Leffel, Timothy and Cremers, Alexandre and Gotzner, Nicole and Romoli, Jacopo
Journal of Semantics
Publication year: 2019


While vagueness and implicature are two of the most studied topics in semantics and pragmatics, their interaction has received little attention in the literature. Part of what makes this an interest- ing domain of investigation is that both implicature and vagueness involve an inherent amount of interpretive uncertainty: with vagueness, there is uncertainty about whether a predicate applies to an object; with implicature, there is uncertainty about whether a candidate inference was intended by the speaker. Gradable adjectives provide an ideal domain for investigating how multiple sources of uncertainty interact in semantic/pragmatic interpretation. In this paper, we argue that interaction between vagueness and implicature is the source of a puzzling and, to our knowledge heretofore unobserved, asymmetry between the interpretations of different classes of intensified gradable adjectives under negation. Specifically, the interpretation of sentences containing the collocation not very ADJ varies depending upon whether ADJ is a relative standard gradable adjective (e.g. tall) or a (minimum standard) absolute gradable adjective (e.g. late), in the sense of Kennedy & McNally (2005); Kennedy (2007); a.o. We first report an experiment investigating the interpretation of not very ADJ and related expressions across these two adjective types. Our results provide strong evidence that not very ADJ gives rise to an ADJ-implicature for minimum standard adjectives like late but not for relative standard adjectives like tall. We propose that this asymmetry follows from an interaction between scale structure and a novel constraint on implicature calculation according to which implicatures are not derived if they lead to “borderline contradictions” (Ripley 2011; Alxatib & Pelletier 2011; a.o.). We then provide evidence for this hypothesis based on reconstructed interpretations of logically complex predicates from the experimental data. In sum, the interpretive asymmetry is reducible to differences in scale structure and vagueness: relative adjectives are vague in a way that absolute adjectives are not, which we argue affects the relative likelihood that an implicature will be generated.