Our purpose was to test a four-domain predictive model of recurrent falls developed for this study. In this model, limited mobility is considered a necessary but not sufficient element in risk of recurrent falls. Three other domains, attitudinal, social, and environmental, are proposed to influence fall risk only in persons with impaired mobility. Prospective cohort study. Veterans Affairs Ambulatory Care Service serving rural and urban central North Carolina. Male Veterans aged 70 or older (n = 306) were monitored prospectively for falls. At baseline, 159 screened as high-risk mobility status and 147 as low-risk mobility status. The primary outcome was recurrent falls. The mobility screen used for risk assignment defined immobile as unable to sit without support for 60 seconds, mobile and stable as meeting criteria for normal ambulation and stair climbing, and mobile but unstable as those who met neither of the above criteria. The high-risk subjects were further assessed in their homes for mobility in more detail, attitude toward risk, social supports, and environmental status. Other data included demographics, functional status, diagnoses, and medications. Recurrent falls occurred in 37 (23.3%) high-risk subjects and seven (4.8%) low-risk subjects (relative risk = 4.8, confidence interval 2.5 to 9.6, P < 0.001). Within the high-risk group, the probability of recurrent falls was significantly affected by degree of impaired mobility (P < 0.001), attitude toward risk (P = 0.005), and environment score (P = 0.03). A simple mobility screen can identify elders at increased risk for recurrent falls. Risk within this group is further modified by risk-taking behavior and environment.