Computational psycholinguistics: An interdisciplinary approach to the study of language By Matthew W. Crocker (review)

164LANGUAGE, VOLUME 74, NUMBER 1 (1998) (13) The Early Irish diphthongs conventionally spelled (ia) and (Ua) were /ia/ and /ua/, respectively , not /i:a/ and /u:a/. (Passim) The impression a specialist gets after reading this volume is that it would have benefited greatly from careful editing and proofreading. The volume as it presently exists will leave the linguist unfamiliar with the Celtic languages confused and all too frequently misinformed. (A lengthy list of references (306-35) gives the reader access to sounder secondary literature, but this is hardly a saving grace.) A second edition should be edited much more carefully, perhaps in consultation with specialists in the individual languages, and the descriptive presentation, especially in matters of syntax, should be less dense and more user-friendly. For now, the nonspecialist wishing to learn about the Celtic languages would be well advised to turn to the collected volumes mentioned infra. Indeed, perhaps a volume intending to treat the Celtic languages both diachronically and synchronically, both descriptively and theoretically (in parts), and with considerable attention to dialectal and sociolinguistic features, really ought to be prepared by a group rather than an individual. REFERENCES Ball, Martin J. (ed.) 1993. The Celtic languages. London: Routledge. MacAulay, Donald (ed.) 1992. The Celtic languages. Cambridge: Cambridge University Press. McCone, Kim (ed.) 1997. Compendium linguarum Celticarum. Wiesbaden: Dr. Ludwig Reichert Verlag, in preparation. Department of English Virginia Polytechnic Institute and State University Blacksburg, VA 24061-0112 [eska@vtaix.cc.vt.edu] Computational psycholinguistics: An interdisciplinary approach to the study of language . By Matthew W. Crocker. (Studies in theoretical psycholinguistics, 20.) Dordrecht: Kluwer, 1996. Pp. xii, 246. Reviewed by Dominique Estival, University of Melbourne Comparison with another book by the same title which came out at the same time (Dijkstra & de Smedt, 1996) shows that the new field of computational psycholinguistics is as diverse as are both linguistics and psycholinguistics. Crocker follows very faithfully the main tenets of the Chomskyan approach to syntax, and his book presents an implementation of the principles and parameters (P&P) framework, while D&S (a collection of 15 papers by researchers in psycholinguistics , linguistics, and AI) presents a completely opposite approach to language processing, namely the use of connectionist networks. The main limitation of Cs work however, is not the framework it adopts but the fact that while D&S covers both language comprehension and language production, C does not attempt to treat language processing but is limited both to the sentence level and toperception. Nevertheless , within those limitations, C argues very persuasively for the adequacy of the model he has chosen to the task he has undertaken. C attempts to demonstrate that the model of competence underlying the P&P approach to grammar can also serve as the basis for the implementation of a model of performance. He argues strongly: (1) for modularity (by which he really means autonomy of syntax); (2) against an interaction- or constraint-based model; (3) for a strongly principle-based parser (defined in Ch. 1, see below); and (4) against conflating competence and performance. These are central issues to be debated when designing a psychologically valid computational model of language processing, but his answers will not necessarily convince researchers in other lines of work. In particular, while I am sympathetic to Cs views on points 3 and 4, points 1 and 2 must largely be taken on faith, and his arguments have failed to persuade me. REVIEWS165 One problem with this book is that C assumes familiarity with the GB and P&P syntactic formalisms (for instance, the S node of the first figures becomes an IP node without explanation on p. 42), and although Ch. 3 later provides some introduction to the main concepts of the theory, the first two chapters should not presuppose such knowledge on the part of the reader if the audience is intended to be wider than syntacticians or psycholinguists working in this framework. While several approaches to processing are discussed, other syntactic theories are largely ignored. There is some comparison with LFG (in particular between Cs thematic structure and F-structure), but it is mainly dismissive, and there is no mention of either GPSG or HPSG, although the feature-based X-bar theory presented in Ch. 3 clearly owes a large...