Linguistic theory and psychological reality Edited by Morris Halle, Joan Bresnan, and George A. Miller (review)

REVIEWS467 Linguistic theory and psychological reality. Edited by Morris Halle, Joan Bresnan, and George A. Miller. (MIT Bicentennial studies, 3.) Cambridge , MA: MIT Press, 1978. Pp. xvii, 329. $17.50. [Paperback edition, 1981; $7.95.] Reviewed by Thomas Roeper, University ofMassachusetts, Amherst* Halle's preface (x-xv) to this book notes that it grew out of a series of meetings held at MIT in 1976, observing the centennial of Alexander Graham Bell's invention of the telephone—and especially from a symposium entitled 'New approaches to a realistic model of language'. The papers published here are as follows: Joan Bresnan, ? realistic transformational grammar' (1-59); George A. Miller, 'Semantic relations among words' (60-118); Eric Wanner and Michael Maratsos, 'An ATN approach to comprehension' (119-61); Keith Stenning, 'Anaphora as an approach to pragmatics' (162-200); Ray Jackendoff, 'Grammar as evidence for conceptual structure' (201-28); Edgar B. Zurif and Sheila E. Blumstein, 'Language and the brain' (229-45); Michael Maratsos, "New models in linguistics and language acquisition" (247-62); Susan Carey, 'The child as word learner' (264-93); and Morris Halle, 'Knowledge unlearned and untaught: What speakers know about the sounds of their language' (294-303). This is both a fine introductory text, with broad essays on a variety of topics, and a challenging, sophisticated volume. It contains a number of important, though partially implicit, claims. I'll begin with a discussion of these issues, which I think the book as a whole lacks. A review ofthe individual contributions follows. The book presents itself as undoing the estrangement between linguistics and psychology. Nothing is more important if either linguistics or psycholinguistics aims for robust biological implications. However, success will require that certain pitfalls be avoided (with respect to the notion of psychological reality) and that a certain modesty be maintained (with respect to the relation between theories and data). Several perspectives are possible. Chomsky 1981 asserts that linguistic theory has psychological reality. This must be true because the theory is based on psychological data, namely intuitions of grammaticality (see also Bresnan's contribution in the present volume ). But is it in fact the case that no level of generalization exists other than the mathematical formulations of linguistic theory and a future neurological model? Of course we talk about what we think, and we use language as well as know it. Therefore there must be a mapping from ideas onto sentences, and some relation between a grammar and a processor. Still, the 'fit' between linguistic notation, real-time notation, cognitive notation, and a notation for language growth might all be so natural that no alternative formulation of linguistic theory would be needed. The articles by Bresnan, Jackendoff, Zurif & Blumstein , and Maratsos all seem to take this view. * Thanks to Lyn Frazier for discussion of various points in this review. 468LANGUAGE, VOLUME 58, NUMBER 2 (1982) However, there could be a second level of psychological reality (in a variety of ways) where a common notation is expressed for both linguistic theory and other mental functions (see Fodor 1975 for discussion). Wanner, Miller, and perhaps Stenning are seeking it. It is important to realize that any 'second' level of psychological reality cannot replace the first (grammatical theory) unless it also accounts for the intuitional data. This means that facts and arguments derived from language use will not invalidate the 'psychological reality' of formulations based on linguistic intuitions. At the same time, studies of language use (or acquisition) are far from irrelevant to formulations of linguistic theory. The common view is that linguistic theory should form a 'natural' part of a model of language use or acquisition. There is no a-priori limit on what kinds of data are relevant to either linguistic or psycholinguistic theory. If facts about use or acquisition can answer an open question in linguistic theory, or vice versa, then they are relevant. The fact that children first use infinitives without subjects (/ like to sing) supports the notion that subjectless infinitives are unmarked. The opposite pitfall also deserves mention. In the past, people have often equated linguistic with psychological mechanisms. This may seem an obvious error; but actually it is a perfectly natural research strategy, which should undergo progressive...