Does the Marshmallow Test Really Predict Success?

The setup is simple enough to run in a kitchen. A four-year-old sits at a table with a treat on a plate: a marshmallow, a pretzel, a biscuit. An adult explains that they need to leave the room for a while. If the child waits until they return, they get a second treat. If not, they can signal and eat the one on the plate, but that is all they get.

Walter Mischel and his colleagues ran versions of this task with children at the Bing preschool on the Stanford University campus between the late 1960s and early 1970s. A 1972 paper by Mischel, Ebbesen and Zeiss focused on what went on in children's minds while they waited, and how the way they directed their attention shaped how long they could hold out.

The fame came later. Follow-up studies reported that children who had waited longer at four grew into adolescents with higher SAT scores, and whose parents described them as more self-confident and socially skilled. Mischel and colleagues described delay as 'an early indicator of an apparently long-term personal quality'. Popular books and talks turned that into a simple story: self-control at four predicts success in life.

The problem with the original follow-ups

The original programme tested more than 600 children, but the famous follow-ups relied on much smaller groups, between 35 and 89 children depending on the analysis, all from a university community that was far from typical.

In 2018, Tyler Watts, Greg Duncan and Haonan Quan revisited the question with a large American study that had followed children from birth at ten sites across the country. They had delay-of-gratification data at age four and a half, plus outcomes at 15, for 918 children, and focused mainly on the 552 whose mothers had not completed college. In this version, the maximum wait was seven minutes.

An extra minute of waiting at four predicted about one tenth of a standard deviation higher achievement at 15. That association was only half the size of the one in the original studies, and it shrank by two thirds once the researchers accounted for family background, early cognitive ability and the home environment. Most of the difference came from whether children could wait at least 20 seconds. Links with behaviour at 15 were much smaller and rarely statistically significant.

That is not the end of the argument. Armin Falk, Fabian Kosse and Pia Pinger replied that the seven-minute cap pushes correlations towards zero, and that controlling for things such as early cognitive ability may remove part of what self-control itself does. Watts and Duncan answered them in turn. The fairest summary is that the link is real but smaller, and more tangled up with family circumstances, than the famous version suggested.

Even in the original sample the picture has faded. In 2020, Daniel Benjamin and colleagues surveyed 113 of the Bing children in their late forties. Their preschool waiting time did not predict eleven mid-life outcomes such as net worth, debt and education, with an average correlation of 0.02. A combined index of self-regulation measured at 17, 27 and 37 did predict them, with an average correlation of 0.19.

Waiting can be a rational choice

In 2013, Celeste Kidd, Holly Palmeri and Richard Aslin added a twist. Before the marshmallow, 28 children with an average age of four and a half did an art project. A researcher offered them a set of well-used crayons, and said that if they waited the researcher would bring better art supplies. Later came the same offer with a small sticker and a promise of better ones.

For half of the children, the researcher came back without the promised supplies both times. For the other half, the promises were kept. Then came the standard marshmallow task, with a maximum wait of 15 minutes.

Children in the unreliable condition waited on average 3 minutes and 2 seconds. Children in the reliable condition waited 12 minutes and 2 seconds. Only 1 of the 14 children in the unreliable group waited the full 15 minutes, against 9 of the 14 in the reliable group.

The authors' point was not that self-control does not exist, but that wait times also reflect a sensible judgement about whether waiting will pay off. For a child whose adults often fail to deliver, eating the marshmallow now is not a character flaw. It is a reasonable bet. The study was small, so the exact numbers deserve caution, but the direction fits everything else we have seen.

Here is one more surprise. A 2018 study led by Stephanie Carlson compared 840 children tested in the 1960s, 1980s and 2000s. Contrary to what most adults predicted, children in the 2000s waited on average about two minutes longer than children in the 1960s. The samples came from similar middle-to-high income backgrounds, so this is not the whole population, but it does not fit the idea of a generation that cannot wait.

What this does and does not mean

It does not mean self-control is irrelevant. Across studies there is a link between being able to wait and later outcomes. It does mean that a few minutes with a marshmallow at four is not destiny. What the test captures is a mix of self-control, understanding of the task, trust in adults and the circumstances a child is growing up in, and the later life of that child is shaped by the same circumstances.

What this means for you

The research suggests some useful ways to think about your own patience.

• Check the promise, not just the willpower. If a plan to save or train keeps collapsing, ask whether you actually trust the reward to arrive. Making the payoff more reliable can matter as much as trying harder.
• Work with attention. Mischel's early research was about where children put their attention, not about gritting their teeth. Arranging your surroundings so temptation is less present is a legitimate strategy, not cheating.
• Small steps count. In the 2018 data, the biggest difference was between children who could not wait 20 seconds and everyone else.
• Do not write anyone off. Not a child, and not yourself, on the basis of one short test.

If you want to see where you stand, our delay of gratification test and impulsivity test look at different sides of patience, and the self-efficacy test looks at something closely related: how much you trust yourself to follow through.