Why wait? A meta-analysis of task and sample moderators of children’s performance on the marshmallow test

<p>The marshmallow test is the most well-known behavioral assessment of delay of gratification inchildren and performance on the test is associated with several later life outcomes. Yet studiesvary considerably in their samples and implementation of the task. This meta-analysis examinedwhether task and sample characteristics moderate children’s waiting patterns. Multilevel meta-regressions synthesized 119 samples from 86 published and unpublished studies with maximumwaiting periods ranging from 7 to 25 minutes. Outcomes were mean wait time, the proportion ofchildren waiting for the full delay, and the proportion excluded from analyses. Older childrenwaited longer and were more likely to complete the full delay. Samples characterized as lowersocioeconomic status had shorter mean wait times, although socioeconomic status was notassociated with waiting the full delay period. Allowing children to choose the reward andproviding an explanation for the experimenter’s delayed return were unrelated to waiting, andBayesian model comparisons provided moderate evidence for the null for these moderators.Most other coded procedural characteristics were also unrelated to waiting. Longer maximumtask durations were associated with longer mean waits but lower odds of completing the fulldelay and greater exclusion. Publication year was positively associated with mean wait time inseveral models, consistent with prior research, although this association varied by sample ageand maximum task duration. Overall, children’s waiting was relatively consistent across manyprocedural variations, while age, socioeconomic context, and task duration emerged as importantsources of variation.</p>