Positive correlations between diversity and stability have been reported for a number of ecosystems and are thought to be caused by a stabilizing effect of differential species' responses to environmental perturbation. Empirical field studies in which investigators tested for diversity–stability relationships are lacking for some taxonomic groups and typically have not included tests of the importance of other potential correlates of diversity or assemblage structure. We sampled stream fish assemblages and associated habitat variables at 36 sites over a 10-y period. Quantitative and qualitative measures of stability were correlated with fish diversity at sites. Fish assemblage composition was correlated with a variety of habitat variables, and diversity was correlated with stream size. We used Akaike's Information Criterion to select the models that best predicted qualitative and quantitative stability. Candidate models contained variables describing diversity, stream size, time between samples, and change in habitat variables over time. Models that included diversity and time between samples were the best predictors of stability. Our results support the existence of diversity–stability relationships, and we showed that other predictors of diversity or habitat change were generally poor predictors of stability.