Attributive Wrong Bernhard

In well-behaved cases of adjectival modification, the adjective-noun sequence describes a firstorder property whose content is fully determined by the contents of the adjective and the noun. For example, the truth conditions of They hired a vegetarian dentist, They hired a former linguist, and They hired an electrical engineer, are functions of the property of being a vegetarian dentist, of being a former linguist, and of being an electrical engineer, respectively. I will refer to adjectival modification that is well behaved in this sense as “local”. While local modification is the rule, a number of “non-local” cases have been described in the literature. For example, the sentence We saw an occasional raccoon conveys that we saw a raccoon occasionally, and so its truth conditions are obviously not a function of the property of being an occasional raccoon (Bolinger 1967, Larson 1999, Zimmermann 2003). The identity adjective same (Carlson 1987), epistemic adjectives like unknown or unidentified (Abusch and Rooth 1997), and possibility adjectives like possible or conceivable (Larson 2000, Schwarz 2005) are also known to participate in non-local modification. Non-local adjectival modification is a challenge for compositional semantics, as its analysis seems to call for special assumptions about the syntax or semantics of the adjectives in question. In this paper I will present a case study on non-local attributive wrong, commenting on brief discussion in Larson (2000). Larson derives non-locality in this case by equating non-local wrong with clause embedding wrong whose clausal complement has elided. I will argue that non-local wrong cannot in fact be so analyzed and that its semantics should instead be reduced to that of clause embedding wrong in the lexicon.