Rethinking the Marshmallow Experiment
As policymakers begin to pay more attention to “non-cognitive skills” and “grit” there’s been renewed interest in Walter Micshel’s famous marshmallow experiment. In the original experiment young children were left alone in a room with a marshmallow (or a cookie) and told they would receive two marshmallows if they didn’t eat it until the experimenter returned. The ability to hold off on eating the marshmallow was found to be predictive of a variety of positive life outcomes, such as educational attainment and health.
The standard take on the experiment is that it reveals a child’s level self-control. If a child wanted to eat the marshmallow right away, they would do it. The fact that some attempted to wait but couldn’t is a sign that they lack a certain ability possessed by those who were able to wait.
Or perhaps not. In a new paper Penn’s Joseph McGuire and Joseph Kable propose that attempting to wait and then giving in may reflect perfectly rational decision making. Their reasoning is fairly straightforward. Generally the assumption is that as the time you wait increases, the remaining wait time decreases. When you’re in line at the DMV and you think it will be 20 minutes, every minute that passes gets you closer to the front of the line. But that’s not always the case. Perhaps the line is moving so slowly that after two minutes have passed you come to believe it will take a total of 30 minutes. If you were initially willing to wait for 25 minutes, deciding to wait but then giving up after two minutes will be a rational decision that has nothing to do with self-control. This is similar to deciding to wait for both marshmallows but then giving up halfway through. What may initially seem like a lack of self-control could actually be a rational decision based on how much waiting you think is left.
It has often been misreported, but in Mischel’s original experiment he didn’t tell the kids he would be back in 15 minutes, he said he would be gone for a “while” or a “long time.” This means it could have been rational for a child to give in. Fox example, imagine a child believed “a while” meant 10 minutes, but also thought it could mean 45 minutes. If the child was willing to wait for 15 minutes, the decision to not initially eat the marshmallow is rational. But once the 10 minute mark passes, the child may begin to believe it will be an additional 35 minutes, and at that point the rational decision is to eat the marshmallow. Thus giving in does not necessarily signal a lack of self-control.
Of course McGuire and Kable are quick to point out that their theory is not directly applicable to the marshmallow experiments, but more generally to adult situations in which not persisting is rational behavior rather than a sign a person lacks self-control. In fact, they conducted an experiment in which people were asked about reaching certain goals related to dieting, training for a mile run, studying for the LSAT, or practicing piano. In each case McGuire and Kable found that as more time elapsed without the goal being reached, participants increased their predictions of the remaining time it would take to reach the goal. So for example, people who imagined attempting to lose weight thought it would take more additional to time to reach their goal on day 45 than on day 1.
The implication is that attempting to draw conclusions from a person’s persistence or lack thereof is not so clear cut. They may lack self-control, but they could also be making a rational decision based on a revised expectation of how much time remains.
I also think there’s a second, unrelated issue with standard marshmallow test. In many situations the utility of self-control is not merely in how it helps you avoid desirable but costly behaviors (eating the marshmallow), it’s in how it helps you persist in undesirable but beneficial behaviors (fitting parts together at your assembly-line job.) In most practical situations avoiding the distraction is only half the battle. You then need to actively engage in the task you’re fighting the urge to be distracted from. It would be great if everything in life was fascinating, but many things — studying, exercising, cleaning your apartment — require actively engaging in a somewhat dull behavior rather than merely avoiding a behavior.
Imagine that instead of sitting quietly and not eating the marshmallow, participants also had to engage in a dull task. For example, a blue circle and a red circle repeatedly appear on a computer screen. Participants have two seconds to click on the blue circle, and they have to do it until the experimenter returns. Would the results be different? Would there be kids who could wait all day for a second marshmallow, but wouldn’t last five minutes if the task required active participation in a boring activity? I’m not sure. But it seems like such an experiment would do a better job of measuring whether a person can operationalize their self-control to accomplish beneficial tasks.
So here’s the new marshmallow experiment I want to see: People are told exactly how long they’ll have to wait and are given assurances that what they are being told is the truth. That way the initial decision to persist can’t turn into a rational decision to give in. Then, while participants wait they have to continuously engage in an extremely dull task. Will the results be different? What do you think?
McGuire, J.T., & Kable, J.W. (2013). Rational temporal predictions can underlie apparent failures to delay gratification. Psychological Review DOI: 10.1037/a0031910