What the Marshmallow Test Got Wrong
For fifty years, the marshmallow test has been parenting folklore: patient four-year-olds turn into successful adults. A 2018 replication with ten times more children found the famous link almost disappeared. What was left tells a different story about what self-control really is.
Most parents have heard of the marshmallow test. A four-year-old sits at a table. A researcher places a marshmallow in front of them and says: if you can wait fifteen minutes without eating it, you'll get a second one. Then the researcher leaves the room.
Some children eat the marshmallow immediately. Others sit on their hands, sing softly, look at the ceiling. A few make it the full fifteen minutes and earn the second treat.
The Stanford psychologist Walter Mischel began running these experiments in the late 1960s. He followed those children for decades. The original papers reported that the children who waited longer scored higher on standardised exams as teenagers, had healthier weights as adults, and coped better with stress. The story took off. Self-control at four predicts everything that comes after. Practise patience now and reap the rewards later.
It became one of the most repeated findings in psychology. It also turns out to be largely wrong.
When Watts looked at the basic correlation, did waiting longer predict better outcomes later, he did find one. But it was about half the size Mischel had reported. And when he controlled for the child's family income and the mother's education, the effect shrank further. By the time he had controlled for the home environment in the child's first three years of life, the famous predictive power of the marshmallow test had nearly vanished.
What was left was not self-control predicting success. It was background predicting both.
When the marshmallow test followed, the children who had just been let down waited an average of three minutes before eating the marshmallow. The children whose adult had kept the promise waited four times as long.
Children, in other words, were not just exhibiting fixed willpower. They were making a judgement. If experience tells you the adults around you can be trusted to deliver, you wait. If experience tells you the second marshmallow may never come, you eat the one in front of you. That is not a deficit. It is reasonable behaviour in an unpredictable world.
A child growing up in a household where money is short and food sometimes runs out is not a child with poor self-regulation. They are a child whose past has taught them that waiting often costs you the thing you already had.
A whole industry of grit and character curricula grew up around this idea, and many schools still use them. They teach children to count to ten, take deep breaths, and wait. There is nothing wrong with any of that. But it places the problem inside the child, when most of it lives outside the child.
Across multiple large samples, the household conditions a child grows up in turn out to do more of the predictive work than any trait we can measure in the child. Whether plans tend to be kept. Whether resources are stable. Whether the adults are warm. Children become patient when adults give them reasons to be.
That sits uncomfortably alongside one of education's favourite phrases, that we should be raising resilient children. Resilience is real. But it is built more from outside than from within. A four-year-old who waits for the second marshmallow is not flexing some inner muscle of restraint. They are reading the room.
When Mischel was asked in interviews near the end of his life whether he thought the test could predict an individual child's future, he said no. He was clear that the original sample was small and unusual, and that whatever the test was measuring was as much about a child's circumstances as their character. By then the story had escaped him.
The marshmallow lives on in TED talks and parenting books partly because it offers a satisfying narrative. There is a single moment, in childhood, that reveals the kind of adult someone will become. And there is something a parent can do, teach patience, to tilt the odds. Both halves of that promise turn out to be smaller than they sounded.
What we are left with is not nothing. Children do better when the adults around them are predictable and when waiting reliably pays off. Those are not character lessons. They are lessons about the world.
Some children eat the marshmallow immediately. Others sit on their hands, sing softly, look at the ceiling. A few make it the full fifteen minutes and earn the second treat.
The Stanford psychologist Walter Mischel began running these experiments in the late 1960s. He followed those children for decades. The original papers reported that the children who waited longer scored higher on standardised exams as teenagers, had healthier weights as adults, and coped better with stress. The story took off. Self-control at four predicts everything that comes after. Practise patience now and reap the rewards later.
It became one of the most repeated findings in psychology. It also turns out to be largely wrong.
What happened in 2018
In 2018, the developmental psychologist Tyler Watts ran a version of the marshmallow test with 918 children. The original Stanford study had used around ninety, mostly drawn from a campus preschool whose parents were academics. The new sample was more than ten times larger and far more diverse, including children from families across the income range and from a variety of ethnic backgrounds.When Watts looked at the basic correlation, did waiting longer predict better outcomes later, he did find one. But it was about half the size Mischel had reported. And when he controlled for the child's family income and the mother's education, the effect shrank further. By the time he had controlled for the home environment in the child's first three years of life, the famous predictive power of the marshmallow test had nearly vanished.
What was left was not self-control predicting success. It was background predicting both.
The rational marshmallow
The cleanest explanation came from a 2013 study at the University of Rochester. Before doing the marshmallow test, the children did a quick craft activity. An adult told them that more crayons and stickers were coming. For half the children, the adult kept the promise. For the other half, the adult came back empty-handed.When the marshmallow test followed, the children who had just been let down waited an average of three minutes before eating the marshmallow. The children whose adult had kept the promise waited four times as long.
Children, in other words, were not just exhibiting fixed willpower. They were making a judgement. If experience tells you the adults around you can be trusted to deliver, you wait. If experience tells you the second marshmallow may never come, you eat the one in front of you. That is not a deficit. It is reasonable behaviour in an unpredictable world.
A child growing up in a household where money is short and food sometimes runs out is not a child with poor self-regulation. They are a child whose past has taught them that waiting often costs you the thing you already had.
Why this matters at the school gate
The original marshmallow story was often used to explain why children from poorer families do worse at school. The reasoning went: they cannot delay gratification, so they cannot put in the hard work that schoolwork demands. The fix, supposedly, was to train them in patience.A whole industry of grit and character curricula grew up around this idea, and many schools still use them. They teach children to count to ten, take deep breaths, and wait. There is nothing wrong with any of that. But it places the problem inside the child, when most of it lives outside the child.
Across multiple large samples, the household conditions a child grows up in turn out to do more of the predictive work than any trait we can measure in the child. Whether plans tend to be kept. Whether resources are stable. Whether the adults are warm. Children become patient when adults give them reasons to be.
That sits uncomfortably alongside one of education's favourite phrases, that we should be raising resilient children. Resilience is real. But it is built more from outside than from within. A four-year-old who waits for the second marshmallow is not flexing some inner muscle of restraint. They are reading the room.
What the test was actually measuring
This does not mean self-control is a fiction. Children genuinely do vary in temperament, attention and impulse, and those differences matter for school. What the marshmallow test was probably picking up was something messier. A mix of temperament, trust, household predictability, language ability, and familiarity with the kind of testing format the experimenters used. Untangling those is hard. Crediting a single trait called willpower was tidier, which is probably why it stuck.When Mischel was asked in interviews near the end of his life whether he thought the test could predict an individual child's future, he said no. He was clear that the original sample was small and unusual, and that whatever the test was measuring was as much about a child's circumstances as their character. By then the story had escaped him.
The marshmallow lives on in TED talks and parenting books partly because it offers a satisfying narrative. There is a single moment, in childhood, that reveals the kind of adult someone will become. And there is something a parent can do, teach patience, to tilt the odds. Both halves of that promise turn out to be smaller than they sounded.
What we are left with is not nothing. Children do better when the adults around them are predictable and when waiting reliably pays off. Those are not character lessons. They are lessons about the world.
Sources & Further Reading