Posted on October 22, 2008

Marshmallow Temptations, Brain Scans Could Yield Vital Lessons in Self-Control

Carey Goldberg, Boston Globe, October 22, 2008

It is a simple test, but has surprising power to predict a child’s future. A 4-year-old is left sitting at a table with a marshmallow or other treat on it and given a challenge: Wait to eat it until a grown-up comes back into the room, and you’ll get two. If you can’t wait that long, you’ll get just one.

Some children can wait less than a minute, others last the full 20 minutes. The longer the child can hold back, the better the outlook in later life for everything from SAT scores to social skills to academic achievement, according to classic work by Columbia University psychologist Walter Mischel, who has followed his test subjects from preschool in the late 1960s into their 40s now.

From church sermons to parenting manuals, “the marshmallow test” has entered popular culture as a potent lesson on the rewards of self-control. It has also raised deep psychological research questions: What is involved in delaying gratifica tion? Why does it correlate with success in life? Why do people fail at it?

Now neuroscientists, using high-tech brain scans, are seeking to answer these questions by examining what goes on in the brain when a person aces or flunks marshmallow-type tasks. They aim to use their findings to figure out how to train people to control themselves better, whether that means focusing on the potential pitfalls of a mortgage broker’s pitch or concentrating on the calorie count of a brownie.


Most recently, Yale University researchers found that delaying gratification involves an area of the brain, the anterior prefrontal cortex, that is known to be involved in abstract problem-solving and keeping track of goals. {snip}

The brain scan findings from 103 subjects suggest that delaying gratification involves the ability to imagine a future event clearly, said Jeremy Gray, a Yale psychology professor and coauthor of the study in the September edition of the journal Psychological Science. {snip}

In the coming months, researchers plan to perform brain scans on 40 of the original subjects of Mischel’s marshmallow test, said John Jonides, a psychology professor and brain imager at the University of Michigan who is working with Mischel on the project.

If brain differences are found between good and poor delayers, he said, they could suggest effective avenues for training. {snip}


In the marshmallow test, [Mischel] said, “the same child who can’t wait a minute if they’re thinking about how yummy and chewy the marshmallow is can wait for 20 minutes if they’re thinking of the marshmallow as being puffy like a cotton ball or like a cloud floating in the sky.”

Neuro-economists, who use brain scans to shed light on economic decision-making, are also exploring the “cool brain hot brain” theory, said Daniel Benjamin, an assistant professor of economics at Cornell University.


Mischel emphasized that though intelligence is related to doing well on marshmallow tests, it is by no means the whole answer.

As many people know, “It’s quite possible to be very smart and not able to inhibit your impulses,” he said.

1 Department of Psychology, Yale University; 2 Department of Psychology, University of Minnesota; 3 Interdepartmental Neuroscience Program, Yale University; 4 Department of Psychology, Princeton University; 5 School of Psychology, Georgia Institute of Technology; and 6 Department of Psychology, Washington University

Address correspondence to Jeremy R. Gray, Box 208205, New Haven, CT 06520-8205, e-mail: [email protected]

Copyright © 2008 Association for Psychological Science

Lower delay discounting (better self-control) is linked to higher intelligence, but the basis of this relation is uncertain. To investigate the potential role of working memory (WM) processes, we assessed delay discounting, intelligence (g), WM (span tasks, 3-back task), and WM-related neural activity (using functional magnetic resonance imaging) in 103 healthy adults. Delay discounting was negatively correlated with g and WM. WM explained no variance in delay discounting beyond that explained by g, which suggests that processes through which WM relates to delay discounting are shared by g. WM-related neural activity in left anterior prefrontal cortex (Brodmann’s area 10) covaried with g, r= .26, and delay discounting, r=-.40, and partially mediated the relation between g and delay discounting. Overall, the results suggest that delay discounting is associated with intelligence in part because of processes instantiated in anterior prefrontal cortex, a region known to support the integration of diverse information.