The Stanford Marshmallow Test is a famous psychology experiment from the 1970s where children were offered one marshmallow now, or two if they waited. It tested self-control and delayed gratification. Early studies suggested that kids who could wait tended to have better life outcomes, but later research showed the results were also shaped by factors like family background and environment.
Key Takeaways
- Experiment: The Stanford Marshmallow Test asked children to choose between one treat now or two treats if they waited, measuring their ability to delay gratification. It became one of psychology’s most famous studies on self-control.
- Purpose: The experiment aimed to explore how early self-control skills might relate to future success in areas like academics, health, and relationships. Researchers followed participants over decades to see long-term patterns.
- Findings: Initial results suggested that children who could wait tended to have better life outcomes later on. These outcomes included higher test scores, healthier lifestyles, and stronger social skills.
- Limitations: Later research showed that family background, environment, and economic stability strongly influenced waiting times and life outcomes. This challenged the idea that self-control alone drives success.
- Relevance: The study continues to shape conversations about willpower, motivation, and child development. It is often used in education and popular culture to illustrate the value—and complexity—of self-control.
In a 1970 paper, Walter Mischel, a Stanford psychology professor, worked with his graduate student Ebbe Ebbesen. They studied preschoolers waiting up to 15 minutes for a preferred treat, a pretzel or a marshmallow.
Children waited much less time when either treat was within sight than when neither treat was in view.
Children with treats present waited 3.09 ± 5.59 minutes; children with neither treat present waited 8.90 ± 5.26 minutes.
The study suggested that gratification delays in children involved suppressing rather than enhancing attention to expected rewards.
Some children who waited with both treats in sight used tricks to avoid fixating on the pretzel or marshmallow. Not looking helped. They stared at a mirror, covered their eyes, or talked to themselves instead.
Mischel, Ebbesen, and Antonette Zeiss, a visiting faculty member at the time, set out to investigate whether attending to rewards cognitively made it more difficult for children to delay gratification.
Stanford Marshmallow Experiments
Mischel, Ebbesen and Zeiss (1972) designed three experiments to investigate, respectively, the effect of overt activities, cognitive activities, and the lack of either, in the preschoolers’ gratification delay times.
Experiment 1
Fifty-six children from the Bing Nursery School at Stanford University were recruited. To build rapport with the preschoolers, two experimenters spent a few days playing with them at the nursery.
Design
Children were randomly assigned to one of five groups (A – E).
The children were individually escorted to a room where the test would take place. Each child was taught to ring a bell to signal for the experimenter to return to the room if they ever stepped out.
Treat vs. No Treats Condition
Children in groups A, B, and C were shown two treats (a marshmallow and a pretzel) and asked to choose their favorite.
They were then told that the experimenter would soon leave for a while. If they waited for the experimenter to return, without signaling, they would get their preferred treat.
Children could also signal to end the wait. But doing so meant getting only the treat they had not chosen as their favorite. Both treats stayed in plain view throughout.
Children in groups D and E were given no such choice or instructions.
Children in groups A, B, or C who waited the full 15 minutes were allowed to eat their favored treat. Those in groups A, B, or C who didn’t wait 15 minutes were allowed to have only their non-favored treat.
Children in groups D and E weren’t given treats. All children got to play with toys with the experiments after waiting the full 15 minutes or after signaling.
Distraction vs. No Entertainment Condition
Children in groups A and D were given a slinky and were told they had permission to play with it.
Children in groups B and E were asked to “think of anything that’s fun to think of.” They were given examples: singing songs and playing with toys.
Controls
Each child’s comprehension of the instructions was tested. Six children didn’t seem to comprehend and were excluded from the test. The remaining 50 children were included.
All 50 were told that whether or not they rang the bell, the experimenter would return, and when he did, they would play with toys.
Waiting time was scored from the moment the experimenter shut the door. The experimenter returned either as soon as the child signaled or after 15 minutes if the child did not signal.
