The Relapse Replication and Extension Project (RREP) has failed to provide empirical support for Marlatt's relapse taxonomy. Neither the reliability of the original Marlatt coding system nor its predictive or construct validity was supported by this group of studies. The present commentary explores a number of possible reasons for the generally negative outcomes. These findings should certainly lead to a re-evaluation of Marlatt's relapse taxonomy and its operationalization. Nevertheless, despite the negative results, there are a number of reasons why the general Relapse Prevention concept is likely to survive in some form: it has been widely adopted and imitated clinically, key elements of the taxonomy are often focal points of treatment, and clinical research studies have repeatedly supported some elements of the taxonomy (e.g. negative emotional states, social pressure, interpersonal conflict, positive emotional states and temptations/urges). The RREP also evaluated some modifications of the original taxonomy as well as the use of more structured assessment instruments, and some of these provided more promising results. Further developments will need to take into account both research needs for greater precision, most likely through the use of more structured assessment instruments, and clinical needs for richness of detail and sensitivity to a wide variety of life circumstances.