This study was designed to identify correlates of change in walking for exercise. Respondents to a random sample mailed exercise survey were asked to complete a follow-up questionnaire 24 months after the first. Responses were obtained from 1,739 adults, reflecting an 86.6% return rate. Nonrespondents did not differ from respondents for baseline level of walking for exercise. Respondents overrepresented Caucasian and middle to high education (or income) adults. Residualized measures of change in walking for exercise served as the dependent variable, adjusted for baseline walking. Similarly, key independent variables, subject to change over time, were included after residualization. At both baseline and 24 months follow-up, adults walked for exercise an average of just under 1 hour per week, and 23.5% of the initially active adults ceased walking for exercise at 2 years. Multiple regression analyses explained more variance in walking for exercise among women than men. Change in dynamic variables such as friend's support and vigorous physical activity accounted for most of the explained variance. Results suggest that dynamic variables may need to be sustained to maintain walking for exercise. The limited explained variance suggests that more precise measures and additional determinants of walking for exercise need to be identified.