In this talk, I will discuss context dependence in two classes of scalar adjectives: relative adjectives (RA) like tall, and (maximum standard) absolute adjectives (AA) like empty.
Both are context sensitive: in sentences like 'The candle is tall' or 'The glass is empty', the standard for what counts as tall or empty varies with aspects of the context of use.
It is largely agreed that interpreting RAs requires a comparison set supplied by the context; the comparison set fixes the standard for what counts as tall. However, there is ongoing debate about the source of variability for AAs (Kennedy 2007; Qing & Franke 2014; Lassiter & Goodman 2015).
Unlike RAs, AAs can be interpreted without reference to a comparison set (e.g. a 100% empty glass); the availability of precise interpretations also means that AAs lack some of the signature properties of vague predicates.
Previous research has argued that interpreting AAs instead involves setting a standard of precision, i.e. asking how much ‘pragmatic slack’ can be tolerated given the goals of the conversation, the norms of the situation, or what is known about the speaker.
I will present some experiments – two offline truth value judgment experiments and one Visual World eye-tracking experiment – that address one prediction of (a particular variety of) such accounts: that standards of precision should vary by properties of the speaker, while standards of comparison for RAs should only be a function of the contextually-supplied comparison set.
The findings suggest an account of AAs that involves pragmatic reasoning about precision, by showing that inferences about speaker intent influence judgments about AAs. They also raise questions about the grain of the variability permitted by existing theoretical accounts and observed empirically.