Purpose. To review critically the evidence for three contemporary theories of delusions.

Methods. The theoretical approaches to delusions proposed by Frith and colleagues (‘theory of mind’ deficits), Garety and colleagues (multi‐factorial, but involving probabilistic reasoning biases) and Bentall and colleagues (attributional style and self‐discrepancies) are summarised. The findings of empirical papers directly relevant to these proposals are critically reviewed. These papers were identified by computerised literature searches (for the years 1987‐1997) and a hand search.

Results. The evidence does not unequivocally support any of the approaches as proposed. However, strong evidence is found to support modifications of Garety and colleagues and Bentall and colleagues theories. Studies have replicated a ‘jumping to conclusions’ data‐gathering bias and an externalising attributional bias in people with delusions. There is preliminary evidence for a ‘theory of mind’ deficit, as proposed by Frith, although possibly related to a more general reasoning bias. Evidence for an underlying discrepancy between ideal and actual self‐representations is weaker.

Conclusions. A multi‐factorial model of delusion formation and maintenance incorporating a data‐gathering bias and attributional style, together with other factors (e.g. perceptual processing, meta‐representation) is consistent with the current evidence. It is recommended that these findings be incorporated into cognitive therapy approaches. However, there are limitations to existing research. Future studies should incorporate longitudinal designs and first episode studies, and should not neglect the co‐morbidity of delusions, including affective processes, or the multi‐dimensional nature of delusions.