Eight national surveys conducted in the United States and Canada between 1972 and 1983 are reviewed for evidence of leisure-time physical activity patterns in the population. The authors' major conclusion is that it is difficult to make reliable generalizations when definitions of exercise used in the surveys vary so widely. Nevertheless, the young and persons of relatively high socioeconomic status are definitely more active than average in their leisure time; this is probably also true of westerners and suburbanites. Males and females are about equally likely to be involved in conditioning activities, but males are more likely to participate in vigorous exercise and sport. It appears likely that exercise prevalence has increased in recent years, and a maximum of 20 percent of the population exercises at a level frequently recommended for cardiovascular benefit. Major areas of uncertainty and ignorance remain, and the authors identify 15 such areas. Secondary analysis is recommended to help resolve several questions currently impeding a complete description of the exercise patterns of the population. For future surveys, five recommendations are offered on definitions and essential data items. Existing time series studies are generally inadequate. The authors recommend that detailed surveys of exercise patterns be conducted every 5 years to supplement the continual monitoring that is also essential to detect shifts in this important health behavior.