I'm reading Raz'
The Truth in Particularism—I find the sections on Dancy kind of unconvincing. (There's one part that I really hope isn't just making the claim that if you think reasons are "generic features of action-types", then you will have problems with particularism, because—of course. And if you take a different view of reasons—or, for that matter, carve up action-types sufficiently finely—the point of that section against Dancy seems to vanish.) But here's a different part. One of Dancy's examples is this: if you have borrowed a book from a friend, and return it, your reason in doing so is that you borrowed it. But that you borrowed it would not be a reason to return the book to your friend if you discovered that he had stolen it from the library, because in that case you should do something else. (Return it to the library, I guess. We don't get that part.) So the very same consideration—that you had borrowed the book—is a reason for action in one case but not the other, according to Dancy. And indeed people will tend to cite just the fact that the book was borrowed as their reason for returning it in the normal case. Raz observes that this doesn't necessarily touch on reasons at all, but just on someone's understanding of a reason; someone who just cites that the book was borrowed may simply incompletely understand the real reason for which he acts. And:
[T]ake the book loan example moentioned above. Most likely when asked people would say their reason for returning the book was that they borrowed it, or promised to return it. But if asked at the time would the fact that the person from whom it was borrowed had the right to possess it, that he did not steal it, etc. be relevant to their reason (i.e. was their reason that they borrowed from someone entitled to lend them the book), most people would say yes. Regarding those people the example fails. Their reason was not one which applies in cases of a borrower who stole the book.
And indeed the real reason (R) to return the book has the form r ∧¬(d1 ∨ … ∨ dn), where r is the seemingly main reason, that the book was borrowed, and each d is a defeater (that the book was stolen; that the book describes how to make bombs and you suspect your friend has a more than academic interest in the subject, etc). But even if that is the reason, you can't infer from the results of asking someone about to act "isn't your reason also this?" that it was or wasn't; the effect of asking a question like that is to make salient a possibility that may never have entered the person's mind—why should we not think that, rather than illuminating the reason on which the person was going to act all along, asking the question changes that reason?
I also wonder, to engage in a little slantwise ipsedixitry, if Raz would endorse a parallel to R in the case of justification for belief—that you're justified in believing something to the extent that you can eliminate every possible ground of doubt. Further: how can R serve either the guiding or the evaluative functions of reasons? The odds that anyone will know that R obtains are very small; in fact, Raz acknowledges that there is reason to think that one could not know all the defeaters for an action. And there seems to be no way to tell that someone acted for R, rather than for the extremely similar reason S—which isn't the reason to act, in this case—r ∧¬(d1 ∨ … ∨ dn-1), where dn happens not to obtain in this case. Why think that someone acted for R, but with a limited understanding that only grasped S, instead of thinking he acted for S, with the correct understanding that that was the reason?