In this paper I will discuss Michael Williams’s inferential contextualism – a position that must be carefully distinguished from the currently more fashionable attributer contextualism. I will argue that Williams’s contextualism is not stable, though it avoids some of the shortcomings of simple inferential contextualism. In particular, his criticism of epistemological realism cannot be supported on the basis of his own account. I will also argue that we need not give up epistemological realism in order to provide a successful diagnosis of scepticism.