Mischel's marshmallow test explained

Few psychology studies have captured the public imagination as completely as Walter Mischel's marshmallow test. In the late 1960s and early 1970s at Stanford's Bing Nursery School, Mischel and colleagues placed young children in front of a single marshmallow (or other preferred treat) and offered them a choice: eat it now, or wait for the researcher to return and receive two marshmallows. Most four-year-olds struggled. Some grabbed the treat within seconds; others used elaborate strategies — covering their eyes, singing to themselves, turning away — to delay. The longest waiters held out for up to 15 minutes.

01

The basic paradigm

A child is left alone with one marshmallow and told they can eat it now or wait for the researcher to return and receive two. Waiting time (up to 15–20 minutes) is the primary measure.

02

Mischel's strategy research

Rather than treating self-control as a fixed trait, Mischel's later work focused on cognitive strategies — particularly 'cool' abstract mental representations versus 'hot' vivid sensory ones — that children deploy to delay successfully.

03

Long-term outcome correlations

In the original Stanford follow-up studies, longer delay correlated with higher SAT scores, better social functioning, and lower BMI in adolescence — generating enormous popular interest.

04

Replication narrows the story

A 2018 pre-registered replication with 900+ participants found that the correlations between delay time and later outcomes largely disappear when family background and cognitive ability are controlled — suggesting earlier results were partly an artefact of sample composition.

Walter Mischel

1930–2018

Austrian-born American psychologist at Stanford and Columbia who designed the marshmallow test and conducted the foundational delay of gratification studies. His later theoretical work on hot-cool self-regulation and 'if-then' personality signatures moved the field away from static trait models toward dynamic, context-sensitive accounts of behaviour.

Celeste Kidd

1981–present

Cognitive scientist at Berkeley who showed that children's marshmallow waiting time is sensitive to their experience of adult reliability — making apparent impulsivity partly a rational inference about environmental trustworthiness rather than a failure of self-control.

Tyler Watts

1985–present

Educational researcher at NYU who led the 2018 pre-registered replication of the marshmallow test with a large nationally representative sample, finding substantially smaller and largely SES-explained correlations — triggering the reappraisal of the original findings' generalisability.

How the marshmallow test works

Participants

Original studies: children aged 3–5 at Stanford's Bing Nursery School — predominantly white, middle-class children of Stanford faculty and graduate students. This sample homogeneity became a central criticism when follow-up results were extrapolated to broader populations.

The delay of gratification procedure

A researcher shows the child a preferred reward (marshmallow, pretzel, or cookie) and a larger reward (two marshmallows). The researcher says they have to leave the room and offers the child a choice: eat the single reward now by ringing a bell to summon the researcher back, or wait until the researcher returns on their own to receive two. Waiting time (up to 15–20 minutes) and strategies used are recorded.

Mischel's 'hot-cool' manipulation

Mischel's experimental focus was on cognitive strategies that extend waiting time. He found that telling children to think of the marshmallow in abstract terms ('think of it as a cloud') or covering the treat increased delay time dramatically, while thinking of it in vivid sensory terms ('imagine how chewy and sweet it is') reduced waiting. This became the theoretical centrepiece of the 'hot-cool' framework of self-regulation.

Follow-up measurements

Mischel and colleagues followed up the original Bing cohort in adolescence and early adulthood, using parent questionnaires and SAT scores to assess outcomes. Later, brain imaging studies compared high- versus low-delay individuals in adulthood on neural measures of self-control.

What Mischel found

Large individual differences in delay time

Children showed striking variation: some ate the marshmallow within seconds; others waited the full 15–20 minutes using a range of self-distraction strategies. Average waiting time was approximately 6–7 minutes, with substantial variability.

Cognitive strategies determine success

The most successful delayers were not those who stared at the marshmallow and resisted; they were those who cognitively transformed the reward — thinking about it abstractly, focusing on its non-consummatory features, or distracting themselves entirely. Mischel argued this showed self-control is a skill, not merely a trait.

Longitudinal correlations (original Stanford cohort)

Children who waited longer at age 4 had significantly higher SAT scores (~200 points higher for the longest vs. shortest waiters), higher parental ratings of social competence, better ability to cope with frustration and stress, and lower BMI in their mid-30s. These correlations were widely reported as evidence that delay of gratification is a stable, predictive trait.

Hot-cool framework

Mischel's theoretical contribution was the hot-cool framework: the 'hot' emotional-motivational system drives impulsive responses toward rewarding stimuli; the 'cool' cognitive system supports strategic delay. Self-control depends on which system dominates in a given context — and can be trained by practising cool, abstract thinking strategies.

