This study is an extension of an experiment where the reliability of children's environment was manipulated before children completed the Marshmallow Task (Cognition, 2013, Vol. 126, pp. 109-114). In that experiment, Kidd, Palmeri, and Aslin found a significant difference in waiting time between two conditions in which the experimenter demonstrated reliability (by returning with promised reward) or unreliability (by not returning with rewardP). Children who had an unreliable experimenter did not wait as long during the Marshmallow Task, suggesting that delay gratification performance may be, in part, based on a rational decision. Due to the important theoretical and practical implications of this finding, we repeated the procedure of this experiment with 60 3- to 5-year-old children (twice as many as in the original study), but in a more familiar context (e.g., children's school instead of a lab). Using Bayesian analyses, we found an effect (albeit smaller than in the original study) of experimenter reliability as well as a significant gender by condition interaction effect.