Diachronic Prototype Semantics: A Contribution to Historical Lexicology (review)

Abstract

Reviews1 83 Diachronic Prototype Semantics: A Contribution to Historical Lexicology. Dirk Geeraerts. Oxford: Clarendon Press, 1997. Pp. viii + 207. $65.00. D1 iiachronic Prototype Semantics is a well-written and useful book. I caught myself reading it both as a lexical semanticist (concerned with prediction and the nature of lexical semantic representation) and as a lexicographer (concerned with accuracy over exhaustiveness and guides to lexical reconstruction, but unworried about prediction) . The tension between semantics and lexicography, it turns out, is part and parcel of Geeraerts's very approach because he has worked in historical lexicography, the source of his data, and lexicology, the source of his theory. Geeraerts's goal is to show the value of cognitive linguistics (Langacker 1987 and 1991; Lakoff 1987) in studying lexical-semantic change. The semanticist in me disagrees in principle with a number of his major claims, but overall I am struck by the book'sjudicious combination of old-school and newschool historical linguistics and by its terrific examples. Perhaps even more impressive is Geeraerts's courageous and instructive run at the prediction problem , a serious challenge to both historical and cognitive linguistics. The former is retrodictive at best, and the latter, while claiming predictive power, most often settles for description over explanation and marshals its evidence in retrospect (see, e.g., the grammaticalization "predictions" in Heine, CIaudi, and Hünnemeyer 1991). But then I think, "What difference does it make?" Dictionaries do not need to predict, and even historical linguistics has a dubious tie to this requirement. Still, the completeness of theory looms large, and I wonder, as does Geeraerts in chapters 4 and 5, if diachronic prototype lexical semantics can ever be an explanation in the canonical sense. Chapter 1, "Cognitive Semantics and Prototype Theory," outlines cognitive linguistics, which sees language as part of the general cognitive system and grounds linguistic categories in conceptual categories organized by fuzzy, prototype structure (see Taylor 1995). This research program translates into four parameters for historical lexical change from a semasiological (form to meaning) standpoint: extension vs. intension and non-equality vs. non-discreteness (see table, p. 184). This typology makes explicit the conditions of lexical change and settles certain debates. As to the latter, it shows that transient meanings surface not because they are somehow "there all the time" and fade in and out, but because they actually appear and disappear discretely as a consequence of structural changes in the referential range of a term. For example, verduisteren 'abduct' occurs in both 17th century and 20th century Dutch, apparently with no form filling the temporal hiatus between. This sort of information would be important for lexicography where decisions must be made about whether a form-meaning pair attested at two different times persists in the interval between the attestations. But then the semanticist's worries surface, and I wonder if a prototype view, where the structure of the lexical system is intrinsically continuous, with 184 Reviews Table PARAMETERS OF HISTORICAL LEXICAL CHANGE Meaning Effects —> Category Effects i Extension (set members) Intension (senses describing the coherence of set members) Non-Equality (core —» periphery; salience, and typicality) (1) changes in the referential range of terms, especially at the core (2) changes in the radial organization of senses Non-discreteness (fuzzy category boundaries, family resemblances, and lack of necessary and sufficient conditions) (3) changes at the edges of referential range, where boundary overlaps occur: appearance of transient meanings (4) changes via encyclopedic knowledge: meaningworld knowledge overlaps meanings scaling into each other, rather than discrete, with meanings carved off determinately from each other, is needed to get the same effects. For example , encyclopedic knowledge may influence the organization of senses not because semantic information scales into world knowledge, but because all languages have systematic tradeoffs between truth and implicature, both of which are discretely and modularly defined. The classic example here is but, which has the same truth conditions as and but differs from and on the pragmatic inference (implicature) it triggers about contrast with presumed information . The truth and implicature of but are quite distinct and do not scale into each other, as a model of continuous lexical meaning would have it. Clustered senses with ostensibly fuzzy interactions...