Marlatt's typology of relapse precipitants has had a major influence on clinical research and practice in the substance use disorders. The purpose of this study was to evaluate the construct validity of the typology, which was done in several ways. First, precipitants of the relapse coded at the baseline assessment and for the first drink post-treatment admission were compared. In addition, baseline relapse precipitants were compared with the highest factor scores recorded in the baseline Inventory of Drinking Situations data. A third set of analyses concerned the relationship among Marlatt precipitant codes, psychiatric diagnoses, and Alcohol Dependence Scale scores. The subjects were 142 men and women who were recruited for this study upon their admission to inpatient or outpatient alcohol treatment programs. Participants completed a baseline assessment covering substance use, including information on the precipitants of a pretreatment "relapse", and other areas of functioning. Subjects then completed bimonthly follow-up assessments during the course of 1 year. The results showed support for only one of the predictions regarding the construct validity of the Marlatt typology. Possible explanations for these findings and their implications for clinical research and practice are discussed.