Five methods for assessing physical activity were evaluated in a population of 255 white, postmenopausal women participating in a randomized trial on the effects of walking on bone loss. Methods were the Paffenbarger survey, a modified Paffenbarger, the large-scale integrated activity monitor (LSI), caloric intake, and the Baecke survey. Significant increase in physical activity was observed in the intervention group. Activity measures were weakly related to each other. Results of factor analyses suggest that factor I reflected voluntary leisure-time pursuits (such as walking) and factor II, activities of daily living. Participation in sports was not a significant contributor to overall activity of the women. Research on physical activity must define the particular dimension of activity measured. Several types of instruments should be used because different instruments reflect different activity patterns which, in turn, may be differentially related to disease.