Note: It is with great sadness that we recognize the passing of Walter Mischel on September 12, 2018. Dr. Mischel, the acclaimed psychologist who created the Marshmallow Test, has had a profound impact on the field of psychology, on how personality is understood and studied, as well as on the professionals and families who have learned so much from him.
By Ellen Galinsky Chief Science Officer, Bezos Family Foundation
In the late 1960s, a Stanford University professor, Walter Mischel, pioneered a simple and visual way to measure “delayed gratification in young children” — the Marshmallow Test. Over time, as evidence accumulated about the long-term impact of this test, it inspired great fascination and continues to generate interest, even debate, from numerous studies and recent articles, including “Why Rich Kids Are So Good at the Marshmallow Test,” in the June 2018, Atlantic.
What’s the big deal?
In the classic study (as described by Mischel for my book Mind in the Making), 4-year-olds are taken into a lab room with a one-way mirror so adults can see into the room but the children can’t see out. The experimenter puts a plate in front of each child being tested. On one side of the plate is a single marshmallow (or whatever treat the child chooses) and on the other side are two marshmallows or treats.
The experimenter explains that she must leave the room, but if the child can wait, the child can have the two marshmallows or treats. If the child doesn’t want to wait (15 minutes or sometimes longer in various experiments), he can ring the “bring-me-back” bellfor the experimenter to return and he can have the single marshmallow.
When the child makes a choice—two marshmallows later, or one now—he inevitably wants two, thus setting a goal. Children are told that there is no right or wrong way to play this game so they aren’t playing to please the adult.
This study began in the nursery school Mischel’s children attended: the Bing School at Stanford University. Years later, as he heard his children talk about their friends in high school, Mischel began to detect a pattern: those children who had waited longer for the marshmallows seemed to be doing better. That prompted Mischel and others to re-contact children from the experiment and to follow as many as possible over time. He found that:
The longer the young children were able to wait at age 4, the better the SAT scores, the better their ability to control themselves and to pursue their academic and other goals successfully—[as reported] by parents and by teachers, and the better their own reports about how they were doing as people in their early 30s.
Children who did better on the original Marshmallow Test had more lifelong success but…
In this video Mind in the Making made of the Marshmallow Test, Mischel, explains his experiment. Note: he says that while the Marshmallow Test shows strong correlations between delayed gratification and school and life success, it doesn’t mean that a child who doesn’t do well on the Marshmallow Test is in any way doomed. These are skills that can be taught.
Any study that becomes iconic is bound to trigger replications. Some researchers have tried to dig deeper into what the experiment really measures; others try to see if these skills have declined or increased over time. Still others try to replicate or topple the findings using an array of statistical controls.
Below are some key findings.
The Marshmallow Test measures self-control
In 2013, Angela Duckworth and colleagues at the University of Pennsylvania analyzed two longitudinal datasets that used the Marshmallow Test: one with 5thgraders using a 30-minute version of the test, and one with 54-month-olds using a 7-minute version from a national dataset.
After they applied statistical controls in their first analytic model for variables including children’s intelligence, family income and maternal education in the version with the 54-month-olds, thefindings weren’t as strong as the original Mischel study. However they held up: children’s performance on the Marshmallow Test at 54 months predicted participants’ high school grades, standardized achievement test scores and their physical health as 15-years-olds.
These mixed results are not surprising, given that the sample was much more diverse than the original Mischel sample of more advantaged children.
A primary purpose of the Duckworth study was to ask: What is the Marshmallow Test really measuring? They found that how long children waited for the larger treat was consistently associated with how teachers (in study 1 with 5thgraders), and parents and teachers (in study 2 with 54-month-olds), rated children’s self-control, assessed by the child’s ability to concentrate on an activity, to plan and to use self-control to resist temptation. In fact, in the final analytic model, self-control itself was significantly predictive of later adolescent outcomes.
Interestingly, the researchers found that the amount of time children spent waiting was not associated with how hungry the children were (perhaps because the treats weren’t enough food to over-ride children’s self-control), or how attracted the children were to rewards. They also found that waiting time wasn’t related to the child’s personality except for conscientiousness. As the researchers say, their analyses suggest that the power of the Marshmallow Test “derives primarily from its assessment of self-control.”
