A number of recent accounts of our first-person knowledge of our attitudes give a central role to transparency - our capacity to answer the question of whether we have an attitude by answering the question of whether to have it. In this paper I raise a problem for such accounts, by showing that there are clear cases of first-person knowledge of attitudes which are not transparent.

I argue two things. One is that conceptual considerations about the nature and identification of psychological systems suggest that these recent empirical findings, being based on manipulated conditions, are not relevant to the issue of what emotions are and thus do not underwrite noncognitivism. The other is that these same considerations lend support to the idea that paradigm emotions, including the purported noncognitive basic emotions, are in fact cognitive. Central to these claims is the concept of proper function, particularly as (...) laid out by Ruth Millikan, which has received insufficient attention in work on the nature of emotion. (shrink)