Objective: We aimed to elucidate why HIV-positive African Americans (AAs) demonstrate poorer antiretroviral therapy (ART) adherence than HIV-positive Caucasians. We tested whether intellectual/reading abilities and health beliefs mediate the association of AA ethnicity with suboptimal ART adherence. Method: HIV-positive adult participants (238 AA, 53 Caucasian) were recruited. Mean education levels were 13.08 years (SD = 1.96) for AAs and 13.58 (SD = 2.39) for Caucasians. Exclusion criteria included history of traumatic brain injury with >60-minute loss of consciousness and active seizure and psychotic disorders, other non-HIV-related neurological diseases, and central nervous system opportunistic infections. We measured intellectual/reading abilities with the American National Adult Reading Test (AMNART), health beliefs with the perceived treatment utility subscale of the Adherence Determinants Questionnaire, and 4-week ART adherence with the Medication Event Monitoring System. Results: Path models demonstrated excellent goodness-of-fit statistics (e.g., χ2/df = 0.54) and revealed that the influence of AA ethnicity on suboptimal ART adherence was mediated by poorer AMNART performance. AA ethnicity was directly related to AMNART errors (β = 0.29), which in turn had a direct (β = −0.14) and indirect (β = −0.21) effect on suboptimal adherence via perceived treatment utility. Conclusion(s): Results suggest that poorer premorbid intellectual/reading abilities play a key mediating role in suboptimal HIV medication adherence in AAs, and its effect is partially mediated by perceived treatment utility. These variables can help identify AAs at risk for suboptimal adherence, and interventions can be tailored to increase understanding of medication regimens and treatment utility.1