The Marshmallow Test Built a Myth About Self-Control. A Larger Study Tore It Down.

What the Original Study Actually Claimed

Starting in the late 1960s, Stanford psychologist Walter Mischel ran a now-famous experiment on preschool children at the university's on-campus school. A child was left alone in a room with a marshmallow and told they could eat it immediately or wait for the researcher to return and receive a second one. Mischel's team then tracked those children as they grew up.

The 1990 follow-up study, described by The Atlantic, claimed that kids who waited scored higher on standardized tests and performed better by other measures later in life. The conclusion that followed — that self-control is a foundational trait predicting success — became one of the most repeated findings in popular psychology.

Parenting experts cited it, schools taught it, and it shaped real policy discussions about how to help disadvantaged kids.

The Replication Problem

There was one significant flaw in the original design: the sample. Mischel's studies included fewer than 90 children, all enrolled at a preschool on Stanford's campus, a population that skewed heavily toward the children of university faculty and staff. This was not a representative sample of American kids.

In 2018, researchers Tyler Watts of NYU and Greg Duncan and Haonan Quan of UC Irvine re-ran the experiment with a sample of more than 900 children that was broadly representative of the U.S. population by race, ethnicity, and parental education level, according to The Atlantic's coverage of the study.

The result: the marshmallow test's predictive power nearly vanished once researchers controlled for socioeconomic background.

What the New Data Actually Shows

Among children whose mothers had college degrees, those who waited for the second marshmallow did no better in long-term outcomes than those who ate immediately. The same finding held among children of non-college-educated mothers: waiting produced no measurable advantage.

The capacity to delay gratification, the new study found, is itself shaped by a child's economic and social circumstances. A child who has grown up in an unstable environment, where promised rewards sometimes don't materialize, has a rational reason not to wait. Eating the marshmallow now is the smart play if you have learned that adults don't always follow through.

What predicted success, the Watts et al. study concluded, was that background. Not the willpower to wait.

The Broader Replication Crisis

The marshmallow study is not an isolated casualty. The Atlantic's reporting explicitly names a broader "replication crisis" in psychology, where foundational studies, many of them conducted on small, unrepresentative samples, have failed to hold up when tested on larger and more diverse populations.

This is a real methodological problem that cuts across ideological lines. Small convenience samples, publication bias toward positive results, and pressure to produce clean narratives have produced a body of research that is shakier than textbooks suggest.

Psychology as a field has been working through this problem for roughly a decade, and the marshmallow test is one of the higher-profile casualties.

The Strongest Defense of the Original Work

Mischel's original research was never designed as a definitive, population-wide study. It was exploratory work, and the 1990 follow-up was a longitudinal look at a specific cohort. Critics of the Watts replication argue that controlling for socioeconomic status may actually wash out a real effect. If self-control partly mediates the relationship between background and outcomes, controlling for background could statistically erase a genuine signal.

Some researchers also contend that the ability to delay gratification still matters at the margins, even if it's not the dominant force Mischel's popularizers claimed. The argument isn't frivolous. But the burden of proof falls on the original extraordinary claim: that a single behavioral measure in a preschooler predicts life outcomes. That claim did not survive a more rigorous test.

What This Means Outside the Lab

The cultural damage from the oversimplified marshmallow narrative is worth naming. It gave intellectual cover to a framework that located the roots of inequality inside children's heads rather than in material circumstances. If poor kids fail to delay gratification, the thinking went, that's a character problem. Teach them self-control and the outcomes will follow.

The Watts study suggests the causality runs the other direction: give families more stable economic footing, and children's capacity for patience largely takes care of itself.

This does not mean self-discipline is worthless. It means that treating it as the master variable in child development was bad science, poorly tested on too few people, and then broadcast to the world as settled fact.

Whether any reliable behavioral measure in early childhood predicts adult outcomes independent of socioeconomic background remains unresolved. Whether the replication crisis in psychology has finally produced enough institutional pressure to change how small, single-site studies get reported to the public before they've been independently confirmed is also an open question.