be modified. Thus, the growth of a scientific

theory is marked not so much by the correctness of facts in it as the properties of it

being refutable and modifiable. The generative linguistic theory is a scientific theory in

that its claims are refutable and the theory modifiable.

As we will see in the course on Advanced Phonology (Paper 5), the premises continue

to hold grounds, even though the theory has been substantially modified.

4 The Main Tenets of Classical Generative Phonology

Based on the premises of the theory as discussed above, classical generative phonology,
as presented in SPE, came to have a set of tenets. Goldsmith & Lacs point out four
main tenets of classical generative phonology, as summarized below:

([Link]

(2)

a. First, the goal of the theory of generative phonology, in accordance with the
theory of generative grammar, was to develop an explicit phonology that
generates all and only the surface forms of a language.
b. Second, the phonological representations in the explicit phonologies “were
linear sequences of matrices of feature values.”
c. Third, in order to generate the forms of a language, phonology “should
employ derivational means (that is, sequential, processual analyses)…”
d. Fourth, a high priority of the theory was the discovery and empirical
establishing of explicit rule ordering which was consistent with the data.

The first tenet is a restatement of the premise about phonology being a part of

generative grammar. It continues to hold today. Not so the other three tenets, which

were considered to be according to the goal of the development of a scientific

7
generative theory of phonological competence. They provided the underpinnings of the

classical generative theory. We will take up each of these tenets and explain them in the

present and the following two modules. We will begin with discussing the model of

generative phonology, and take up the other tenets for elaboration

• Levels of representation and rule in generative phonology

• Phonological alternation

• Rule formalism in generative phonology

• Rule ordering in generative phonology

Of these, the last two topics are taken up for discussion and elaboration in the following

modules.

5 The Model of Generative Phonology

As discussed above, generative phonology is part of the theoretical model of generative
grammar. Since its first proposal in Syntactic Structures (1957), the model has
undergone several modifications. The first model of generative grammar is presented
below:
(3)
Lexicon

Syntax

Phonology

The Syntactic Structure model of generative grammar

In the above model, the lexicon is the repository of all lexical information including all
irregularities- phonological, morphological and syntactic. Syntax is the central

8
component of grammar that includes all morphological and syntactic operations. The
phonological component follows syntax and all syntactic operations.

The generative model has undergone modifications; the most prominent ones are the
Government-Binding (GB) and the Minimalist theoretical models, presented in (4) and
(5).
(4)
DS

SS

PF LF

(5)

LEXICON

SPELL-OUT

PF LF

The Minimalist model of Generative Grammar

In the minimalist model word-level morphological and phonological operations take

place in the lexicon. Syntax is the mapping through spell-out operations between the
phonetic form and the logical form. Whereas the phonetic form includes both language-
specific and universal audio-perceptual properties of sentences, logical form is
necessarily universal; it has “properties of semantic representation that are strictly
determined by grammar, abstracted from other cognitive systems”. (See more on this in
Paper 6, Module 9 on Minimalist syntax.)

9
The earliest model of generative phonology to which we are introducing ourselves is the
model in (3), in which the entire phonological component follows the syntactic
component (which included morphological operations): It is represented in (6).
(6)
Underlying representation
R1
R2
:
:
Rn
Surface phonetic representation
The Classical Model of Generative Phonology

The main feature of the model of generative phonology was the claim, in consonance
with generative grammar as a whole, that there are two levels of phonological
representation- the underlying (phonemic) level and the surface phonetic level. The
features of the model were the tenets presented in (2). These are expressed simply
below

(7)

• The phonological representations in the explicit phonologies “were linear

sequences of matrices of feature values.”
• In order to generate the forms of a language, phonology had to employ
derivational means, with sequential operations of rules;
• Phonological analyses had to discover and empirically establish explicit rule
ordering that was consistent with the data;

Phonological analyses were subject to the Evaluation Metric of generative grammar,

which made crucial use of the criteria of simplicity and generality. These criteria
determined the formalization of phonological representations and rules.

10
6 Two levels of phonological representation

Generative Phonology, proposed that phonological representation of linguistic forms,

words and sentences are at two levels- the underlying level and the surface level. The
main motivating for positing two levels of phonological representation is phonological
alternation. An example of alternation is the morpheme {/bil/} with two alternant forms
[biɫ] in bill and [[bil] as in billing. As the two forms are alternant shapes of the a single
morpheme, the morpheme is represented in its underlying form as /bil/, and its alternant
forms are derived with the help of a phonological rule that velarizes the lateral in the
coda position of a syllable.
The basic difference between structural phonology and generative phonology is that in
generative phonology the underlying form is not restricted to the level of
morphologically simple words, but applies to complex words as well. (This point is
mentioned several times in this course, and you will thus be well familiar with it). As a
result of the unified account of simple and complex words, phonological rules are
assumed to be a single set applying throughout word phonology. We will address this
aspect after having looked at the issue of alternation.
An objection to the claim that there is only one level of representation can be made by
arguing that the different alternants of a morpheme are stored in the mental lexicon of
the native speaker as such, and that they are not derived from a single alternant form.
This can easily be found view can easily be shown to be unrevealing of a native
speaker’s competence if we consider that the native speaker applies regular
phonological rules to novel forms as well. For example, the rule of l-Velarization can be
predicted to apply in the speech of a native speaker of English when he pronounces a
new word, such as the Hindi word /kәl/ ‘tomorrow’. Since the speaker of a language applies
a rule even in new forms, the only way to explain this aspect of his phonological knowledge is
to assume that the rule is part of his grammar.
Phonological rules are the main link between the underlying representation and the
surface representation of forms. As we will see below, the phonological grammar of the

