In this article, I argue that for risk/recidivism research to achieve its full potential, it must continue to grow beyond its empirical roots, a process that began more than 20 years ago. The theoretical model highlighted in the current article can explain one of the brute facts of criminology: namely, that past offending is one of the best predictors of future offending. By understanding how cognitive variables mediate the past crime–future crime relationship, we improve our ability to more effectively manage recidivism risk. Toward this end, six considerations for mediation analysis are discussed: variable selection, model creation, research design, data analysis, effect size estimation, and sensitivity testing. An illustrative example using attitude toward fighting as a mediator of the past physical aggression -> future physical aggression relationship is presented.