The Marshmallow Test: What Replications Found and Critics Say

Key Points

  • A 1989 review by Mischel, Shoda and Rodriguez said that four-year-olds who delayed gratification longer developed into more competent adolescents.
  • A 2018 conceptual replication, in children whose mothers had not completed college, found a correlation between waiting and later achievement that was half the size of the original correlations and shrank by two thirds with controls for family background.
  • A 2013 experiment with 28 children found that those in a reliable environment waited significantly longer, suggesting wait times can reflect beliefs about whether waiting pays off.
  • A 2020 paper that surveyed 113 people in their late 40s found that a preregistered preschool delay measure did not predict 11 mid-life outcomes, while a composite self-regulation index predicted 10 of them.

Published .

A close-up heap of small pastel marshmallows in pink, orange, yellow and green against a dark background.
Colored marshmallows. Photo: Smith609, CC BY-SA 3.0, via Wikimedia Commons, resized, photo source.

Mischel, Shoda and Rodriguez summarized a research program in 1989, reporting that four-year-olds who waited longer for a reward in laboratory tasks developed into more competent adolescents, with higher scholastic performance and better coping with frustration and stress1. The line of work became known as the marshmallow test. Later work has disputed how much the waiting predicts and what it measures, and the sections below follow that debate.

How the test works

A Wikipedia summary of the Stanford experiment describes a child offered a choice between one small reward right away and two small rewards if the child waited, with the researcher leaving the child alone with a marshmallow for about 15 minutes7. Mischel and his colleagues found differences in self-control as early as the preschool years, and later follow-ups tracked the children into adolescence1.

The 2018 replication

In 2018, Watts, Duncan and Quan replicated and extended the 1990 Shoda, Mischel and Peake study, concentrating on children whose mothers had not completed college. An additional minute waited at age 4 predicted a gain of about one tenth of a standard deviation in achievement at age 152. That correlation was only half the size of those in the original studies, and it shrank by two thirds once the authors controlled for family background, early cognitive ability and the home environment2.

Most of the variation in adolescent achievement came from being able to wait at least 20 seconds, and associations with behavioral outcomes at age 15 were much smaller and rarely statistically significant2.

A reanalysis of the same data

In 2020, Michaelson and Munakata ran a preregistered secondary analysis of the same data set the 2018 replication used, and found a relationship between delay and problem behavior that the 2018 paper had not. Using the analytic approach of the original studies, associations were significant for three of the five outcomes they tested, and the relationship between delay and problem behavior held in bivariate, multivariate and multilevel models4. They argued that the relationships were better explained by social support than by self-control, which they read as meaning that the test reflects aspects of a child's early environment that matter over the long term4.

Waiting can be a reasonable choice

In 2013, Kidd, Palmeri and Aslin tested 28 children, average age 4 years 6 months, in an environment demonstrated to be either unreliable or reliable. Children in the reliable condition waited significantly longer3. The authors concluded that wait times may reflect beliefs about whether waiting will pay off, and not only differences in self-control3.

Adult outcomes

A 2013 paper followed a subset of 164 of the preschoolers about 30 years later, using self-reported height and weight, and found that each additional minute of delay at age 4 predicted a 0.2-point lower body mass index, a correlational result the authors said does not allow causal conclusions6.

A 2020 paper by Benjamin and colleagues, including Mischel and Shoda, surveyed 113 participants of the 1967 to 1973 Bing preschool studies in their late 40s, a sample the authors describe as small. A preregistered measure built from preschool delay did not predict 11 mid-life outcomes such as net worth and educational attainment, with a sign-adjusted average correlation of 0.025. A composite self-regulation index that combined the preschool measure with surveys at ages 17, 27 and 37 did predict 10 of the 11, with an average correlation of 0.19, and adding the preschool measure did not change its predictive power5.

Where this leaves the test

The sources above point in different directions. The original follow-ups found links between early waiting and adolescent outcomes, the 2018 replication found a smaller link that shrank with controls, and a reanalysis of the same data found support for some outcomes. The 2013 Kidd paper raises questions about what a wait time measures. The 2020 Benjamin paper found that preschool delay did not predict mid-life outcomes in a small sample, though a composite self-regulation index did, while the 2013 Schlam paper found a link to adult body mass index. This article describes published findings and gives no guidance for any child or family.

Trivia

  1. In what year did Watts, Duncan and Quan publish their conceptual replication? 20182.
  2. In the 2013 reliability experiment, how many children were tested? 283.
  3. How much lower was body mass index in adulthood for each additional minute of delay, per the 2013 Schlam paper? 0.2 points6.
  4. How many participants did Benjamin and colleagues survey in their late 40s? 1135.
  5. The 2018 replication focused on children whose mothers had not completed what? College2.

Sources

  1. Mischel, Shoda and Rodriguez, Delay of Gratification in Children, Science, 1989 (abstract record). Accessed October 26, 2026.
  2. Watts, Duncan and Quan, Revisiting the Marshmallow Test: A Conceptual Replication Investigating Links Between Early Delay of Gratification and Later Outcomes, Psychological Science, 2018 (abstract record). Accessed October 26, 2026.
  3. Kidd, Palmeri and Aslin, Rational Snacking: Young Children's Decision-Making on the Marshmallow Task Is Moderated by Beliefs About Environmental Reliability, Cognition, 2013 (abstract record). Accessed October 26, 2026.
  4. Michaelson and Munakata, Same Data Set, Different Conclusions: Preschool Delay of Gratification Predicts Later Behavioral Outcomes in a Preregistered Study, Psychological Science, 2020 (abstract record). Accessed October 26, 2026.
  5. Benjamin and colleagues, Predicting Mid-Life Capital Formation with Pre-School Delay of Gratification and Life-Course Measures of Self-Regulation, Journal of Economic Behavior and Organization, 2020 (abstract record). Accessed October 26, 2026.
  6. Schlam and colleagues, Preschoolers' Delay of Gratification Predicts Their Body Mass 30 Years Later, The Journal of Pediatrics, 2013 (abstract record). Accessed October 26, 2026.
  7. Stanford marshmallow experiment, Wikipedia. Accessed October 26, 2026.