Skip to main content

The marshmallow test: Self-discipline, poverty, and what really predicts success

You've probably heard the marshmallow test. Researchers put a child in a room with one marshmallow and tell them: "You can eat this now, or if you wait 15 minutes, you get two marshmallows." Most children eat the marshmallow immediately. Some wait.

Decades later, the children who waited have better life outcomes: higher income, better health, less crime, better relationships. The narrative is simple: Self-discipline predicts success. Rich people delayed gratification. Poor people don't. If poor people would just resist marshmallows, they'd be rich.

This test has become gospel in personal finance and child development. It's been cited millions of times. It's informed parenting advice, education policy, and how we judge poor people. It sounds nice. It's also deeply misleading.

Quick definition: The marshmallow test is a 1970s study in delayed gratification where children choose between eating a treat immediately or waiting for a larger reward later. The original study suggested self-discipline predicts life success, but later research revealed confounding variables that change the interpretation entirely.

The original marshmallow test and its misleading conclusions​

The original marshmallow test was conducted in the 1970s by Walter Mischel at Stanford with primarily affluent, educated children. Most of the participants were from wealthy families. The study followed them into adulthood.

The original conclusion was straightforward: Children who delayed gratification had better life outcomes. Higher income. Better health. Better relationships. Better credit. The interpretation was obvious: "Self-discipline matters. Delaying gratification is the path to success."

This narrative stuck because it was clean, satisfying, and ideologically convenient. It suggested that poverty was a character issue, not a system issue. It meant that if poor people would just be more disciplined, they'd be rich. It justified inequality: some people have the discipline to wait for marshmallows; some don't.

The problem is that the original study had a massive confound. All the children were from similar backgrounds. They were mostly white, educated, middle-to-upper class. They had lived experience with a fundamental truth: when an adult promises you something later, it happens.

That promise is credible because their parents have always followed through. When they say "eat your vegetables now and you can have dessert after dinner," dessert actually comes. When they say "do your homework now and you can play after," playtime actually comes. For these children, delayed gratification is rational because promises are kept.

The replication study that changed everything​

Decades later, researchers conducted the marshmallow test again with a broader sample. This time, they included children from low-income backgrounds, children whose parents had experienced broken promises, children whose lived experience taught them that "later" doesn't always come.

The results were completely different.

Self-discipline explained almost none of the outcome differences. The real predictor was whether the child believed the researcher. Children from affluent backgrounds took the deal (wait 15 minutes, get two marshmallows) because they had evidence that delayed gratification works. Adults keep their promises. When you wait, you get rewarded.

Children from low-income backgrounds ate the marshmallow immediately. Not because they had poor impulse control. Not because they lacked self-discipline. But because their lived experience taught them that waiting doesn't pay off. They'd been told "later" and later never came. They'd been promised things that didn't materialize. Their rational brain said: "A marshmallow now is better than a promise of two marshmallows later if promises get broken."

This is crucial: A marshmallow now is better than a promise of two later if you have evidence that promises don't work.

The replication study showed that when researchers pre-demonstrated that they keep their promises (by following through on small promises before the marshmallow test), low-income children who previously ate the marshmallow immediately were now willing to wait. The self-discipline didn't change. The trust in the system changed.

The original marshmallow test didn't measure self-discipline. It measured trust. Specifically, it measured trust in institutions, in promises, in delayed gratification systems. And trust is not evenly distributed. It's a function of experience and history.

Why this matters for understanding poverty​

This reframing is essential for understanding poverty psychology and money behavior. The original marshmallow narrative suggests "poor people are impulsive." The research-backed reality is more complex: "People who have lived through scarcity and broken promises are rationally skeptical of delayed gratification because their history taught them it doesn't work."

This is not a character issue. It's a rational adaptation.

Someone who grew up in poverty has learned through direct experience that:

  • Saving money might be lost to emergency expenses (car breaks, medical crisis)
  • Future promises (pensions, social security) might disappear
  • Delaying spending might mean going without (missing opportunities)
  • Certainty now is more valuable than promises later

This is not impulsiveness. This is rational risk assessment based on a legitimate history of broken promises.

When financial advisors tell poor people "You need to save more and delay gratification," they're often ignoring this lived reality. They're telling someone to trust a system that has not been trustworthy for them. They're asking them to ignore their own evidence and adopt the trust posture of someone who has experienced a reliable world.

The limitations of the marshmallow test​

Even if self-discipline did matter (and the evidence is much weaker than originally claimed), the marshmallow test only explained a tiny portion of outcome variance.

The original longitudinal study showed that self-discipline explained about 10% of the variance in adult outcomes. Everything else mattered way more:

  • Family wealth and socioeconomic status
  • Access to education
  • Luck and timing
  • Network and connections
  • Health
  • Race and discrimination
  • Geographic opportunity
  • Family stability

The marshmallow test became famous because it was the simple, clean, satisfying story. But it was wildly overstated. A 10% effect size is not nothing, but it's not determinative either. It's one factor among many.

