Linguistic structure in language processing Edited by Greg N. Carlson and Michael K. Tanenhaus (review)

REVIEWS627 REFERENCE Comrie, Bernard. 1981. Languages of the Soviet Union. (Cambridge Language Surveys .) Cambridge: Cambridge University Press. Department of Linguistics[Received 4 February 1991.] University of Southern California Los Angeles, CA 90089-1693 Linguistic structure in language processing. Edited by Greg N. Carlson and Michael K. Tanenhaus. (Studies in theoretical psycholinguistics, 7.) Dordrecht : D. Reidel, 1989. Pp. viii, 413. Cloth $89.00. Reviewed by Joseph F. Kess, University of Victoria This collection of papers is intended to exemplify what psycholinguistic investigations in the comprehension and production of natural language by adults have attended to in the 1980s. A brief sketch of recent history in psycholinguistics is followed by nine experimentally-oriented papers, organized into a sequence which presents phonology first and semantics last, with papers on morphology, syntax, and lexical structure in between. A singular feature of the collection is that it represents the renewed interdisciplinary nature of psycholinguistics , an enterprise in which the two disciplines of psychology and linguistics once again interact in some meaningful way. For example, in this volume joint works are authored by a linguist and a psychologist, and even the single-authored papers exhibit research goals that are not confined to a single discipline. In their sketch of recent history ('Introduction', 1-26), editors Greg N. Carlson & Michael K. Tanenhaus note that psycholinguistics is more vibrant than it has been since the 1960s, and that this is accompanied by a renewal of interest in the role of linguistic structure in language behavior. By the mid1970s , psychological investigations of language processing made little reference to linguistic theory, owing to the perceived failure of transformational grammar as having explanatory value in processing matters. In turn, many linguists saw psycholinguistics as not germane to explaining language structure, and typically paid little attention to the results of psycholinguistic experimentation, especially ifthey were contradictory. The Chomskyan suggestion that linguists were really cognitive psychologists did not square with the impenetrability of linguistic theory to psycholinguistic findings, and psycholinguistics was simply reabsorbed into cognitive psychology. Carlson & Tanenhaus suggest a number of reasons, like the robust evidence for the role of surface structure in processing , but the relative lack of evidence for transformations. Serious competitors within linguistic theory itself began to minimize or eliminate transformations altogether; computational work on language was also unnecessarily complicated by transformations and often excluded them altogether . The drive to make linguistic theory explicit in a computational sense now prompts clarification of the relationship between linguistic theory and models of language processing, and methods and goals in linguistics and psychology are again sympathetic in their attempts to understand language be- 628LANGUAGE. VOLUME 67, NUMBER 3 (1991) havior. Such developments are compatible with cognitive science aspirations to provide a formal science of the mind, and analysis of the interaction of linguistic structure and language processing is once again in favor, benefiting as it does from the double rigors of linguistic analysis and psychological experimentation . But linguistics certainly does not occupy center-stage in this new rapprochement, and there is even some question as to where linguistics fits in all this—and specifically, whether the psychologists out there are even paying attention to the vast wealth of crosslinguistic data we are prepared to offer as counterpoint to proposed universal processes (see Anderson 1989 and Miller 1990). One of the central questions these days is whether the language processing system is modular or interactive, and to what extent. Modularity suggests that the processing system is composed of a series of modules, each handling some specific type of information without reference to the activities of other modules. In contrast, an interactive explanation of processing supposes that there is an active exchange of information in the processing system, and that levels of language knowledge are not cut off from one another. One interactive model that has been attractive to some psychologists lately is connectionism, in which processing units activate units to which they are connected on other levels. Much recent research probes the degree of modularity or interaction in the system, and this issue motivates some of the papers in this volume as well. Given their traditional preoccupation with levels of language knowledge, linguists have favored the modularity hypothesis, while psychologists seem less impressed...