Does the “Marshmallow Test" Really Predict Success?

Intriguing surprises upon repeating the original high-impact research.

Posted May 27, 2018

Last night I dreamt I ate a ten pound marshmallow. When I woke up the pillow was gone.

  —Tommy Cooper

Think of the universe as a benevolent parent. A child may want a tub of ice-cream and marshmallows, but a wise parent will give it fruits and vegetables instead. That is not what the child wants, but it is what the child needs.

 —Srikumar Rao

Marshmallows across time

The original Marshmallow Experiment (Mischel, 1958) was conducted in Trinidad, comparing the capacity of Creole and South Asian childrens to forgo a 1-cent candy in favor of a much nicer 10-cent candy one week later.

In the original study, Mischel is presented as an American gathering information about children in local schools, made up of Creole and South Asian cultural groups. He shows the children the candy options, and tells them: “I would like to give each of you a piece of candy but I don’t have enough of these [better ones] with me today. So you can either get this one [the smaller] right now, today, or, if you want to, you can wait for this one [the better one], which I will bring back next Wednesday [a week later]”.

He found that the Creole children were significantly more likely to take the candy right away, as contrasted with the South Asian kids. He found two predictors for immediate gratification—having a home without a father, and being younger, both presumed to be related to psychological and emotional maturity. Children from homes with fathers (typically the South Asian families), and older children, were able to wait until the following week, and enjoy more candy. Future research explored the ongoing themes of self-regulation strategies geared to delay gratification for future benefit, ego control, and ego resilience

In the early 1970s, Mischel and his colleagues (1972) studied children between the ages of 3 and 5 years old to look at how they handled gratification in the face of temptation to better understand voluntary self-control. There were three experiments. In the first one, distraction from the reward (sitting right in front of the children) prolonged the wait time. In the second, cultivating sad thoughts versus happy thoughts made it harder to take the immediate pay-off, and in the final experiment being encouraged to think about the reward (now out of sight) made it harder to wait. Their research continued to tease apart different regulation strategies, identifying what children who were able to wait did to enable them to delay gratification, whether these skills might be teachable, and looking at how those skills could translate into real-world performance later on in life. The marshmallow test came to be considered more or less an indicator of self-control—becoming imbued with an almost magical aura.

In 1988, Mischel and Shoda published a paper entitled The Nature of Adolescent Competencies Predicted by Preschool Delay of Gratification. In this research, the seminal Marshmallow Experiment paper everyone’s heard about, study authors looked at the relationship between the ability to wait longer to take a desired treat—one marshmallow now or two after 10 minutes—and markers of performance and success measured 10 years after, as reported by the participants’ parents and performance measures including verbal fluency, social success, focus, dependability, trustworthiness, standardized test scores for college application, and a host of other admired qualities most desirable in one’s offspring. Every moment longer that a child had been able to wait appeared to be correlated with how much better they did later in life. As the data diffused into the culture, parents and educators snapped to attention, and the Marshmallow Test took on iconic proportions. Pity the child who couldn’t resist temptation, because that might portend dismal future prospects.

Marshmallows, revisited

Fast-forward to 2018, when Watts, Duncan and Quan (a group of researchers from UC Irvine and New York University) published their paper, Revisiting the Marshmallow Test: A Conceptual Replication Investigating Links Between Early Delay of Gratification and Later Outcomes. Researchers looked at ability to delay gratification at age 5 as related to various benchmarks at age 15. The design was similar to the original experiments in many respects. The children were offered a treat, assigned according to what they said they liked the most, marshmallows, cookie, or chocolate, and so on.  If they were able to wait 7 minutes, they got a larger portion of their favorite, but if they could not, they received a scantier offering.

Researchers were surprised to find that a large proportion of children were able to wait the full time, and the proportion varied with the mother’s level of education. Sixty-eight percent of those whose mothers had college degrees and 45 percent for those whose mothers did not complete college were able to wait the full 7 minutes. This limited the data analysis for the group with more highly educated mothers.

