The Marshmallow Test, Recovered

The part of Mischel’s research that survived the replication crisis — and why it matters more than the part that didn’t

Jun 09, 2026

You’ve probably read about the Marshmallow Test many times — its original premise, the prediction claims that followed, and the 2018 replication that walked most of them back. So I won’t drag you through the full story. Here’s the short version.

In 1972, Walter Mischel sat four-year-olds in front of a marshmallow and told them they could eat it now, or wait fifteen minutes and get two. Some waited. Some didn’t. Over the years that followed, Mischel and his colleagues found that the kids who waited went on to score higher on the SAT, eat better, weigh less, and generally do well in life. The Marshmallow Test became shorthand for the idea that willpower predicts everything.

In 2018, Tyler Watts and colleagues ran the study again with a larger, more diverse sample. The longitudinal prediction held up much weaker than originally claimed. Most of the effect was explained by family income and home environment — not by some inner moral fibre. The headlines flipped. Willpower lives, after all, in the wallet of the parents.

That’s roughly where most articles stop.

Here’s what they leave out.

Walter Mischel, who designed the original experiment and spent the next fifty years studying it, never made the argument the headlines made. He didn’t think the marshmallow test was about willpower as a character trait. He didn’t think kids who waited were morally superior or destined for greatness. He thought something quite different. And the part of his work that holds up — the part the replication crisis never touched — is exactly the part the popular telling skipped.

To see it, you have to watch the children.

The kids who waited didn’t sit there staring at the marshmallow, gritting their teeth, suffering through fifteen minutes of pure restraint. That’s not what the footage shows. The footage shows them doing things.

Some of them physically turned away — swivelling their chairs, putting their backs to the table. Some covered their eyes. Some put their heads down. Some kicked their feet. Some sang songs. Some played invented games with imaginary friends. One famous child — and you can watch this on tape — talked to himself in a low, sustained mutter, narrating his own resistance like a sportscaster covering a slow game.

What Mischel noticed, watching this over and over, was not that some kids had more willpower than others. It was that the kids who waited were doing something different than the kids who didn’t. They were employing techniques. They were redirecting their own attention. They were changing what was in front of them — not in physical reality, but in their own representation.

Mischel called this the difference between “hot” and “cold” cognition. A hot representation of the marshmallow is the gooey-sweet-sticky-delicious version — the version that’s hard to resist. A cold representation is something like: it’s a small round white object on a plate. The kids who made it to fifteen minutes weren’t enduring the hot version harder. They were converting it into a cold version. Imagine it’s a cloud. Imagine it’s a picture, not real. Pretend it’s a stone. The marshmallow that’s a cloud is much easier not to eat than the marshmallow that’s a marshmallow.

Then Mischel ran the next experiment. He took kids who couldn’t wait — the ones who had failed in the first round — and he taught them the techniques. When you start to think about how good it would taste, think about it like a cloud instead. Look away. Sing a song. And those kids, the ones who had failed, started waiting longer. Sometimes much longer.

This is the finding that rarely makes it into the popular retelling.

The delay-of-gratification “ability” was not a fixed trait. It was a learnable cognitive skill. The kids who waited longer hadn’t been born with more virtue. They had — somehow, from somewhere, by themselves or from someone — figured out the techniques. And when the techniques were taught to other kids, those kids could do it too.

It’s worth being clear about what the 2018 replication did and didn’t do.

What Watts et al. tested was the longitudinal prediction claim — that how long a four-year-old waited at the table predicts their SAT scores at eighteen, their BMI at thirty, their salary at forty. That claim was the basis for a thousand parenting books and at least two TED Talks. And that claim, it turns out, doesn’t hold up.

What Watts et al. did not test was Mischel’s actual experimental finding — that teaching the techniques lets kids wait longer. That finding was never about prediction. It was about mechanism. And it was repeatedly replicated by Mischel and his collaborators through the 70s, 80s, and 90s, in studies nobody bothered to challenge because they were never controversial.

The popular debate threw out both findings together. The character-as-destiny version got debunked, and the technique-is-learnable version got buried in the same hole.

If you work in behaviour change — as a coach, a trainer, a therapist, anyone helping people do things they keep failing to do — this matters more than it might seem.

The reason it matters: most of the field still operates on the assumption that “weak willpower” is something approaching a personality trait. The client either has it or doesn’t. If they don’t, you can try to motivate them harder, which is the field’s main lever and the most poorly understood one in it. Or you can quietly conclude that they aren’t a good candidate for change.

What Mischel’s actual finding suggests is something much more useful. The client who can’t resist the cake on day three of their diet may not be lacking willpower. They may be lacking technique. Nobody taught them how to redirect their attention when the craving spiked. Nobody taught them to convert a hot representation into a cold one. Nobody taught them what to do with their hands or where to put their eyes in the moment the system goes loud.

These are not exotic skills. They sit at the core of cognitive behavioural therapy, acceptance and commitment therapy, mindfulness training, addiction recovery, eating disorder treatment. The clinical field has known about them for decades. The fitness industry, by contrast, mostly hasn’t.

The Marshmallow Test, properly told, is not an argument for willpower or against it. It’s an argument that the question is wrong.

We spent fifty years asking which children had more discipline.

The more useful question was: what to do when discipline ran out?

Core Sources

Mischel, W., & Ebbesen, E. B. (1970). Attention in delay of gratification. Journal of Personality and Social Psychology, 16(2), 329–337.

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), 204–218.

Mischel, W., & Baker, N. (1975). Cognitive appraisals and transformations in delay behavior. Journal of Personality and Social Psychology, 31(2), 254–261.

Metcalfe, J., & Mischel, W. (1999). A hot/cool-system analysis of delay of gratification: Dynamics of willpower. Psychological Review, 106(1), 3–19.

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), 978–986.

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.

Hungry Listening

Discussion about this post

Ready for more?