Purpose of the study: To assess the recruitment, adherence, and retention of urban elderly, predominantly African Americans to a falls reduction exercise program. Design and methods: The randomized controlled trial was designed as an intervention development pilot study. The goal was to develop a culturally sensitive intervention for elderly persons who suffered a fall and visited an emergency department (ED). Participants were taught exercises during 4 on-site group classes and encouraged to continue exercising at home for 12 weeks and attend additional on-site monthly classes. The protocol included a specifically designed intervention for increasing retention through trained community interventionists drawn from the participants' neighborhoods. Results: The screening of 1,521 ED records after falling yielded the recruitment of 204 patients aged 65 years and older. Half were randomized into the falls prevention program. Of the 102 people in the intervention group, 92 completed the final 6-month assessment, 68 attended all on-site sessions, but only 1 reported exercising at home all 12 weeks. Those who lived alone were more likely (p = 0.03) and those with symptoms of depression were less likely (p = 0.05) to attend all on-site exercise classes. The final recruitment rate was estimated as 31.8%. The final retention rates were 90.2% and 87.3% for the intervention and control groups, respectively. Implications: Recruitment of frail elderly African American patients is resource intensive. Adherence to the on-site exercise classes was better than to the home-based component of the program. These findings have implications for the design of future community-based exercise programs and trials.