Results
Children in the rewarded groups (A, B, and C) waited much longer than children in the other groups, D and E, who got nothing.
Group A, given a slinky, and group B, told to think of fun things, waited much longer for their treats than group C did. Group C had nothing to distract them, so they fixated on the treats instead.
Experiment 2
This test differed from the first only in the following ways:
- Thirty-eight children were recruited, with six lost due to incomplete comprehension of instructions.
- Thirty-two children were randomly assigned to three groups (A, B, C).
- All children were given a choice of treats, and told they could wait without signalling to have their favourite treat, or simply signal to have the other treat but forfeit their favoured one.
- In all cases, both treats were left in plain view.
- Children in group A were asked to think of fun things, as before.
- Those in group B were asked to think of sad things, and likewise given examples of such things.
- Those in group C were asked to think of the treats.
Results
Children given a distracting and fun task (thinking of fun things, group A) waited much longer for their treats than children given other tasks. Group C, asked to think of the treats, was not distracted. Group B, asked to think of sad things, was distracted but not entertained.
Experiment 3
This test differed from the first only in the following ways:
- Sixteen children were recruited, and none excluded.
- Children were randomly assigned to three groups (A, B, C),
- All children were given a choice of treats, and told they could wait without signalling to have their favourite treat, or simply signal to have the other treat but forfeit their favoured one.
- In all cases, both treats were obscured from the children with a tin cake cover (which children were told would keep the treats fresh).
- Children in group A were asked to think about the treats.
- Those in group B were asked to think of fun things, as before.
- Those in group C were given no task at all.
Results
When treats were obscured, by a cake tin, group C and group B waited about the same length of time. Group C had no distracting task. Group B had a distracting, fun one.
Group A was told to think about the treats, so the task did not distract them. For this group, obscuring the treats made no difference to delay time compared with the second test, where the treats stayed visible.
Final Conclusions: What Helps Children Delay Gratification
The studies convinced Mischel, Ebbesen, and Zeiss that successful delay depended on cognitively avoiding the expected treats during the wait. Children managed this by keeping the treats out of sight, or by thinking of fun things instead.
Children, they reasoned, could wait a relatively long time if they –
- Believed they really would get their favoured treat if they waited (eg by trusting the experimenter, by having the treats remain in the room, whether obscured or in plain view).
- Shifted their attention away from the treats.
- Occupied themselves with non-frustrating or pleasant internal or external stimuli (eg thinking of fun things, playing with toys).
The Hot/Cool System: Why These Strategies Work
Metcalfe and Mischel (1999) later explained why attention and mental representation matter so much for delay of gratification.
They proposed that behavior is guided by two interacting systems. A “hot” emotional system is fast and reflexive, and it generates the urge to grab a tempting reward immediately.
A “cool” cognitive system is slower and more reflective. It supports planning, abstract thinking, and self-control.
Successful delay happens when a child shifts control from the hot system to the cool system.
This reframes each strategy from the earlier experiments as a cool-system tactic. Covering or hiding the treat removes the hot trigger. Reframing a marshmallow as a “round white cloud” engages abstract, cool representation. Playing a distracting game keeps the cool system in charge.
On this account, delaying gratification is not a fixed trait of “strong-willed” children. It is a set of learnable cognitive skills for calming a hot impulse, with direct implications for teaching self-regulation.
This attentional account differs from resource-based models like ego depletion, which explain self-control as a limited store of willpower that can be used up.
The two accounts make different predictions. The hot/cool model says that changing what a child attends to, or how they picture the reward, should change waiting time immediately. No prior act of self-control is needed.
A pure resource model instead predicts that waiting should shorten mainly after an earlier act of self-control has used up the resource.
The resource model has also run into trouble. A large, multi-lab, pre-registered replication found the ego-depletion effect close to zero (Hagger et al., 2016). Mischel’s own attentional manipulations, tested and retested since the 1970s, have not faced a comparable replication crisis at the level of basic mechanism.
