Beta diversity can be defined as the variability in species composition among sampling units for a given area. We propose that it can be measured as the average dissimilarity from individual observation units to their group centroid in multivariate space, using an appropriate dissimilarity measure. Differences in beta diversity among different areas or groups of samples can be tested using this approach. The choice of transformation and dissimilarity measure has important consequences for interpreting results. For kelp holdfast assemblages from New Zealand, variation in species composition was greater in smaller holdfasts, while variation in relative abundances was greater in larger holdasts. Variation in community structure of Norwegian continental shelf macrobenthic fauna increased with increases in environmental heterogeneity, regardless of the measure used. We propose a new dissimilarity measure which allows the relative weight placed on changes in composition vs. abundance to be specified explicitly.