In multivariate analyses of the effects of both natural and anthropogenic environmental variability on community composition, many species are interchangeable in the way that they characterise the samples, giving rise to the concept of structural redundancy in community composition. Here, we develop a method of quantifying the extent of this redundancy by extracting a series of subsets of species, the multivariate response pattern of each of which closely matches that for the whole community. Structural redundancy is then reflected in the number of such subsets, which we term "response units", that can be extracted without replacement. We have applied this technique to the effects of the Amoco-Cadiz oil-spill on marine macrobenthos in the Bay of Morlaix, France, and to the natural interannual variability of macrobenthos at two stations off the coast of Northumberland, England. Structural redundancy is shown to be remarkably high, with the number and sizes of subsets being comparable in all three examples. Taxonomic/functional groupings of species within the differing response units change in abundance in the same way over time. The response units are shown to possess a wide taxonomic spread and, using two different types of randomisation test, demonstrated to have a taxonomically and functionally coherent structure. The level of structural redundancy may therefore be an indirect measure of the resilience or compensation potential within an assemblage.
Keywords: Compensation potential; Key words Multivariate analysis; Marine macrobenthos; Oil pollution; Randomisation test.