On standard accounts of modal expressions, sentences like (1) and (2) have been taken to express the same propositions, (2) making explicit the epistemic nature of the modality left implicit by (1).
(1) Jones might be dead.
(2) For all we know, Jones might be dead.
A problem for such accounts, however, is the fact that (1) and (2) do not support the same counterfactual continuations. (3), for example, is an acceptable follow-up to (2) but not to (1).
(3) But that’s just kind of a fluke, since we could have investigated his disappearance much more thoroughly.
This sort of problem does not generalize to other, non-epistemic modals, as (4) and (5) show.
(4) You can get a license in Georgia when you’re 16. But that’s just kind of a fluke, since Georgia could have had the laws New Jersey did.
(5) Given its laws, you can get a license in Georgia when you’re 16. But that’s just kind of a fluke, since Georgia could have had the laws New Jersey did.
Why should this implicit-explicit distinction be important for epistemic modals but not for non-epistemic ones? Some have argued on independent grounds that implicit epistemic modals exhibit idiosyncratic behavior (Yalcin (2007)), but such accounts are insufficient to handle the contrast exemplified by (1) and (2). I argue, instead, that the nature of epistemic modality has been misunderstood, that (1) does not contain an epistemic modal, and that this fact explains the difference between (1) and (2). Getting clear on the nature of epistemic modality also potentially help clears up a host of other problems (presumed) epistemic modals have posed for standard semantic theories.