The children who waited for marshmallows grew up in families with money, stability, education, and optimism about the future. Those families also paid for better schools, better neighborhoods, better healthcare, better opportunities. The children who delayed gratification had other advantages that made their outcomes predictable.

When you separate self-discipline from those other factors, the effect size shrinks dramatically.

Key takeaways​

  1. Trust, not discipline, predicted marshmallow test results: Children from families where promises were kept waited for the reward. Children from families where promises were broken ate immediately. Same test, opposite outcomes, based on rational trust assessments.

  2. Delayed gratification is rational only when you have evidence it works: If promises are kept and savings compound, delaying is smart. If promises break and savings disappear, delaying is foolish. The behavior isn't irrational; it's context-dependent.

  3. Self-discipline is necessary but not sufficient for financial success: You need to delay gratification, but you also need opportunities, information, luck, circumstances. The person making $35,000 and never buying coffee is not saving their way to wealth. They're just poor and thirsty.

  4. The marshmallow test measures background, not character: Self-discipline was only one small factor. Family history, trust, resources, opportunity, and luck mattered far more.

  5. Poverty teaches rational skepticism about delayed gratification: This isn't a flaw. It's an adaptation to a world where promises don't materialize reliably.

  6. Scarcity makes everyone high time preference: When money is tight, immediate consumption becomes more valuable (and rightly so). In scarcity, present bias is rational. Criticizing poor people for spending now ignores the logic of their situation.

Scarcity mindset vs. abundance mindset​

The reframing requires understanding scarcity versus abundance psychology.

When you're in scarcity (money is tight, resources are uncertain), your brain is optimized for immediate needs. You focus on the present because the future is uncertain. Delayed gratification is a luxury good—it assumes security. In scarcity, a bird in hand really is better than two in the bush because certainty is rare.

When you're in abundance (basic needs are secure, resources are stable, tomorrow is guaranteed), delayed gratification becomes more rational. You can afford to wait because you know the future is coming.

The same person can be "impulsive" in scarcity and "disciplined" in abundance, depending on their nervous system's state and the actual material conditions they face, not on their character.

This is documented in behavioral economics research. Studies show that thinking about scarcity reduces cognitive capacity and impulse control. It's not that poor people have less discipline; it's that scarcity itself reduces the capacity to delay gratification. The brain is responding rationally to material conditions.

Marshmallow test research: The fuller picture​

The most important follow-up to the marshmallow test came from researchers who looked at a broader sample and controlled for socioeconomic background. When they did, self-discipline still predicted outcomes, but the effect size dropped dramatically. And when they looked at children who came from unstable or low-trust environments, waiting for the marshmallow didn't predict positive outcomes at all. It was actually slightly worse—more waiting, worse outcomes—because the environment had taught them correctly that promises don't materialize.

Mischel's own later work acknowledged these limitations. He recognized that the original study had selected for a particular population and that the replication studies with more diverse samples told a different story.

The institutional response was slower. The narrative of "self-discipline predicts success" was too appealing to be questioned quickly. It confirmed existing biases about poverty (it's a character issue), about wealth (it's earned), and about the rightness of inequality (the disciplined are rich, the undisciplined are poor).

The research actually says: "Self-discipline explains a small portion of outcome variance. Background, opportunity, family stability, and trust in institutions explain most of it."

Common mistakes about the marshmallow test​

  1. Thinking poor people lack self-discipline: They often lack trust in systems that have failed them repeatedly. That's rational, not undisciplined.

  2. Using the marshmallow test to justify inequality: It explains ~10% of variance. The other 90% is opportunities, background, luck, networks.

  3. Applying delay-gratification advice to people in scarcity: Telling someone struggling paycheck-to-paycheck to "delay gratification" ignores that immediate consumption is the rational response to uncertain futures.

  4. Assuming that because waiting "works" it works everywhere: Waiting for the marshmallow works if promises materialize. In environments where they don't, waiting is foolish.

  5. Confusing correlation with causation: The children who waited also had stable families, educated parents, and secure futures. Those factors did as much (or more) work as self-discipline.

  6. Forgetting that self-discipline has costs: Some people are genuinely happier spending on experiences now rather than deferring everything. That's not a failure of discipline; it's a value judgment.

The context matters: When delayed gratification doesn't work​

Some of the most important research comes from studies of actual saving behavior in low-income communities. Researchers found that when saving systems had risk (money might be stolen, lost to emergency, taken by predatory institutions), people didn't save. It wasn't lack of discipline—it was rational risk assessment.

When researchers introduced secure, protected savings accounts with predictable returns, the same people saved at high rates. Their behavior changed not because they developed discipline, but because the system became trustworthy.

In other words: delayed gratification only works if you have rational evidence that it will pay off. Poor people who don't save are often making the correct decision given the unreliable systems they have access to.

Mermaid: Trust and delayed gratification​

Real-world examples and research​

Mischel's own follow-up research: In later work, Mischel found that the delay ability was related to trust in the experimenter and in the system. When he told children "I have a big marshmallow here for kids who wait," more waited. When he said "I also have a small one," the percentage who waited dropped. Children rationally evaluated their probability of getting the reward.

