PURPOSE: Few studies have explored how relationships of perceived environment and physical activity vary across different activity domains and populations. This question was explored in five physical activity intervention trials funded by the National Institutes of Health Behavior Change Consortium. DESIGN: Observational. SETTINGS: San Francisco peninsula, California (N = 94); Eugene, Oregon (N = 122); Atlanta, Georgia (N = 256); Kingston, Rhode Island (N = 109); Memphis, Tennessee (N = 64). SUBJECTS: Ethnically diverse community adults ages 18 to 85 years. MEASURES: The Neighborhood Environment Walkability Scale and CHAMPS physical activity questionnaire. Response rate among those invited to complete these measures was 90%. RESULTS: Cross-sectional pooled signal detection analysis indicated that people who reported living in neighborhoods with more attractive scenery and ease of walking were more likely to meet national physical activity recommendations (67%) compared with those without these neighborhood attributes (36%; chi2 = 13.04, p = .0003). Within-site multiple regression identified two additional variables--seeing others when walking and encountering loose dogs that make it difficult to walk--as correlates across multiple sites and activity domains (i.e., minutes of weekly moderate or more vigorous activity, walking for errands, walking leisurely) (incremental R2 = 2.0-7.5; p < .05). Analyses of covariance suggested that traffic safety might be particularly important in facilitating or impeding physical activity in response to a formal intervention (for traffic-arm assignment interactions, F = 3.8-7.0, p < or = .05). CONCLUSIONS: Relationships between perceived environments and physical activity may differ depending upon population groups and activity domains and merit investigation by using stronger prospective designs.