It used to be fashionable for compatibilists to analyze "x can A" as something like:
- x would A if x wanted to
- You would run a five minute mile if you really wanted to.
- You can run a five minute mile.
I submit that for the same reason that (1) does not entail (2), (0) does not entail that x can A. Just as (1) raises the question whether you could get yourself to really want to, so too (0) raises the question whether you could get yourself to want to.
Objection 1: There is a difference between conditioning on "wanted to" and conditioning on "really wanted to". The latter condition requires a particular degree of desire while the former simply attributes the desire. One cannot infer "x can A" from "x would A if x wanted to with degree D", but one can infer it from "x would A if x wanted to".
Response: Actually, the "wanted to" in (0) has to say something about the degree of desire, or else "x can A" would not entail (0). For it can be true that x can A and x wants to A, but x does not A, because x does not sufficiently want to A. So the "wanted to" in (0) has to rule out, for instance, the case where x wants to A, but only just a little.
Objection 2: Claim (0) is a straw man. What really should be said is something like:
- x would A if x wanted to A more than x wanted any alternative.
Response: But now the bear case comes back. For among the alternatives to running a five minute mile, for someone who does not habitually do so, there is the avoidance of severe exhaustion and pain. It could well be that only a threat like a hungry bear could make one want to run a five minute mile more than one wants to avoid these kinds of alternatives. And in that case, (3) could be true, because if one wanted to A more than one wanted any alternatives, that was because one's reason for Aing was at least as motivating as a hungry bear. But it could still be false to say that one could run the five minute mile sans hungry bear (or equivalent).
I should end by saying that I am not that happy with my response to (3). I fear it is finkish.[note 2]