Interactive and additive terms for substance misuse and crime were used to predict recidivism in 1,435 male inmates. Negative binomial regression revealed that although the additive term consistently predicted general recidivism and the presence of new charges, the interaction term consistently failed to predict these same two outcomes. The additive model continued to predict general recidivism and new charges when age and prior convictions were both controlled and when prior substance misuse and crime were entered separately. Both the substance misuse and crime components predicted income (property and drug) offenses, but only the crime component predicted person offenses. These results have implications for assessing and treating substance-involved offenders and suggest that the two components of the "worst of both worlds" hypothesis accumulate rather than interact in their effect on future recidivism. Although the effect depends on both components, one component may be stronger than the other in predicting certain outcomes.