Epistemic Uses of "Might" and "Likely" Are Only Indirectly So
Authors: Quinn Harr and Alexander Williams

We can reason about what makes modal claims true in terms of the contingencies of their modal domains (cf. (2) as a response to (1), and (4) to (3)). Yet for some apparently epistemic modal claims, this sort of explanation is not available. (4) is an anomalous response to (5), though (6) is not.

(1) Wisconsinites can shoot pigeons during hunting season.
(2) That could be because they haven’t been listed as a threatened species.

(3) It’s {plausible/epistemically possible} that Jones is contagious.
(4) That could be because we haven’t asked him his diagnosis.

(5) John {might/is likely to} be contagious.
(6) That could be because he works in the worst hospital ward.

We argue that the best explanation of this fact is that modals like “might” and “likely” are not directly epistemic. Unlike “plausible” and “epistemically possible”, it is not evidence qua evidence that determines their domain. Hence, we cannot reason about what would make such claims true in terms of the contingencies of the evidence. Instead, their domain is determined by propositions construed as circumstances (thus enabling reasoning about these claims in terms of contingencies of the circumstances). Only as grounds for these claims is evidence ever germane, often making them appear epistemic. Hearers often infer that the speaker’s evidence does not enable them to say whether the modal’s complement is true: i.e., that the complement is epistemically possible for them. An epistemic proposition is often communicated when such claims are made, then, but only ever indirectly so.

Reporting Modal Beliefs
Author: Quinn Harr

Modals in belief reports seem to display an important asymmetry. Belief reports like (1)—with non-epistemic modals—report a belief about the modal’s domain. For (1) to be true, Mary must have a belief about what is compatible with the laws. In contrast, belief reports like (2)—with apparently epistemic modals—seem not to report a belief about the modal’s domain. For (2) to be true, Mary need not have a belief about what is compatible with her or anyone else’s beliefs or knowledge.

(1) Mary believes that the president can declare war on his own.
(2) Mary believes that there might be landmines.

The standard response to this observation is that epistemic modals like might interact differently with belief (and other attitude) verbs from non-epistemic ones like can. They contribute not to the content of the reported attitude but instead modify its force. What (2) requires, on this view, is simply that there be landmines in at least one of the worlds compatible with Mary’s beliefs. Call this the compatibility view. I argue there is a better view of (2) which abandons the presupposition that its modal has an epistemic domain. Taking it instead to have a circumstantial one, this view permits a uniform interaction of modals with belief verbs. What (2) requires on this circumstantial view is that Mary have a belief about what is compatible with the relevant circumstances. This view, I argue, avoids problems both of descriptive adequacy and of theoretical economy facing the compatibility view.