The Marshmallow Experiment: Self-Control, Class, and a Crumbling Prediction
Photo: Amel Uzunovic / Pexels
A four-year-old sits alone at a table. On the plate in front of her: one marshmallow. A researcher explains that if she waits fifteen minutes without eating it, she will get two marshmallows when the researcher returns. Then the researcher leaves, and the waiting begins. The child's strategy for those fifteen minutes — whether she stares at the treat, covers her eyes, sings to herself, or simply grabs it — became one of psychology's most famous predictors of adult life outcomes. Fifty years later, that prediction turned out to be far more complicated than anyone advertised.
What the Marshmallow Test Was Actually Measuring
The Stanford marshmallow experiment refers to a series of studies conducted at Stanford University's Bing Nursery School between 1968 and 1974, led by psychologist Walter Mischel. The core question was straightforward: can young children resist an immediate reward in order to receive a larger one later, and does that capacity predict anything meaningful about their futures?
Mischel's team worked with children aged roughly three to five years. A researcher would show the child a snack of their choosing — marshmallows were common, but pretzels and cookies were also used — and explain the deal: wait alone without eating it, and you get two when the researcher returns. Ring a bell and the researcher will come back immediately, but you only get the one.
The test was not designed to measure raw willpower in some motivational-poster sense. Mischel was investigating delay of gratification — the cognitive and emotional strategies children use to bridge the gap between a tempting present reward and a better future reward. That distinction matters, because the experiment later got stripped of its nuance when it entered popular culture.
Source: Bing Marshmallow Studies: 50 Years of Continuing Research — Stanford Bing Nursery School
The original 1972 studies involved approximately 56 children drawn from the Bing Nursery School. That is a small number, and it comes with an important caveat: the children attended a nursery school at one of the wealthiest, most elite universities in the United States. Their families were largely middle- and upper-class, predominantly white, and affiliated with Stanford. This sample profile would become the most important detail critics pointed to decades later.
Source: Marshmallow Test Experiment in Psychology — Simply Psychology
Definition — delay of gratification: The ability to resist a smaller, immediate reward in favor of a larger reward available after a waiting period. In developmental psychology, this is considered a component of executive function, the set of cognitive processes that regulate goal-directed behavior.
The Setup: One Room, One Marshmallow, One Camera
The procedure was carefully controlled to isolate the child's self-regulation strategies. Each child was taken into a room furnished with only a table, a chair, and the chosen snack sitting on a plate. The researcher demonstrated the options — ring the bell for immediate reward, or wait and receive double — and then left, typically for up to fifteen minutes.
Source: The Stanford Marshmallow Experiment: How Self-Control Affects Success — Effectiviology
Mischel and his colleagues observed the children through a one-way mirror and, in later iterations, through video recording. What they saw was a vivid behavioral catalog: children who covered their eyes with their hands, who turned away from the plate, who began singing songs or talking to themselves, who examined the marshmallow with scientific intensity before the temptation broke their resolve, and children who simply ate it within seconds of the researcher leaving the room.
The strategies that worked shared a common feature: they created psychological distance from the tempting object. A child who imagined the marshmallow as a fluffy cloud — a technique Mischel's team experimented with — waited significantly longer than one who was asked to think about how sweet and chewy it was.
Think of it like trying not to think about a pink elephant. The moment your mental architecture places the object at center stage, restraint collapses. The children who succeeded were not exerting heroic self-control; they were reorganizing the situation so self-control was less necessary.
Source: Self-Control and Grit: The Marshmallow Experiment — Mr Sanchez Tutoring
The Protocol: How Mischel Structured the Studies
The experiments evolved across several years. Early versions tested which types of rewards were most effective at motivating children to wait. Later versions introduced variations: some children were shown both the immediate and delayed reward simultaneously; others were given instructions about mental strategies; still others had the reward obscured from sight.
| Year | Key Variable Tested | Main Finding |
|---|---|---|
| 1968–1970 | Reward salience (visible vs. hidden) | Hidden rewards led to longer wait times |
| 1970–1972 | Cognitive strategies (think "cool" vs. "hot") | Abstract/distancing thought extended waiting |
| 1972 | Standard delay paradigm | Baseline data collected; ~56 children |
| 1988 | Follow-up with original participants | Waiting time correlated with parent-rated competence |
| 1990 | SAT score correlation study | Longer waiters had higher SAT scores |
The 1990 follow-up, conducted by Shoda, Mischel, and Peake, tracked participants from the original nursery school cohort. Those who had waited longer as preschoolers scored, on average, 210 points higher on the SAT than those who had waited the shortest times. They also received more positive behavioral ratings from their parents as adolescents — better attention, better social skills, better ability to cope with frustration.
