Few studies have tried to identify the process of risk behavior change. In this paper, we present a method for the analysis of the process of risk behavior change; the methodology involves using time series data to form transition probability matrices. We apply the method to human immunodeficiency virus (HIV) sexual risk behavioral data from a cohort of gay men in Amsterdam. We determine the process of risk behavior change in this cohort, and we demonstrate that it is not possible to deduce the process simply from an examination of the static pattern. Our results imply that the risk behavior change that is observed in this cohort may be viewed as a homogeneous one-step Markovian process. This process of risk behavior change has implications for the long-term prediction of HIV seroconversion rates and for the design and evaluation of HIV behavioral intervention studies and vaccine trials. Our results also have implications for the interpretation of relapse behavior.