The literature on alien animal invaders focuses largely on successful invasions over broad geographic scales and rarely examines failed invasions. As a result, it is difficult to make predictions about which species are likely to become successful invaders or which environments are likely to be most susceptible to invasion. To address these issues, we developed a data set on fish invasions in watersheds throughout California (USA) that includes failed introductions. Our data set includes information from three stages of the invasion process (establishment, spread, and integration). We define seven categorical predictor variables (trophic status, size of native range, parental care, maximum adult size, physiological tolerance, distance from nearest native source, and propagule pressure) and one continuous predictor variable (prior invasion success) for all introduced species. Using an information-theoretic approach we evaluate 45 separate hypotheses derived from the invasion literature over these three stages of the invasion process. Our results indicate that successful establishment is best predicted with our global model that includes all seven variables, suggesting an inherent multivariate nature to the establishment stage of the invasion process. Spread of introduced fishes is best predicted using measures of physiological tolerance and propagule pressure. Species integration and impact is best predicted using a measure of prior invasion success. The results from analyses like this will help guide and inform management decisions regarding current and future species introductions.