How reliable are reconstructions of body mass and joint function based on articular surface areas? While the dynamic relationship between mechanical loading and cross-sectional geometry in long bones is well-established, the effect of loading on the subchondral articular surface area of epiphyses (hereafter, articular surface area, or ASA) has not been experimentally tested. The degree to which ASA can change in size and shape is important, because articular dimensions are frequently used to estimate body mass and positional behavior in fossil species. This study tests the hypothesis that mechanical loading influences ASA by comparing epiphyses of exercised and sedentary sheep from three age categories: juvenile, subadult, and adult (n = 44). ASA was measured on latex molds of subchondral articular surfaces of 10 epiphyses from each sheep. Areas were standardized by body mass, and compared to diaphyseal cross-sectional geometrical data. Nonparametric statistical comparisons of exercised and control individuals found no increases in ASA in response to mechanical loading in any age group. In contrast, significant differences in diaphyseal cross-sectional geometry were detected between exercised and control groups, but mostly in juveniles. The conservatism of ASA supports the hypothesis that ASA is ontogenetically constrained, and related to locomotor behavior at the species level and to body mass at the individual level, while variations in diaphyseal cross-sectional geometry are more appropriate proxies for individual variations in activity level. Am J Phys Anthropol 116:266–277, 2001. © 2001 Wiley-Liss, Inc.