BACKGROUND: Self-report is the most common and feasible method for assessing patient adherence to medication, but can be prone to recall bias and social desirability bias. Most studies assessing adherence to artemisinin-based combination therapies (ACTs) have relied on self-report. In this study, we use a novel customised electronic monitoring device--termed smart blister packs--to examine the validity of self-reported adherence to artemether-lumefantrine (AL) in southern Tanzania. METHODS: Smart blister packs were designed to look identical to locally available AL blister packs and to record the date and time each tablet was removed from packaging. Patients obtaining AL at randomly selected health facilities and drug stores were followed up at home three days later and interviewed about each dose of AL taken. Blister packs were requested for pill count and extraction of smart blister pack data. RESULTS: Data on adherence from both self-report verified by pill count and smart blister packs were available for 696 of 1,204 patients. There was no difference between methods in the proportion of patients assessed to have completed treatment (64% and 67%, respectively). However, the percentage taking the correct number of pills for each dose at the correct times (timely completion) was higher by self-report than smart blister packs (37% vs. 24%; p<0.0001). By smart blister packs, 64% of patients completing treatment did not take the correct number of pills per dose or did not take each dose at the correct time interval. CONCLUSION: Smart blister packs resulted in lower estimates of timely completion of AL and may be less prone to recall and social desirability bias. They may be useful when data on patterns of adherence are desirable to evaluate treatment outcomes. Improved methods of collecting self-reported data are needed to minimise bias and maximise comparability between studies.