The Evaluability Hypothesis : The Syntax and Semantics of Polarity Item Licensing in Swedish

Sammanfattning: This dissertation is concerned with the empirical and theoretical aspects of polarity item licensing in Swedish. I argue that polarity items are semantically sensitive to evaluability, a concept that refers to the possibility of accepting or rejecting a clause as true in a communicative exchange. Clauses are either evaluable or non-evaluable. According to the present hypothesis, non-evaluable clauses constitute natural environments for polarity items, and may host polarity items without any kind of formal (syntactic) licensing. Evaluable clauses, in contrast, are restricted environments, and may only host formally licensed polarity items. Under this proposal, the occurrence of polarity items in such non-evaluable environments as yes/no-questions and conditionals is regarded as the prototypical case, while the occurrence of polarity items in (evaluable) affirmative and negative sentences is the marked, or exceptional, case. It is further argued that evaluability is mirrored syntactically in the left-periphery of the Swedish clause: evaluable clauses have Spec-CP, while non-evaluable clauses lack Spec-CP. I propose that this observation can be accounted for within a Minimalist framework by assuming a fixed (but arbitrary) connection between evaluability and the edge-feature in C (Chomsky 2008). The focus on Spec-CP distinguishes the present work from most minimalist accounts of Swedish clause structure, which tend to be primarily concerned with verb movement to C. All in all, the Evaluability hypothesis should be regarded as an alternative to previous theories of polarity item licensing put forward in the literature. It incorporates the insights of Ladusaw’s (1979, 1980) downward entailing hypothesis, but restricts its applicability to evaluable clauses. Furthermore, a revised version of Linebarger’s (1980) Immediate Scope Constraint is argued to be fully compatible with the present hypothesis. However, the Evaluability Hypothesis challenges the widely acknowledged Veridicality Hypothesis (Zwarts 1993 and Giannakidou 1998), since it is shown that the distribution of polarity items in Swedish is better accounted for in terms of evaluability than veridicality.