This paper examines the relationship between physical fitness and activity among elderly males in the traditional rural community of Chetbesi, Nepal. It takes advantage of the unique character of the Hindu caste system to implement a quasiexperimental research design that approximates random assignment to high and low activity levels. The members of the Sarki caste have lower heart rates and systolic blood pressure, relative to other castes, at each of three submaximal workloads and during recovery from bicycle ergometer exercise. Direct observation and physiologic monitoring show that the Sarkis engage in more frequent and extended periods of heavy labor. Thus intracohort variation in physical fitness and activity patterns among the Chetbesi elderly is a function of birth into a socially defined group rather than of self-selection. This pattern of differential fitness may typify the process of aging in many stratified traditional and modernizing societies where socially delimited segments of the population perform the bulk of the hard work. Intrapopulation differences aside, comparison of Sarkis and non-Sarkis with other samples reveals that both lie within the reported range of variation. The rural, unmechanized, agricultural lifestyle and mountain environment of Chetbesi do not result in exceptional fitness for residents. The Chetbesi data suggest that the popular notion that aging is less debilitating in traditional agrarian societies located in rugged mountain terrains may be a myth. The demonstration of the influence of social forces on physical fitness suggests that future research might concentrate profitably on identifying social structures that produce high levels of physical fitness.