Beyond the Marshmallow Test

What makes some people successful in life and others not? The classical answer to this question is that those who succeed must be the "right sort of person," possessing the quality of excellence (Greek arete). Aristotle suggested that this excellence could be developed through deliberate practice, while Plato proposed that it was inborn. In modern behavioral science, people often propose that excellence is tied to certain behavioral tendencies like self-control or executive functioning, which are in turn supposed to be tied to one's genetic characteristics. This is not too different from Plato: Some people are just born to be more excellent than others. The ancient philosophers had a blind spot when it came to the role of life circumstances or opportunities in producing excellence -- the word arete, in fact, comes from the same root as the word aristocrat, with the implication that the rulers of society also had excellent character. The idea that excellence is also at least in part the result of one's environment was taken seriously for the first time in the Enlightenment era, with John Locke proposing the mind began as a blank slate that would be shaped by one's environment, and Jean-Jacques Rousseau going further to argue that people are innately excellent but then are corrupted by society!  

A key piece of research that is often pointed to as evidence for innate differences in willpower or self-control is called "the marshmallow test," or more properly the "Stanford marshmallow experiment." The basic study paradigm involves putting a preschooler in a room with a marshmallow, telling them to leave it alone, and then leaving the room. The children are told that they will receive two marshmallows if the first one is still there when the experimenter returns. As you might expect, some children immediately gobble the marshmallow as soon as the experimenter's back is turned. Others resist temptation for a while (it's interesting to watch the videos of them attempting to do this) but ultimately fail. And still others are able to hold out the full time and then receive the promised reward. Psychologist Walter Mischel's original study in 1970 was conceived as "delayed gratification" research, and his goal was to find out whether children at this age could delay gratification, and what helped them to do it. Contrary to the dominant psychodynamic drive theory of the time, Mischel found that very young children could keep themselves from eating the marshmallow, and that they did it mainly by way of distraction rather than by reminding themselves of what they stood to gain if they waited. In particular, children had longer successful wait times if the marshmallow was hidden from view than if they looked at it while they waited. Mischel's original study has been replicated many times, including in nonhuman primates, parrots, and even cuttlefish, all of whom show an ability for self-control.

Mischel interpreted the original findings in terms of frustrative nonreward, a lesser-known model that fits better with behavioral learning theories and also has been linked to neurological findings. When people can't access a promised reward, for instance, they show increased activity in the limbic system (Intuitive system) and decreased activation in the prefrontal cortex (Narrative system). Mischel calls these the "hot" and "cool" thinking systems, referring to the predominance of emotion in the Intuitive mind; these are just other terms used in the literature that seem to refer to the same underlying concepts. In the frustrative nonreward literature, people with increased limbic system activation have shown a tendency toward impulsive behavior, anger, or stress. On the other hand, they also may show greater learning and an increase in determination to press on to achieve the delayed reward. Interestingly, children with attention deficit disorder show the same level of limbic activation (not more); they just don't show the corresponding increase in learning. These two competing responses to frustration can quickly take us right back to Plato's "two horses" metaphor -- the human spirit striving for better things while bodily urges drag us backward -- and lead to moralizing explanations in which some people show more virtuous behavior than others in the face of temptation.

None of that, of course, is the reason that you have probably heard about the marshmallow test. The study became famous because of longitudinal results published in 1988, when the original preschoolers were adolescents. Researchers found that these same children were reliably different in adolescence and young adulthood, with the length of time that they waited for the marshmallow when they were preschoolers predicting parents' ratings of their later ability to tolerate frustration, rational planning abilities, verbal fluency, social competence, attentiveness, and academic success. Additionally, the children who waited longer for the marshmallow had higher scores on the SAT college entrance exam, with correlations of r = .42 (SAT verbal) to .57 (SAT quantitative) between scores and wait duration as a preschooler. These are moderately large correlations, and are particularly impressive given that they were based on a child's performance in a single contrived situation 15 years previously. Replications of the study have confirmed associations between preschoolers' marshmallow test results and their later academic achievement, social skills, and absence of behavior problems based on standardized measures. At ages 27-32, the children who delayed gratification in preschool were less likely to be obese. And a study published in 2011 found that the same children had better performance 40 years later on a task that required them to suppress responses (i.e., press this button in response to a happy face on the screen but not in response to a sad face). Brain imaging using fMRI directly confirmed Mischel's hypothesis that those who were able to successfully inhibit responses had more activation in the prefrontal cortex while those who could not had more limbic system activity (ventral striatum) -- a direct neurological demonstration of the brain's two minds at work. 

