Major obstacles to conducting process research with parent education programs exist; new directions for studying how programs interact with participants are needed. In particular, intervention research is time consuming and expensive. The small number of subjects involved are often at risk and under considerable social stress. Self-selection and nonrandom attrition erode the sample. Comparing results from different programs is difficult because of (1) lack of clarity about program goals and effects to be measured, (2) wide within-program variations in program implementation, and (3) inadequate process and outcome measures. An ecological, experimental study of educational support groups for adolescent mothers illustrates these problems, and suggests an approach to learning from them. To explore the role of supportive education groups in the development of maternal and infant competence, 48 pregnant, low-income adolescents were interviewed and assigned to parenting group or control conditions. At 1- and 8-months postpartum, mothers were interviewed, mothers and infants were observed, and infants' developmental status was assessed. Results showed different patterns of participation were based on needs and resources of mothers. Future research should focus on attrition as a nonrandom variable, while studying a broader range of programs, directly comparing alternative forms of programs, and utilizing theoretical concepts of how context affects development. (RH)