Challenges and limitations

Replication failure: sample and SES confounds

Tyler Watts, Greg Duncan, and Haonan Quan (2018) pre-registered a replication with 918 participants from a nationally representative US birth cohort. They found that correlations between delay time at age 4.5 and outcomes at age 15 were much smaller than Mischel's original findings, and largely disappeared when controlling for family background, maternal education, and early cognitive ability. The implication: marshmallow waiting time may primarily reflect background advantages, not a stable self-control trait.

The rationality of not waiting

Celeste Kidd and colleagues (2013) showed that children's decision to eat the marshmallow or wait is partly rational given their experience of adult reliability. Children who experienced an adult as unreliable (in a prior experimental manipulation) waited significantly less than children who experienced the adult as reliable. If you live in an environment where promised rewards often fail to materialise, eating now is the adaptive choice — not a failure of self-control.

Homogeneous original sample

The Bing Nursery School sample was unrepresentative: predominantly white, middle-class children of Stanford affiliates. The strong longitudinal correlations found in this sample may not generalise to more diverse populations with different relationships to delayed reward, institutional trust, and material scarcity.

Measurement of self-control

The marshmallow test is a single measure, taken on one occasion, of one type of delay of gratification. It may not capture the broader construct of self-regulation across contexts (emotional regulation, effortful attention, social self-control). Whether it is the delay of gratification specifically, or underlying cognitive ability or temperament, that predicts later outcomes remains unclear.

Why it still matters

Self-regulation as a trainable cognitive skill

Mischel's theoretical emphasis on cognitive strategies — rather than a fixed capacity for willpower — transformed how psychologists conceptualise self-regulation. The hot-cool framework showed that the ability to delay gratification depends on how a situation is mentally represented, not on a fixed trait. This reframing opened the field to studying self-regulation as a learnable, context-sensitive skill, influencing research in developmental, clinical, and personality psychology.

Structural vs. individual explanations of outcomes

The replication debate productively shifted attention from individual self-control differences to the structural and material conditions that constrain or enable delayed gratification. Policy implications shifted accordingly: rather than targeting children's self-control skills, interventions that reduce family poverty and housing instability may more effectively improve life outcomes.

The hot-cool framework in clinical psychology

Mischel's hot-cool framework — and the distinction between reflexive, emotion-driven responding and reflective, strategy-driven responding — influenced cognitive-behavioural and mindfulness-based treatments for impulsivity, addiction, and emotional dysregulation. The idea that self-control is a skill that can be improved through strategic practice underpins many therapeutic approaches.

Ego depletion and the resource model debated

The marshmallow test contributed to the broader culture of 'willpower' research, including Baumeister's ego depletion model — the claim that self-control depletes a limited resource. Large-scale meta-analyses and pre-registered replications have raised serious doubts about ego depletion, further contextualising the limits of the original marshmallow narrative.

What is the marshmallow test?+

A 4–5-year-old child is offered a single marshmallow (or preferred treat) and told they can eat it immediately, or wait until the researcher returns (approximately 15 minutes) and receive two marshmallows. Waiting time is recorded. Mischel found that children who waited longer in the late 1960s–70s Stanford studies had better outcomes on measures like SAT scores and social competence in adolescence — though large replication studies have found these correlations are much weaker and largely explained by family socioeconomic background.

Does the marshmallow test really predict future success?+

The original Stanford findings suggested it did — but the picture is more complicated. A 2018 pre-registered replication with 900+ participants from a nationally representative US sample found that the predictive correlations largely disappear once family background, maternal education, and early cognitive ability are statistically controlled. The test may primarily reflect the advantages of the middle-class, predominantly white Stanford sample rather than a stable individual trait. Mischel's own emphasis was on teachable strategies, not fixed willpower.

Last reviewed August 2026
  1. 1.

    Mischel W., Shoda Y., & Rodriguez M.I. (1989). Delay of gratification in children. Science, 244(4907), 933–938. https://doi.org/10.1126/science.2658056

    +About this source

    Report of the longitudinal follow-up findings linking preschool delay time to adolescent outcomes.

  2. 2.

    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. https://doi.org/10.1177/0956797618761661

    +About this source

    Pre-registered conceptual replication finding substantially attenuated correlations that largely disappear when controlling for family background.

  3. 3.

    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. https://doi.org/10.1016/j.cognition.2012.08.004

    +About this source

    Shows that marshmallow waiting time reflects rational inference about adult reliability, not only self-control capacity.