At some point when I wasn't paying attention Velleman's "The Way of the Wanton" was published; rereading it now I'm struck by two (very Vellish) points about reflexive desires that seem completely unjustified. One is offered apparently as an obiter dictum: Velleman identifies the two desires that the famous "willing addict" has as "his addictive urge to take the drug and a desire that he take the drug because of that urge, which is a desire for himself to take the drug and hence a reflexive desire to take it." (p 175 of Practical Identity and Narrative Agency). If the referent of "which" is the "addictive urge to take the drug", then one wishes to say of the reflexive pronoun in the later exposition of its content what Baier said of Chisholm's ever-present agent, namely, that it is somewhat over-advertised, or anyway too much is being made of it; if I have the urge to take a drug, then indeed the urge is for me to take it, not for you to. If that is sufficient to make it a reflexive desire then all desires that aren't targeted at someone else are reflexive. (Whereas it seems more natural to me to call only the desires that involve oneself as another reflexive. One half suspects that here the absence of something like Latin's ipse is felt: I want to take the drugs myself, not, I want myself to take the drugs. (Where the "myself"s would be Latinized respectively as ipse and me.) Velleman's phrasing is more in line with the latter but also more in line with the oneself-as-another reading, and that reading is not at all forced by the example.) If, on the other hand, the "which" refers to the desire that he take the drug because of the addictive urge, it is still unclear where the reflexivity comes in; a desire that I φ because of p is indeed a desire that I φ; why (hence) it should be a reflexive desire to take it is beyond my ken.
Later on the same page Velleman writes:
This explanation necessitates a subtle clarification about the content of second-order volitions. A second-order volition that one be effectively moved by a first-order desire cannot have the content that one be effectively moved by the first-order desire alone. The content of a second-order volition must be that one be effectively moved by the first-order desire as reinforced by this very volition. Otherwise, the volition would tend to be self-frustrating.
It would be self-frustrating because if I desired to be effectively moved by the desire to φ alone, either (a) the first desire would not be effective, or (b) it would be effective in which case I would not have φed owing to the desire to φ alone. This does not at all establish that a second-order volition must be one that runs "φ because effectively moved by the first-order desire to do so as reinforced by this very volition", because I could have a second-order volition that runs "φ because effectively moved by the first-order desire to do so". We can agree that a second-order volition can't have the content "be effectively moved by the first-order desire alone", but it can have the content "be effectively moved by the first-order desire" alone.
(Of course my head swims when the topic of self-referential mental states comes up; how—I wonder—do they get their content at all? One is brought to imagine some kind of mental fixed-point combinator, but even those only work by stages, and what could account for the first stage? Harman has addressed this at various points but nowhere in a way that rids me of my confusion.)