Patterns of spatial autocorrelation of biota may reveal much about underlying ecological and biological forces responsible for generating the patterns. Operationally, ecological work and many applied problems (e.g., impact detection, ecosystem health assessment using reference sites) require statistical knowledge of autocorrelation patterns. Here, we report on assemblage-level autocorrelation in the benthic-invertebrate assemblages of riffles in two adjacent, relatively pristine rivers in south-eastern Victoria, Australia (40 km reaches of the Wellington and Wonnangatta rivers). The assemblages of the Wellington River were strongly autocorrelated, but those of the Wonnangatta River showed a distance-independent pattern. There was no effect of taxonomic resolution, rarity protocols or whole-assemblage surrogates on the inferred levels of autocorrelation. We conclude that there is little evidence that one can assume the pattern of spatial relationships among invertebrate faunas within a river, and this probably holds true for the usual set of taxonomic resolutions and subsets used to discern changes wrought by human impacts.