Critical Evaluation
- Undisclosed Sample Size: The researchers did not report how they determined their sample size.
- Unrepresentative Population: Stanford’s Bing Nursery School, whose population the original paper did not characterize, was mostly attended by middle-class children of Stanford faculty and alumni, so findings may not generalize to the wider population.
- Involuntary Delay Only: The task only tested delay when the immediate reward was unavailable, so findings might not extend to voluntary delay of gratification.
- Socio-Economic Confounds: Watts, Duncan, and Quan (2018) showed that family income and home environment predict both childhood delay and later achievement, so the original correlations were substantially confounded with affluence rather than reflecting delay ability alone.
- Deterministic Misreadings: The study is often mischaracterized as showing that one childhood trait fixes a person’s fate. This popular reading ignores the modest effect sizes and the confounds above, and wrongly implies that success is decided at age four.
Taken together, the original experiments and the modern evidence point in different directions. The controlled manipulations of reward visibility and distraction give strong, repeatable evidence that attention and mental representation drive delay: that mechanism finding still holds up well.
The longitudinal predictions built on top of it are weaker. Watts et al.’s (2018) large, pre-registered replication found the delay-achievement link much weaker, and largely gone once family background was controlled.
The test’s lasting value lies in explaining attention control, not in predicting life success.
Contemporary Research
Since 2015, the strongest evidence has shifted from celebrating the marshmallow test’s predictive power to re-examining what it actually measures. The large, pre-registered replication by Watts, Duncan, and Quan (2018), described above, found that early delay ability barely predicts later achievement once family background is taken into account.
The predictive claim shrank. A separate line of research recasts waiting itself as a rational judgement about the world rather than a fixed store of willpower.
Kidd, Palmeri, and Aslin (2013) showed that children calibrate how long they wait to whether their environment has proven reliable. McGuire and Kable (2012) found the same pattern in adults.
What survives this reappraisal is Mischel’s original mechanistic finding: attention and how a reward is mentally represented govern self-control, and these are teachable skills. What does not survive is the idea that a single childhood measure can predict life success regardless of context.
Longitudinal Studies Using Stanford Data
Delayed Gratification and SAT Scores
In 1990, Yuichi Shoda and Philip Peake, graduate students at Columbia University and Smith College respectively, worked with Walter Mischel, then a professor at Columbia. Together they examined the relationship between preschoolers’ delay of gratification and their later SAT scores.
Design
Six-hundred and fifty-three preschoolers at the Bing School at Stanford University participated at least once in a series of gratification delay studies between 1968 and 1974.
Four hundred and four of their parents received follow-up questionnaires. One hundred and eighty-five responded. Ninety-four parents supplied their children’s SAT scores.
Children were divided into four groups. The groups differed on whether a cognitive activity, like thinking of fun things, was suggested before the delay, and whether the treats stayed in sight throughout it.
Control
Children of parents who responded to the survey were compared with children of parents who didn’t. Their mean waiting times did not differ significantly (p = 0.09, n = 653).
Results
Preschoolers’ delay times correlated positively and significantly with their later SAT scores when no cognitive task had been suggested and the expected treats had remained in plain sight.
Other correlations were not significant.
Limitations
Shoda, Mischel, and Peake (1990) urged caution in extrapolating their findings since their samples were uncomfortably small.
Delayed Gratification and Positive Functioning
In a 2000 paper, Ozlem Ayduk, then a postdoctoral researcher at Columbia, and colleagues explored a key question. Did preschoolers’ ability to delay gratification predict their later self-worth, self-esteem, and ability to cope with stress?
Design
Five-hundred and fifty preschoolers’ ability to delay gratification in Prof. Mischel’s Stanford studies between 1968 and 1974 was scored.
Each preschooler’s delay score compared their own result with their experimental group‘s average delay time. It was the difference between the two.
