The purpose of this study was to determine the conditions in which self-reports provide an accurate assessment of adherence to HIV antiretrovirals. In a sample of 58 participants, self-reported and electronically monitored adherence to antiretroviral therapy were compared over a three-day period. Of the 16 who reported missed doses, only six (38%) accurately reported the number of doses they took, although the electronic monitoring data confirmed that all but one (94%) did in fact miss at least one dose. In contrast, 25 (60%) of the 42 participants who reported no missed doses were accurate. Nearly all (96%) participants who actually did take all their doses accurately reported their adherence compared to only 20% of those who missed at least one dose (P<0.001). Cognitive functioning was marginally associated with self-report accuracy, but all other baseline factors were not associated with accuracy. Our findings highlight the need for more effective self-report methodology and a better understanding of the circumstances in which self-reports are valid measures of adherence.