Semantic and conceptual development: An ontological perspective By Frank C. Keil (review)

REVIEWS457 tions; but the consequences of these within a formal theory are not discussed, so that they are really no more than suggestive notations. Dov Gabbay and Julius Moravcsik, 'Negation and denial' (251-65), begins by motivating the ideas (a) that negation in natural language is intended, more often than not, to be understood as applying to sub-sentential constituents rather than whole sentences, and (b) that negations are used to deny, with the understanding that some alternative to the negated constituent would make the sentence true; then the authors briefly sketch a formalism to capture these ideas. The rules suggested are to apply to surface structures, although no discussion of the origin and properties of surface structure is supplied; this makes it a little difficult to figure out how the rules are actually supposed to work. The authors use a notion of incompatible lexical elements to help formalize the idea that the negation of a lexical element invites the truth of the sentence formed when some other, incompatible lexical element is substituted for it. There is a tantalizing appendix containing the beginnings of a study of incompatibilities within lexical fields; I would have enjoyed more discussion of these. The papers are best thought of as work-in-progress. None of the admittedly very difficult linguistic questions dealt with can be said to have been solved. In addition, it is too early to tell how much of what is done in these pages will ultimately form part of our conception of the nature of meaning in natural language, and how much will be rejected as irrelevant or wrong. In particular, much more work than is found in this collection will have to be done concerning how the kinds of systems discussed can be fitted into a grammar of a language as a whole. [Received 6 July 1981.] Semantic and conceptual development: An ontological perspective. By Frank C. Keil. (Cognitive science series, 1.) Cambridge, MA: Harvard University Press, 1979. Pp. xv, 214. $16.50. Reviewed by Charles E. Caton, University ofIllinois, Urbana-Champaign Keifs purpose is to try to identify what constraints exist on a certain kind of human knowledge, and how they guide its acquisition; this cognitive domain is that of ontological knowledge (OK), i.e. 'one's conception of the basic categories of existence, of what sorts ofthing there are' (p. 1). OK is unique, is highly structured, and 'constrains the nature of semantical and conceptual knowledge', which is Keil's main reason for studying it. Finding little psychological literature on this topic, Keil makes important use of philosophical and linguistic sources. 'My goal,' he says, 'is ... to develop a formal model specifying the nature of ontological knowledge and its relation to more overt psychological phenomena, and to use this model to explore the development of ontological knowledge and its influence on semantic and conceptual development ' (p. 2). The four main phenomena dealt with as 'surface manifestations' of OK (p. 3) are anomalous (meaningless) sentences, natural classes, similarity between pairs of natural classes, and natural co-predicability of predicates (Chap. 2). Other phenomena that may be similarly approached are said to include metaphor construction, some forms of entailment, and possible lexical items (p. 6). 458LANGUAGE, VOLUME 58, NUMBER 2 (1982) Keil's approach to a treatment of these phenomena is guided by work in the philosophical literature on predicability, 'the phenomenon that only certain predicates can [meaningfully] be combined with certain terms in a natural language' (7). However, 'predicability itself is a reflection of an even more basic kind of human knowledge', viz. OK (10). The fundamental idea relating to predicability is drawn from the work in ontology of Sommers 1959, 1963, 1965, 1971: it is the M-constraint (MC), viz. the requirement that predicability trees (PTs) and their underlying ontological trees (OT's) must not contain an M- (or W-)shaped structure. That is, two predicates can never meaningfully, whether truly or falsely, apply to (or span) respective sets of terms that intersect , each spanning some but not all the terms spanned by the other. Thus a given predicate, A, will either span a superset or subset of those of another, B, or else share no terms with...