How women's social roles such as parenthood, marital status and employment status affect their exercise participation, is important for the rational planning of suitable and effective exercise programs for women in different social contexts. A secondary analysis of a large data-base of urban women aged 20-49 (n = 5939) was carried out to determine if social roles, their characteristics and their combinations were related to women's exercise participation. Parenthood and marital status were significantly related to whether women exercise, and among those exercising, all three roles were significantly related to the amount of exercise. Statistical modelling, including all three roles and their interactions, controlling for age and education, showed that with respect to whether women exercise, parenthood affected exercise participation most and that its effect was dependent on age. Of the role characteristics only the number of children and overall satisfaction with daily activities improved the prediction of exercising. With respect to how much women exercise, the effect of parenthood interacted with marital status. Married parents were least likely to exercise very much. Of the role characteristics only overall satisfaction with daily activities improved the prediction of the amount of exercise.