The most widely cited road block to successful treatment outcomes for psychological and substance use disorders has been described as the return to the previous behavior, or "relapse." The operational definition of "relapse" varies from study to study and between clinicians, but in general the term is used to indicate the return to previous levels of symptomatic behavior. One explanation for the variation in the operationalization of relapse is the wide variety of behaviors for which the term is applied, including (but not limited to): depression, substance abuse, schizophrenia, mania, sexual offending, risky sexual behavior, dieting, and the anxiety disorders. A second explanation for the multitude of definitions for relapse is the inherent complexity in the process of behavior change. In this paper we present the most recent treatment outcome research evaluating relapse rates, with a special focus on the substance use disorders. Following this review of the literature we present an argument for the operationalization of relapse as a dynamic process, which can be empirically characterized using dynamical systems theory. We support this argument by presenting results from the analysis of alcohol treatment outcomes using catastrophe modeling techniques. These results demonstrate the utility of catastrophe theory in modeling the alcohol relapse process. The implications of these analyses for the treatment of alcohol use disorders, as well as a discussion of future research incorporating nonlinear dynamical systems theory is provided.