Abstract Two sets of data were used to evaluate the procedure for rapid assessment of rivers described by Chessman (1995): (i) 72 samples from four habitats at 27 sites on the Nepean River and tributaries; and (ii) 41 riffle samples from the Blue Mountains. In the Nepean system all the sites had moderate anthropogenic disturbance but none had gross organic pollution. There were, however, conductivity differences related to mixed sandstone and shale lithology. The sites differed widely in natural physical attributes such as stream order (2–6) and altitude (40–600 m). The Blue Mountains data included sites with little or no anthropogenic disturbance and sites greatly affected by organic pollution. There was again a wide range of altitudes (15–1000 m) but stream orders were more restricted (1–4). The occurrence of macroinvertebrate families was analysed using cluster analysis (TWINSPAN) and ordination (semi-strong hybrid multidimensional scaling). Relationships between the patterns observed, the SIGNAL biotic index, and physical and chemical data were investigated. The Nepean data showed that the sample size of 100 animals used in the rapid procedure was sufficient to reveal natural distribution patterns in the communities, and that SIGNAL was essentially independent of these patterns. In the Blue Mountains, water pollution had a greater effect on macroinvertebrate communities than the physical habitat, and SIGNAL distinguished sites with differing levels of pollution.