Evidence for a mixed approach to E-type anaphora

Abstract

The program of the field of semantics is to model or formally represent the meaning in natural language of sentences and discourses and of different constructions and interpretations. E-type anaphora constitutes a family of constructions and interpretations that have proved especially difficult to model formally. Previous approaches generally have aimed to analyze all E-type anaphora using similar semantic tools. This has resulted in models or representations that are complex in order to deal with the full range of E-type anaphora. We can trade off complexity of representation with number of different semantic tools that we are allowed to use to analyze different categories of E-type anaphora. In this work, I argue that E-type anaphora should receive a “mixed approach” in that different semantic tools should be applied to different categories of E-type anaphora to arrive at simpler analyses. Simpler analyses are preferable a priori. Though mixed approaches are not preferable a priori, a mixed approach to E-type anaphora is justified because different sorts of expressions in natural language are more aligned with certain subcategories of E-type anaphora. The use of choice functions to model a subset of E-type anaphora is one of the main novel theoretical contributions of this work. This subset of E-type anaphora includes an interpretation of donkey anaphora introduced here using the term “coreferential existential”. This interpretation has received little attention in the literature. The main empirical contribution of this work is to show that indefinite syntax and semantics has a larger effect on interpretation of donkey sentences than definite or pronoun syntax and semantics. Therefore, an approach that traces differences in interpretation to the indefinite antecedent NP/DP in an E-type anaphoric dependency is best supported, contra previous works that focus only on anaphoric definite or pronoun.