We used descriptive assessment information to generate hypotheses regarding the function of destructive behavior for 2 individuals who displayed near‐zero rates of problem behavior during an experimental functional analysis using methods similar to Iwata, Dorsey, Slifer, Bauman, and Richman (1982/1994). The descriptive data suggested that destructive behavior occurred primarily when caregivers issued requests to the participants that interfered with ongoing high‐probability (and presumably highly preferred) behaviors (i.e., a “don't” or a symmetrical “do” request). Subsequent experimental analyses showed that destructive behavior was maintained by contingent termination of “don't” and symmetrical “do” requests but not by termination of topographically similar “do” requests. These results suggested that destructive behavior may have been maintained by positive reinforcement (i.e., termination of the “don't” request allowed the individual to return to a highly preferred activity). Finally, a treatment (functional communication training plus extinction) developed on the basis of these analyses reduced destructive behavior to near‐zero levels.