BACKGROUND: The devastating impact of AIDS in the world especially in sub-Saharan Africa has led to an unprecedented global effort to ensure access to antiretroviral (ARV) drugs. Given that medication-taking behavior can immensely affect an individual's response; ART adherence is now widely recognized as an 'Achilles heel' for the successful outcome. The present study was undertaken to investigate the rate and predictors of adherence to antiretroviral therapy among HIV-infected persons in southwest Ethiopia. METHODS: The study was conducted in the antiretroviral therapy unit of Jimma University Specialized Hospital. A prospective study was undertaken on a total of 400 HIV infected person. Data were collected using a pre-tested interviewer-administered structured questionnaire at first month (M0) and third month (M3) follow up visits. RESULTS: A total of 400 and 383 patients at baseline (M0) and at follow up visit (M3) respectively were interviewed. Self-reported dose adherence in the study area was 94.3%. The rate considering the combined indicator (dose, time and food) was 75.7%. Within a three month follow up period, dose adherence decreased by 2% and overall adherence rate decreased by more than 3%. Adherence was common in those patients who have a social support (OR, 1.82, 95%CI, 1.04, 3.21). Patients who were not depressed were two times more likely to be adherent than those who were depressed (OR, 2.13, 95%CI, 1.18, 3.81). However, at the follow up visit, social support (OR, 2.42, 95%CI, 1.29, 4.55) and the use of memory aids (OR, 3.29, 95%CI, 1.44, 7.51) were found to be independent predictors of adherence. The principal reasons reported for skipping doses in this study were simply forgetting, feeling sick or ill, being busy and running out of medication in more than 75% of the cases. CONCLUSION: The self reported adherence rate was high in the study area. The study showed that adherence is a dynamic process which changes overtime and cannot reliably be predicted by a few patient characteristics that are assumed to vary with time. Adherence is a process, not a single event, and adherence support should be integrated into regular clinical follow up.