Between 1993 and 1995, researchers mailed questionnaires to 444 parents of the original preschoolers, for themselves and their now adult-aged children. A hundred and eighty-seven parents and 152 children returned them.
The questionnaires used nine-point Likert-scale items to measure self-worth, self-esteem, and ability to cope with stress. Scores were standardized into a single positive functioning composite.
Results
The positive functioning composite, derived either from self-ratings or parental ratings, was found to correlate positively with delay of gratification scores.
Preschoolers who delayed gratification better went on to show higher self-worth and self-esteem in adulthood. They also coped better with stress than preschoolers who were less able to delay gratification.
Delayed Gratification and Body Mass Index
In a 2013 paper, Tanya Schlam, a doctoral student at the University of Wisconsin, and colleagues explored a possible link. Could preschoolers’ ability to delay gratification predict their later Body Mass Index?
Design
Prof. Mischel’s data were again used. Of the 653 preschoolers in his studies, researchers had valid addresses for 306. They were mailed in December 2002, January 2003, and again in May 2004.
Of these, 146 individuals responded with their weight and height. Individual delay scores were derived as in the 2000 Study.
Results
Preschoolers’ ability to delay gratification accounted for a significant portion of the variance seen in the sample (p < 0.01, n = 146).
Specifically, each additional minute a preschooler delayed gratification predicted a 0.2-point reduction in BMI in adulthood.
Delayed Gratification and Adult Brain Function
In a 2011 paper, B.J. Casey and colleagues traced a subset of the original Bing preschoolers into their forties. They asked whether early delay ability still predicted self-control decades later, and what its neural signature was.
Design
Fifty-nine former participants, now in midlife, completed a go/no-go task that required them to respond to some cues and withhold responses to others.
Twenty-six of these adults performed a version using emotionally “hot” social cues, happy and fearful faces, during functional magnetic resonance imaging, or fMRI. This brain-scanning technique shows which brain regions are active in the moment.
Results
Adults who had been low delayers as preschoolers made more errors of commission. They failed to suppress responses to the alluring “hot” cues.
This showed that a difficulty inhibiting responses to attractive stimuli persisted across four decades.
The imaging added a neural basis to the hot/cool account. Resisting the tempting cues recruited greater activity in the right inferior frontal gyrus, a prefrontal “cool” region that supports inhibitory control, in the high delayers.
Low delayers instead showed heightened activity in the ventral striatum, a “hot”, reward-valuing region, when confronted with the alluring cues.
Conclusion
Individual differences in delay ability reflect, in part, the relative dominance of a prefrontal control system over a striatal reward system. These differences are partly stable from preschool into adulthood.
Marshmallow Test Replication Study
In a 2018 paper, Tyler Watts, an assistant professor and postdoctoral researcher at New York University, worked with Greg Duncan and Haonan Quan, doctoral students at UC Irvine. Together they set out to replicate longitudinal studies based on Prof. Mischel’s data.
The study used data on 918 individuals. These data came from a longitudinal, multi-center study on children run by the National Institute of Child Health and Human Development, part of the NIH.
Sample
The sample was split into two groups –
- Data on children of mothers who had not completed university college by the time their child was one month old (n = 552);
- Data on children of mothers who had completed university college by that time (n = 366).
The first group (children of mothers without degrees) was more comparable to a nationally representative sample (from the Early Childhood Longitudinal Survey—Kindergarten by the National Center for Education Statistics). Even so, Hispanic children were underrepresented in the sample.
Design
A variant of the marshmallow test was administered to children when they were 4.5 years old. An interviewer presented each child with treats based on the child’s own preferences.
Children were then told they would play the following game with the interviewer –
- The interviewer would leave the child alone with the treat;
- If the child waited 7 minutes, the interviewer would return, and the child would then be able to eat the treat plus an additional portion as a reward for waiting;
- If the child did not want to wait, they could ring a bell to signal the interviewer to return early, and the child would then be able to eat the treat without an additional portion.
