Every philosopher should wish for commentators as thoughtful and rigorous as Mario G6mez-Torrente and Delia Graff. Their papers contribute significantly to our understanding of the fine structure of vagueness, but not by undermining its epistemicist interpretation. According to epistemicism, there is an unknowable truth of the matter in borderline cases. Either is bald or he is not, but we cannot know which. Why not? Suppose that in fact is bald; the argument is similar if he is not. Then we speak truly if we say Harry is But since the case is borderline, we might easily have uttered the same words even if our overall pattern of use had been slightly different so as slightly to shift the boundary of 'bald' in that context, making 'Harry is bald' express a different and false proposition. Thus even if we actually make the true judgement that is bald in this case, we do so on an insufficiently reliable basis to count as knowing that is bald. This is a conceptual analogue of more familiar kinds of ignorance, in which limits on our powers of perceptual discrimination prevent us from knowing the exact values of physical quantities by ordinary perception. Such limits on our knowledge can be expressed as margin for error principles (Williamson 1992, 1994: pp. 216-47). Gomez-Torrente points out that in perceptual cases we can discover the values of the quantities exactly (or at least far less inexactly) by scientific measurement; the unknowability is relative to limited cognitive means. He suggests that the epistemicist explanation establishes at most a similarly limited ignorance in borderline cases, thereby missing the depth of the phenomenon to be explained. The epistemicist explanation is as general as its underlying assumptions. In this context the relevant ones are: (a) the boundary of a vague term is unstable with respect to small changes in use because there is no natural line for it to follow; (b) because our use of such a term is not perfectly co-ordinated internally, a small change in one aspect of use (enough to shift the boundary a little) can easily be combined with preservation of another aspect