Although studies have shown that patients can distinguish probable adverse drug reactions (ADRs) from adverse clinical events (ACEs) caused by other factors, it is not known whether these attribution judgments add any independent validity to other accepted methods of identifying ADRs, such as physician assessments or epidemiologic data. Data from 2487 patients receiving fluoxetine and 815 receiving trazodone were used to see whether such information was redundant when added to standard statistical analysis directed toward detecting ADRs. Relative risk values for 14 trazodone or fluoxetine ADRs were selected because each was significantly identified by an innovative postmarketing surveillance system. In one analysis, all patient reports were used to compute relative risk; in the other, only reports attributed by patients to the study drug were included. Results indicate that taking into account patient attribution judgments results in a consistent, albeit modest, increase in the discriminatory power of this monitoring method.