To determine the validity of the aspirin/Reye's syndrome association, we developed an epidemiologic investigation to assess the effects of five potential sources of bias. A case-control study incorporated procedures to avoid temporal precedence and susceptibility bias. These included classifying cases as having monophasic or biphasic patterns of illness and matching for severity of symptoms at zero-time. To evaluate the effect of a potential recall bias, an "alternatecondition" control group was enrolled. A medical record review study was conducted to assess the potential for diagnostic bias, and a blanket surveillance of all hospitals in a region was conducted to evaluate reporting bias. Twenty-four case subjects and 48 matched controls were enrolled. Eighty-eight percent of case subjects and only 17% of controls had received aspirin prior to the onset of Reye's syndrome (matched odds ratio, 35; 95% confidence interval, 4.2 to 288). Further analyses demonstrated that the association could not be attributed to the five potential sources of bias.