Oh, if only I knew. I've been reading Kolodny's "How Does Coherence Matter?" and while I've been somewhat sympathetic to his error theory, I'm starting to have some doubts. It seems that there's something bad about incoherence, say, believing p and p's negation. We might try to explain the violation intuition by saying that subjects who believe p and p's negation violate (N):
(N) In any given case, one is required by rationality (either not to believe p or not to believe ~p).
Or, we might try to explain the intuition by appeal to (R1):
(R1) In any given situation, either it will be the case that (one is required by reason not to believe p), or it will be the case that (one is required by reason not to believe ~p).
By believing p and ~p, you ensure that you'll believe something or other that you are required by reason to refrain from believing. Kolodny offers some reaosns for thinking that we need more than (R1) to account for our various violation and satisfaction intuitions, but here's a worry that I think isn't yet on his list. Suppose in some given situation it is the case that one is required by reason not to believe p but you believe p anyway. Suppose, however, that your attitudes are coherent. That's bad. However, isn't it always worse to believe what one is required by reason not to believe when one does so by having incoherent beliefs? It seems worse from the subject's point of view that she's believed against reason and done so incoherently. But, then could the demands of coherence really be nothing but the demands of the reasons on the attitudes?
I'm not quite sure that this is an intuition easily accommodated by someone who is willing to deny that there are demands of formal coherence as such. (It's late, I'll add more after bedtime)