Many of California's native populations of freshwater fish are in serious decline, as are freshwater faunas worldwide. Habitat loss and alteration, hydrologic modification, water pollution, and invasions have been identified as major drivers of these losses. Because these potential causes of decline are frequently correlated, it is difficult to separate direct from indirect effects of each factor and to appropriately rank their importance for conservation action. Recently a few authors have questioned the conservation significance of invasions, suggesting that they are "passengers" rather than "drivers" of ecological change. We compiled an extensive, watershed-level data set of fish presence and conservation status, land uses, and hydrologic modifications in California and used an information theoretic approach (Akaike's information criterion, AIC) and path analysis to evaluate competing models of native fish declines. Hydrologic modification (impoundments and diversions), invasions, and proportion of developed land were all predictive of the number of extinct and at-risk native fishes in California watersheds in the AIC analysis. Although nonindigenous fish richness was the best single predictor (after native richness) of fishes of conservation concern, the combined ranking of models containing hydrologic modification variables was slightly higher than that of models containing nonindigenous richness. Nevertheless, the path analysis indicated that the effects of both hydrologic modification and development on fishes of conservation concern were largely indirect, through their positive effects on nonindigenous fish richness. The best-fitting path model was the driver model, which included no direct effects of abiotic disturbance on native fish declines. Our results suggest that, for California freshwater fishes, invasions are the primary direct driver of extinctions and population declines, whereas the most damaging effect of habitat alteration is the tendency of altered habitats to support nonindigenous fishes.