The world is not enough : situations, laws and assignments in counterfactual donkey sentences

Abstract

In this thesis, I explore the phenomenon labeled counterfactual donkey sentences in the literature (van Rooij 2006, Wang 2009). Counterfactual donkeys combine the features of classical (indicative) donkey sentences (Geach 1962) with features of counterfactual conditionals: They feature an indefinite noun phrase in the antecedent which is optionally anaphorically picked up by a pronoun in the consequent, as well as the morphological marking of counterfactual conditionals, e.g., an additional past tense in the antecedent and "would" in the consequent in English.(1) If John had owned a donkey, he would have beaten it.Counterfactual donkeys present a problem for similarity-based approaches to counterfactuals (Lewis 1973), because they sometimes require us to consider possibilities beyond the most similar ones, as the classical theory would have it: (1), for example, allows us to draw conclusions about John's attitude towards all donkeys, not just the ones he owns in the most similar worlds. In this thesis, I compare three different solutions to this problem: First, I discuss a combination of Dynamic Predicate Logic (Groenendijk & Stokhof 1991) with a variably strict semantics (Lewis 1973) and a dynamic strict semantics respectively (von Fintel 1999), concluding that only the latter accounts for the full set of data. Second, I consider whether this solution can be extended to D-type theories (Elbourne 2005) and conclude that this requires us to make stipulations that are not required in the dynamic framework. Third, prompted by additional data that shows the same behaviour in the absence of an overt indefinite, I explore the option of accounting for counterfactual donkeys by switching to a situation-based framework that standardly considers a larger set of possibilities (Arregui 2009), optionally restricting this set by law-like regularities in the context. In evaluating all three proposals, I suggest that there are two possible further avenues: Exploring the empirical evidence for positing covert indefinites, allowing us to continue with a dynamic framework, or working out detailed accounts of pragmatic notions like relevance and law-like regularities, in order to precisify the situation-based framework.