We assessed relationships between environmental characteristics and macroinvertebrate assemblages in lotic habitats of California's Central Valley with community metric and multivariate statistical approaches. Using canonical ordination analyses, we contrasted results when assemblage structure was assessed with macroinvertebrate metrics, as suggested for use in indices of biotic integrity, or with genera abundances. Our objectives were to identify metrics or taxa diagnostic of lotic environmental stressors and compare the capacity of these approaches to detect stressors in order to suggest how they might be used to diagnose stressors. For macroinvertebrate metrics, redundancy analysis (RDA) extracted three axes correlated with channel morphology and substrates. For genera abundances, canonical correspondence analysis (CCA) extracted two axes correlated with soluble salts and with channel morphology and substrates but did not separate these gradients onto different axes. Cluster analyses identified five RDA and five CCA site groups, which exhibited differences for environmental variables, metrics, or genera abundances, and agreement between the analyses in partitioning of sites was greater than if sites were partitioned randomly. These approaches differ in their ability to detect environmental stressors, because they measure different aspects of assemblages and would be complementary in design of new metrics diagnostic of stressors.