Against the standard semantics for ‘all’ and ‘every’

Abstract This paper presents a new problem for the standard semantics for ‘all’ and ‘every’ on which ‘every F is G’ and ‘all Fs are Gs’ are true just in case the set of things that satisfy ‘F’ is a subset of the set of things that satisfy ‘G’. I argue that the standard semantics makes highly counterintuitive predictions about what credences it is rational to adopt in a generalisation ‘every F is G’ when we are confident but not certain that nothing is F. I show how, unlike the better known counterintuitive result that if nothing is F, ‘every F is G’ is true, this problem cannot be avoided by appealing to pragmatic principles, nor by rejecting some orthodox principles about the relationship between credences and propositions.