Invariantism about ‘might’ says that ‘might’ semantically expresses the same modal property in every context. This paper presents and defends a version of invariantism. According to it, ‘might’ semantically expresses the same weak modal property in every context. However, speakers who utter sentences containing ‘might’ typically assert propositions concerning stronger types of modality, including epistemic modality. This theory can explain the phenomena that motivate contextualist theories of epistemic uses of ‘might’, and can be defended from objections of the sort that (...) relativists mount against contextualist theories. (shrink)

Infinitives are known to encode covert modality in certain environments including infinitival relatives and questions. Beyond these environments, however, the precise distribution and interpretation of infinitival modality remains poorly understood. In that light, this paper investigates infinitive-embedding _enough_/_too_ sentences like _Pat is tall enough to be the thief_ or _Lee is too old to drive_. These sentences have a modal semantics whose compositional source is contested: on one approach, the infinitive encodes the modality, and on another approach, the _enough_/_too_ morpheme (...) itself is modal. To adjudicate this debate, I consider heretofore largely overlooked _finite_-clause-embedding _enough_ sentences like _Pat is tall enough that she might/must be the thief_ or _Lee was fast enough that she won the race_. They provide, I argue, novel support for the view that the modality is in the embedded clause (whether nonfinite or finite) and not in _enough_/_too_. I then compare the covert modality of nonfinite _enough_ clauses to the covert modality of infinitival relatives, questions, and complements to attitude predicates and content nouns. I generalize that covert modality in nonfinite clauses never encodes epistemic necessity, and I tentatively hypothesize that this constraint reflects the marked status of nonfiniteness in the finite/nonfinite opposition. (shrink)