During the study of benthic communities, the choice of a sampling programme is often not given enough attention. Common practice consists of estimating the number of replicates through standard error or coefficient of variation of individual species abundance, but when the object of the study is the whole community, this method is not relevant. We considered the effect of a different number of replicates on the result of community differentiation using different approaches, in particular cluster analysis. Previous studies have shown that five samples did not guarantee a reliable assessment of heterogeneity of benthic biocenosis for long-term monitoring at three control sites in the waters of the Keret archipelago (the White Sea). In 2008 the number of replicates per station was increased to 10. It was shown that the communities that were studied should be identified as the typical biocenosis in the White Sea littoral zone. We estimated the number of replicates that are sufficient to achieve biocenosis differe