The role of experimental analyses in guiding treatment is well established. However, not all experimental analyses yield conclusive results. Outcomes may be inconclusive due to time limitations that preclude extended observation and detailed experimental manipulations, or may result from interactions across experimental conditions, multiple control, or other unknown factors. In this study, we describe an assessment sequence that moves through four phases beginning with relatively brief (1 to 2 hr) analyses and culminating in extended analyses that may control for experimental confounding effects (e.g., interaction effects). Data illustrating the model are presented for 20 individuals referred for severe behavior problems including self-injury, aggression, stereotypy, and tantrums. Analyses were considered to be complete only when clear and replicable response patterns emerged. Results showed that clear and replicable response patterns emerged for 85% of the participants.