Marconi's main question is what it means for a person (or computer system) to be seman-tically competent, especially with respect to the meaning of open-class words. His claim is that there are two components to lexi­cal competence: referential competence and inferential competence. The different types of knowledge required by each correspond in some ways to T-box and A-box knowl­edge (terminological and assertional knowl­edge) in KL-ONE-like knowledge represen­tation systems; but Marconi's distinction is principled, not merely a stipulation bolstered by intuition as it has been in research in knowledge representation. Marconi's theory is not a semantic theory, although it is closely related to semantic theories; rather, it is a cognitive theory—a theory of what it is that people know when they use words cor­rectly. But this obliges Marconi to take on the philosophers of linguistic semantics— Putnam, Bürge, Fodor, Searle, and others— and carefully distinguish his approach from many semantic theories.

Marconi begins with a critical historical re­view. Most philosophers of language have been concerned more with linguistic struc­ture and composition than with lexicons or content. (Wittgenstein saw no lexical prob­lems other than naming!) To the extent that there was any emphasis in this research on lexical matters, it was on questions such as "meaning as reference," and the distinctions between extension and intension or between sense and reference. This led to approaches that were based on truth rather than con­tent, with meaning postulates as constraints on possible worlds. Marconi complains that meaning postulates are quite inadequate: they don't give a unique interpretation; they are still just uninterpreted symbols; and they say almost nothing about what we really know about a word, such as the associated encyclopedic or world knowledge. Semantic nets and other knowledge representations, decompositional semantics, Lakoff's proto- types, and Putnam's stereotypes fare no bet­ter.

Marconi then introduces his basic idea: Lexical competence, with respect to any par­ticular word, is knowledge of true sentences in which the word is used plus knowl­edge of its application in the world (p. 58). The latter component is the usual kind of competence: referential. We can identify a primrose without needing to know ev­erything about primroses. But conversely, most people know quite a bit about ura­nium and yet would not be able to recog­nize it if they saw it; they have only the first component. An expert has both com­ponents of competence. And the two com­ponents interact: one can be obtained from the other. But, while the two are not rigidly separated, Marconi spends a lot of time showing that they are indeed distinct and independent—and that studies of aphasia show that the distinction is cognitively real. He spends considerable effort defending his approach against objectivist theories of refer­ence.

The problem of determining, through ob­servation of its actions, whether or not an entity has understood some linguistic in­put brings Marconi to Searle's well-known "Chinese-room" problem. Marconi's answer to Searle is that, yes, the Chinese room is merely doing symbol processing without un­derstanding, but this does not preclude there being some linguistic competence. The Chi­nese room is still not linguistically competent in Marconi's sense, as it has at best only infer­ential, not referential, competence. This leads Marconi to a discussion of the role of visual perception and recognition in linguistic com­petence, and the prospects for AI systems with vision and language. Marconi's posi­tion, contra Searle, is that for his purposes, such robots would indeed be equivalent to humans. —Graeme Hirst, University of Toronto

The contents of the volume are the following: "Recent issues in logic, language and com­putation" by Seiki Akama "Restricted quantification" by M. J. Cresswell "The epsilon calculus' problematic" by B. H. Slater

by Klaus von Heusinger "Spatio-temporal structure in text" by

"Displaying as temporalizing: Sequent sys­tems for subintuitionistic logics" by Hein­rich Wansing

"Quantification and dependence logics" by L. Farinas del Cerro and V. Lugardon

"Relevant conditionals, and relevant applica­tion thereof" by Richard Sylvan