Arnold Kling  

Stanford Marshmallow Experiment

PRINT
What Racism Is... My Contigency Plan...

I went to hear Greg Clark give a noontime seminar at Cato today. He is fairly persuasive in person. One thing about pursuing a line of thought for a long time when most people disagree with you is that your arguments tend to get pretty sharp.

During the discussion afterward, somebody brought up the Stanford Marshmallow Experiment. You can Google that subject to find out much more, but here is one site that mentions it.


In one of the most amazing developmental studies ever conducted, Walter Michel of Stanford created a simple test of the ability of four year old children to control impulses and delay gratification. Children were taken one at a time into a room with a one-way mirror. They were shown a marshmallow. The experimenter told them he had to leave and that they could have the marshmallow right then, but if they waited for the experimenter to return from an errand, they could have two marshmallows. One marshmallow was left on a table in front of them. Some children grabbed the available marshmallow within seconds of the experimenter leaving. Others waited up to twenty minutes for the experimenter to return. In a follow-up study (Shoda, Mischel, & Peake, 1990), children were tested at 18 years of age and comparisons were made between the third of the children who grabbed the marshmallow (the "impulsive") and the third who delayed gratification in order to receive the enhanced reward ("impulse controlled").

The third of the children who were most impulsive at four years of age scored an average of 524 verbal and 528 math. The impulse controlled students who scored 610 verbal and 652 math! This astounding 210 point total score difference on the SAT was predicted on the basis of a single observation at four years of age! The 210 point difference is as large as the average differences between that of economically advantaged versus disadvantaged children and is larger than the difference between children from families with graduate degrees versus children whose parents did not finish high school! At four years of age gobbling a marshmallow now v. waiting for two later is twice as good a predictor of later SAT scores than is IQ.


The quoted source sees the Marshmallow experiment as showing the importance of emotional intelligence. The economists at the Clark seminar just thought of it as a remarkably reliable indicator of general intelligence--apparently even better than IQ. Also, the economists see this as a measure of time preference, rather than impulse control. That may be more than a mere matter of terminology.

Clark's view of the world is that differences in average individual characteristics, such as time preference, account for a lot of differences in the standard of living, even holding things like institutions constant. He acknowledges that places like North Korea and Zimbabwe are poster children for the role of institutions, but he warns us not to get caught up in confirmation bias. There are, he argues, also quite a few poster children for countries where institutions appear to be decent, but poverty is still severe.

Anyway, if you, like me, were unaware of what clearly must be a classic in empirical psychology, consider yourself enlightened.


Comments and Sharing





COMMENTS (13 to date)
Lord writes:

So, is all IQ is, is time preference? Is all time preference is, is IQ?

Trent McBride writes:

Did they control for hunger?

Unit writes:

Arnold,

instead of Industrial Revolution, should we start calling it the Capitalist Revolution? Since capital accumulation has something to do with time.

PrestoPundit writes:

These experiments were filmed, and it's something to watch these kids when you know the SAT results years down the road. The film of this study is standard fodder in PBS type psychology documentaries.

I hate to get all dystopian, but I bet if you cross-referenced the impulsive kids from the Marshmallow Experiment with children with low resting heart rates and high-scorers in the Stanford Prison experiment, it would be an excellent predictor of criminal behavior.

Lord writes:

Some of the questions are what do they comprehend and what do they interpret it to be (what does later mean for example), is the decision based on reason or emotion, once decided is the action consistent with their decision, do they change their mind, or is it a matter of lack of control over their actions. If hungrier to begin with, previously sated or more easily sated, expecting other food soon afterward, or more or less inclined towards marshmallows, may all modify their desire. How much thinking enters into their decision and how much is it worth investing into it? How conscious are they of their own state and situation and what assumptions and extrapolations are they willing to make? Is a decision made and committed to, made and remade, or is it protracted? How much is it subject to commitment, impulse, whim, and willpower, and is it overcome or reconsidered? Does reason exist or is it only an ex post rationalization?

TGGP writes:

"Emotional intelligence" is just pop-psych. Same with Martin Gardner's "multiple intelligences"

Do you know what the sample size was in the Stanford Marshmallow Experiment?

Patrick Molloy writes:

The marshmallow experiment is also described in the book The Happiness Hypothesis (pp. 17-18) by Jonathan Haidt whose work has been noted here in the last few days.

Since the four year old subjects of 1970 are now forty there may be other intriguing findings from this cohort.

FanthonyB87 writes:

The marshmallow experiment can easily be related to production periods in economics because demand is much more elastic in the long-run. The one who delays gratification now will be benefit more by waiting in the long-run production period because one can develop the plant size and obtain sustainable labor specialization, managerial specialization, and efficient capital. By doing so, one will have more economies of scale with a better chance of producing the best plant size and lowest unit cost.

This all comes back to opportunity costs because one must always think about producing capital goods now instead of directly satisfying human wants of consumer goods. A firm cannot survive off impulsive decisions but can survive by thinking of the long-run goods for the future. If a firm does not have a constant return to scale then it will produce too many diseconomies thus destroying the firm swiftly. In all, patience is a virtue.

m.edwards writes:

What if the kids didn't like marshmallows?

John Bullock writes:

Do you know what the sample size was in the Stanford Marshmallow Experiment?

The original sample size was 653; these were the children with whom the impulse control studies were done (Mischel, Shoda, and Peake 1998). In the follow-up, parents of 94 children reported SAT scores for their children. ETS was able to locate records for 69 of the children; the correlation between the self-reported scores and the scores that ETS had on file was .94.

In a follow-up study (Shoda, Mischel, & Peake, 1990), children were tested at 18 years of age and comparisons were made between the third of the children who grabbed the marshmallow (the "impulsive") and the third who delayed gratification in order to receive the enhanced reward ("impulse controlled").

The third of the children who were most impulsive at four years of age scored an average of 524 verbal and 528 math. The impulse controlled students who scored 610 verbal and 652 math!

This claim is nowhere in the Shoda, Mischel, and Peake 1990 article -- the one in Developmental Psychology, which I think is the only one that they did in 1990. The authors do analyze the connection between the impulse control results and SAT scores, and there does seem to be a connection. But they never compare the top third to the bottom third, nor do they report average SAT scores for any group –– not unless I'm overlooking something. I wonder where these numbers are coming from. And whether the dataset is publicly available.

deborah manson writes:

Where might I be able to obtain a copy of the film - The Stanford Marshmallow Experiment? Any information is truly apppreciated

Comments for this entry have been closed
Return to top