The Marshmallow Test Is Only Half True — From the Original Paper to the 2018 Replication

Introduction — Can One Marshmallow Predict a Whole Life?

You have probably heard this story. You place a marshmallow in front of a 4-year-old and say, "You can eat it now, but if you wait 15 minutes, you will get two." Decades later, the story goes, the children who waited had better grades, higher salaries, and better health. The conclusion is usually compressed into one line: "Self-control in early childhood determines your life."

Replayed endlessly in self-help books and lectures, this story is half fact and half myth. And to find out which half is which, you have to open the original papers yourself.

This post is the first installment of a new series, "Psychology, Straight from the Papers." The rules of the series are simple: take a famous psychology story, explain in plain language what the original paper actually did, follow the replications that came after, and keep only the lessons we can genuinely take home.

What the Original Experiment Actually Did — 1972, Bing Nursery School

The experiments by Walter Mischel and his team were run from the late 1960s to the early 1970s at Bing Nursery School, attached to Stanford University. The actual concern of the 1972 paper (Mischel, Ebbesen, Zeiss) is different from the popular image. The paper was not trying to sort out "which children are the patient ones"; it tested which conditions make waiting easier or harder.

The design goes like this. Children were split across conditions: the reward (a marshmallow or a pretzel) visible versus covered, guided to think fun thoughts while waiting versus not, and so on. The results were clear. When the reward sat in plain view, children on average gave up much faster; covering the reward or steering attention elsewhere made the waiting several times longer.

So the conclusion of the original experiment is not that "patience is an innate temperament" — rather the opposite. Delay of gratification is a function of situation and strategy. The children who waited well were not the ones who gritted their teeth and endured; they were the ones who looked away from the marshmallow, sang songs, and imagined the marshmallow as "a picture of a cloud." This is also what Mischel himself emphasized for the rest of his life. He saw self-control as "the skill of handling hot stimuli with the cool system" — a matter of learnable strategy.

Where the Famous Prophecy Came From — the Actual Size of the Follow-up Studies

The claim that "the kids who waited turned out well" comes from the later longitudinal follow-ups (most notably Shoda, Mischel, Peake 1990), which reported that waiting time in preschool correlated with adolescent academic performance (SAT), parent-rated coping ability, and so on.

Here is a fact you only learn by reading the papers themselves: the sample size. The full Bing cohort numbers in the hundreds, but the group actually tracked for decades and entered into the correlation calculations shrinks to a few dozen, depending on the analysis. The subsample behind the SAT correlation, for example, is under 100 children. On top of that, all of these children came from the Stanford University community (mostly children of professors and graduate students) — an extremely homogeneous and unusual group.

A small sample, a special population, and an attractive story. Every condition that calls for a replication was in place.

The 2018 Replication — What Happened When Family Background Was Controlled

In 2018, Tyler Watts, Greg Duncan, and Haonan Quan published a replication study in Psychological Science. Using a large child-development cohort from the US National Institutes of Health, they asked the same question with a sample far larger (about 900 children) and more diverse than the original: does delay of gratification at age four and a half predict academic achievement and behavior problems at 15?

The result can be summarized in three lines.

One detail is especially interesting. Most of the predictive power lived within the first 20 seconds. There was a difference between children who waited 20 seconds and children who ate immediately, but beyond that point, how many extra minutes a child held out predicted almost nothing more about the future.

The interpretation of this result matters. A child who waits is not displaying a temperament in a vacuum; the waiting is also a product of the environment. For a child whose home reliably has snacks tomorrow, "wait and you get two" is a credible promise; for a child in an environment where adult promises are routinely broken, eating now can be a rational choice rather than an impulse. Indeed, Celeste Kidd's 2013 experiment showed that when the experimenter first demonstrated keeping a promise, children's waiting times grew several times longer. The marshmallow test, it turns out, was measuring not only self-control but also how much a child trusts the world.

What Remains — Not Temperament, but Strategy and Environment

To sum up, here is what to discard and what to keep from the marshmallow story.

Discard: the fatalism of "patience at age 4 determines your life." The evidence was a correlation in a small, unusual sample, and in a large sample the effect shrank sharply once family background was controlled. Reading the self-control of a child (or of yourself) as a fixed grade is a view that even the original research does not support.

Keep, number 1 — the strategies are real. The core finding of the original experiment — redirecting attention weakens the pull of temptation — has been confirmed again and again in later self-regulation research, and it is a sturdy principle we will keep meeting in this series. The fight against temptation is not a contest of willpower but a matter of where you point your eyes. It is why leaving your phone in another room beats resolving not to look at it, and it is also the theoretical root of building habits through environment design.

Keep, number 2 — trust is the precondition for patience. Waiting is rational only in a world where promises are kept. Translated to teams: before blaming a member for a "short-term attitude," first ask whether the organization has kept its promises — evaluations, rewards, schedules.

Reading Guide

• Original experiment: Mischel, W., Ebbesen, E. B., & Zeiss, A. R. (1972). Cognitive and attentional mechanisms in delay of gratification. Journal of Personality and Social Psychology, 21(2), 204-218.
• Longitudinal follow-up: Shoda, Y., Mischel, W., & Peake, P. K. (1990). Predicting adolescent cognitive and self-regulatory competencies from preschool delay of gratification. Developmental Psychology, 26(6), 978-986.
• Replication: Watts, T. W., Duncan, G. J., & Quan, H. (2018). Revisiting the marshmallow test. Psychological Science, 29(7), 1159-1177.
• Trust condition: Kidd, C., Palmeri, H., & Aslin, R. N. (2013). Rational snacking. Cognition, 126(1), 109-114.

The recommended order is 1972 → 2018. In the 1972 paper, the table of experimental conditions (how many minutes children waited under which condition) alone gives you half the value; in the 2018 paper, the highlight is the table comparing the coefficients before and after the control variables enter. The next installment of the series covers the evidence on growth mindset.