A data set of Attributional Style Questionnaire responses collected by Kaney & Bentall (1989) from persecutory deluded, normal and depressed subjects was added to a similar data set collected by the first author. The attributions made for hypothetical positive and negative events by the combined pool of subjects were then blind rated for internality by five independent judges. Whereas the internality ratings made by the subjects for their own attributions showed evidence of an exaggerated self-serving bias in the case of the deluded subjects, and an absence of such a bias in the case of the depressed subjects, the independent ratings showed no such group differences. These findings suggest that the differences in internality for positive and negative events observed on the ASQ between deluded, normal and depressed subjects may not reflect differences in the types of causal statements made by these subjects but differences in their attributions about their attributions. The relevance of this observation for attribution theory in general and attributional accounts of psychopathology in particular are discussed.