Abstract: It is customary for philosophers to distinguish between epistemic and metaphysical modality. This distinction tends not to be central in linguistic discussions of modality, but a nearby one is: this is the distinction between epistemic and circumstantial (or 'dynamic') modality. Circumstantial modality concerns what an agent or object can do or be, independently of our knowledge of it. In this sense it has the alethic flavor of the philosopher's metaphysical modality: it is, one might say, the 'best deserver' in natural language for the metaphysical modality role.

In light of this, it is helpful for those interested in metaphysical modality to investigate the semantics of circumstantial modality, often expressed in English by the modal auxiliary 'can.' I do this, and note an interesting feature of 'can': sentences of the form 'x can F' always express circumstantial (or deontic) modality, and never express epistemic modality. This observation is supported by corpus searches and also appears to be cross-linguistically robust. I propose an explanation for this phenomenon. This is that (contrary to the standard semantic story of Kratzer and Lewis) 'can F' in such sentences is functioning as a predicate, one which expresses a power or potentiality that the sentence ascribes to x. On this view (again contrary to the view of Kratzer and Lewis) possible worlds never come into the semantic story. Having argued for this, I revisit the question of what bearing, if any, all of this has on the topic of metaphysical modality, in particular its alleged connection to possible worlds.