The Famous Marshmallow Test Was Not Really Measuring Willpower

The marshmallow test is one of the most cited findings in all of psychology: leave a small child alone with one marshmallow, promise a second one if they wait, and see how long they last. The original 1990 study found that preschoolers who waited longer went on to have better outcomes years later, and the story that spread from it was simple and appealing. Self-control in a four year old predicts the whole trajectory of a life. A larger, more careful attempt to replicate the finding tells a considerably more complicated story.

In 2018, researchers Tyler Watts, Greg Duncan, and Haonan Quan repeated the marshmallow test with a much larger and more representative sample than the original study used, including far more children from families without a college educated parent. They found a real correlation between waiting longer and doing better later, but it was roughly half the size originally reported, and shrank by a further two thirds once the researchers statistically accounted for the child's family background, early cognitive ability, and home environment. Most of what predictive power remained came simply from being able to wait at least 20 seconds, not from the longer, more dramatic waits the original study became famous for.

This matters because it points to something other than raw willpower doing a lot of the original work. A child's willingness to wait for a second marshmallow depends heavily on whether that child has actually learned, through real experience, that adults reliably follow through on promises. Children growing up in less stable or less resourced environments have often, entirely reasonably, learned that a promised future reward does not always materialize, which makes waiting a rational response to real uncertainty rather than a simple failure of self-control. Later research examining group influence found something similar: children delay gratification more readily when they see people around them consistently doing the same, suggesting trust and social context shape the behavior as much as any individual trait does.

This matters because it points to something other than raw willpower doing a lot of the original work.

Walter Mischel, who ran the original study, actually anticipated part of this nuance decades before the 2018 replication, describing effective waiting less as raw self-denial and more as a strategy, distracting yourself, reframing the marshmallow as something less tempting, looking away from it entirely, rather than simply gritting through the full waiting period by force of will. That detail rarely made it into the popular retelling of the study, which tended to compress a nuanced finding about strategy and trust into a much simpler, more dramatic story about a fixed trait some children apparently had and others did not.

None of this means delay of gratification is meaningless, and the correlation with later outcomes, smaller as it turns out to be, did not disappear entirely. It means the popular version of the story, that raw individual willpower in a four year old determines the whole shape of an adult life, was always an oversimplification of what the original data could actually support, and a more careful, larger replication has since made that oversimplification difficult to defend.

The practical takeaway runs in the same direction as most of the newer research on discipline. Whether someone waits, resists, or follows through has as much to do with trust, stability, and the reliability of their surrounding environment as it does with some fixed, internal reserve of grit. Building discipline in your own life, or helping a kid build it, probably has less to do with demanding more willpower and more to do with building an environment reliable enough that patience actually pays off the way it is supposed to.