Binding into Character

Abstract

Since Kaplan's “Demonstratives,” it has become common to distinguish between the character and content of an expression, where the content of an expression is what it contributes to “what is said” by sentences containing that expression, and the character gives a rule for determining, in a context, the content of an expression. A tacit assumption of theories of character has been that character is autonomous from content— that semantic evaluation starts with character, adds context, and then derives content. One consequence of this autonomy thesis is that the rules for character can contain no variables bound by content-level operators elsewhere in the sentence. Tacit appeal to this consequence features essentially both in Jason Stanley's recent argument, in “Context and Logical Form,” that all contextual ambiguity must be linked to “elements in the actual syntactic structure of the sentence uttered”, and in my arguments against character-based theories of complex demonstratives in my “Complex Demonstratives.” However, I argue here that the autonomy thesis is unmotivated, and show that we can separate Kaplan's notion of character into two independent components: an aspect of meaning which is context-sensitive, and an aspect of meaning that is exempted from scopal interactions with other operators. The resulting semantic framework allows constructions similar to Kaplan's rejected notion of “monsters begat by elegance,” but which are both more empirically adequate and more theoretically versatile. Having made the distinction between context-sensitivity and autonomy from scopal interaction, I show how that distinction allows binding into the character of expressions and hence undermines the immediate success of both Stanley's argument and my previous argument against character-based theories of complex demonstratives, and discuss briefly the prospects for reinstating modified versions of those arguments. Finally, I show how that same distinction allows a defusing of Kripke's modal argument against a descriptive theory of names. Once autonomy from semantic interaction is separated from context-sensitivity, the first of those two alone can be used to capture the modal rigidity of proper names. I argue that the resulting semantic account, which bears important resemblances to (but is, I think, a significant generalization on) Recanati's use of the REF feature, captures some of the core intuitions of wide-scope responses to the modal argument without incurring the weaknesses of those responses.