Conducted 3 studies with 158 3rd graders to examine a cognitive model of how children terminate self-imposed delay of gratification when a preferred but delayed reward is pitted against an immediately available nonpreferred reward. According to this model, delay maintenance is primarily a function of the perceived value of the preferred reward, which is assumed to vary as a function of delay duration. In the 1st study, Ss devalued a physically present, delay-contingent preferred reward after a 10-min delay. Results of Study 2 reveal that the devaluation of the preferred reward observed in Study 1 occurred only when the preferred and nonpreferred rewards were initially highly similar in perceived value. Study 3 examined the effect of reward similarity and salience on children’s actual delay behavior and found that delay times were shortest under those conditions in which reward devaluation had been most pronounced. Results support a general model of self-control in which interruptions in self-control are assumed to be preceded by temporally based changes in outcome evaluation.