This paper presents a method for studying between-group differences in physical activity patterns through the analysis of musculo-skeletal stress markers (MSM). The specific aim was to develop a method that could overcome the problems related to confounding variables such as body size by shifting the analytical focus from the comparison of differences in intensity to the comparison of differences in kinds of activity. Instead of testing whether there are differences in the measures of central tendency of MSM scores between two groups, we proposed looking for differences in MSM covariation patterns. Formally, this is achieved through the statistical comparison of group-specific MSM intercorrelation matrices. The null hypothesis is that the matrices are equal, and the statistical significance of the test statistic is obtained with a permutation test. In this way, the problem of confounding variables such as body size is mostly avoided because the procedure is based on the comparison of group-specific sets of correlation coefficients which are scale-free. The method was employed in a case study to investigate the differences in activity patterns between men and women from the Early Bronze Age population buried at the necropolis in Mokrin, Serbia. The null hypothesis of no difference between the male and female MSM correlation matrices was rejected (p = 0.0135). After the statistical significance of the difference was established, further investigation of male and female activity patterns was undertaken by means of principal component analysis (PCA) with varimax rotation. PCA results suggest that covariation between MSMs is stronger in the male sample. Success in demonstrating differences in activity patterns between sexes implies that the proposed method can be used to test for differences in physical activity between groups of individuals defined by criteria other than sex. Copyright © 2011 John Wiley & Sons, Ltd.