VI
Mark học AIThe Marshmallow Test · Walter Mischel and colleagues, Stanford University
Can a 4-year-old's ability to wait predict their future? The current answer: a little, far less than the legend says, and it depends largely on circumstances and on how children are taught to wait.
The original goal was not to predict the future but to find out what children do to hold themselves back.
This is Mischel's main finding: the ability to wait depends on attention strategies, and strategies can be taught.
10 common claims about the experiment, checked against current research. Tap each row to see why.
Mischel's series of experiments in the 1970s gave consistent results and has been replicated many times. This is the most solid part, and it was the original purpose of the experiment.
The replication by Watts, Duncan and Quan (2018) still found a link, but a small one: 0.24, dropping to 0.08 after accounting for family background, home environment and early cognitive ability.
The Dunedin study (Moffitt and colleagues, 2011) followed about 1,000 people from birth to age 32. But it measured self-control in many ways over many years, not with a single treat test.
Shoda, Mischel and Peake (1990) managed to reach only a small share of the children; the highest correlations came from about 35 children, almost all from the same social class. Samples like that easily inflate the numbers.
Kidd, Palmeri and Aslin (2013) had only 28 children, but the effect was very large (3 minutes versus 12). Later studies on trust and social groups point the same way, with smaller effects.
Carlson and colleagues (2018) compared different samples at different times, so the cause is hard to isolate. An interesting result, but not yet certain.
Casey and colleagues (2011) scanned the brains of about 26 people from the original sample, grouped by how they did on self-control tests over many years. A very small sample for a conclusion about the brain.
Yanaoka and colleagues (2022) compared Japanese and American children: one study, two countries, a clear result. Plausible, but it needs many more cultures.
The popular reading in self-help books, TED talks and adverts. Larger data show that a single test at age 4 says very little, and most of that comes from family background. Mischel also pushed back against this reading many times.
Waiting time changes with how the reward is presented, which strategies are suggested, trust in adults and culture. It is behaviour that depends on circumstances, not a fixed trait.
A good experiment about how children hold themselves back, turned into a prophecy. Keep the waiting tricks, drop the prophecy.
The story of "one marshmallow predicts the future" has everything it needs to spread: cute children, a simple choice, and a lesson about willpower that suits a culture that prizes individual effort. It appeared in Daniel Goleman's books, on TED and in many skills-training programmes.
The 2018 replication does not say the original experiment was wrong. It says the link is real but small, and most of the predictive power comes from family circumstances that sit behind both the waiting and the good grades. It is a familiar example of psychology's "replication crisis": small sample, pretty result, spreads fast; large sample, modest result, rarely mentioned.
Mischel, in his own 2014 book, stressed that self-control is a skill that can be learned, and that the test result is no child's destiny.
A 4-year-old's tricks still work for adults: change how you see the temptation, don't white-knuckle it.
Every number in this piece and its Facebook posts comes from the studies below.