The techniques currently available for studying drug self-administration in animals offer the unique opportunity to carry out micro-analysis of initial episodes of drug use which are extremely difficult to obtain for human subjects. Nonetheless, traditional self-administration techniques do not allow a cost-effective output of large sample sizes needed for genetic analysis. Additionally, the statistical techniques that allow the integration of within-subject temporal data with genetic information are scant. We therefore propose a two-stage method for analyzing strain differences in dynamic phenotypes for a high-throughput version of the self-administration procedure. On a first phenotype-refinement stage, a change-point algorithm (Gallistel et al. (2004) Proc. Natl Acad. Sci. USA 101:13124-13131) was used to separate individual drug self-administration response curves into three distinct components. In a second stage, strains differences in these indexes were assessed. This two-stage approach is illustrated with drug self-administration data and through a computer simulation.