Additional longitudinal work found that each additional minute a preschooler delayed gratification predicted a 0.2-point reduction in adult BMI. Delay time also correlated with lower rates of substance abuse, higher educational attainment, and better stress regulation in adulthood.
Source: 'Willpower' over the life span: decomposing self-regulation — PMC
These findings entered the popular imagination rapidly. By the 2000s, the marshmallow test was shorthand for character: children who waited were destined for success; children who grabbed were marked for struggle. Parenting books, business seminars, and motivational speakers all reached for the same conclusion — self-control is a trait that predicts everything, and therefore it is a trait you should cultivate in your children.
The Results That Made the Marshmallow Famous
The headline finding from the original longitudinal work was striking enough to anchor a decade of self-control research and one bestselling book (Mischel's own The Marshmallow Test, published in 2014). The claim, stated plainly: a single behavioral measure at age four could predict meaningful differences in life outcomes well into adulthood.
Source: New Study Disavows Marshmallow Test's Predictive Powers — UCLA Anderson Review
The effect was real. The data was real. The correlations existed. What the popular retelling consistently omitted was the sample: roughly 90 children from one nursery school on one university campus, drawn from a narrow socioeconomic band. The study never claimed to represent all children, but that context fell away as the finding spread.
Source: What the Marshmallow Test Teaches About Weight Loss — Michael Moody Fitness
The interpretive leap — from "self-control correlates with outcomes in this sample" to "self-control is the key driver of lifetime success" — was the science journalism layer, not the scientific finding. But it was the leap that made the marshmallow test famous, and it was that leap that the 2018 replication would directly target.
The Controversy: Does Background Explain Everything?
In 2018, Tyler Watts, Greg Duncan, and Haonan Quan published a conceptual replication in Psychological Science that used a sample more than ten times larger than Mischel's original: 918 children, drawn from a study designed to represent the diversity of American families.
Source: Revisiting the Marshmallow Test: A Conceptual Replication — Psychological Science (PubMed)
Their key move was to control for socioeconomic variables — parental income, maternal education level, and the quality of the home environment (measured by the HOME scale, a validated assessment of cognitive stimulation and emotional support in the household). After applying those controls, the correlation between delay of gratification at age four and academic achievement at age fifteen dropped by roughly two-thirds.
Among children whose mothers had not completed college, each additional minute waited predicted a gain of about one-tenth of a standard deviation in achievement at age fifteen. That sounds small — and it is small — but it was still statistically significant before controls. Add in background variables, and it effectively disappeared.
The interpretation the 2018 team offered was not "self-control doesn't matter." It was more precise and more unsettling: if you come from a stable, well-resourced home, you have both the cognitive slack to deploy self-regulation strategies and the life-experience evidence that promises actually get kept. A child who has grown up in an environment where adults reliably follow through — where waiting is worth it because the second marshmallow actually arrives — will wait longer. A child whose experience teaches that promised rewards frequently evaporate has rational grounds for eating the marshmallow now.
Think of it this way: the marshmallow test measures not just self-control but also trust. Children who grab immediately are not necessarily impulsive; some of them have learned, correctly, that delayed rewards are not always delivered.
Source: What the Marshmallow Test Got Wrong About Child Psychology — Psyche
This reframing has significant policy implications. The original finding, as popularly understood, suggested that teaching self-control to children was a high-leverage intervention — get kids to wait, and they will do better in life. The 2018 finding suggests that the causal arrow may run differently: address family economic security and environmental stability, and self-control will follow, along with the outcomes that come with it.
Replication and Reanalysis: What the Science Actually Shows
The 2018 Watts study prompted a wave of commentary and pushback. Supporters of the original Mischel findings pointed out that the two studies were not direct replications: Mischel's delay paradigm used the child's own preferred snack, while the Watts study used different procedures and measured different outcomes. Critics of the Watts reanalysis argued that controlling for so many background variables could be overcorrecting — self-control and socioeconomic background are not independent, so removing the variance from background conditions might inadvertently remove some of the variance attributable to self-control itself.
Source: Revisiting the Marshmallow Test: On the Interpretation of Replication Results — ResearchGate
The methodological debate is genuine, and it has not fully resolved. What both sides agree on is that the original claim — one measure at age four predicts a wide range of adult outcomes — was overstated in its popular form. The original studies used a homogeneous, advantaged sample. The effect sizes were large because the range of outcomes in that sample was large. When the sample expanded to include families from lower income brackets, the effect shrank.
Source: The Marshmallow Study Revisited — Social Science Space
What remains of the original finding after a decade of reanalysis:
- Self-control in early childhood does correlate with some later outcomes.
- That correlation is considerably weaker than originally reported.
