On the alleged impossibility of coherence

If coherence is to have justificatory status, as some analytical philosophers think it has, it must be truth-conducive, if perhaps only under certain specific conditions. This paper is a critical discussion of some recent arguments that seek to show that under no reasonable conditions can coherence be truth-conducive. More specifically, it considers Bovens and Hartmann’s and Olsson’s “impossibility results,” which attempt to show that coherence cannot possibly be a truth-conducive property. We point to various ways in which the advocates of a coherence theory of justification may attempt to divert the threat of these results. According to coherentism, a person is justified in holding a belief if, roughly speaking, the belief coheres well with most (or even all, depending on the particular version of coherentism) of her other beliefs, where the notion of coherence is typically circumscribed, at least for starters, in terms of beliefs’ hanging well together, or dovetailing with each other, or supporting each other, or in similar metaphorical terms. It is generally thought that a minimal requirement for the tenability of this position is that there be some positive correlation between coherence and truth. Coherence, in other words, should be truth-conducive in the sense that, even if perhaps only under certain specific conditions, one set of beliefs’ being more coherent than another entails its being also more probable than that other. In this paper we will call the claim that coherence is truth-conducive at least in this sense the “Truth-Conduciveness Claim” (TCC). TCC has faced a number of criticisms, the most recent of which are by Bovens and Hartmann (2003, 2005) and Olsson (2005a, 2005b), who purport to show that it is inconsistent with a number of straightforward and seemingly uncontentious background assumptions; following their usage, we will refer to the arguments they mount to this end as “impossibility results.” These impossibility results are explicitly meant to apply to all definitions of coherence, so that, if correct, it follows that TCCmust be abandoned however exactly the notion of coherence is spelled out. Olsson (2005a:4) even goes so far as to conclude that his result “puts us in an excellent position to explain why coherence theorists have been unsuccessful in defining their central notion: coherence is in a sense not definable.” Although these authors are not the first to argue against the truthconduciveness of coherence, they are to be commended for raising the discussion to a much higher level of precision. Among many other things, they have done pioneering work by pointing the way toward a formal treatment of the key traditional discussions