Phonetics & Phonology 101: Generative Phonology and The Sound Pattern of English

Foreword

Phonetics and phonology are two essential parts of linguistic competence. While both are related to the processes of production and perception of language sounds, they differ in their approach and goals. Phonetics is concerned with the physical properties of sounds, i.e., articulatory, acoustic, and perceptual, and aims at a scientific description and classification of all sounds attested in human languages. Phonology, on the other hand, is more focused on how these sounds are organized in single languages, i.e., how they constitute their phonological inventory through distinctive oppositions and how they relate to morpho-syntax and semantics. This course aims at introducing the reader to both disciplines, focusing on the following aspects: (a) the basic concepts of phonetics and phonology, their interrelation, and differences; (b) the use of IPA to write phonetic and phonological transcriptions; (c) the major issues and debates in modern phonological theory.

This series is divided into the following chapters:

1. Phonetics & Phonology 101: The Role of Phonetics and Phonology in Linguistics

2. Phonetics & Phonology 101: The International Phonetic Alphabet (IPA)

3. Phonetics & Phonology 101: Phonology and its Relation with Language Theory

4. Phonetics & Phonology 101: Generative Phonology and The Sound Pattern of English

5. Phonetics & Phonology 101: Phonological Opacity, a Debate in Modern Phonology

6. Phonetics & Phonology 101: Phonology in Language Acquisition

Phonetics & Phonology 101: Generative Phonology and The Sound Pattern of English

This chapter focuses on one of the most relevant contributions to modern phonology, namely generative phonology. The structure of this article is the following: first, an introduction to the underlying theory, generative grammar, will be given; second, there will be a focus on the phonological component of such a theory; third, the main features of generative phonology exposed in The Sound Pattern of English will be described.

Generative Grammar

While the notion of a generative procedure for describing the language faculty was explicitly formulated by Noam Chomsky in his early work (Chomsky, 1965), the idea behind such a procedure had been expressed before in the history of linguistics. But what is generative grammar? By generative grammar, linguists mean a formalization that explicitly describes the way all sentences in a language are produced. In order to do this, a grammar needs specification of the lexicon of that language and the rules that combine the words of the specified language in well-formed sentences, i.e., acceptable to the speaker (John ate an apple is grammatically correct but *Ate apple an John is not). The following is a description of basic English grammar, where A stands for article, N for noun, V for verb, S for sentence, and ‘-->’ indicates the symbol on the left must be rewritten as the symbol(s) or string(s) on the right:

S --> A + N + V + N

A --> {the}

N --> {dog, cat}

V --> {bites}

Following the indicated procedure, this grammar is able to generate the following (grammatical) sentences: the dog bites the cat, and the cat bites the dog. The syntactical representation of grammatical sentences like this one is often that of a tree structure (Figure 1).

Figure 1. Basic english syntax tree (Cadr, 2017)

Of course, this simple implementation is insufficient to capture the complexity of most grammatical structures of English (what about adjectives, prepositions, tenses?); but it is a proper approximation of the notion of generative grammar. Generative grammar is also used to describe the theory which exploits such formalism. This theory sees the language as a formal system describable in terms of generative procedures like the one exposed above. The result of such a conception is a somewhat mathematized approach to languages. Not only this, but according to generative grammar, all languages share a series of constraints and mechanisms which are thus called universals.

A similar intuition was expressed way before Chomsky. Indeed, the ancient Indian grammarian Pāṇini, who probably lived around the first half of the IV century BC (Cardona, 1976), in his Aṣṭādhyāyī (‘the eight books’) described a procedure for generating words and sentences which astonishingly resembles Chomsky’s formalization. This is also true for the description of the phonological mechanism, the focus of the current article.

The Role of Phonology in Generative Grammar

Significant developments in phonology came only centuries after Pāṇini. Nikolai Trubezkoy was the first who constructed a discipline with the specific goal of describing not simply the different sounds across languages but also the underlying structures which constitute phonologic systems of languages, traditionally described as that aspect of languages that put in contrast different sounds relying on their capacity to distinguish the words' meaning. Trubezkoy's theory considers different ways by which languages create these kinds of sound contrasts. But while Trubetzkoy's analysis is only about common phonological contrasts, such as that between voiced and voiceless sounds or between different heights of vowels, Roman Jakobson's work (Jakobson, 1957) and that of him and two of his collaborators (Jakobson, Fant, & Halle, 1963) had greater ambitions. They wanted to describe all and only the phonological rules for attested languages. According to Jakobson, every phoneme creates an opposition with another one through the presence/absence (+/-) of single features. This kind of regularity was somewhat advanced by Trubetzkoy, but it was considered inapplicable to vowels, which were considered to feature gradual, and not binary, distinctions. Jakobson realized that the fact that vowels present phonetically diverse degrees of height has no relevance from a phonological point of view. On the phonological level, every vowel is distinguished by the others for properties such as height, tongue position and roundness; but these properties are perfectly expressible in relevant binary features' terms. Consider the vowels of the Italian language. In pure phonetic terms, they are seven: a low vowel [a], two high vowels [i]-[u], and four medial vowels [ε]-[ɔ]-[e]-[o]. The articulatory description does not distinguish vowels relying on binary features because it is supposed to analyze precisely the articulatory realization of sounds (in terms of active articulator, way, and place of articulation). A phonological description, in Jakobson's view, instead, must describe just the features that are distinctive in the language taken into account. First, consider that in Italian, the opposition [ε]-[e] has no phonological value (the cases in which this couple is in real phonemic opposition are rare, then irrelevant); then, suppose that this kind of opposition exists phonologically. Let us take a case in which the distribution of the vowels [ε]-[e] creates a phonological distinction. Among the few cases attested, let's take the words [ˈpεska] `peach' and [ˈpeska] `fishing (n.)'. We can describe the difference between [ε]-[e] in binary terms. If we take Vowel Traits according to Jakobson (Figure 2), then we will see that, phonologically, [e] and [E] differ only in the feature [tense], since [e] is [+tense] but [ε] is [-tense]. Jakobson’s binary theory is then valid for both consonantal and vocalic sounds, and it works for every language (Hyman, 1975).

