Objective: To compare the costs of 2 atypical drug therapies (olanzapine and risperidone) with one another and with a conventional antipsychotic (haloperidol) in the treatment of schizophrenia. Design and setting: The analysis is based on a simulation model with parameter values taken mainly from clinical trial data in patients with schizophrenia, and was conducted within a UK context. Results: The 3 therapies are approximately cost neutral over a 5-year period (olanzapine (L)35 701, risperidone (L)36 590 and haloperidol (L)36 653). There is evidence of greater efficacy with the atypical drugs [average percentage of 5 years with Brief Psychiatric Rating Scale (BPRS) scores <18: olanzapine 63.6%, risperidone 63.0% and haloperidol 52.2%]. The cost and efficacy differences between the 2 atypical drugs are too small to rank them in terms of cost effectiveness. Extensive sensitivity analysis does not change any of the main conclusions. Conclusions: Given evidence of efficacy gains to the atypical drugs, these represent cost- effective treatment options. Prospective data from nontrial treatment settings would help substantiate the model findings.