Mischel’s marshmallow doesn’t predict success: the replication broke it

The marshmallow experiment was run by Walter Mischel at Stanford between 1968 and 1974: children aged 3 to 5 were left with a marshmallow on the table and told that if they waited a few minutes without eating it, they would get two. Those who held out, according to the follow-ups published decades later, got better grades, earned more money and had fewer health problems. Self-help books, weekend magazines and schools have spent forty years selling the idea: ‘childhood self-control equals a bright future’.

In 2018 Tyler Watts, Greg Duncan and Haonan Quan repeated it at New York University with 918 children from a nationally representative U.S. sample, using a cleaner instrument. Without controls, the effect was still there, weaker but visible. Once they controlled for the mother’s income and education and for six household features, the effect dropped by more than half and stopped being statistically significant for behaviour and health at age fifteen.

What predicted the future was not holding out for five minutes in front of a marshmallow, but being born in a home where promises are kept and food is guaranteed. Practical advice, for you or for a child: before demanding willpower, fix the environment where you will decide. Put the marshmallow in a locked cupboard, leave the phone in another room, delete the banking app. The discipline you buy with your environment costs less than the one you extract from your will.