As indicators of the intensity and frequency of muscle activity, the rugosity and size of muscle insertion areas in human skeletal remains may provide a record of average work effort in past populations. In this paper a diet breadth model derived from optimal foraging theory was used as a heuristic means of exploring the utility of musculoskeletal stress markers (MSM) for determining subsistence labour intensity in prehistoric populations. The model was used to make predictions about relative muscle scar rugosity and size in three samples of pre-European-contact Khoisan skeletons from distinct biomes that vary in primary productivity and biomass structure (thus requiring different average subsistence work efforts). Equality of MSM scores between groups could only be rejected for the upper limb among males. In this case, the between-group differences were also in the direction predicted by the diet breadth model (forest>fynbos>savanna). The same pattern obtained for the lower limb in males, but the groups were not significantly different in median scores. Female samples did not differ significantly in mean MSM scores in either the upper or lower limb. Results suggest that ecological differences between biomes may have had a greater impact on the labour costs of male-foraged rather than female-foraged food items. Correlations between variables and analysis of additional measures of activity further suggest that MSM may reflect certain types of muscle activity (loading intensity) better than others (loading frequency and duration), which may account in part for the obtained results. These results invite further study of the ecological correlates of muscle scar rugosity and robusticity in the post-cranial skeleton of foraging peoples. © 1998 John Wiley & Sons, Ltd.