While hunger doesn’t necessarily matter, the reliability of the experimenter does
A study conducted by Celeste Kidd published in 2013, now of the University of California, Berkeley, and her colleagues, experimented with children’s responses to reliable or unreliable adults. Prior to the Marshmallow Test, the experimenter gave children used crayons in a tight jar for an art project and asked them to wait, promising to provide new art materials soon.
In the “unreliable” condition, the experimenter didn’t follow through, saying, “We don’t have any other art supplies after all,” suggesting that children open and use the jar of well-worn crayons. In the “reliable” condition, the experimenter returned with exciting art supplies. That experience was repeated with promises of stickers, which children in the reliable condition got and children in the unreliable condition didn’t get.
When the experimenters then conducted the Marshmallow Test, children in the reliable conditions waited 12 minutes and 2 seconds on average versus 3 minutes 2 seconds in the unreliable group.
Children in the 2000s can’t/can wait longer than children in earlier decades
When I talk about the Marshmallow Test, I inevitably hear that “kids today” can’t wait as long for the larger treat as children in the past. Stephanie Carlson of the University of Minnesota and her colleagues have heard the same thing. When theytested this common wisdom with a survey of adults, they found that most adults believe that children wouldn’t wait as long as children 50 years ago and that children’s ability to control themselves would have decreased.
These researchers were in a unique position to then compare this common wisdom with delay of gratification data collected with 3- to 5-year-olds in the 1960s by Mischel and his colleagues, in the 1980s by Larry Aber and his colleagues at Barnard College, and in the 2000s by Stephanie Carlson and colleagues at the University of Washington and at the University of Minnesota.
In the Carlson’s 2018 article, findings illustrated that children in the 2000s waited on average 2 minutes longer than children in the 1960s, and 1 minute longer than children in the 1980s. The children in these studies were predominately more advantaged children, like the children in the original study at the Bing School at Stanford.
Poverty is not destiny
Another article was published in 2018 in which Tyler Watts of New York University and his colleagues revisited the same longitudinal dataset used by Angela Duckworth and her colleagues in 2013. In this re-analysis, the research team concentrated on children whose mothers had not completed college when their children were 1-month-old, using the 7-minute variation of the longer Marshmallow Test with children at 54 months. Like the Duckworth team, they found that delay time was predictive of children’s standard achievement scores at age 15. The findings weren’t as strong as the Mischel study, but they held up.
When they applied statistical controls for the child’s background and early home environment, some of the findings remained but when they further controlled for the child’s current cognitive abilities and behavioral problems, the findings became non-significant.
Does that mean that The Marshmallow Test has been toppled or it is only predictive for children of privilege? Does it mean that poverty is too hard to overcome, as some of the coverage of this study stated?
I think that would be a huge misreading of this and the other Marshmallow Test studies conducted over the years.
- Not all children are the same: neither low-income children nor children of privilege.
- The control variables in the Watts study included more than family income. They included factors that can be changed, like the child’s home environment.
- That measure they used, in fact, is designed to capture aspects of the home environment known to support positive cognitive, emotional and behavioral functioning, all of which can vary and can change.
- Poverty itself, though very difficult to manage, is not destiny, and low-income children can be in environments that positively support their development.
The Marshmallow Test is a measure of executive function skills
In a recent video, Stephanie Carlson and Philip Zelazo co-directors of the Carlson and Zelazo Lab at the University of Minnesota, describe how the Marshmallow Test calls on executive function skills, a specific set of attention-regulation skills involved in conscious goal-directed problem solving, including cognitive flexibility, working memory and inhibitory control.
In the Marshmallow Test, the children set a goal: to have two marshmallows or treats rather than one. They have to remember the rules (working memory), to think flexibly and to use inhibitory control and stop themselves from grabbing the treat now. This task calls on both cool and hot EF skills, cooler thinking strategies within a high-stakes, hot emotional situation. This is important because EF skills have been found to be predictive of school readiness as well as school and life success and because they are malleable and can change.
The Marshmallow Test remains iconic
Watts and his colleagues conclude their article by arguing that if we are to develop interventions to improve children’s school and life success, we should target more than delay of gratification. I agree we need stronger interventions, especially those that target parenting skills andchildren’s skills—importantly including self-control but other skills as well.
The Marshmallow Test has stood the test of time. The debates in the field and the re-analyses it has triggered, have led the way to a more solid basis on which to help children and families thrive. As Mischel said: “I think in summary, what we’ve found is a very simple and direct way of measuring a competence that seems to make an important life difference.”