Experimenting with every American king

The standard contemporary semantics for 'every' predict the truth of occurrences of sentences with restrictors that denote the empty set, such as 'Every American king lives in New York'. The literature on empty restrictors has been concerned with explaining a particular violation of this prediction: many assessors consider empty-restrictor sentences to be odd rather than valued, and they are apparently more likely to do so when such sentences include determiners like 'every' as opposed to those like 'no'. Empirical investigation of this issue is overdue, and I present the results of three experimental surveys. The first unexpected outcome is that there is no evidence of a contrast in assessors' tendencies to judge sentences to be odd based on determiner type. An additional surprising result is that those assessors who assign a truth value to sentences where 'every' combines with an empty restrictor overwhelmingly assign the value false. The full results do not fit straightforwardly with any existing account.