Two measurements of adherence, patient self-report and electronic measurement by the Medication Event Monitoring System (MEMS), were compared in a 3-month adherence study of 44 HIV-infected patients who had been placed on regimens that included protease inhibitors (PIs). The dose percentage and degree of clinically significant dosing time fluctuation were calculated monthly. The mean dose percentage by self-report versus MEMS was 97.5% versus 90.3% during month 1 of adherence monitoring, 96.5 versus 90.1% during month 2, and 98.4% versus 92.8% during month 3. Thirty-two percent of patients taking PIs and 21% of patients taking nucleoside analogues demonstrated clinically significant dosing time fluctuation. Our data confirm that self-reports of adherence overestimate true adherence behavior, and patients' self-reports of dosing times may not accurately reflect their deviation from those times.