Gradients in the distribution of riverine benthos

SUMMARY. 1. Numerous hypotheses regarding the controls on species distributions in streams depend on longitudinal analyses of collections of stream organisms. Techniques for the detection of trends and for the detection of station groupings among biological collections have been widely reported, but rarely compared with each other. We have conducted an analysis of ten or eleven samples from each of thirteen stations along a stream abiotic gradient, and have compared several techniques. 2. An analysis of longitudinal distribution must first assess the existence of a gradient in biotic composition. Random skewers on species proportions were the most sensitive in demonstrating such a gradient. Q and M statistics were most useful in further describing the trend due to their additive properties; i.e. the importance of one or more stations in dictating the trend may be determined. Cluster analysis, as well as percent faunal similarity, was most efficient at isolating station groupings. Techniques such as faunal replacement and species loss calculations are generally more useful in testing hypotheses about controls on species distributions, and should be employed where potential causal factors may be isolated.