OBJECTIVE: Married individuals have lower rates of problem drinking, but little is known about this relationship in the context of other factors. This longitudinal analysis examines marital status with other individual predisposing, problem severity and social predisposing characteristics to understand its strength in predicting alcohol consumption over 5 years. METHOD: A probability sample of dependent (n = 600) and problem (n = 992) drinkers was recruited through consecutive adult intakes from a Northern California county’s alcohol and drug treatment programs and through a general population survey. Annual volume of drinks consumed over a 5-year period-measured at baseline, and at 1, 3 and 5 years later-was estimated in four nested models using maximum likelihood estimation via PROC MIXED. RESULTS: In simpler models that examined only marital status, married individuals drank significantly less than those never married (p < .01 for problem drinkers and p < .05 for dependent drinkers); however, when our models added individual predisposing, problem severity and social predisposing characteristics, marital status was no longer significant in predicting a trajectory of decreased drinking. For problem drinkers, the following characteristics were more important than marital status in predicting alcohol consumption over the 5 years: individual predisposing characteristics (age, p < .001; income, p < .001; education, p < .001; and age of initiation of regular alcohol use, p < .001), problem severity (number of alcohol dependence symptoms, p < .001; number of alcohol-related social consequences, p < .001; and higher drug severity, p < .05) and social predisposing characteristics (family member with an alcohol problem, p < .05; the size of heavy alcohol- and drug-using social network, p < .05; and chemical dependency treatment in the prior year, p < .001). For those who were alcohol dependent, income (p < .05), number of alcohol dependence symptoms (p < .001), higher drug severity (p < .05) and a heavy alcohol- and drug-using social network (p < .05 ) were more important than marital status in predicting consumption. CONCLUSIONS: In longitudinal models, individual predisposing, problem severity and social predisposing characteristics are more important than marital status in predicting alcohol consumption.