Data from samples of the macrobenthic Abra alba– Pectinaria koreni community of the eastern Bay of Seine (English Channel) collected in winter 1986 are analysed to illustrate the advantages of a novel method of multivariate analysis of spatial patterns described by Thioulouse et al. (Environ. Ecol. Stat., 2 (1995) 1–14), consisting of local and global approaches. Multivariate ordination procedures are applied that take spatial components into account explicitly through the construction of a neighbourhood graph between closely placed sampling sites, which is then used to weight the data. The result is a decomposition of spatial structure on local and global scales. This method is for the first time applied to macrobenthic data of this region. It shows the underlying importance of spatial scaling in analysis and proves to offer more information than classical ordination methods such as correspondence analysis, which may confuse the two different spatial scales. Global analysis is proposed as a powerful tool to define species assemblages and local analysis as an additional instrument to define partitions resulting from biological interactions. Additionally, this method appears capable of incorporating rare species (which influence classical analyses, often resulting in their elimination from datasets) by minimising their effects on the global scale and conversely maximising them on the local scale. This analysis demonstrates the importance of explicitly incorporating spatial information into the detection and interpretation of patterns in a macrobenthic community.