Embedded Exhaustification: Evidence from Almost.

Recent work in the domain of scalar implicatures argues for the existence of a grammatical device of covert exhaustification exh (Krifka 1995; Chierchia 2006; Fox 2007; Fox and Spector 2009; Chierchia et al. 2011; Sauerland 2012). This covert operator strengthens sentence meanings in a way similar to neo-Gricean quantity inferences, but is crucially different in that it can strengthen embedded propositions. Embedded exhaustification can interact with higher operators, and many arguments for exh rest on these effects. This article presents another argument for the presence of a covert exhaustification operator in the grammar from its interaction with the approximative modifier almost. Intuitively, almost α might be paraphrased as 'close to but not quite α'. I motivate, and take as a starting point, a denotation for almost which derives both its proximal ('close to') inference and its negative ('not quite') inference as entailments (Horn 2002, 2005, 2011; Penka 2006). I present a new instance of a puzzle noted in Nouwen (2006) that poses a challenge to such an analysis: in certain environments, almost α can have an interpretation in which it seems that the negative inference has disappeared, and instead a monotonic inference arises (i.e. α is inferred rather than negated). I revisit and spell out the previously suggested solution (Horn 2005; Penka 2006; Spector 2015) that almost may take logical scope higher than other operators which c-command its surface position. The analysis of the novel case that I present assumes this scopal mobility, but crucially also relies on covert exhaustification. Under this proposal, the negative inference does not actually disappear, but is interpreted in a position where its contribution is obscured. [ABSTRACT FROM AUTHOR]