Motivating the Relevant Alternatives Approach

I Introduction

Knowing that p requires being able to 'rule out' the relevant not-p alternatives. Such is the core claim of the Relevant Alternatives (RA) theorist. Of course, to endorse the core claim is not to have a complete and satisfactory account of knowing: any RA theorist has some explaining to do. Most obviously, anyone who endorses the core claim must ultimately provide an account of 'ruling out' and 'relevance'. And some who've been critical of the whole RA approach have done so because of a scepticism about the prospects of cashing these notions out in a satisfactory way.

But it's not the mere fact that the RA theorist needs an account of 'ruling out' and 'relevance' that has tended to lead people to regard the RA approach with suspicion. In itself, this simply means that the RA theorist has some further work to do; and what theorist doesn't? No; the principal source of scepticism regarding the ability of the RA theorist to come up with a complete and satisfactory account of knowing stems, rather, from an unhappiness with the specific elaborations of the core RA claim that various theorists have offered; for these elaborations have typically involved some rather controversial claims and/or assumptions, and it is against these claims/assumptions that the bulk of the criticism of the whole RA approach has been directed.

For example, the majority of RA theorists (e.g., Cohen, 1988, 1999; Heller, 1999a, 1999b; Lewis, 1979, 1996; and Stine, 1976) are also contextualists — their preferred standard(s) of relevance are importantly context-sensitive. Indeed, it is sometimes implied that RA theories are inherently contextualistic. Thus, for instance, Vogel introduces RA as 'a systematic articulation' of the idea that 'the requirements for knowledge are limited and context-dependent' (1999, 155); and Sosa writes, '[c]ontextualism has gained center stage in epistemology mainly through its way with the skeptic, from the early days of 'relevant alternatives' to important recent publications' (2000, 1). But, as contextualism is controversial to say the least,¹ the RA approach, construed as an instance thereof, is fated to be so as well.

In the same way, the RA approach is often associated with the denial of the closure principle for knowledge, whereby if S know that p and that p entails q, and as a result forms the belief that q, then S knows that q.² For instance, Dretske claims that if we adopt RA 'we must live with the failure of closure' (1991, 190), and Pritchard that 'the RA approach to knowledge entails the denial of the closure principle' (2000, 275). Whereas most epistemologists take the closure principle to be obviously correct. And not without reason: denying closure opens up the possibility of embracing the truth of 'abominable conjunctions' such as, 'S knows [p] that he has hands, but not [q] that he's not a handless BIV' (DeRose 1995, 27-9). To the extent that such results strike us as unacceptable, we will regard closure as 'axiomatic' (Cohen 1999, 68) and any view that entails its rejection as inherently flawed.

However, as others have pointed out, it is entirely possible to distinguish, in principle, between the core RA claim and these more controversial, and perhaps strictly-speaking-optional, theoretical add-ons. Thus, DeRose (1992; 1995, 13; 1999, 193) notes that RA theorists don't have to be contextualists; and Cohen (1988, 1991), Stine (1976), and Williams (1996) argue — contra, e.g., Dretske (1970), Heller (1999a), and Pritchard (2000) — that the RA approach per se does not support the abandonment of closure.

But if RA is, in principle, separable from those elements which people have found most problematic about typical RA theories, why do those elements dominate discussions of the RA approach? And why do they so persistently form part of most actual RA theories?

Well, if it's true and unsurprising that the plausibility of a particular development...