Accurately assessing nonadherence is a necessary first step toward improving adherence to highly active antiretroviral therapy (HAART). Patient self report is the most practical method for assessing adherence in clinical settings, but may produce unreliable and invalid results unless optimally performed. A computer-assisted, self-administered interview (CASI) may improve the disclosure of medication nonadherence by providing a neutral and seemingly private interview. One hundred and ten patients completed a computer program which assessed their understanding of and adherence to HIV medications and produced a report for their providers. Eleven providers of these patients completed a questionnaire describing their patients' medication regimens and estimating adherence. Patients completed a written exit survey and providers completed an exit interview to assess the acceptability of our CASI-based assessment. More than half of patients (54%) made at least one error in reporting their medication regimen. Providers tended to overestimate their patients' adherence and correctly classified only 24% of nonadherent patients at the 80% adherence level. Computerized HIV medication adherence assessment is feasible and acceptable to patients and providers. Clinical tools that can accurately and efficiently detect important medication errors and nonadherence, and alert providers to these problems, will help ensure the health of HIV-seropositive patients.