The "waiting room" effect: A crucial detail from later replications: some children were in a room with reliable adults (teachers, researchers, parents who kept promises). Others were in rooms with less reliable actors. The children in reliable rooms waited. The children in unreliable rooms ate. Same kids, different behaviors based on context.

Fels Longitudinal Study data on delayed gratification: Long-term follow-up data shows that patience in adolescence predicted earnings, but the effect was much smaller when controlling for parental education, family income, and family stability. The advantages came from the background, not just the patience.

Behavioral economics research on savings in scarcity: Studies of microfinance clients show that when you protect savings from risk and guarantee returns, saving rates increase dramatically—even in populations that were "undisciplined" before. The discipline was always there; the system was unreliable.

FAQ: Marshmallow test questions​

Q: Does the marshmallow test measure anything meaningful?

A: It measures trust in promises and systems. Which is meaningful, but not what we thought it measured. It's a proxy for background, not self-discipline per se.

Q: Should I teach my kids to delay gratification?

A: Yes, but contextualize it. Teach them that delaying works when you have evidence it will pay off. Build that evidence by keeping promises. When you say "wait and you get two," actually deliver two. Create a reliable system.

Q: If self-discipline only explains 10% of outcomes, why does it matter?

A: Because 10% is real. The issue is not whether discipline matters; it's how much it matters relative to everything else. And one of the ways to develop discipline is through systems (automation, commitment devices) rather than through willpower alone.

Q: How do I know if I'm being impulsive or being rational about immediate needs?

A: Ask: "Do I have evidence that delayed gratification will pay off?" If yes and you're still choosing now, that's impulsiveness. If no and you're choosing now, that's rationality. This varies by context.

Q: Why do some wealthy people spend everything immediately?

A: Because they have evidence that they can. They know they'll earn more. For them, immediate spending on experiences is rational. Discipline isn't about never spending; it's about making conscious choices based on evidence.

Q: Is it wrong to spend money now when financial advisors say I should save?

A: Not necessarily. If you've lived through systems where saving didn't work, or if your present needs are pressing, spending now is rational. The advice to save assumes you have security and reliable systems. If you don't, the advice might be misaligned with your actual situation.

Q: How do I talk to my partner about their spending if they grew up poor?

A: Don't frame it as discipline. Frame it as building trustworthy systems together. "We both experienced different things with money. Let's build systems where we both feel secure—some savings, some spending." Acknowledge that their skepticism of delayed gratification makes sense given their history.

Q: Can scarcity mindset change?

A: Yes, but slowly. When someone experiences real, sustained security—money in the bank, predictable income, kept promises from institutions—their time preference gradually shifts. They start naturally delaying gratification more. It's not willpower changing; it's material conditions changing.

The neurobiology of trust and time preference​

Recent neuroscience research shows that trust and time preference are neurologically linked. The same brain regions that process social trust also process time preference. When you trust someone or trust a system, your brain is more willing to delay gratification.

This is not metaphorical. The neural pathways are literally connected. Your ability to delay gratification isn't isolated to willpower—it's embedded in your assessment of trustworthiness.

This explains why the replication study with pre-demonstration of trustworthiness worked. By demonstrating that the researcher keeps promises, they activated the neural pathways associated with both trust and patience.

What the marshmallow test actually teaches about money psychology​

The real lesson of the marshmallow test is this: Behavior is rational given the person's history and context. Children who ate the marshmallow weren't failing. They were succeeding at a different optimization—maximizing certainty in an uncertain world.

For money psychology, this means:

  • If someone doesn't save, don't assume laziness. Ask whether the savings systems they have access to are reliable.
  • If someone spends immediately, don't assume impulsiveness. Ask whether they have evidence that delayed gratification pays off.
  • If someone is suspicious of financial institutions, don't assume irrationality. Ask what past experiences taught them to be suspicious.

The person who is "undisciplined" about spending might be correctly responding to a system that hasn't earned their trust. The person who is "disciplined" about saving might be responding to systems that have always been reliable.

Summary​

The marshmallow test is not what we thought it was. It doesn't measure self-discipline predicting success. It measures trust in systems predicting outcomes. Children from families where promises are kept waited. Children from families where promises break ate. Same test, opposite outcomes, based on rational trust.

Self-discipline matters for financial success—about 10% of the variance. Everything else (background, opportunity, resources, luck, networks) matters more. Using the marshmallow test to justify inequality or to blame poverty on character is not supported by research.

The lesson for personal finance: Build reliable systems so that delayed gratification actually works. Then, with evidence that delayed gratification pays off, people naturally become more disciplined. You don't need willpower if the system is trustworthy.

And if you grew up in scarcity or broken promises, your skepticism about delayed gratification makes sense. You've been rationally adapting to an unreliable world. As you build more reliable systems (or move into more secure circumstances), your behavior will naturally shift.

Next​

→ Couples and money fights — why