Same Data Set, Different Conclusions: Preschool Delay of Gratification Predicts Later Behavioral Outcomes in a Preregistered Study.
Level 3 - non-randomized controlled study
Secondary analysis of longitudinal cohort data (by design analogy, non-clinical)
PubMed 31961773 · doi:10.1177/0956797619896270
What was done
The authors conducted an independent, preregistered secondary analysis of an existing longitudinal dataset to test whether preschool delay of gratification predicts adolescent outcomes. They evaluated five outcomes using the analytic framework of original marshmallow test studies, alongside bivariate, multivariate, and multilevel models.
What was found
The abstract reports no numerical values, sample sizes, or test statistics. Preschool delay of gratification was significantly associated with three of the five tested adolescent outcomes. The association with problem behavior remained significant across bivariate, multivariate, and multilevel models, and was better explained by social support than by self-control.
Why it matters
This study illustrates how different analytic choices on the same dataset can yield conflicting conclusions about the marshmallow test. It suggests that delay tasks may predict long-term outcomes because they reflect early social support and environmental factors rather than self-control alone.
Limits
Sample size, participant characteristics, effect sizes, and specific outcome measures are not reported in the abstract. Two of the five tested outcomes showed no significant relationship. As a secondary observational analysis, causal inference is limited.
Cited by
- supports A re-analysis of the marshmallow replication dataset by Yuko Munakata's team using theory-driven covariates found that delay of gratification still predicted problematic behavior in children.