Our objective was to compare antiretroviral adherence questions to better understand concordance between measures. Among 53 methadone maintained HIV-infected drug users, we compared five measures, including two single item measures using qualitative Likert-type responses, one measure of percent adherence, one visual analog scale, and one multi-item measure that averaged responses across antiretrovirals. Responses were termed inconsistent if respondents endorsed the highest adherence level on at least one measure but middle levels on others. We examined ceiling effects, concordance, and correlations with VL. Response distributions differed markedly between measures. A ceiling effect was less pronounced for the single-item measures than for the measure that averaged responses for each antiretroviral: the proportion with 100% adherence varied from 22% (single item measure) to 58% (multi-item measure). Overall agreement between measures ranged from fair to good; 49% of participants had inconsistent responses. Though responses correlated with VL, single-item measures had higher correlations. Future studies should compare single-item questions to objective measures.