Here’s a follow-up on my reference to Keith’s split-finger fastball (you’ll have to have been reading the comments diligently to get that reference…). I bet most you have thought, like I have, that Keith’s theory is a variant of a sensitivity theory of knowledge, one that makes central use of the subjunctive conditional that goes by that name. I’ve been reading the Blackwell volume on Sosa, having agreed to review it for NDPR, and discovered an important fact here. Keith claims that his account of knowledge is better characterized in terms of strong enough true belief, where strength is measured by how remote are the possibilities in which one goes wrong about p. Of course how strong is strong enough depends on the context, but the interesting point to note is that in spite of the amount of space devoted by Keith to subjunctive conditional accounts of knowledge, his own view is significantly different from them. The primary difference is that the strength account pays attention both to worlds in which one believes p and p is false and to worlds in which p is true and one believes ~p. On the subjunctive conditionals account, only the former worlds are relevant (and the same is true of the safety view Sosa once endorsed).