BACKGROUND A possible neuropsychological basis for poor insight in psychosis has been proposed. Consistent supporting evidence for this is lacking. METHOD Seventy-four consecutive acutely psychotic in-patients who were recruited for a randomised controlled trial of compliance therapy were given a battery of neuropsychological tests, along with a comprehensive clinical assessment, before and after the intervention. RESULTS Performance on neuropsychological tests improved during the patients' admission and treatment in hospital, as did symptoms and levels of insight. Cognitive function showed no relationship to insight and compliance initially, and very little after the intervention. Factors related to insight and compliance prior to discharge included: diagnosis, attitudes to medication, side-effects, being a detained patient, and whether or not compliance therapy was given. CONCLUSIONS Clinical variables and attitudes to treatment appear to be more relevant to compliance and the development of insight in acute psychosis than neuropsychological impairment.