The Conditional Analysis

I have a small remark/complaint about the literature that came up in passing as I was refereeing a paper, not for the first time. I wasn’t sure how to get this out of my craw, so a blog post seemed appropriate.

Various authors, including myself, have inveighed against “the conditional analysis” of ability. Indeed some authors (mainly in the free will literature) taken it as given that this analysis is bankrupt. And yet. In the recent literature, there are several proposals that defend a sophisticated version of the conditional analysis. This recent, elegant, paper in the Philosophical Review is perhaps the best example. So a cursory reading of the literature might suggest that reports of the demise of the conditional analysis were premature.

In fact I think this is not so. There are at least two things that might be called the “conditional analysis”, or, better, there are two roles that conditionals might be thought to play in an account of ability and the ascription of ability. Once we distinguish these, the issues in the literature become much clearer.

First, the conditional analysis might be thought to give a reduction of agentive modality to some more general variety of modality, namely that involved in the subjunctive conditional. So it is proposed that: S is able to A <-> if S were to try to A, S would A. If a reductive semantics of the subjunctive conditional could be given – perhaps in terms of possible worlds a la Stalnaker and Lewis – we would then have gone some way towards the reduction of agentive modality to the amodal. In this sense the conditional analysis of ability is of a piece with the conditional analysis of dispositions, the best-systems account of laws of nature, and other redoubtable members of the Humean repertoire.

I think there’s a pretty good case that this reductive understanding of the conditional analysis is not going to work. (I give a few reasons here and elsewhere). But . . .

Second, the conditional analysis might be deployed to address a quite different issue. This is a problem not of reduction but of extension. Here’s the issue. The things that agents are able to do – what I call their options – are somewhat different from the complements in ability ascriptions. For instance, I might have truly said in 2007: “Obama is able to become President.” But Obama perhaps didn’t have that as an option. Maybe he had a bunch of smaller options – giving this speech, shaking that hand, etc. – that together would add up to his becoming President. And there is some plausibility to understanding that as follows: there was something/somethings Obama could do such that, if he did it/them, he would become President. That is in some sense an “analysis” of “Obama is able to become President,” and it is one that makes essential appeal to a conditional.

Now I’m not sure if that proposal is right, but it’s certainly worth considering as a treatment of the extension problem and it certainly is in some sense a “conditional analysis.” It’s the kind of proposal that’s advanced in the article mentioned above. But it’s important to distinguish it from the traditional conditional analysis. Crucially, a notion of agentive modality figures as a primitive in the account, whereas the whole point of the traditional conditional analysis was to analyze that away. This rather different kind of conditional analysis involves deploying the conditional in a much different, and more modest, way.

So: the conditional analysis, understood as a reductive account of agentive modality, did not work. But, once we admit some primitive agentive modality into our account, we may introduce subjunctive conditionals into our account to solve other problems, such as the extension problem sketched above.