High level adherence to antiretroviral therapy (ART) is required to achieve and maintain suppression of HIV replication. Although directly observed therapy (DOT) has been suggested as an intervention to improve adherence, there is a paucity of data describing the attitudes and beliefs regarding DOT for ART among HIV-infected individuals. This study was designed to evaluate the acceptability and psychometric properties of a survey instrument for use in assessing barriers and facilitators of adherence to ART DOT in prison. From July 1, 1999 to April 1, 2000, we piloted an interviewer-administered questionnaire to assess health beliefs and attitudes regarding HIV treatment among 65 HIV-infected prison inmates receiving one or more of their antiretrovirals via directly observed therapy (DOT). The first 24 participants were administered the questionnaire to determine the feasibility of surveying prisoners in a correctional setting. There were no adherence data collected on these participants. The remaining 41 participants had their adherence measured in addition to receiving the questionnaire. Thirty-one were included in the final analysis because 10 did not complete the study. Multiple antiretroviral adherence measures (electronic device medication monitoring [eDEM] caps, medication administration records [MARs], and pill counts) were assessed among a subset of the participants (n = 31) and correlated to the instrument response items. The median internal consistency reliability coefficient for the multi-item scales was 0.79. The strongest correlation between inmates' beliefs and their adherence was between "positive beliefs about protease inhibitors" and the MAR adherence measure (r = 0.72; p < 0.001). This study provides preliminary support for the psychometric properties of the survey in this correctional setting.