11
native speaker consists of a set of phonological rules that apply to yield a complex
pattern of interaction.

7 Types of phonological alternation

Phonological alternation is of two main types- allophonic alternations and neutralizing
alternations

Allophonic alternations. Allophonic alternations involve, as the term suggests,

allophones, for example, the [l] ~ [ɫ] alternation in English. Among the well-known
allophonic alternations in English are the aspirated and unaspirated voiceless stops, e.g.
[p] ~ [pʰ] in speak and peak, oral and nasalized vowels, e.g. [ε] ~ [ɛ̃ ] in bet and Ben,
etc. WE have looked at many such allophonic alternations in the modules on phonemic
analysis. Examples of allophonic alternations can also be found for many languages in
the modules on phonological processes, e.g. [p] ~ [b] ~ [ß] alternation in Tamil. Take a
re-look at them to gain closer familiarity with them.

Neutralizing alternations. Neutralizing alternations involve alternation between

contrasting segments or phonemes. These are of three main types - phonologically
conditioned , morphologically conditioned and lexically conditioned. We
look at them one by one below.

An example of a phonologically conditioned neutralizing alternation is /n/-

Assimilation in Hindi. Look at the following examples (Allophonic nasalization on the
vowels preceding the nasals is not shown):

(8)

sundәr beautiful/ handsome ɖәŋka beat

sәntәra: orange sәŋgi:t music
әɳɖa: egg sәmbhәʋ possible
ghәɳʈi:: bell kәmpәn shiver
gәɲʤa: bald kәɲʧәn gold

12
There are five nasal segments in (8): [m n ɳ ɲ ŋ]. Of these /m/ and /n/ contrast. In a
restricted number of words, / ɳ/, too, contrasts with /m/ and /n/, as in the examples
below:

(9)

ma:ta: mother na:ta: ‘relation

kʰa:n (a) mineN kʰa:ɳ molasses

The segments [ŋ ɲ] are allophones of /n/; they occur before velar ([p b]) and palatal ([ʧ
ʤ]), respectively. In the examples in (8), we find that the contrast between /m/ /n/ and /
ɳ/, too, is neutralized, so that the nasals occur before homorganic stops. The rule can be
informally stated as follows:

(10) Nasal Assimilation in Hindi

Nasal consonants assimilate in place to the following stop.

Neutralizing alternations can be both phonologically and morphologically conditioned.

For example, the English regular plural morpheme /z/ has two other allomorphs- /s/, /z/
and /ɪz/. The alternation between them is determined phonologically in that /ɪz/ occurs
after strident obstruents /s z ʃ ʒ ʧ ʤ/ and /s/ occurs after voiceless obstruents other than
/s ʃ ʧ/.

(11)

[s] [z] [ɪz]

cats mobs busses

tapes bags buzzes
packs robes garages
boys judges
dishes
churches

13
However, the phonological rule of assimilation must in itself have the inflectional
morphological boundary + before the plural morpheme in order for it to undergo
assimilation. The two rules required to predict the correct pronunciations of the Plural
allomorphs are
(12)

a. /ɪ/- epenthesis

ø → [ɪ] / [-son, +str] + __ #

b. Voicing Assimilation

/z/ → [s] / [-voice] + __ #

Thus the alternations among English plural allomorphs are both phonologically and
morphologically conditioned.

Neutralizing phonological alternations can also be phonologically, morphologically and

lexically conditioned. For example, English has a rule of /k/- Spirantization that
accounts for /k/ ~ /s/ alternation of certain morphemes:

(13)

electric electricity
specific specificity
public publicity
medical medicinal

The context for /k/ ~ /s/ alternation in the words in (13) is the vowel /ɪ/ following the
derivational morpheme boundary.

(14)

/k/ → [s] / __ + /ɪ/

14
The alternation is prohibited if /ɪ/ is not preceded by a morpheme boundary, for
example, in kit and king. It is also lexically restricted to certain forms, but not others,
for example, rollicking and lucky.

8 Summary

In the present module, we have looked closely at the foundational notions of generative
phonology as part of generative grammar- the main assumptions, the principal tenets
and the model of classical generative phonology. We will continue the introduction to
the theory in other aspects in the following three modules.

###

15