Restrictive vs. nonrestrictive modification and evaluative predicates

Evaluative adjectives have often been claimed to manifest a strong, and even exclusive, preference for the nonrestrictive reading (henceforth the 'nonrestrictive bias' of evaluative adjectives). For those languages like French that allow both the post and pre-head positions for at least a subset of their adjectives, a frequent observation reported in support of this claim is that evaluative adjectives are often odd in post-nominal position. The argument relies on what has been called the complementarity hypothesis, namely the hypothesis that pre-head modifiers receive a nonrestrictive interpretation in Romance, while post-head modifiers receive a restrictive interpretation. An immediate problem for this argument is that evaluatives do appear in postnominal positions in corpora. One of the goals of this paper is to reconcile these data with the nonrestrictive bias and the complementarity hypothesis. The idea pursued is that a modifier can either be (non)restrictive according to the standard definitions, which are purely extensional, or be (non)restrictive with respect to a particular modal base α (thus, α-restrictive vs. α-nonrestrictive). Being restrictive or α- restrictive (respectively nonrestrictive or α-nonrestrictive) allows the modifier to appear in the post-head (respectively pre-head) position. In section 2, we recall the standard (purely extensional) definitions of (non)restrictivity. We then show that these definitions cannot distinguish between restrictive and nonrestrictive modifiers in a number of contexts, e.g. in non-partitive indefinites. In section 2.3, we introduce modal definitions of (non)restrictivity that solve the problem. Section 3 identifies the contexts in which evaluative adjectives can appear in post-nominal position and explains why, on the basis of the definitions of (α)-(non)restrictivity built in previous sections. The analysis proposed is compared with two previous accounts of the nonrestrictive bias of evaluative predicates.