Phonological Knowledge: Conceptual and Empirical Issues (review)

What is the place of phonology in the mind? Is phonology, like syntax, compartmentalized from the rest of our cognitive abilities? And why are some phonological processes common while others are unattested? These are the threads—the ‘conceptual and empirical issues’, as the subtitle says—running through this book.

In their introduction, the editors present the questions which the invited articles address. The main question stems from the fact that phonology is an interface to the nonlinguistic (articulatory and acoustic phonetics). If phonology is grounded in the nonlinguistic, can this be reconciled with the idea that phonology is part of language, under the Chomskyan view that language is modular?

The editors sketch several answers and situate most of the articles in light of these answers. Other issues include questions of types vs. tokens and the importance of variation. The papers themselves range from the philosophical to the empirical.

In a paper originally given as a talk to philosophers, Silvain Bromberger and Morris Halle claim that while phonologists often speak as if phonology is ‘about’ types, it is in fact about tokens, at least according to Bromberger & Halle’s definition. Another issue concerns the steps of phonological (rule-based) derivations; while in the last line of a derivation, the phonetic symbols represent an intention to produce the sounds so represented, in earlier stages the symbols are merely representations to be manipulated by the computational machinery.

Noel Burton-Roberts’s chapter is easy to summarize: Phonology is not linguistic. Since this will come as a shock to most phonologists, the author explains that by ‘linguistic’ he means having to do with a language faculty which is ‘“completely internal” … neither internalized [during growth] nor externalizable’ (42). He leans on quotes from Chomsky to make his claims about the meaning of the term ‘linguistic’. Those who do not agree with Chomsky’s views will probably skip this chapter, and even those who do agree may question the details. For example, Burton-Roberts holds that for phonology to be wholly formal, the symbols it manipulates would have to be ‘substance free’. This last term is used by Mark Hale and Charles Reiss (see below) to mean that phonology manipulates symbols without regard to their phonetic interpretation. Burton-Roberts, however, holds that if phonology is wholly formal in this sense, ‘its content and rationale are obscure’ (47–48). I am not sure what to make of this. If by ‘rationale’, he means the reason phonology exists, not knowing why is not a reason to doubt whether. (The claim that the content would be ‘obscure’ surfaces again in Philip Carr’s paper; see below.)

Philip Carr’s paper attacks the issue (raised by Janet Pierrehumbert, Mary E. Beckman, and D. R. Ladd) of whether generative phonology is scientific. Carr’s answer is that generative phonology is testable and therefore ‘scientific’. However, Carr questions whether one can learn anything about Universal Grammar from E-languages. (The latter term was introduced in Chomsky & Lasnik 1993 and refers to language as external and extensional—roughly, the sociopolitical notion of ‘a language’.) Carr argues that one cannot, for ontological reasons. But generative phonologists have usually reasoned from I-languages, that is, the internal, intensional knowledge of individual speakers; hence the argument seems beside the point.

Carr charges generativists with inconsistency: While denying a realist interpretation of mental representation and phonological rules (by which Carr means the on-line processing of those rules), they employ such a view in practice. Carr’s only citation denying the realist approach is from Chomsky 1965, Aspects of the theory of syntax, although he alludes to later debate. In any case, for many phonologists constraints have superseded rules (although analogous criticisms of real-time constraint processing are certainly possible). Apart from this, Carr’s claim that a realist interpretation of rule-based derivations requires real-time rule application...