The Marshmallow Test Mostly Measured How Much Money Your Family Had

Walter Mischel's preschoolers, in the late 1960s, were given a marshmallow and a deal: don't eat it for fifteen minutes and you'll get two. The ones who waited turned out, decades later, to score higher on the SATs and have lower BMIs and fewer behavioral problems. The conclusion that became gospel — that early self-control predicts adult success — ran on a sample of about 90 children, mostly from a Stanford-affiliated nursery, mostly from professional families.

In 2018 Tyler Watts, Greg Duncan, and Haonan Quan ran a conceptual replication using 918 children from a national longitudinal dataset, ten times Mischel's follow-up sample and far more diverse. The basic correlation showed up: kids who waited longer at age 4 did somewhat better in adolescence. Then they added controls — family socioeconomic status, the home environment, cognitive ability at age 4 — and the relationship mostly evaporated. What had looked like the predictive power of self-control was largely the predictive power of growing up with money and stability.

The finding makes more sense once you ask why a kid would wait. A child whose parents reliably keep promises has good reason to expect a second marshmallow; a child whose household runs on scarcity has good reason to grab the one in front of them. The behavior at age 4 is partly a measurement of self-control and partly a measurement of trust in the experimenter — which is itself shaped by the world the child is already living in.

Mischel had been careful in his original papers; the strong predictive claims came mostly from journalists and self-help writers who picked up the study. The 2018 work didn't kill the result so much as rescale it: delayed gratification matters, but as one ingredient in a stew, not the whole flavor.