The marshmallow test, continued

In 2013 Celeste Kidd and colleagues at the University of Rochester ran an experiment that rarely comes up alongside the marshmallow test, and should.¹

Twenty-eight children, four and a half years old on average, were seated at tables with a set of battered crayon stubs and promised a good new art kit. The experimenter left the room. A couple of minutes later he came back: half the children got what they had been promised, and to the other half he spread his hands — sorry, nothing turned up, use these. Then the same thing was repeated with stickers, and again half were left with nothing.

After that every child was offered a marshmallow on the usual terms: eat one now, or wait until I come back and get two.

The children who had been let down held out for a little over three minutes on average. The ones whose promises had been kept waited twelve. Four times longer, and it was arithmetic rather than willpower: the adult in this room had already broken his word twice, so why bet on a third time.

What the marshmallow test actually showed

Walter Mischel ran the original series in the late 1960s at the Bing Nursery School at Stanford.² The mechanics were exactly as they are usually retold: a child was seated in front of a treat, given the rules, and left alone. Some ate it at once; others fidgeted, turned away, hummed to themselves and made it to the end. What that work measured was not character: when both rewards sat in plain view, a child gave in after about a minute; when they were taken out of sight, the wait stretched to eleven.

The fame came from what followed. In 1990 Mischel and colleagues went back to the same cohort, sent questionnaires to parents and gathered data on roughly 185 of the 653 children.³ Those who had held out longer scored better on the SAT on average — though only the thirty-five who had been left alone with the reward in plain view and given no strategy for waiting. From there came the whole popular literature about willpower as the great predictor of success in life, and with it the parental advice to train patience.

In 2018 Tyler Watts, Greg Duncan and Haonan Quan repeated the study with 918 children drawn from a far wider range of family incomes.⁴ The association survived, but it shrank: once family income, home environment and early cognitive scores were taken into account, a fifth of it was left, statistically indistinguishable from zero.

What the test appears to have measured is circumstance: how predictably food appears in a child's home, whether promises are kept, whether tomorrow is worth counting on. A child from a household where a word is kept waits because he has grounds to.

The child who never got the promised art kit ate the marshmallow in three minutes. That was a sensible conclusion: the adult in this room had already broken his word, and the marshmallow was going nowhere.

Which leads to a practical point about trust rather than patience: a child waits exactly as long as he believes your word is good for. So keep what you promise, and the parts that are easy to forget — interest on what has been put aside, for instance — hand to the app: in PocketPal the rule is written down once and runs by itself after that.