Delay of gratification was recorded as the number of minutes the child waited.
Academic achievement was measured at grade 1 and age 15. Measures included mathematical problem solving, word recognition and vocabulary (only in grade 1), and textual passage comprehension (only at age 15). Scores were normalized to have a mean of 100 ± 15 points.
Behavioral functioning was measured at age 4.5, grade 1, and age 15. Mothers were asked to score their child’s depressive and anti-social behaviors on 3-point Likert-scale items.
Controls
For intra-group regression analyses, the following socio-economic variables, measured at or before age 4.5, were controlled for –
- Demographic characteristics like gender, race, birth weight, mother’s age at child’s birth, mother’s level of education, family income, mother’s score in a measure-of-intelligence test;
- Cognitive functioning characteristics like sensory-perceptual abilities, memory, problem-solving, verbal communication skills; and
- Home environment characteristics are known to support positive cognitive, emotional, and behavioral functioning (the HOME Inventory by Bradley & Caldwell, 1984).
Results
- Watts, Duncan, and Quan (2018) did find statistically significant correlations between early-stage ability to delay gratification and later-stage academic achievement, but the association was weaker than that found by researchers using Prof. Mischel’s data.
- In addition, the significance of these bivariate associations disappeared after controlling for socio-economic and cognitive variables.
- There were no statistically significant associations, even without controlling for confounding variables, between early gratification delay and later behavioral functioning at age 15.
Conclusions
These results further complicated the relationship between early delay ability and later life outcomes.
Prof. Mischel’s findings, from a small, non-representative cohort of mostly middle-class preschoolers at Stanford’s Bing Nursery School, were not replicated in a larger, more representative sample of preschool-aged children.
Increasing Delayed Gratification
The following factor has been found to increase a child’s gratification delay time –
Trust in rewarders:
Trust matters. Children who believe they will be rewarded for waiting are significantly more likely to wait than those who don’t.
Kidd, Palmeri, and Aslin (2013) tested this by replicating Prof. Mischel’s marshmallow study with 28 four-year-olds, tested twice. In the first test, half went without their promised treat.
In the second test, children who had experienced this unreliable experimenter waited a median of only about 3 minutes for the second marshmallow. Children who had experienced a reliable experimenter waited far longer: a median of about 12 minutes.
Waiting, in other words, was partly a rational response to how trustworthy the situation seemed.
The following factors may increase an adult’s gratification delay time –
Knowledge of time-to-reward:
Individuals who know how long they must wait for an expected reward are more likely to continue waiting for said reward than those who don’t.
McGuire and Kable (2012) tested 40 adult participants. One group was given known reward times, while the other was not. The first group was significantly more likely to delay gratification.
This calibration to experience is the same rational strategy children showed above: both groups adjust how long they persist based on what they have learned to expect.
Probability of the expected reward materializing:
When the individuals delaying their own gratification are also the ones creating their own reward, the size of that reward matters.
Consider someone dieting to reach a goal weight. Those who set realistic rewards are more likely to keep waiting than those who set unrealistic or improbable ones.
Gelinas et al. (2013) studied this in 219 adults awaiting weight-loss surgery, looking at the link between unrealistic weight-loss expectations and weight gain beforehand.
The correlation was statistically significant for male participants (r = 0.377, p < 0.008, n = 53) but not for female participants (n = 166).
Applications
The hot/cool account gives teachers concrete tools. Reduce tempting cues where possible, and teach children to look away, reframe the temptation, or break a hard task into smaller steps.
These are trainable skills, not fixed traits. Social-and-emotional-learning programmes build on exactly this idea.
The same attentional logic extends to emotion regulation. Ayduk and colleagues (2000) applied it to rejection sensitivity, the anxious expectation of being rejected, and found that strategically shifting attention buffers people against its social costs. Related reasoning informs treatments for impulse-control difficulties, addiction, and disordered eating.
That is real, useful science.