Researchers used a battery of assessments to look at a range of factors: the Woodcock-Johnson test for academic achievement; the Child Behavior Checklist, to look for behavioral issues (internalizing e.g. depression vs. externalizing e.g. acting out); and the Home Observation for Measurement of the Environment (HOME), a highly detailed roster of important factors related to the home environment, along with a variety of demographic variables. HOME looks at the early childhood environment, including factors such as the quality of the learning environment, the approach to languages, the physical environment, responsivity of those around the child, academic resources, the availability of role models, and other crucial influences not previously included in studies of confectionary fortitude.

Surprising results

When all was said and done, their results were very different from those of the original Marshmallow Experiment. First of all, when they controlled for all the additional variables, especially the HOME measures, they did not see a significant correlation with how long kids had been able to wait and future success and performance. They found that for children of less educated parents, waiting only the first 20 seconds accounted for the majority of what was predicted about future academic achievement. Waiting longer than 20 seconds didn’t track with greater gains. For the children of more educated parents, there was no correlation between duration of delaying gratification and future academic or behavioral measures, after controlling for the HOME and related variables.

Notably, the uncontrolled correlations did seem to show a benefit for longer delayed gratification, appearing to mirror the original experiment's findings, but that effect vanished with control of variance. Another notable—it would have been interesting to see if there were any effects observed if the waiting period had been longer than 7 minutes. Maybe if you can wait at least 12 minutes, for example, you would do much better than those who could only wait 10 minutes—but presumably the researchers did not expect that many would be able to wait longer, and so used the shorter time-frame.

Much to ponder

I can’t help but wonder if kids have learned to be able to wait longer because of the Marshmallow Experiment, the broad exposure it has had, and potential effects on education and child-rearing. This “Marshmallow Effect”, one of the propeller blades of helicopter parenting, might very well be stronger for the "Marshmallow Kids" of highly educated parents. Educated parents might be more familiar with parenting research and recommendations, consumers of popular psychology, and highly motivated to provide the most enriched environments for their offspring (thus driving up the HOME scores for positive influences). From this point of view, next time you are frustrated with a Millennial, you might consider whether you are feeling aftershocks from the Marshmallow Experiment.

Most importantly though, this research suggests that basic impulse control, after correcting for environmental factors and given the right context, may turn out to be a big predictor of future success. The more nuanced strategies for self-regulation, tools which presumably take longer than 20 seconds to implement, may not be as clearly implicated in success as earlier research would suggest. Let's see what the next round of research shows, no easy feat given the time spans involved and the foresight to have a good research design.

The Marshmallow Test may not actually reflect self-control, a challenge to the long-held notion it does do just that. Moreover, the study authors note that we need to proceed carefully as we try to better understand how children develop self-control and develop cognitive abilities. It’s a good idea to resist the temptation to over-generalize or even jump to conclusions about what to do to give children a competitive advantage, and look more closely at a variety of developmental influences. In a culture which brainwashes us to "fail fast and fail often", delaying gratification also may not be as adaptive as it once was. Time will tell.


Mischel, W. (1958). Preference for delayed reinforcement: An experimental study of a cultural observation. The Journal of Abnormal and Social Psychology, 56(1), 57-61.

Mischel, W., Ebbesen, E. B., & Raskoff Zeiss, A. (1972). Cognitive and attentional mechanisms in delay of gratification. Journal of Personality and Social Psychology, 21(2), 204-218.

Mischel W & Shoda Y. The Nature of Adolescent Competencies Predicted by Preschool Delay of Gratification, Journal of Personality and Social Psychology, 1988, Vol. 54, No. 4, 687-696.

Watts TW, Duncan GJ & Quan H. Revising the Marshmallow Test: A Conceptual Replication Investigating Links Between Early Delay of Gratification and Later Outcomes. Psychological Science, 1-19, 25 May, 2018.