Marshmallow Study Roasted?

[david_myers]

While on break in St. Andrews (Scotland) last week, I enjoyed a dinner conversation with a celebrated M.I.T. developmental psychologist and a similarly brilliant University of St. Andrews researcher. Among our dinner topics was an impressive recent conceptual replication of Walter Mischel’s famous marshmallow test.

Mischel and his colleagues, as you may recall, gave 4-year-olds a choice between one marshmallow now or two marshmallows later. Their long-term studies showed that those with the willpower to delay gratification as preschoolers went on as adults to have higher college-completion rates and incomes and fewer addiction problems. This gem in psychology’s lore—that a preschooler’s single behavioral act could predict that child’s life trajectory—is a favorite study for thousands of psychology instructors, and has made it into popular culture—from Sesame’s Street’s Cookie Monster to the conversation of Barack Obama.

In their recent replication of Mischel’s study, Tyler Watts, Greg Duncan, and Hoanan Quen followed a much larger and more diverse sample: 918 children who, at age 4½, took the marshmallow test as part of a 10-site National Institute of Child Health and Human Development child study. Observing the children’s school achievement at age 15, the researchers noted a modest, statistically significant association “between early delay ability and later achievement.” But after controlling for other factors, such as the child’s intelligence, family social status, and education, the effect shriveled.

“Of course!” said one of my dinner companions. Family socioeconomic status (SES) matters. It influences both children’s willingness to await the second marshmallow, and also academic and economic success. As other evidence indicates—see here and here—it is reasonable for children in poverty to seize what’s available now and to not trust promises of greater future rewards.

But my other dinner companion and I posited another factor: Any predictive variable can have its juice drained when we control for myriad other variables. Perhaps part of a child’s ability to delay gratification is intelligence (and the ability to contemplate the future) and experience. If so, controlling for such variables and then asking what’s the residual effect of delay of gratification, per se, is like asking what’s the real effect of a hurricane, per se, after controlling for barometric pressure, wind speed, and storm surge. A hurricane is a package variable, as is delay of gratification.

I put that argument to Tyler Watts, who offered this response:

If the ability to delay gratification is really a symptom of other characteristics in a child's life, then interventions designed to change only delay of gratification (but not those other characteristics) will probably not have the effect that you would expect based on the correlation Mischel and Shoda reported. So, if it’s the case that in order to generate the long-term effects reported in Mischel's work, interventions would have to target some combination of SES, parenting, and general cognitive ability, then it seems important to recognize that.  

This major new study prompts our reassessing the presumed predictive power of the famed marshmallow test. Given what we’ve known about how hard it is to predict to or from single acts of behavior—or single items on a test or questionnaire—we should not have been surprised. And we should not exaggerate the importance of teaching delay of gratification, apart from other important predictors of life success.

But the new findings do not undermine a deeper lesson: Part of moral development and life success is gaining self-discipline in restraining one’s impulses. To be mature is to forego small pleasures now to earn bigger rewards later. Thus, teacher ratings of children’s self-control (across countless observations) do predict future employment. And parent ratings of young children’s self-regulation predict future social success. Self-control matters.