Because the original marshmallow research was conducted with such young children, commentators have often used it to argue that self-control has a genetic basis. There is some justification for this view, although the data don't come directly from the original marshmallow test participants. Both of the brain areas implicated in self-control are responsive to serotonin, and in fact some research has shown a direct correlation between ability to delay gratification and low serotonin levels in the ventral striatum area of the brain's limbic system. Additionally, differences in the serotonin transporter gene (5HTT-LPR) have been shown to correlate with differences in children's impulsive behavior. In addition to a direct genetic effect on behavior, greater impulsivity might lead children to be more engaged with delinquent peers, with low self-control and number of delinquent peers being two of the strongest predictors of antisocial behavior. There is also some evidence that more impulsive boys elicit different parenting responses from their mothers, so that even a mother's parenting response (which we might assume to be a modifiable environmental factor) can be predicted by the child's genetic profile on the serotonin transporter gene!

This seems like a pretty conclusive chain of evidence, which might be illustrated as follows:

5HT transporter gene -> self-control in preschool -> environmental consequences -> later impulsivity -> academic success, social success, behavior problems -> ultimate success in life

In fact, many people have made exactly this argument, which could be summed up as "self-control is genetic" or "success in life depends on having good genes." Plato might have agreed; but Aristotle, Locke, and Rousseau certainly would not have.

Walter Mischel's own view, as summarized in his 2014 book The Marshmallow Test, is still that self-control is a learned skill. He worked with the producers of Sesame Street to teach it in a series of episodes where Cookie Monster learns to quit gobbling his food and instead joins the "Cookie Connoisseur Club." He insists that the brain's neuroplasticity is sufficient to overcome any pre-existing genetic differences, and that children can do better on self-control tasks if they are taught "metacognition" strategies like distracting themselves, labeling distressing thoughts, and consciously reminding themselves of why it is advantageous to delay gratification. In a 2009 New Yorker article, Mischel expressed a fundamental lack of interest in the fMRI findings and instead suggested that we should be focusing research resources on the development of early childhood programs that teach fundamental life skills -- essentially a return to the Aristotelian "virtue ethics" approach. Unfortunately, Mischel's efforts didn't pan out very well -- he contributed to the development of the KIPP charter schools that have been critiqued for their harsh disciplinary practices and high dropout rates, and his co-investigator Angela Duckworth became famous for promoting the idea of "grit," another concept that has been widely interpreted as genetically determined.

Despite these challenges, Mischel's own perspective in his 2014 book is very much in line with others' critiques of the marshmallow test research. First, participants in the original study were part of preschool housed at Stanford University, often the children of faculty and other high-achieving individuals who were likely already predisposed to succeed. Second, research also consistently finds that children from more affluent families do better on the marshmallow test. One possible explanation is that when someone comes from a low-resource environment, it's simply more adaptive to take any reward as soon as it is offered because it might not be there later on. And certainly research shows that early environment also contributes to impulsiveness and later behavior problems, with any initial genetic differences that are expressed in childhood becoming weaker and weaker in later life. 

Other recent studies have confirmed Mischel's focus on learning as opposed to innate differences. A 2013 study showed that children were more likely to grab the marshmallow when the experimenter had previously broken a promise to them. This finding provides a striking alternate explanation for why children from low-income or minority families might show less "self-control" in an experimental procedure, one that's based on fundamental considerations of justice rather than on genetic predispositions. A 2018 study found smaller relationships between marshmallow test results and later achievement than Mischel's original results, and showed that two-thirds of those relationships could be alternately explained by parents' education, home environment, and early cognitive ability. Both of these studies suggest explanations for the marshmallow test results based on social determinants rather than innate differences -- a better fit with Locke and Rousseau than with Plato.

Finally, some research on the marshmallow test specifically supports Mischel's (and Aristotle's) contention that self-control is a learned skills. The original result showing that children waited longer when they were able to distract themselves has been replicated, and Mischel was involved in later research testing structured practice and modeling interventions, both of which improved self-control. These intervention strategies are consistent with a behavioral learning approach, which is how Mischel interpreted his original results back in 1970. Some other modifications have shown other ways for children to improve their marshmallow test performance: A 2020 study found that children waited longer on the marshmallow test when they were told that their teacher would find out how long they waited -- essentially activating the Narrative system by bringing up social perception concerns. And another recent study found that children waited longer when they worked in teams rather than alone. 

What can we take from all this? First of all, when someone says the marshmallow test "proves" some point, be skeptical. The original findings have been questioned on methodological grounds, they have had weak replications, and they can be interpreted in multiple ways. Second, when someone says the marshmallow test shows that self-control is genetically determined and immutable, be especially skeptical. There are some brain-imaging studies with the original marshmallow test cohort, but the genetic links currently rely on a complicated chain of inference that draws from multiple lines of research rather than having been directly shown with the kids in the marshmallow study. Third, don't assume that any particular policy implication necessarily results from the marshmallow test findings. The innate-differences perspective isn't helpful because some studies have shown that kids' marshmallow test performance can be improved in various ways. Yet Mischel's own idea that self-control can be taught in a particular way is also challenged by the failings of the KIPP schools that he was involved in developing. As usual, it seems likely that human behavior is the product of many weak factors. We can get valid insights from Locke and Rousseau as well as from Plato or Aristotle. And various interventions can likely help us to strengthen the Narrative mind and overcome the weaknesses of the Intuitive mind as we continue the quest for self-control and success.