It's a great story. Is it true? Tyler W. Watts, Greg J. Duncan, and Haonan Quan call it into doubt in their study "Revisiting the Marshmallow Test: A Conceptual Replication Investigating Links Between Early Delay of Gratification and Later Outcomes," just published in the journal Psychological Science (2018).
They go back to the original 1990 study: Shoda, Y., Mischel, W., and Peake, P. K. (1990). "Predicting adolescent cognitive and self-regulatory competencies from preschool delay of gratification: Identifying diagnostic conditions." Developmental Psychology, 26(6), 978-986. The original sample for this study was collected over a period of six years (1968-1974) among preschool students at the Bing Nursery School at Stanford University, described in the study as "mostly middle-class children of faculty and students from the Stanford University community." The studies included 653 preschool children, average age about four years. About 10 years later, surveys were mailed to parents whose addresses could be located, ending up with follow-up data on 185 children.
The study found that there was a positive correlation between children who were more likely to defer gratification at age 4, and those who were later rated by their parents as "more likely to exhibit self-control in frustrating situations, less likely to yield to temptation, more intelligent, and less distractable" compared to their peers. There was also some mild evidence (because there wasn't data for many of the students on this point) that SAT scores were higher for those who deferred gratification at age 4.
The more recent study used data from the National Institute of Child Health and Human Development (NICHD) Study of Early Child Care and Youth Development (SECCYD). This study draws on 10 different sites around the country, and tracks and studies a group of children up to age 15. Using this data, the researchers could look at children who had done a delay-of-gratification test by the age of four years, six months, and where they had follow-up data on behavior and educational achievement at age 15. The total sample size was 918. Of that group, the mothers of 552 of the children had not completed college when the child was one month old, which allows the researchers to split the sample into children whose mothers had completed college, and those who had not.
Before describing the results, just consider the samples. The second study is not a nationally representative sample. But it's larger in size and more representative than a single nursery school on the Stanford campus.
The follow-up study did find positive correlations between deferred gratification and some later measures, but the correlations were small, when they existed at all. In addition, the follow-up study was able to study whether the differences in deferred gratification might instead be picking up other factors. For example:
"[D]elay of gratification was strongly correlated with concurrent measures of cognitive ability ... This implies that an intervention that altered a child’s ability to delay but failed to change more general cognitive and behavioral capacities would likely have limited effects on later outcomes. If intervention developers hope to generate program impacts that replicate the long-term marshmallow test findings, targeting the broader cognitive and behavioral abilities related to delay of gratification might prove more fruitful."The follow-up study did have problems of its own. For example, the study asked children to defer gratification for 7 minutes, rather than the 15-20 minutes in the earlier studies. But for the group of children whose mothers had completed a four-year college degree, most of the children waited the full seven minutes. Thus, it wasn't possible within this group to draw meaningful conclusions about deferred gratification and later behaviors.
Of course, this finding suggests that a higher education for the mother can be relevant to whether a four-year-old can defer gratification, but even in this case, "most of the achievement boost for early delay ability was gained by waiting a mere 20 s." In other words, in the part of the sample for mothers who had not completed college, children who barely waited at all did perform less well, and waiting even 20 seconds was mildly associated with later gains for this group.
The short lesson here is not to freak out if your four-year-old gobbles some candy. The longer lesson is that level of mother's education is relevant to children's development, and that improving cognitive skills at younger ages can matter. Fpr some additional discussion of the results, see these short pieces in the Atlantic and the Guardian.