1991; Fauconnier and Turner 1994 propose a somewhat similar view in terms of a“generic space”). Here the
polysemy is the result of a feature variation, with no fully specified sense as core.
However these disputesare resolved,allmajor schools ofthoughtin conceptual and cognitivesemantics agree that the
linguistic parallelisms shown in (13)–(16) reflect substantive parallelisms in the concepts these sentences express, and
thereby reveal fundamental organization in human conceptual structure.
11.8 Function-argument structure across semantic elds
11.8.1 Some basic state- and event-functions
Every theory of semantics back to Frege acknowledges that word meanings may contain variables that are satisfied by
arguments expressed elsewhere in the sentence. In Chapter 5, for example, we treated the meaning of eat as a two-
placefunctionEAT(X,Y), where X is expressed as the subjectofeatand Y is optionallyexpressed as its object. (And we
recall that such organization constitutes a challenge for neural network theories of language). Many approaches,
including most of formal semantics, leave it at that, simply expressing verb meanings as unanalyzed functions written
in capital letters.
However, other approaches take the decomposition of verb meanings seriously, asking if there is a set of primitive
functions out of which the semantic argument structure of verbs (and other argument-taking words) can be built.
Schank (1973) was an earlyattemptin theAI tradition;Millerand Johnson-Laird (1976) offeredextensiveanalysis with
an attempt to ground verb meanings in psychological primitives; Wierzbicka (1985; 1987; 1996) has offered extensive
analyses of verbs in terms of a small set of English words (the latter two in fact deal with far more than just verbs).
My own approach to verb decomposition (Jackendoff 1976; 1983; 1990a) grew initially out of the insights of Gruber
discussed in the previous section; a similar approach in quite a different notation appears in cognitive grammar
(Langacker 1987). The most basic unit is a two-placefunctionBE(X,Y), supplemented by afield feature.Thefieldfeature
determines the character of the arguments of BE and the sorts of inferences that can be drawn. If thefield feature is
Spatial, X is an object and Y is a location where X is located. This configuration in CS interfaces with the SpS
conceptualization of space and the associated inferences; it is the configuration expressed, for instance, in our old
favorite,The little star is beside a big star. If thefield feature onBE(X,Y)isPossession, X is an object and Y is a person who
owns it; this configuration invokes inference