The test also gets misused. Because early delay appeared to predict SAT scores, BMI, and adult adjustment, it was widely used to argue that teaching willpower early could improve life outcomes across the board.
That argument leaps from a correlation in a small, privileged sample to a sweeping policy claim. It also neglects the family and environmental conditions that shape both childhood delay and later outcomes (Watts et al., 2018).
FAQs
What is the marshmallow test?
The Marshmallow Test is a psychological experiment conducted by Walter Mischel in the 1960s. In this study, a child was offered a choice between one small reward (like a marshmallow) immediately or two small rewards if they waited for a short period, usually 15 minutes, during which the tester left the room.
What does the marshmallow test measure?
The marshmallow test measures a child’s ability to delay gratification by offering them a choice to eat a marshmallow immediately or wait for a reward for an additional marshmallow after a short period.
It assesses self-control, impulse control, and the capacity to delay instant gratification, which is connected to future success and self-regulation skills.
References
Ayduk, O., Mendoza-Denton, R., Mischel, W., Downey, G., Peake, P. K., & Rodriguez, M. (2000). Regulating the interpersonal self: strategic self-regulation for coping with rejection sensitivity. Journal of Personality and Social Psychology, 79 (5), 776.
Bradley, R. H., & Caldwell, B. M. (1984). The HOME Inventory and family demographics. Developmental psychology, 20 (2), 315.
Casey, B. J., Somerville, L. H., Gotlib, I. H., Ayduk, O., Franklin, N. T., Askren, M. K., Jonides, J., Berman, M. G., Wilson, N. L., Teslovich, T., Glover, G., Zayas, V., Mischel, W., & Shoda, Y. (2011). Behavioral and neural correlates of delay of gratification 40 years later. Proceedings of the National Academy of Sciences, 108 (36), 14998-15003.
Gelinas, B. L., Delparte, C. A., Hart, R., & Wright, K. D. (2013). Unrealistic weight loss goals and expectations among bariatric surgery candidates: the impact on pre-and postsurgical weight outcomes. Bariatric Surgical Practice and Patient Care, 8 (1), 12-17. https://doi.org/10.1089/bari.2013.9999
Hagger, M. S., Chatzisarantis, N. L. D., Alberts, H., Anggono, C. O., Batailler, C., Birt, A. R., et al. (2016). A multilab preregistered replication of the ego-depletion effect. Perspectives on Psychological Science, 11 (4), 546-573. https://doi.org/10.1177/1745691616652873
Kidd, C., Palmeri, H., & Aslin, R. N. (2013). Rational snacking: Young children’s decision-making on the marshmallow task is moderated by beliefs about environmental reliability. Cognition, 126 (1), 109-114.
McGuire, J. T., & Kable, J. W. (2012). Decision makers calibrate behavioral persistence on the basis of time-interval experience. Cognition, 124 (2), 216-226.
Metcalfe, J., & Mischel, W. (1999). A hot/cool-system analysis of delay of gratification: Dynamics of willpower. Psychological Review, 106 (1), 3-19.
Mischel, W., & Ebbesen, E. B. (1970). Attention in delay of gratification. Journal of Personality and Social Psychology, 16 (2), 329.
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.
Schlam, T. R., Wilson, N. L., Shoda, Y., Mischel, W., & Ayduk, O. (2013). Preschoolers” delay of gratification predicts their body mass 30 years later. The Journal of pediatrics, 162 (1), 90-93.
Shoda, Y., Mischel, W., & Peake, P. K. (1990). Predicting adolescent cognitive and self-regulatory competencies from preschool delay of gratification: Identifying diagnostic conditions. Developmental psychology, 26 (6), 978.
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.
Keep Learning
- Cohort Effects in Children’s Delay of Gratification (Carlson et al., 2018)
- Predicting adolescent cognitive and self-regulatory competencies from preschool delay of gratification: Identifying diagnostic conditions
- Delay of Gratification as Reputation Management
