Lexical & conceptual semantics Edited by Beth Levin and Steven Pinker (review)

REVIEWS Lexical & conceptual semantics. Edited by Beth Levin and Steven Pinker. Cambridge, MA, & Oxford: Blackwell, 1991. Pp. 244. Cloth $21.95. Reviewed by Mary Laughren, University of Queensland What is particularly nice about Lexical & Conceptual Semantics (L&CS) is that its six articles report on very different types of research with essentially the same goal: to discover the underlying principles governing the relationship between components of meaning and syntactic structures. Included are revealing investigations of specific classes of lexical items and construction types (by Ray Jackendoff, James Pustejovsky, Beth Levin & Malka Rappaport Hovav, and George A. Miller & Christiane Fellbaum), crosslinguistic language acquisition data (by Soonja Choi & Melissa Bowerman), and data obtained through psycholinguistic experimentation (by Jess Gropen, Steven Pinker, Michelle Hollander, & Richard Goldberg). In their introduction, the editors define their concept of lexical semantics as the study of those semantic components of lexical items which relate to their syntactic properties. They place current research in lexical semantics in its modern historical context, referring to work from the 1960s and 1970s before reviewing more current work and giving an overview of the articles in the book. The assumption 'that syntactic properties of phrases reflect, in large part, the meanings of the words that head them' (3) is shared by each of the authors in this volume and motivates their research goals and methodology. Notions of semantic fields, into which lexical items are classified on the basis of shared semantic components defined in terms of the item's referential domain, are not relevant to the research program ofthe contributors, with the possible exception of Miller & Fellbaum. Rather, the aim is to discover syntactically relevant semantic elements or properties of lexical items over which important generalizations can be made, in order to further our understanding of how conceptual elements are syntactically encoded and how language speakers manipulate both conceptual and syntactic categories and constructions in order to create and interpret meaning. Jackendoffs 'Parts and boundaries' (9-45) furthers an investigation that he has been vigorously pursuing for many years into the nature of 'conceptual structure', which he claims to be common to all natural languages, while assuming there to be no isomorphism between syntactic and conceptual structure.1 Jackendoffs conceptual structure, which has generative X-bar properties, classifies 'conceptual constituents' into 'conceptual categories' likened to 'semantic parts of speech'. His model of linguistic representation posits 'correspondence 1 Hale (1986) postulates a very similar theory in which his 'World View-2' corresponds to Jackendoffs 'conceptual structure'. Hale claims that World View-2 is 'universal—a part of the innate linguistic capacity of human beings—albeit instantiated in different ways and in different proportions in different languages' (235). Like Jackendoff, Hale implies that languages vary, not in relation to World View-2, but in how its constituent elements are mapped onto syntactic structures. 546 REVIEWS547 rules' (CRs) that map conceptual structures onto syntactic structures. CRs take the form of lexical entries, morphological affixation, construction meanings, and rules of construal. Jackendoff argues that if conceptual categories are decomposed into a small number of constituent features, greater generalizations about the conceptual structure component of lexical items crosscategorially and those associated with various construction types can be achieved. For example, in terms of whether they are conceived as bounded ( + b) or unbounded ( - b), and/or having internal structure ( + i) or lacking internal structure (-i), members of his conceptual category Matferial Entity] can be 'individuals' [ + b, -i] (a pig), 'groups' [ + b, -i] (a committee), 'substances' [-b, -i] (water), or aggregates' [-b, +i] (buses, cattle).2 Jackendoff discusses a range of semantic functions that map between values of b and i in the NP domain, including PL(ural), which maps its argument, a member of the Mat category which is [ + b, -i] (a dog), into one that is [ — b, +i] (dogs). Jackendoff shows how functions such as PL are also involved in the conceptual structure of aspectual properties of verbs and VPs, which are characterized by these same conceptual constituents in addition to others such as DIMfensionality] and DIR[ectedness], features of conceptual categories such as Time and Situation]. When added to a VP headed by a V linked to an event type or Sit(uation) which is [ — b] in Jackendoffs terms, a...