Little is known about the life course of gambling problems and there have been no prospective studies of problem gambling. This article describes a study of 77 problem gamblers and 66 nonproblem gamblers recruited from a national prevalence survey in New Zealand in 1991 and reassessed in 1998. While most 1991 problem gamblers were nonproblematic at follow-up, a significant minority had developed more serious problems. Multiple logistic regression analysis identified more severe gambling problems, hazardous drinking, and a preference for track betting as the strongest independent predictors of subsequent problem gambling. These findings contradict conventional notions that pathological gambling is invariably a chronic or chronically relapsing disorder. The findings have implications for the interpretation of previous research, conduct of future research, and problem gambling policy and treatment.