Gilbert Harman has argued that the common-sense characterological psychology employed in virtue ethics is rooted not in unbiased observation of close acquaintances, but rather in the 'fundamental attribution error'. If this is right, then philosophers cannot rely on their intuitions for insight into characterological psychology, and it might even be that there is no such thing as character. This supports the idea, urged by John Doris and Stephen Stich, that we should rely exclusively on experimental psychology for our explanations of behaviour. The purported 'fundamental attribution error' cannot play the explanatory role required of it, however, and anyway there is no experimental evidence that we make such an error. It is true that trait-attribution often goes wrong, but this is best explained by a set of difficulties that beset the explanation of other people's behaviour, difficulties that become less acute the better we know the agent. This explanation allows that we can gain genuine insight into character on the basis of our intuitions, though claims about the actual distribution of particular traits and the correlations between them must be based on more objective data.