On this page
You may have heard the story: a four-year-old is offered one marshmallow now or two if they wait, and the children who waited longer went on to score higher on exams years later. The marshmallow test is often used to argue that self-control in early childhood decides a child's future. In 2018 a team tested this with a much larger and more varied group of children. The result was more modest, and more useful for parents. Here is the detail.
What did the original studies claim?
A well-known follow-up was Shoda, Mischel and Peake (1990) in Developmental Psychology. It linked preschool waiting time to adolescent abilities among children from a university community. According to the 2018 paper's summary of that work, the longitudinal samples were small, between 35 and 89 children, all from a Stanford University preschool. In the key condition, the correlation between waiting time and later SAT maths scores was 0.57, and 0.42 for verbal scores, without adjusting for family background.
What did the 2018 study do differently?
Tyler Watts, Greg Duncan and Haonan Quan published Revisiting the Marshmallow Test in Psychological Science. According to the full text on PubMed Central, they used a long-running US study that recruited children at birth from 10 sites. Key details:
Children took a version of the test at 54 months (four and a half). They could eat the treat any time by ringing a bell, or wait up to 7 minutes for a bigger portion.
918 children had test scores and outcomes at age 15. The main analysis focused on the 552 whose mothers had not completed college, a group about 10 times larger than the original sample.
The team measured achievement in Grade 1 and at age 15, and behaviour problems at both ages.
Crucially, they then added controls: family background and the home environment, and then children's early maths, reading, vocabulary and behaviour at 54 months.
In the main group, children waited 3.99 minutes on average. 45% waited the full 7 minutes and 23% lasted less than 20 seconds.
What did it find?
The raw link was smaller. The correlation-style coefficient between minutes waited and age-15 achievement was 0.24, against 0.57 for maths in the original. The abstract describes it as half the size.
Controls shrank it further. Adding family background and home environment cut it to 0.08. Adding early cognitive skills and behaviour at 54 months cut it to 0.05, which was not statistically significant. The abstract summarises this as a reduction of two thirds.
Most of the difference came from the first 20 seconds. Children who could wait at least 20 seconds did better than those who could not, but waiting the full 7 minutes added little more.
Behaviour showed almost no link. Waiting time barely predicted behaviour problems at age 15, even before controls.
The test tracked early maths. In the full sample, waiting time correlated most strongly (0.37) with an early maths problems test, more than with measures of attention or self-control.
What does it not show?
It does not say self-control is useless. It says one short test at age four is a weak predictor once other factors are counted.
It is not causal either way. Neither the original nor the replication changed children's waiting ability and tracked what happened.
The 7-minute cap was shorter than in the original studies. For children of college-educated mothers, 68% hit the cap, so results for that group were imprecise.
US children only, from a sample the authors describe as geographically diverse but not nationally representative.
What does this mean for parents and students in India?
Single tests in early childhood get turned into labels: "he has no focus", "she is a natural". This study is a reminder that early differences are bound up with home environment and early skills, which parents and teachers can influence.
Don't treat one early measure as destiny. A child who cannot wait at four is not set on a path.
Invest in the things that carried the effect. Early maths talk, reading and a stable home routine were among the factors that absorbed most of the link. Our post on maths talk with young children has practical ideas.
For students: self-control is built by setup, not willpower alone. Putting the phone in another room is a strategy, not a character test.
Be wary of one-study stories. The same caution applies to famous claims about grit; see what critics found about grit research.
For an older student building study habits, structure helps more than willpower. Fermi's AI tutor can work out what you already know and build a plan for the session, so you sit down to a clear next step rather than relying on willpower to get started.
How to read a study like this
Watch what happens to an effect when controls are added. A big raw correlation that shrinks sharply once you account for family income, parents' education and earlier skills is telling you the measured trait was partly a marker for those things. Here, 0.24 shrank to 0.05, and that drop is the main finding.
Next step: if you are a parent, swap one "you have no patience" comment this week for ten minutes of a number or reading game together.
