(Note: this post is in large part a reprint of a couple posts I wrote on the subject a few weeks back on my own blog.)
1. The case
I take it that most people are familiar with what Williams’ petrol case (“petrol” cuz he was British lol), so my exposition will be cursory. Wilfrid wants a gin and tonic and has good epistemic reason to believe that the glass of clear liquid on the counter is a gin and tonic. Actually, though, it’s a glass of petrol. Consider the following two sentences:
(1) Wilfrid has reason to drink what’s in the glass.
(2) Wilfrid has reason to drink the petrol.
My intuition in the case is that (1) is true while (2) is false. In the comments to my first post on this topic, Pitt grad student Shawn Standefer offered the following diagnosis:
I asked someone knowledgable and he said that the problem lies more with the opacity of descriptions of intentional actions than with the opacity of reason ascriptions (there is a roughly non-existent literature on reason ascriptions apart from general propositional attitude ascriptions). The opacity of descriptions of actions goes back to Anscombe, through Davidson, and on. He said to look at Carl Ginet’s “On Actoin” [sic] for an introductory discussion.
My familiarity with the opacity of description of intentional action is cursory. I take it that Shawn’s point is supposed to be something like this. Generally speaking, A intentionally φ’s only if (i) A φ’s and (ii) either φ’ing is the description under which A φ’s, or ψ’ing is the description under which A φ’s and A is in some sense aware that by ψ’ing she will φ. (I’m sure this needs to be corrected. Suggestions?)
2. A proposal
I’m happy to grant that the reason we can’t move from (1) to (2) is that reasons-ascriptions inherit their opacity from the description of intentional action. It would be extremely interesting to see a worked-out account of what exactly the connection is. As a rough first pass, we might say that A has reason to φ only if it is possible (in some to-be-specified sense) for A to intentionally φ. The hesitation in moving from (1) to (2) makes me think that the relevant sense of possibility will be rational and relativized to (what’s presupposed to be) the content of A’s doxastic state. That is to say: it is possible in the relevant sense for A to intentionally φ iff it would be rational, given what A believes, for A to intentionally φ. (I realize this isn’t exactly an argument, though it seems like a plausible starting point.) So the revised account of the connection would be something like what’s given in (O):
(O) A has reason to φ only if it would be rational, given what A believes, for A to intentionally φ.
(O) has the virtue of explaining (in what’s to my mind an intuitively satisfying way) the divergent intuitions about (1) and (2). In Williams’ case, it would be rational, given what Wilfrid believes, for Wilfrid to intentionally drink what’s in the glass. It would not be rational, given what Wilfrid believes, for Wilfrid to intentionally drink the petrol. I think this is an interesting result, and I haven’t really seen it discussed anywhere in the literature on normative reasons.
3. A word on intuitions and the folk
Ethicists seem to be put off by my intuitions about (1) and (2), as they seem to universally converge on the intuition that each sentence is false in this context of use. (This was, of course, Williams’ reaction as well.) Interestingly, though, my intuitions about (1) and (2) seem to be fairly standard among the folk. Steve Campbell, bless his heart, sent out a survey to some of his non-philosopher acquaintances, describing Williams’ petrol case and asking them about their intuitions about (1) and (2). There seems to be some kind of consensus that, in Williams’ case, (1) is true and (2) is false. (Caveats: the sample was very small and the consensus wasn’t perfect. Still, having done a bit of my own informal intution pumping, my impression is that this response will tend to be the standard one.)
I take this to be prima facie evidence that, despite the ordinary co-referentiality of ‘what’s in the glass’ and ‘the petrol’, these two expressions are, in fact, not co-referential in this context of use.[*] (You can resist this, but you’ll need some kind of error-theory to go with your resistance. This doesn’t, to me, seem like a promising direction, given that only those corrupted by the ethics literature seem to converge on the intuition that (1) and (2) are both false.) Here, then, are some lessons.
First, you can’t argue — as I suspect many ethicists will be tempted to do — for the falsity of (1) by citing supposed entailment relations between (2) and (1). If ‘what’s in the glass’ and ‘the petrol’ are not co-referential in this context of use, then (2) does not entail (1) and (1) does not entail (2).
Second, we need some kind of explanation for the non-co-referentiality of ‘what’s in the glass’ and ‘the petrol’ in this context of use. It seems to me that the best explanation will be that we have a specific sort of intensional context on our hands, one where referring expressions behave a lot like they do in the clausal complements of belief-ascriptions. A more complete explanation might be wanted, in which case the (O) criterion proposed in the prior section seems to be adequate.
[*] I should make clear that I’m bringing along a lot of my own belief-ascriptions baggage here. For independent reasons, I think it’s best to assimilate intensional contexts to a special kind of extensional context where the semantic values of the constituents are determined by a kind of pragmatic accommodation. Even if you don’t go this route, if you think there are entailment relations between (1) and (2) (as most ethicists do), you’ll need to give an error theory that explains why most people are willing to affirm (1) and reject (2).