The Atlantic's account of the marshmallow test is pretty overzealous. In particular, Calarco describes it as a "failed replication", but in reality the new stu

The Atlantic's account of the marshmallow test is pretty overzealous. In particular, Calarco describes it as a "failed replication", but in reality the new study (with 10x the participants and more diversity) also found a significant correlation between childhood wait time and teen test performance. I think calling this a failure to replicate is simply wrong. The new result is that adding two large sets of controls makes the effect disappear. One of those control sets was environmental, including family income - but it also included traits about the individual child up through age three. The other was the child's score on a battery of other psych tests alongside the marshmallow test. Data for individual control variables isn't available, only these two aggregate scores. And the environmental control alone wasn't sufficient to erase the effect; it weakened below significance for some but not all groups studied, but only with both controls did the effect drop to near-zero for all groups. So "wealth causes patience" is one explanation, but "controlled for the variable" is another. And the age-15 test assesses metrics like reading comprehension, math skill, and general knowledge. The age-4 battery included metrics like "memory for sentences", "applied problems", and "complete words". Is it really surprising that controlling for vocabulary at age 4 makes it hard to find differences in vocabulary at age 15? Even if this isn't controlling for the variable, we might have controlled away the cause instead. Wealth causing lower discounting is one obvious candidate, but we haven't ruled out the possibility that bad circumstances directly lower patience, which then causes bad outcomes. This study controlled for infant temperature, birth weight, and cognitive functioning at age two; if any of those things are a cause of or close correlate with lowered patience, the study loses all of its sensitivity. And even if that isn't the case, I think the 'wealth causes patience' explanation is just permitted by the data, rather than demonstrated. I'm not sure exactly how you'd test for reverse causation on wealth/patience, but it would probably involve assessing patience at least twice to see whether kids gained/lost it corresponding to wealth. But as far as I can tell, no one has even checked it a second time without controls; wouldn't it be fascinating if age 4 patience correlated with SAT scores better than age 15 patience? (For the record, I think the real answer is probably that wealth does increase patience in teens/adults, but the early-childhood 'patience' with marshmallows is a secondary symptom of any of numerous factors actually driving these results.) Source for the best breakdown I've found: https://ifstudies.org/blog/did-the-marshmallow-test-fail-to-replicate https://ifstudies.org/blog/did-the-marshmallow-test-fail-to-...