Investigation of problem gambling has relied heavily on retrospective and cross-sectional studies of problem gamblers in clinical and community settings. While making a useful contribution, studies of this type have inherent limitations with respect to the examination of change and the determination of risk and protective factors for problem onset and progression. This article critically reviews general and special population studies that have employed prospective designs. While few in number, recent in execution and typically methodologically compromised, findings from these studies significantly challenge core assumptions about the nature, development and measurement of problem gambling and raise important questions for future research. These substantive matters are considered, along with identification of conceptual and methodological shortcomings that require remedy if prospective research is to fulfil its potential. This potential is profound - to serve as both catalyst and vehicle to move the field from its rather disjointed preoccupation with description and distribution to become a theory driven, cumulative science of problem gambling determinants and consequences.