- Socioeconomic and family environment variables account for much of the apparent relationship.
- The delay strategies Mischel identified — cognitive distancing, attention redirection — are real mechanisms with practical applications.
- No single behavioral measure at age four is a destiny.
A meta-analysis of self-control research broadly supports the view that self-regulation is a meaningful predictor of outcomes, but the marshmallow test specifically has a weaker track record than the popular narrative suggested.
Where It Stands Today: Self-Regulation Without the Mythology
Walter Mischel, who died in 2018 — the same year the Watts replication was published — spent the last years of his career clarifying what the marshmallow studies actually showed. He was not a proponent of the character-is-destiny interpretation that his work had spawned. He emphasized that the studies were about strategies: if children learn to create distance from a tempting object, they can extend their waiting time. That is a skill, not a fixed trait.
Source: Stanford marshmallow experiment facts for kids — Kiddle
Current developmental psychology treats self-regulation as a multidimensional construct — not a single dial that goes from "impulsive" to "patient," but a suite of cognitive and emotional capacities that develop differently in different contexts. Hot executive function (the ability to regulate emotion in the face of temptation) is distinct from cool executive function (planning and problem-solving). The marshmallow test mainly measures the hot variety.
Source: Bing Marshmallow Studies: 50 Years of Continuing Research — Stanford Bing Nursery School
The practical implications today look different from the original 1990s enthusiasm. Interventions designed purely to train self-control in low-income children without addressing the environmental stressors driving their behavior have shown limited effectiveness. More promising are programs that address both the child's regulatory skills and the family's material stability — outcomes improve when children have both the capacity and the rational expectation that waiting is worth it.
The marshmallow test also spawned useful lines of research that do hold up: the work on how cognitive reappraisal and mental distancing reduce impulsive behavior has direct applications in behavioral economics, addiction treatment, and cognitive behavioral therapy.
Source: Self-Control and Grit: The Marshmallow Experiment — Mr Sanchez Tutoring
What the Marshmallow Test Should Be Replaced With
The marshmallow task measures one narrow slice of self-regulation under one specific set of conditions. The field of developmental psychology has since built a substantially richer toolkit for assessing the broader construct that the marshmallow test was trying to reach — and several of those instruments predict outcomes better, across more populations, without the confounds that the single-task paradigm baked in.
The modern consensus, shaped in large part by Adele Diamond's landmark 2013 review in Annual Review of Psychology, treats self-regulation as part of executive function — a cluster of interrelated cognitive capacities that includes working memory, cognitive flexibility, and inhibitory control. Diamond's review distinguishes "cool" executive function (deliberate, cognitive, relatively emotionally neutral problem-solving) from "hot" executive function (regulation under conditions of motivation and emotional arousal). The marshmallow test is primarily a hot executive function task: it loads on inhibitory control in a highly motivating context. It misses most of cool executive function entirely.
Source: Diamond, A. (2013), "Executive functions," Annual Review of Psychology, 64, 135–168. doi:10.1146/annurev-psych-113011-143750.
Better predictive instruments exist. Mary Rothbart's Children's Behavior Questionnaire (CBQ), a parent-report measure validated across large samples, assesses Effortful Control — the ability to suppress a dominant response and activate a subdominant one — along with attention focusing, inhibitory control, and perceptual sensitivity. The CBQ and its derivatives (the Early Childhood Behavior Questionnaire for children under three, the Adult Temperament Questionnaire) are not single-sitting tasks. They aggregate behavior across many contexts and many days, which makes them substantially less sensitive to the environmental and trust confounds that the Watts 2018 replication exposed in the marshmallow paradigm. Parent-report temperament measures have shown predictive validity for school-readiness, social competence, and behavioral outcomes that equals or exceeds single-task performance measures in several longitudinal datasets.
Source: Rothbart, M. K., Ahadi, S. A., Hershey, K. L., & Fisher, P. (2001), "Investigations of temperament at three to seven years: The Children's Behavior Questionnaire," Child Development, 72(5), 1394–1408.
The 2018 Watts, Duncan, and Quan replication — which used 918 children and found the marshmallow effect shrank by two-thirds once socioeconomic controls were applied — carries a direct policy implication that goes beyond "the test was overhyped." If a child's delay time reflects their trust in promise-keeping adults as much as their internal inhibitory control, then a single task administered by a stranger in an unfamiliar room is measuring the intersection of regulatory capacity and environmental experience. Pulling those apart requires longitudinal observation across multiple contexts, which is what parent-report and composite instruments provide.
Source: Watts, T. W., Duncan, G. J., & Quan, H. (2018), "Revisiting the marshmallow test: A conceptual replication investigating links between early delay of gratification and later outcomes," Psychological Science, 29(7), 1159–1177.
