Between “Cost” and “Default” of Scalar Implicature: the Cost of Embedding

Abstract

Two main points constitute a matter of debate concerning the phenomenon of Scalar Implicatures (SIs): the place of their derivation, which opposes a “recursive”/grammatically driven approach such as Chierchia’s (Chierchia, 2002&2006; Fox, 2003; Landman, 1998; Levinson, 2000) to traditional Neo- Gricean approaches that view SIs as genuinely post-grammatical/pragmatic processes that are added “globally”, independently of compositional semantics (Russell, 2006; Sauerland, 2005; Spector, 2003 a.o.); and the question of the processing cost of SI computation, which most of the experimental works on SIs have recently been focused on (Bott & Noveck, 2004; Breheny, Katsos, & Williams, 2005; Noveck & Posada, 2003). Orthogonal to this debate, our contribution is based on the assumption that SIs are derived locally (following Chierchia, 2006) and tests the effect of logical abstract properties of the context (e.g. monotonicity) on the computation of implicatures and their cost. Our main finding is that a “cost” is found only when implicatures are added despite the fact that they lead to a weakening of the overall assertion (namely, in Downward Entailing contexts): this loss in informativity, and not implicature computation per se, is interpreted as the source of this “cost”.