The child is not waiting on character

A treat lies in front of a preschool child. They may eat it right away — or wait until the researcher comes back and then receive two. The camera records averted eyes, hands held over faces, small improvised songs. Few psychology experiments have produced images as human as the marshmallow test.

Those images were later attached to a large claim: a four-year-old who can wait will be more successful, healthier and more dependable decades later. A situation became a character test; minutes became a forecast for a life. Memorable as that story is, it has little to do with the question the original experiments actually asked.

What Mischel actually studied

In the 1970s Walter Mischel and his collaborators — Ebbe Ebbesen, Antonette Raskoff Zeiss and others — were not asking whether children “possess willpower”. They varied what children could direct their attention to while waiting, and measured how long the children then held out. In the first of these studies, published in 1970, 32 preschoolers sat in front of the reward — sometimes visible, sometimes covered. When it was visible, waiting became markedly harder. Two years later the same group showed that more than the field of vision is at stake: children who thought about something pleasant during the wait held out longer than children who were prompted to think about the reward.

Two later experiments sharpened the point. In 1975 Mischel and Nancy Baker found that what matters is how a child handles the treat in its mind: children instructed to picture it as something inedible waited longer than children who dwelt on taste and texture. In 1976 Bert Moore, Mischel and Zeiss added a further turn — children held out longer when a picture of the reward lay in front of them instead of the reward itself. What decided the outcome was not the object but its mental representation.

This is a practical and still plausible result: delay of gratification is not merely gritted teeth. Situations can be arranged so that temptation holds less attention. Someone who leaves the phone in another room does not have to say a heroic “no” every minute.

It belongs to the ironies of the reception that Mischel of all people was enlisted as the star witness for strength of character. His 1968 book Personality and Assessment had startled the field with the opposite thesis: personality scores predicted behaviour across different situations only weakly, and the situations themselves explained most of it. The debate over situationism and traits that grew out of that argument shaped a decade. Mischel’s waiting room was meant as a contribution to it — as evidence that self-control can be produced by arranging the situation.

Where the life forecast came from

The famous prediction does not come from the experiments but from follow-up studies. In 1990 Yuichi Shoda, Mischel and Philip Peake published data on children from the Bing Nursery School on the Stanford campus — mostly the children of the university’s faculty and students. Between 1968 and 1974, 653 children there had taken part in at least one waiting experiment. By 1990, follow-up data were available for 185 of them and SAT scores for only 94. Longer waiting was associated with later test scores and parental ratings — but above all in those conditions where the reward was visible and the children were given no strategy. Where the experiment told children what to think, the association largely disappeared.

That alone is a qualification the popular version does not carry along. The second one is more fundamental: a correlation does not answer what produced the waiting. Language, family income, dependable routines and early cognitive skills can shape both the behaviour in the test room and the later educational path. Anyone who does not adjust for this confounding credits waiting with what belongs to background.

What the larger replication found

In 2018 Tyler Watts, Greg Duncan and Haonan Quan did exactly that. They analysed 918 children from a US longitudinal study, 552 of them children of mothers without a college degree — that subgroup alone was ten times the size of the 1990 sample. Their sample was geographically broader, though expressly not nationally representative. The result: one additional minute of waiting at age four predicted roughly 0.1 standard deviations of achievement at age 15. Their abstract puts it this way — the bivariate association was

Associations with behavioural measures at age 15 were small and rarely statistically meaningful.

In 2024 a group led by Jessica Sperber followed 702 people up to the age of 26. Without controls, weak associations appeared — around r = 0.17 for educational attainment and r = −0.17 for body mass index. Almost all regression-adjusted coefficients were not meaningful. A single waiting task in preschool did not reliably reveal an adult life.

This reading is not uncontested. In 2020 Laura Michaelson and Yuko Munakata reanalysed the same data set in a preregistered study, this time along the analytic path of the original studies rather than with the broad set of controls used by Watts and colleagues. For three of five adolescent outcomes they found associations: achievement (β = 0.27), less problem behaviour (β = −0.23) and social competence (β = 0.18). How much the marshmallow test predicts therefore also depends on what one controls for. The quarrel runs over analytic decisions, not only over data — and anyone who settles it by leaving one side out has not understood it.

That does not mean self-control is meaningless. Broader, repeated measures of self-regulation can predict development. It means that a brief task with a four-year-old is not a personality scan, and that statistical adjustment does not turn poverty into deficient willpower.

Why waiting is also a calculation

A child who waits is making a decision under uncertainty — and that decision depends on whether it believes the promise. In 2013 Celeste Kidd, Holly Palmeri and Richard Aslin ran a craft session with 28 children of average age four and a half before the waiting task. In one condition the researcher kept her promise to fetch better materials; in the other she came back empty-handed. Children who had just experienced reliability then waited markedly longer. A child who has learned that promises do not hold eats immediately — and in doing so acts not impulsively but sensibly.

The habits a child grows up with shift the result as well. In 2022 Kaichi Yanaoka and colleagues tested in a preregistered study what children in Japan and in the United States are able to wait for. The 80 Japanese children held out longer with food, the 58 American children with a gift. In Japan, waiting before eating is the practised rule; in the United States, waiting before unwrapping. What decided the outcome was not a general inner muscle but what the child had rehearsed a hundred times.

That also names the measurement problem. A trait such as self-regulation is meant to denote a reasonably regular pattern across different situations. A single waiting time, by contrast, is one data point that contains hunger, mood on the day, comprehension of the rule, the appeal of the reward, trust and cultural habit all at once. The same score can therefore arise for very different reasons. A robust assessment would require several tasks, observations from everyday life and reports from different caregivers. Even then the result would not be a number standing for a destiny, but an uncertain profile with room to change.

What remains

What the marshmallow test shows is well documented and smaller than its reputation promises: attention can be steered, and children who steer it wait longer. What it does not show is a character score that fixes a life. The forecast drawn from a single minute of waiting shrinks as soon as background and early ability are counted in — how far it shrinks is the real quarrel between the research groups, and it is unsettled. What is not in dispute is that the Stanford sample was small and socially unusual, and that the follow-up reached only a fraction of the children originally tested.

In everyday life the fallacy can be recognised by a turn of phrase. To say that a child “cannot wait” is to turn a behaviour into a trait and leave out the circumstances: whether promises have held so far, whether scarcity has been experienced, whether the child knows a technique, whether it has ever practised this kind of waiting at all. For education, nothing in the research recommends testing children with sweets. More useful is to keep promises, to make waiting comprehensible, to practise strategies of distraction, and not to read material insecurity as a flaw of character. The child who looks away has found a technique. The child who eats has not given away its future.