Abstract

Recent challenges to Kratzer’s canonical contextualist semantics for modal expressions are united by a shared methodological practice: Each requires the assessment of the truth or warrant of a sentence in a scenario. The default evidential status accorded these judgments is a constraining one: It is assumed that, to be plausible, a semantic hypothesis must vindicate the reported judgments. Fully assessing the extent to which these cases do generate data that puts pressure on the canonical semantics, then, requires an understanding of this methodological practice. Here I argue that not all assessments are fit to play this evidential role. To play it, we need reason to think that speakers’ assessments can be reasonably expected to be reliable. Minimally, having such grounds requires that assessments are given against the background of non-defectively characterized points of evaluation. Assessing MacFarlane’s central challenge case to contextualism about deontic modals in light of this constraint shows that his judgments do not have the needed evidential significance. In addition, new experimental data shows that once the needed scenario is characterized non-defectively, none of the resulting range of cases provides data that cannot be accommodated by a Kratzer-style contextualism.