The social reasoning of patients suffering from persecutory delusions and matched groups of depressed and normal controls was investigated using the framework of Kelley's (1967) theory of social attribution. Subjects were required to choose between person, circumstances and stimulus attributions for a series of social vignettes describing interactions between two persons. The vignettes varied in terms of the distinctiveness, consistency and consensus information supplied, as well as in respect of whether the actions described were positively or negatively valued. It was found that the deluded patients made excessive person attributions for negative events. It was also found that the deluded patients were excessively certain about their judgements compared to the depressed controls. Normal subjects tended to rate their certainty in their own judgements midway between the deluded and depressed patients.