The Marshmallow Test Revisited: What Delayed Gratification Actually Predicts

About this psychology article

A four-year-old alone with a marshmallow, promised a second one if they can wait, became one of psychology's most famous images of willpower. Decades of follow-up research tell a more complicated a...

A preschooler sits alone at a table with a single marshmallow in front of her. She's been told that if she can wait, without eating it, until the researcher returns, she'll get a second one. If she can't wait, she can ring a bell and eat the one in front of her immediately, no second marshmallow.

What follows, captured on video in dozens of now-famous clips, is a small child engaged in an intense, visible struggle: covering her eyes, turning her chair away, singing to herself, smelling the marshmallow without touching it, sitting on her hands. Some children last the full waiting period. Others eat the marshmallow within seconds of being left alone.

This is Walter Mischel's marshmallow test, conducted at Stanford beginning in the late 1960s, and it became one of the most widely cited studies in developmental psychology — not primarily because of what happened in the room, but because of a follow-up finding, published decades later, that appeared to show something remarkable: how long a four-year-old waited predicted meaningful outcomes in their life years afterward.

That follow-up claim shaped an enormous amount of parenting advice, self-help content, and popular psychology. It also turned out to be considerably less solid than its popularity suggested, and the more accurate story underneath is, in several ways, more useful than the original headline version.

What Mischel Was Actually Trying to Understand

Mischel's original research question wasn't really about willpower as a fixed trait. He was interested in a more specific, more mechanistic question: what cognitive strategies allow a person to delay an immediate reward in favor of a better, later one, and can those strategies be identified and taught, rather than treated as an innate quality some children simply have more of than others.

His theoretical framework proposed that the mind operates through two competing systems when facing temptation: a "hot" system, emotional and reflexive, that reacts to an immediately available reward as urgently desirable, and a "cool" system, more cognitive and reflective, capable of reframing the situation in ways that make waiting easier. The key experimental manipulation, run across several variations of the study, directly tested this framework by teaching children specific cognitive strategies and observing whether those strategies changed how long they could wait. The methodology and full experimental design are covered in The Marshmallow Test Explained: What Delayed Gratification Really Predicts .

The Manipulation That Mattered More Than the Trait

In several key variations, Mischel's team didn't just observe which children waited longer. They directly instructed different groups of children to use different mental strategies while waiting. Some were told to think about the marshmallow's abstract, non-tempting qualities — imagining it as a round white cloud, for instance, rather than something edible. Others were instructed to focus on its tempting sensory qualities: how sweet and chewy it would taste.

The results were clear and, at the time, genuinely important: children instructed to focus on the treat's abstract qualities waited substantially longer than children instructed to focus on its tempting qualities. This demonstrated something considerably more actionable than "some children have more willpower than others." It showed that a child's ability to delay gratification wasn't a single fixed trait, but was meaningfully shaped by the specific cognitive strategy available to them in the moment. A four-year-old with no reframing strategy might struggle badly. The same child, taught to mentally reframe the marshmallow as something non-tempting, might wait considerably longer.

The Follow-Up Study That Made It Famous

In 1990, Mischel and colleagues Yuichi Shoda and Philip Peake published a longitudinal follow-up, tracking down a number of the original preschool participants years later as adolescents. They reported a correlation between how long a child had waited in the original test and various positive outcomes reported years afterward, including higher SAT scores and better parent-rated social and cognitive competence.

This finding was popularized widely, often stripped of its statistical nuance, into the idea that a four-year-old's self-control on a single afternoon could meaningfully predict their success years later. This version of the story became a staple of parenting books and popular psychology content, generally without much discussion of sample size, the specific and fairly narrow population studied, or the size of the actual correlation being reported.

The Replication That Changed the Story

In 2018, psychologists Tyler Watts, Greg Duncan, and Haonan Quan published a much larger and more methodologically rigorous replication, using a substantially bigger and more socioeconomically and ethnically diverse sample than Mischel's original, mostly Stanford-affiliated preschool population.

The results complicated the popular version of the story considerably. When controlling for family background, particularly socioeconomic status and the mother's education level, the relationship between marshmallow-test performance and later outcomes shrank dramatically, and in some analyses, became statistically negligible. In other words, a substantial portion of what the original study appeared to show as an effect of individual self-control was better explained by the child's broader environment: children from more stable, higher-resource backgrounds tended to both wait longer in the original test and have better long-term outcomes, but the waiting itself wasn't doing nearly as much independent predictive work as the popular narrative suggested.

