Data suggest that low-income and minority individuals are less physically active than the general population. We conducted a constituency-based physical activity promotion program, the Physical Activity for Risk Reduction (PARR) project, from 1988 to 1991, with residents of rental communities administered by the Housing Authority of the Birmingham District in Birmingham, Alabama. Data collected through focus groups and by survey of randomly selected residents' exercise practices, beliefs, barriers to and facilitators of physical activity were used to develop specific intervention programs that were implemented and evaluated in six intervention and two control communities through surveys and process evaluations. Trained community residents conducted data collection and interventions. Ninety-nine percent of respondents in the baseline survey were African Americans. Thirty-one percent reported no participation in any of 13 physical activities in the previous year. Significant differences (P < .05) in average attendance by community at group exercise sessions indicated substantial variability in effectiveness of the intervention, due at least in part to factors related to both the communities and the leaders. This variability appeared to explain patterns of change in preintervention and postintervention physical activity scores in intervention communities. Data and experiences from PARR will be applicable to the delivery of similar types of health-related programs in undeserved communities.