We examined attrition from an HIV/STD group counselling intervention in two ways: quantitative analyses of the entire sample (n=287) and structured interviews of a subset of 30 women with low attendance. In the interviews, the most common reasons for low attendance were time conflicts and enrolling primarily to obtain the monetary incentives given for completing research questionnaires. Latent class statistical analysis of the full sample identified two subgroups that differed from each other in the number of psychosocial problems recently experienced. Relative to the 'non-distressed' class, the 'distressed' class members had higher probabilities of psychological distress, low incomes, heavy substance use, sex for trade, relationship violence, and unstable housing. This group had higher HIV/STD risk, but lower intervention attendance. Members also had less education and less knowledge about HIV/STD. A higher proportion of the distressed group was African-American. Study findings suggest that in interventions for women at HIV/STD risk, it is not necessarily enough to ensure cultural relevance and to provide food, childcare, and transportation. To increase retention, interventionists should consider (1) the use of strategies to support attendance (such as monetary incentives and attention to group process factors) and (2) intervention formats that are brief, matched to participants' stage of change, population-specific, and/or maximally accessible.