This study was designed to identify correlates of walking for exercise in adults. Over 2,050 (43.4% response rate) responses to a mailed questionnaire were analyzed. Possible correlates of walking were based on learning theory and previous empirical observations. Respondents averaged less than 1 hr of walking for exercise per week. Women and older adults (greater than or equal to 50 years) reported significantly (P less than 0.05) greater walking than men or younger respondents. Multiple regression analyses were conducted for selected subgroups of respondents. Analyses were conducted on subjects who reported no regular vigorous exercise. Multiple correlation coefficients ranged from 0.32 to 0.48, and most reached significance (P less than 0.05). For the most sedentary subgroups, self-efficacy, family and friend support, and consumption of a heart-healthy diet were repeatedly associated with walking for exercise. It was concluded that an economically secure and well-educated Caucasian sample performs an inadequate amount of walking to ensure benefits such as prevention of cardiovascular disease. Longitudinal analyses are required to confirm the influence of social learning variables. Tentatively, interventions that increase family and friends' support for walking and that enhance perceived self-efficacy should be developed.