Digital adaptations are also reshaping the measurement landscape. The NIH Toolbox Cognition Battery includes computerized flanker tasks and dimensional change card sort tasks — validated, normed, and more sensitive to executive function profiles than a single delay task. Computerized Stroop variants, which measure inhibitory control by asking participants to name ink colors of color-words printed in incongruent ink, are now standard in large cohort studies. These instruments separate the components that the marshmallow test collapses together. A child who scores poorly on a delay task but well on a flanker task has a different profile from one who scores poorly on both — and they likely require different interventions. The marshmallow test, by design, cannot make that distinction.
FAQs
Q: How many children participated in the original marshmallow experiment? A: The 1972 core studies involved approximately 56 children aged three to five at Stanford's Bing Nursery School. Later longitudinal follow-ups tracked a larger subset of children who had participated in the delay paradigm studies between 1968 and 1974.
Q: What did the 2018 replication find exactly? A: Watts, Duncan, and Quan tested 918 children and found that the correlation between delay of gratification at age four and academic achievement at age fifteen was roughly half the size of the original study's reported effect, and that it shrank by about two-thirds once they controlled for family socioeconomic background, maternal education, and home environment quality.
Q: Does the marshmallow test measure intelligence or self-control? A: Primarily self-control and strategy use — specifically the ability to shift attention away from a tempting object. However, cognitive ability and self-control are correlated in this age group, and the 2018 replication found that early cognitive ability was among the variables that reduced the apparent self-control effect when controlled for.
Q: Can children be taught to wait longer on the marshmallow test? A: Yes. Mischel's own research showed that teaching children to mentally reframe the marshmallow — thinking of it as a picture of a marshmallow, or as a fluffy cloud, rather than as a sweet, chewy treat — significantly extended waiting times. The strategies are real; the question is what the strategy improvements actually predict in the long run.
Q: Was the marshmallow test only done with marshmallows? A: No. The snack varied by child preference. Marshmallows, pretzels, and Oreos were all used. The "marshmallow test" name comes from the most commonly cited version of the experiment and from the image that entered popular culture, not from any scientific primacy of marshmallows over other snacks.
Q: Does the finding that socioeconomic background matters mean self-control is irrelevant? A: No. The 2018 study found that self-control still showed a modest predictive relationship even before controls, and the debate in the field is about magnitude and causation, not about whether self-regulation matters at all. The consensus is that it matters — but less dramatically than advertised, and with socioeconomic context doing a large share of the explanatory work.
The Takeaway
The marshmallow experiment produced real findings about how cognitive strategies affect the ability to wait for delayed rewards. What it did not produce was a destiny test. The predictive power that made the study famous turned out to depend heavily on the advantaged, homogeneous sample used in the original work. A broader, more representative sample reveals that family economic stability and environmental reliability explain much of what the waiting time appeared to predict. That does not make self-regulation irrelevant — the strategies Mischel identified are genuine, teachable, and useful. It means that a child grabbing the marshmallow may be demonstrating rational behavior shaped by experience, not a character defect. The experiment endures as a landmark in developmental psychology, but its legacy is better understood as a question about trust and environment than as a proof that self-control determines your future.
Read next:
Weekly digest
Stay ahead of the curve
Get the best science, AI, and tech stories delivered to your inbox every week. No spam — unsubscribe any time.
About the Author
SciHub101 is an independent science and technology publication founded and edited by Buğra Sözeri, a software engineer and app developer based in Türkiye. Articles are drafted against peer-reviewed literature and primary sources, fact-checked, and edited by a human before publication. Where we compare products, anything built by our own team is labelled as such — see the editorial guidelines and About page for how corrections and disclosures work.
Related Articles
The McGurk Effect: When Eyes Override Ears
The McGurk effect proves your brain merges sight and sound into a single perception. The 1976 Nature paper that revealed how speech is never purely auditory.
The Loftus Misinformation Effect: How Memory Gets Rewritten
Elizabeth Loftus's 1974 car-crash study showed a single word can alter memory. Here's how she proved it, what it means for eyewitness testimony, and why courts still struggle with the findings.
The Marshmallow Test Explained — Self Control
Walter Mischel's famous marshmallow test shaped how we think about self-control. New research shows the story is more complicated than willpower.
The Stanford Marshmallow Replication: What the 2018 Study Actually Found
In 2018, Watts, Duncan, and Quan retested the marshmallow experiment with 900 children. The self-control prediction largely collapsed once family background was controlled for.
The Bobo Doll Experiment: How Bandura Proved We Learn Aggression by Watching
In 1961, Albert Bandura showed that children copy adult aggression toward an inflatable doll. The Bobo doll experiment built social learning theory — and still shapes media violence debates.