Inconceivable Support Relations: Reply to Stanford

Philosophers are drawn to the Atomic Theory like a dog to an old shoe, but my results about realism and anti-realism in Tracking Truth, and the distinctive position I carved out on their basis, are independent of the fate of my comments about that historical case. I will defend those comments against Stanford’s objections below, but first I will explain the argument he ignored, because its results undermine not only historically important anti-realist positions, but also the approach via unconceived conceivables that Stanford’s criticisms (and his own pessimistic induction) depend on. My argument is not based on lack of imagination for alternative hypotheses. The issue is claims about equal evidential support that empiricist defenses of epistemological anti-realism, including Stanford’s, must and do appeal to. In Chapter 6 of Tracking Truth, I show that the best known probabilistic definitions of evidential support imply that the anti-realist’s key equal-support claims are either false or inconsistent with the view he wants to base on them. There are good reasons to think that it is not even possible to define a probabilistic notion of support that will do the job the antirealist needs. Due to the role of auxiliary assumptions in determining evidential support relations, the only way of maintaining an anti-realist stance would be to adopt a purely syntactic criterion of confirmation, which most of us regard as a reduction ad absurdum. It follows from this, as I explain below, that the anti-realist has not defended the claim that we are in principle, or even probably, unable to confirm high-level theories – for this, I say, we must wait and see, and make our best estimates, as I do with Perrin’s evidence for the Atomic hypothesis. I do not endorse the kind of realist claim that says we generally have a right to believe our successful hypotheses that go beyond observables. Some successful hypotheses are better confirmed than others, and others are not at all. However, since the anti-realist’s general skeptical arguments fail, we also do not need such general claims in order to have a right to confidence in those particular hypotheses for which we do have good evidence. The interesting questions that remain are how we do and should make such judgments. The nub of the issue between the two camps in the epistemological realism-antirealism debate has always been whether we have a right to believe our scientific theories – in some sense that includes at least some of their claims about unobservable matters – on the basis of our evidence. Answers to this question depend on what our evidence actually is, and on the definition or criteria for evidential support. The former is a factual matter, which can be discussed by making substantive claims in individual cases. But philosophers in this dispute have often also made general assumptions about what our evidence is, such as the empiricist assumption that our evidence includes only observations. The latter, the criteria for what counts as evidential support, is an issue barely even noted in discussions of realism