Good adherence is critical for antiretroviral therapy (ART) in sub-Saharan Africa. We report on the characteristics of medicine companions (MCs) chosen by Ugandan patients enrolling on ART, and on how MCs were chosen, and what roles they played. Baseline data on MCs of 1453 participants in a randomized controlled trial comparing facility and home-based delivery of ART in Jinja, Uganda were analyzed. Textual data on experience with MCs were collected through in-depth interviews among a subsample of 40 trial participants equally divided by sex and trial arm. Significantly more women (71%) than men (29%) were recruited. The majority (75%) of women participants were either widowed (51%) or separated or divorced (24%), whereas most of the men (66%) were married. Women were most likely to choose a child as their MC while men were most likely to choose their spouse; 41% of women chose an MC under 21 compared with only 14% of men. Only 31% of married women chose their husband, compared with 66% of married men who chose their wife. Qualitative interviews suggested MCs proved useful for reminding and other supportive tasks in the first three months but were generally less essential by six months and beyond. Convenience, reliability, and trust were key considerations in choosing an MC. Children provided the only alternative for many unmarried women, but even some married women felt children made more reliable MCs than husbands. Participants who had disclosed their serostatus usually received drug-taking reminders from multiple household members. One participant in the qualitative sample with poor family relations delayed starting treatment due to unwillingness to identify an MC. MCs were generally welcome and useful in supporting early adherence. However, disclosure to an MC should not be a condition of obtaining treatment.