This matters because it points toward a more accurate underlying explanation. A four-year-old's ability to delay gratification isn't purely an internal trait, fixed and predictive on its own. It's substantially shaped by environmental reliability. A child who has consistently experienced adults following through on promises has good reason to trust that a second marshmallow will actually materialize. A child whose environment has taught them, through repeated experience, that promised rewards often don't arrive has a rational, evidence-based reason to take the reward available right now rather than wait for an uncertain future one.

Why Trust in the Environment May Matter More Than Willpower

A separate and, in some ways, more illuminating study, conducted by researchers Celeste Kidd, Holly Palmeri, and Richard Aslin in 2013, tested this directly. Before the marshmallow task, children were first placed in an interaction with an adult who either reliably followed through on a small promise (providing promised art supplies as stated) or unreliably failed to follow through (promising supplies that never arrived). Children who had just experienced an unreliable adult waited dramatically less time in the subsequent marshmallow test than children who had just experienced a reliable one — in this study, the reliable-environment group waited roughly four times longer on average.

This reframes the entire delayed-gratification narrative in an important way. Waiting for a later, better reward is only a rational strategy if the environment has given a person reason to believe the later reward will actually arrive. A child, or an adult, who has learned through repeated experience that delayed rewards frequently don't materialize is not displaying a self-control deficit by taking the immediate option. They're making a reasonable inference based on real evidence about how trustworthy their environment has been.

How This Connects to Broader Patterns of Trust and Behavior

This distinction has meaningful implications well beyond a preschool experiment. It connects to patterns explored in Why Do Traumatized People Forget Their Childhood? The Science of Amnesia and related research on how early, unpredictable environments shape adult behavior in ways that often look, from the outside, like poor judgment or impulsivity, but frequently reflect learned, reasonable adaptations to genuinely unreliable circumstances.

It also offers a useful corrective to popular narratives around willpower and success more broadly, similar to the corrective offered in The Sunk Cost Fallacy Explained regarding decision-making under uncertainty: behavior that looks irrational or undisciplined from a distance often makes considerably more sense once the actual conditions and history behind it are accounted for.

What the Research Actually Supports

Delay of gratification is teachable, not fixed. The original strategy manipulations remain among the most solid findings from this line of research: specific cognitive reframing techniques measurably improve a person's ability to wait for a better later reward, regardless of their baseline tendency.

Environmental reliability shapes the rationality of waiting. Whether waiting for a delayed reward is a wise strategy depends heavily on how trustworthy the environment providing that reward has proven to be. This has direct implications for parenting, workplace management, and relationships: consistently following through on commitments measurably increases others' willingness to delay gratification and trust future promises.

A single childhood measure doesn't determine a life outcome. The 2018 replication's most important contribution was demonstrating that broader environmental and socioeconomic factors account for much of what the original correlation appeared to show, meaning a four-year-old's marshmallow-test performance is far from a fixed verdict on their future.

Self-control and trust in others are intertwined, not separate. Kidd, Palmeri, and Aslin's research suggests that what looks like a self-control difference is frequently, at least in part, a trust difference, shaped by prior experience with how reliably promised rewards actually arrive.

Understand Your Own Self-Control Patterns

The marshmallow test's real legacy isn't a fixed verdict on willpower. It's a reminder that self-control is shaped by strategy, environment, and trust, not fixed trait alone.

→ Take the Self Awareness Test

Free. No account required.

Want a deeper look at how you weigh immediate versus long-term rewards under real decision-making pressure?

→ Explore the full PsychOS Assessment Lab

Frequently Asked Questions

What was the Marshmallow Test?

The Marshmallow Test was a series of studies led by psychologist Walter Mischel beginning in the late 1960s, in which preschool-aged children were offered a treat and given the choice to eat it immediately or wait alone for a period to receive a second treat as a reward. It became one of the most famous demonstrations of delayed gratification in psychology.

Does the Marshmallow Test really predict future success?

The original 1990 follow-up reported a correlation between waiting time and later outcomes, but a larger, more rigorous 2018 replication found that this relationship shrank considerably once family background and socioeconomic status were accounted for, suggesting the original effect was substantially overstated in popular retellings.

Why do some children wait longer than others?

Multiple factors are involved, including specific cognitive strategies (such as mentally reframing the treat as non-tempting), individual differences in temperament, and, importantly, the child's prior experience with whether adults in their environment reliably follow through on promises.

Is delayed gratification a fixed trait?

No. Mischel's own research demonstrated that specific cognitive strategies could measurably improve a child's ability to delay gratification, indicating the capacity is teachable and context-dependent rather than a fixed, unchangeable trait.

What does the 2013 reliability study add to the original findings?

It found that children who had just experienced an adult failing to follow through on a promise waited dramatically less time in a subsequent marshmallow test, suggesting that willingness to delay gratification is heavily influenced by learned trust in whether delayed rewards will actually materialize.