Figure 2. Features of the vowels /e/ - /ɛ/ according to Jakobson (Piersigilli, 2020).

As seen in previous chapters, phonology is concerned with the way sounds are produced and contribute to different inventories in different languages. The generative conception of phonology does not differ with respect to this general definition, although it uses different terminology. The innovations of generative phonology are mainly the following: (a) the distinction between underlying and surface forms; (b) the notion of a phonological feature.

In generative grammar, phonology is that component of the language capacity which converts underlying features into surface forms. Traditionally, phonology describes the inventory of language sounds in terms of phonemes, i.e., the smallest units of sounds, like /a/, /p/, /j/, and so forth. Instead of considering these phonemes, the atomic elements of phonology, the generative approach decomposes them into features. Although these features are somewhat articulatory-oriented, generative features for phonemes, as it will be clarified below, do not match exactly with features described in articulatory phonetics. Thus, there is a difference between underlying features and the surface realization of these features. This conception of phonology was introduced by Noam Chomsky and Morris Halle in The Sound Pattern of English (SPE), the first major contribution to the phonological side of generative grammar. Despite its title suggesting a focus on the English language, this text deals with a wide array of phonological phenomena not limited to those which characterize English. SPE conceives phonology as a subsystem of language that converts underlying phonemes into phonetic sequences. Phonemes are not atomic units, but they are further decomposable into features. The conversion from underlying features to surface forms is the result of the application of a phonological rule. A phonological rule in SPE has the following shape:

A --> B / X_Y which is interpreted as ‘when preceded by X and followed by Y, rewrite A as B’. Let us consider an example that can be represented by both the distinction between underlying and surface forms and how a phonological rule is applied. Figure 3 shows the features according to SPE for a set of English phonemes.

Figure 3. Features of consonants and vowels according to Chomsky and Halle (Piersigilli, 2020).

Each feature loosely reflects an articulatory aspect of the phoneme (e.g., anterior/posterior refers to the place of articulation) and is expressed through a system of binary oppositions: a phoneme cannot be [+anterior, +posterior]! Binary features may also be used in phonological rules. Consider plural suffixation in English, a grammatical phenomenon that has been mentioned in the previous chapter. As it has been said, the phoneme /s/ surfaces as [s], [z], [ɪz] according to the sonority of the preceding segment and to the presence of sibilants. The phonological rule which derives the plural forms of English may be formalized in terms of binary features as follows:

/s/ --> [+voice] / __ [+cons] [+voiced]

This kind of formalization is economical and descriptively satisfactory: it captures all the possible phonetic variations of the plural suffix /s/: keep in mind that the realization [ɪz] is conditioned by a general constraint of English which prohibits two adjacent sibilants.

Later Developments in Generative Phonology

As it has been clarified, in order to formulate a phonological theory that reflects the speaker’s competence, rules must be integrated with constraints. Indeed, while rules are a representation of processes the sounds of a language undergo, constraints represent the speaker’s intuition that some combinations of sounds are not possible in a given language. Quite interestingly, every possible phonological rule is perfectly convertible in the relevant constraint. With regards to the plural suffixation rule in English, it could be substituted by a constraint that prohibits the sequences which do not share the same sonority feature. So, is there a sense in which a system based only on constraints is better? Yes, there is. For example, some phonological properties of languages are explained in a more economical way with a system of constraints. Consider the case of Tiberian Hebrew (Figure 4), where the nasal /n/ is deleted when it is in the body of the word and is followed by a consonant. Crucially, this rule does not apply when the consonant following /n/ is guttural. So, a phonological system based on SPE needs both a rule deleting the nasal and a constraint that prohibits the ungrammatical sequence of geminate gutturals. On the contrary, a constraint-based approach can exploit only one constraint which allows /n/-deletion with consonant duplication in all the cases except when it is followed by a guttural. This constraint may be formulated as follows:

*nC: sequences of a nasal and a non-guttural consonant are prohibited

This aspect led to an important debate in the phonological community. Some scholars decided to abandon a theory of phonology based on SPE, or at least they underlined the fact that such a theory would not be optimal. The more conservative phonologists decided to stick with SPE, claiming that although rule-based phonology is less economical with respect to constraint-based approaches, the latter suffers from further problems which are even worse. Some of these problems will be explained in the next chapter of this 101 series (Gussenhoven & Haike, 2011).

Figure 4. The title page of "A Grammar of the Hebrew Tongue" (Monis, 1735)

Conclusion

Although ancient grammarians had interesting insights about how sounds are articulated and organized in languages, phonology is a relatively young discipline. It has received its current role in linguistics, i.e., to explain the processes and properties of attested phonological systems, only in the first half of the 20th century. The first major contribution to the birth of phonology was from Trubetzkoy. Later developments of his work, such as that of Jakobson and his collaborators, Fant and Halle, constituted the basis for generative phonology, formulated at the end of the 60s. Although SPE by Chomsky and Halle represented a ground-breaking work in the field of linguistics for its formal rigor and its elegance, it has been partly criticized for its anti-economical nature. This led to a new theoretical development that separated SPE-inspired systems from the so-called constraint-based approaches. This debate will be described in the next chapter of this 101 series.

Bibliographical References

Visual Sources