The marshmallow test for adults

Cost Me Research Desk · May 26, 2026

A four-year-old sits at a table with a marshmallow in front of her. The researcher says: I'm leaving the room. If you wait until I come back, you get two marshmallows. If you can't wait, ring this bell and you get one now.

Then he leaves for fifteen minutes.

That experiment, run by Walter Mischel and Ebbe Ebbesen starting in 1970, became the most famous study in psychology (Mischel & Ebbesen, 1970).¹ It also became the most overinterpreted. Both facts matter, and the honest version of the story is more useful for your spending than the headline version.

What the original studies actually found

Mischel's team ran the test on hundreds of children at Stanford's Bing Nursery School in the late 1960s and 70s. The headline finding was about the strategies children used. The kids who waited longest weren't the ones with iron willpower; they were the ones who successfully distracted themselves.

They covered their eyes. They turned their chairs around. They sang songs. They pretended the marshmallow was a cloud, not food. Children who stared directly at the marshmallow and tried to resist it almost always rang the bell within a few minutes.

Willpower didn't win the marshmallow test. Distraction did.

This is the part of the research that often gets lost in popular retellings. The lesson was never “some children have better self-control.” The lesson was that self-control is a skill of attention management, not a moral attribute.

The longitudinal follow-ups

Mischel, Shoda, and Rodriguez published a 1989 paper in Science reporting that children who had waited longer in the original 1970s tests had, as adolescents, higher SAT scores and were rated by their parents as more competent at handling stress (Mischel et al., 1989).²

The correlation was real, modest, and frequently exaggerated. In subsequent retellings it became a sweeping claim: delay-of-gratification at age four determines success at age forty. The original data never supported anything that strong.

The 2018 replication that complicated everything

In 2018, Tyler Watts, Greg Duncan, and Haonan Quan published a conceptual replication in Psychological Science using a much larger, more diverse sample — about 900 children, compared to the original ~90 (Watts et al., 2018).³

Their results: the delay-of-gratification effect on later outcomes was much smaller than the original studies suggested, and most of it disappeared once you controlled for family socioeconomic status and the child's cognitive ability.

In other words: the four-year-old who waited longer wasn't demonstrating a deep trait that predicts adult success. She was, in large part, demonstrating that she came from a household where waiting was safe, where promises tended to be kept, and where she had the cognitive scaffolding to deploy distraction strategies in the first place.

Crucially: the Watts replication doesn't refute that self-control matters. It refutes the simplistic version of the story where a single childhood test forecasts a life. The bias is more about context and learnable skill than innate trait.

The adult version of the test

Every impulse purchase you've ever resisted — and every one you haven't — is a version of the marshmallow test. One marshmallow now, in the form of the purchase. Two marshmallows later, in the form of whatever the money would have done if you'd kept it.

The ratio is far better than the original test, by the way. At a 7% real return over 30 years, a $50 purchase you skip becomes roughly $380. The choice isn't one marshmallow now vs two later. It's one marshmallow now vs seven later.

And yet adults fail this test far more reliably than the Bing nursery kids failed theirs. Why?

The future marshmallow is invisible

The nursery school children could literally see the marshmallow. The reward for waiting was tangible. The adult financial version offers no such visibility — the future return is abstract, statistical, and contingent.

This is why showing the compound future value of a purchase decision works. It converts the invisible second marshmallow into a visible one. (See also: hyperbolic discounting.)

The waiting interval is decades, not minutes

Fifteen minutes is hard for a four-year-old. Thirty years is mostly meaningless for an adult — the time horizon is so long that the present-bias circuitry treats the second marshmallow as essentially nonexistent.

The countermeasure is to artificially shrink the wait. A 48-hour cooldown on a purchase is not 30 years — but it's a deliberate inserted delay that gives the deliberative system time to weigh in before the impulsive system commits.

Distraction strategies still work

Mischel's deepest finding wasn't about who passed. It was about how. The children who passed routed attention away from the marshmallow.

The adult equivalent is brutally direct: stop visiting the website. Stop following the influencer. Unsubscribe from the sale email list. Don't install the store's app. The marshmallow doesn't tempt you if you're not looking at it.

The honest takeaway

The marshmallow test is not a personality assessment. It's a demonstration that self-control is a function of three things: the visibility of the reward, the length of the wait, and the strategies available for managing attention during it.

All three are manipulable. None of them require an innate trait you either have or don't. The question of whether you can wait isn't about who you are; it's about the environment you've built around the decision.

The Watts replication, properly read, is good news. If delay-of-gratification were a fixed trait set in childhood, there'd be nothing to do. Because it's contextual and learnable, the design of your environment matters more than your willpower.

How Cost Me applies this research

Cost Me is built around the three manipulable variables Mischel identified.

Visibility of the reward. The calculator's entire job is to make the second marshmallow visible. The future value of a not-spent dollar is shown alongside the price of the item being considered — not as a guilt trip, but as a comparable number to weigh against the purchase.

Length of the wait. The 48hrs button compresses the decision horizon to a workable size. Two days is long enough to defuse the immediate impulse, short enough that the user's attention can be successfully managed in the interim. By the time the timer ends, the typical user finds the urgency has evaporated — and we have the data to show that a meaningful fraction of items in cooldown never come back out.

Attention management. The Amy coach reads your behavior patterns and surfaces the right intervention at the right moment — a notification at a late-night browsing session, a question before a repeated lookup of the same item. This is the modern version of covering your eyes or turning the chair around. Cost Me doesn't require you to have iron self-control. It requires you to follow a few attention-management cues at the right moment, which the app provides.

The Watts replication is the version of the marshmallow finding we take seriously: the goal isn't to identify high-willpower users and congratulate them. It's to give every user the environment that makes delay possible in the first place.

References

1. Mischel, W., & Ebbesen, E. B. (1970). Attention in delay of gratification. Journal of Personality and Social Psychology, 16(2), 329–337. https://doi.org/10.1037/h0029815
2. Mischel, W., Shoda, Y., & Rodriguez, M. L. (1989). Delay of gratification in children. Science, 244(4907), 933–938. https://doi.org/10.1126/science.2658056
3. 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

Related reading: why the immediate reward wins and the 48-hour rule explained. Back to costme.io.