Generative Phonology

Iggy Roca

• Publisher: Routledge
• Year: 2003
• Print ISBN: 9780415041416
• eBook ISBN: 9781134952526
• Topic: Languages & Linguistics
• Subtopic: Linguistics
• Index: Languages & Linguistics

"Generative Phonology" offers an overview of the post-SPE theory of generative phonology and is suitable for linguists not specializing in phonology, who want to keep abreast of the latest developments in the subject. It deals with all the major trends in what has come to be known as "non-linear" phonology, including: particle phonology; dependancy phonology; government and charm phonology. Iggy Roca guides the reader through the developments of the various approaches, justifying their rationale against the background of SPE machinery, and providing the reader with the basic tools necessary to penetrate current problems and debates. This text aims to integrate the modules and proposals of what can seem a fragmentary field, into a cohesive body of living theory.

1 Phonological representations

1.1 THE ARRANGEMENT OF ELEMENTS IN SPE

We shall start our discussion with a brief summary of the basic formal fabric underpinning the SPE model of phonology (SPE = Chomsky and Halle 1968), with which we assume a certain familiarity.¹ This will give us a useful basis of comparison with subsequent developments, with which the body of this book is primarily concerned.

In SPE a phonological representation was made up of a linear array of feature matrices. Let us consider the simple word pen. Each of the three segments of this word was broken down into more elementary particles, called ‘distinctive features’, each essentially representing one aspect of its pronunciation. Formally, each SPE feature was assigned one of two values, positive or negative. Simplifying somewhat, the formal SPE representation of /p/ will therefore be as follows:

(1) SPE representation of /p/

The feature matrix in (1) is of course incomplete, but will suffice at this point in the exposition.

Given such decomposition of segments into features, the word pen will be formally represented as in (2):

(2) Formal representation of pen in SPE

As we shall see in Chapter 3, distinctive feature theory has evolved substantially since the time of SPE, but this is not important for our immediate purposes. Also negligible at the moment is the exact interpretation of each of the feature labels in terms of articulation, acoustics, etc., which we will introduce step by step as we go along (see Chapter 3 for a more thorough discussion). What is significant at the present point is the arrangement of the matrices into unilinear strings characteristic of SPE. In fact, from this perspective, the representation in (2) is equivalent to that in (3):

(3) /pen/

Both (2) and (3) are made up of a linear array of elements, strung together like beads in a string. There is an obvious formal claim implicit in this arrangement, namely, that phonological representations consist of sequences of elements (whether distinctive feature matrices, as in (2), or phonemes, as in (3)) drawn out of a basic vocabulary and capable of entering linear relations of precedence with each other. This is a particularly simple and commonsensical conception of phonological structure, one that also runs through IPA transcription and conventional orthographic writing, where phonetic symbols or letters follow each other in a linear fashion. This mode of representation could therefore easily be thought to be directly rooted in phonetic facts, and thus to constitute the null hypothesis for the representation of sound. As is well known, however, it is in fact not possible to draw up precise boundaries between sounds, for the simple reason that in real life both articulatory and acoustic events form a continuum.

For greater explicitness, we shall now review the relevant formal aspects of the model under scrutiny. Let us say that, for a hypothetical language L, the vocabulary of phonological elements consists of three members, as follows:

(4)	Vocabulary of phonological elements for L
	{P, e, n}

Moreover, let us imagine that no restrictions are imposed in L on the precedence relations between these elements. The following sequences will thus be licensed:

(5)	pen
	pne
	enp
	epn
	nep
	npe

All these strings will be possible phonological words in this language. The number of well-formed phonological objects could of course be easily increased by bringing optionality and recursion into the picture (p, pe, etc.; penpen, penpne, etc.), but such additional complexity is unnecessary for our present purposes. What is important to realise is that the language under consideration has an inventory of three phonological elements, as in (4), and a grammar which allows these elements to enter into unrestricted unilinear relations with each other. A graphic representation of this grammar is given in (6):

(6) x_l, x₂, x₃,.. ., x_n

The xs in (6) stand for phonological elements, and the subscripted digits indicate sequential ordering. The precedence relationship (call it ‘<’) has the properties of transitivity, antisymmetry and irreflexivity (see Sagey 1988a). Thus, in (6), (x₁ < x₂) and (x₂ < x₃) entails x₁ < x₃ (transitivity). Given x₁ < x₂, it follows that it is not the case that x₂ < x₁ (antisymmetry). Finally, it is also not the case that x_j < x_j, i.e. no element can precede itself (irreflexivity). This is the gist of the formal representational machinery made available by SPE.

As we shall now see, a revolution that started in the 1970s has profoundly transformed the formalisation of phonological structure in generative phonology. Specifically, the unilinear geometry which characterises SPE has now been abandoned in favour of multilinear representations.² In order to understand the meaning and implications of this far-reaching change, we shall first briefly examine its basic rationale.

1.2 COMPLEX SEGMENTS

1.2.1 Affricates

The analysis of affricates has long puzzled phonologists, because of the Janus-like nature of these sounds, which behave simultaneously as mono- and bisegmental. Thus, consider the affricate sound [t ∫], which occurs twice in the word church. The following question arises: does such a sound correspond to one or two phonological elements? Specifically, is there in English a phonological element /t∫/, directly corresponding to the affricate, or should this sound be analysed as a sequence of the individual elements /t/ and /∫/, which are attested in tut and shush, respectively. As it happens, there is evidence supporting either stand.

On the one hand, English only allows one obstruent word-initially, with the notable exception of s + obstruent clusters.³ Thus, alongside tar [t

] we have star [st

], but orthographic tsar is not pronounced *[ts

], at least by linguistically unsophisticated speakers, but rather [z

]. Note that, irrespective of their pronunciation, words like tsar and tsetse are felt as exotic, in clear contrast with such everyday non-oddities as church, chat, much, and so on, all containing [t∫]. So, from the point of view of the ...

Table of contents

• Front Cover
• Generative Phonology
• Linguistic Theory Guides
• Title Page
• Copyright
• Dedication
• Contents
• Series editor’s preface
• Preface
• List of abbreviations
• 1 Phonological representations
• 2 Lexical redundancies
• 3 Basic elements
• 4 The sonority fabric
• 5 Domains and modes of application
• Notes
• References
• Indexes