The Marshmallow Test Was Mostly About Trust

Walter Mischel ran the first marshmallow experiment at the Bing Nursery School at Stanford in the late 1960s. A child was shown one marshmallow, told the experimenter was stepping out, and promised a second if the first was still there on their return. Mischel's later work correlated wait time with SAT scores, self-control measures, and life outcomes. The study became one of the most-cited experiments in developmental psychology and the basis for a thousand self-help lectures.

Celest Kidd's group at Rochester ran a 2013 variant with a critical tweak. Before the marshmallow test, they staged an art project. Half the children worked with an experimenter who made promises and kept them (she said she'd return with better supplies and did). The other half worked with an experimenter who made promises and broke them (she said she'd return with better supplies and came back empty-handed). Both groups then did the classic marshmallow task.

Kids in the reliable condition waited, on average, four times as long. The median kid who'd seen promises kept made it to the 12-minute cap. The median kid who'd seen a broken promise gave up around three minutes. "Self-control" was swamped by reasonable inference: if the adult wasn't good for their word, eating the marshmallow now was the smart move.

A 2018 replication by Watts, Duncan, and Quan in Psychological Science using a larger, more diverse sample found that once socioeconomic status was controlled for, the famous correlation between wait time and later achievement shrank dramatically — to roughly half its reported size. Self-control may still matter. It's also measuring the child's accurate model of the adult world around them.