The Marshmallow Test myth crumbles under a larger sample, and the results point to what actually matters for children.
Introduction: “That Experiment” We All Know
In the 1970s, a Stanford psychologist placed a marshmallow in front of children and said, “Eat it now, one marshmallow. Wait 15 minutes, two.” Some ate right away; some squirmed but waited. Years later, follow-ups showed that the waiters had higher SAT scores, performed better at work, and were less likely to be obese. Walter Mischel’s “Marshmallow Test” moved beyond psychology textbooks into parenting and self-help.
For parents, the story is sweet. If “a 4-year-old’s patience” predicts a lifetime, the job is clear — build self-control in your child.
Is that really so?
In 2018, Watts and colleagues re-tested the claim with a sample more than 10 times larger than the original. Tracking into adolescence, they found the predictive power shrank to about one-third of the original claim, and mostly disappeared after controlling for family background. In 2024, the same research team published a follow-up — tracking into adulthood, age 26. This is the most rigorous test to date.
The Core Research Question
Researchers from Columbia University and UC Irvine asked simply:
“Does the patience shown at 54 months on a marshmallow test predict the life of that child at age 26?”
They examined three domains:
- Achievement — years of education, annual income, debt
- Health — BMI, depression, substance use
- Behavior — impulsivity, risky behavior, police contact
How the Study Was Done
Participants
702 children recruited across the U.S. and tracked from 54 months (age 4.5) to age 26. Starting from 1,364, the analysis included only those who participated at both time points.
- 46% male, 54% female
- 83% White, 8% Black, 5% Hispanic
- Regionally and economically diverse but not a nationally representative sample
The Marshmallow Test
At 54 months, each child chose a preferred treat, then had up to 7 minutes to wait while the experimenter stepped out. 56% waited the full 7 minutes; average wait was 4.64 minutes.
Age-26 Measurements
- Educational attainment, annual income (log-transformed), non-housing debt over $10,000
- BMI, depression scale (CES-D), substance use index
- Impulsivity (Weinberger Adjustment Inventory), risky behavior, police contact
Controls
To disentangle family background, the team controlled for race, mother’s age and education, family income, early home environment quality, and cognitive ability and problem behavior at 54 months.
Results: The Expected Predictive Power Nearly Vanished
Before Controls — Only Two Links
Without controlling for family background, only two of 11 adult outcomes correlated with the marshmallow test.
- Years of education: r = .17 (p < .001)
- BMI: r = −.17 (p < .001) — children who waited longer had lower adult BMI
The other 9 — income, debt, depression, substance use, risky behavior, impulsivity, police contact — weren’t statistically significant.
After Controls — Only BMI Survived
Controlling for family background and early cognitive ability, even the link with education disappeared (β = −0.01). Nearly every adult outcome’s regression coefficient collapsed to near zero.
The one exception was BMI. Children who waited the full 7 minutes had meaningfully lower adult BMI than those who didn’t (β = −.17, p = .03).
An Unexpected Finding: The Middle Group
When children were split by wait time, an odd pattern emerged. The “middle group” that waited 2–7 minutes had lower educational attainment than the group that couldn’t wait 20 seconds (β = −.23, p = .05). A similar pattern in annual income (β = −.44, p = .01).
The researchers explicitly noted they couldn’t explain this. It could be statistical noise, a byproduct of the sample, or something else. What’s clear is that “more waiting = better” — a simple linear relationship — doesn’t appear in the data.
Why the Famous Result Didn’t Replicate
This is the study’s most interesting moment.
Not “Patience Itself” but “What Creates Patience”
The researchers explain:
The marshmallow test doesn’t measure pure self-control. It’s a blend of the quality of a family where waiting is possible, experiences of whether adult promises are reliable, and overall stability of daily life.
A hungry child, or a child in a home where adult promises often break, is rationally choosing “eat now.” Not because they lack patience, but because they’ve learned that waiting isn’t rewarded. When the team controlled for family background, the marshmallow test’s predictive power melted away.
Why BMI Survived
The team also interpret BMI cautiously. The marshmallow test is fundamentally about “impulse control in front of food.” It’s natural that it connects to a narrow trait like appetite regulation, not general self-regulation.
In other words, what the marshmallow test predicted wasn’t “overall success” but only “self-regulation around food.”
Practical Implications
Don’t Judge a Child by Patience at Age 4
The most important message. A child who eats the marshmallow first isn’t “in trouble.” The age-26 data says they’re not. Conversely, waiting doesn’t guarantee a future either.
Environment Beats Narrow Training
The researchers’ conclusion:
Interventions narrowly targeting a child’s delay of gratification are unlikely to produce long-term effects unless other aspects of the child’s environment change with them.
Instead of “patience drills for my child,” the relationships and environment a child grows up in are a far stronger predictor. Consistent parenting, predictable routines, trustworthy adults — these shape a child’s self-regulation.
Don’t Turn One Test Into a Myth
The appeal of a single measurement predicting a lifetime is strong, but it isn’t scientifically supported. A child’s development is many factors weaving together over time. No single moment explains everything.
Limitations
Honestly reported by the team:
- Though diverse, the sample skews White and middle-class; patterns among low-income and minority populations may not be fully captured
- 56% hit the ceiling at 7 minutes — “extraordinarily patient” children couldn’t be differentiated. Extending to 15–20 minutes might yield different results
- Age 26 is early adulthood; midlife could show different patterns
- Observational, so causal claims can’t be confirmed
Final Thoughts
“The marshmallow of age 4 predicts the life of age 30” is now closer to a myth of the past.
The truth Sperber and colleagues found, tracking 702 people to age 26, is more complex and more hopeful. A child’s future isn’t set by one test. It’s built, layer upon layer, by the relationships and environment around them.
So the parent’s job isn’t to drill patience — it’s closer to building a world where patience gets rewarded.
Source: Sperber, J. F., Vandell, D. L., Duncan, G. J., & Watts, T. W. (2024). Delay of gratification and adult outcomes: The Marshmallow Test does not reliably predict adult functioning. Child Development, 95(6). https://doi.org/10.1111/cdev.14129