require for their evaluation.It misses a generalization to set their treatment off as special. They don't require different
resources, just a subset of the ones needed to evaluate sentences in context.
The upshot is that there is little to be gained in theoretical economy or explanatory adequacy by making a strict cut
between logical and non-logical properties. Logical properties alone don't make enough distinctions among lexical
items; the semantic features involved in logical properties don't form a disjoint set from those involved in non-logical
properties; and any syste mthat can deal with non-logical properties can easily deal with logical ones as well. So logical
properties are just some heterogeneous subset of semantic properties, and there seems no point in recovering them
“first”fro mlinguistic for mthen adding the other properties on“later.”^143
9.7.3 Grammatically realized vs. grammatically irrelevant content
Proposal (12C) seeks to segregate as“linguisticsemantics”just those features of meaning thatplay a role in syntax and
morphology, for instance the distinction between singular and plural but not that between red and green. Various
versions of this idea have been proposed by Bierwisch (e.g. Bierwisch and Lang 1989), Pinker (1989), and Grimshaw
(1993). Of course, different languages grammaticalize different semantic distinctions. For instance, many European
languages base grammatical gender distinctions in part (but in part only) on semantic gender distinctions, but English
makes no such grammatical distinction. Similarly, many languages make a mood distinction based in part on
evidentiary status (indicative vs. subjunctive), but English does not. We could therefore envision two versions of this
proposal, one in which the elements of linguistic semantics are language-specific, and one in which they are universal,
but differentlanguages happentogrammaticalize differentsubsetsof linguisticsemantics. The former positionwill fall
under the proposal to be discussed in the next subsection, so we will restrict ourselves to the latter here.
Accordingtothisproposal, thedifferencebetweenredandgreen, betweendogandcat, and betweentwenty-sevenandthirty-
eightmake nogrammaticaldifference, so they are not part of linguistic semantics. Hence this version of linguistic
semantics is even narrower than the previous two. It provides no basis
SEMANTICS AS A MENTALISTIC ENTERPRISE 289
(^143) Levinson (2000) is correct in pointing out that the necessary conditions must still be distinguished from the defeasible ones, say by notatingthem in“different colors”or
underdifferent logicaloperators. But hethinks thatthisdifferencecomesfrom thekind ofrulethatleads totheconditionin question—semanticor pragmatic.The position
here is that the difference simply is a lexicalproperty of the conditionin question; e.g.‘upwardness’is simply marked necessary in the lexicalentry ofriseand defeasible in
that ofclimb. See section 11.6.2 for more discussion.