Bioassessment is based upon the premise that biological assemblages have predictable relationships with the surrounding natural and human environments. As the nature of these relationships can vary from region to region, it is important that environment–biota relationships be established prior to the initiation of any bioassessment program. In this study, multivariate analysis was used to establish how fish and benthic macroinvertebrate (BMI) assemblages in southwestern Ontario streams vary across natural and human activity gradients. The use of canonical correspondence analysis allowed us to determine that changes in community composition of both fish and BMI are strongly correlated with variation in the extent of human activity. The primary source of variation in community composition across activity gradients appeared to reflect a shift from intolerant to tolerant taxa as the extent of human activity increased. Habitat and feeding traits, for BMI and fish respectively, accounted for a secondary source of variation primarily attributable to differences in the extent of human activity at the reach scale. However, variation in human activity, especially at the basin scale, covaried with the dominant natural gradient of surface geology, making interpretation of the results difficult. Implications for bioassessment studies are discussed.