Carol Dweck’s growth mindset theory has shaped curricula at hundreds of school districts, training programs at Fortune 500 companies, and a generation of parenting advice. Large meta-analyses now show effects of r = 0.10 --- about a tenth the size the popular version suggests. The honest story is more interesting than either “growth mindset works” or “growth mindset is fake.” Here is the most nuanced case in the replication crisis, and what to do when the field is genuinely divided.

If you have spent any time in an American school district in the last fifteen years, you have probably encountered growth mindset. The poster on the classroom wall. The teacher training session. The professional development workshop. The parent newsletter explaining why we say “you worked hard” instead of “you’re smart.” The whole curricular and pedagogical movement, drawn from the work of Stanford psychologist Carol Dweck, that has reshaped how a generation of American schools thinks about effort, intelligence, and academic achievement.

Growth mindset is one of the most consequential pieces of behavioral-science research of the last twenty-five years. It has shaped public education at enormous scale. It has shaped corporate training programs at Fortune 500 companies. It has shaped how millions of parents talk to their children about learning. Carol Dweck’s 2006 book Mindset has sold over a million copies. The TED talks have been viewed tens of millions of times. The framework has been adopted, in some form, by educational institutions across the United States, the United Kingdom, and dozens of other countries.

The empirical evidence behind all of this is real. It is also much smaller than the popular versions suggest.

This article is about the most uncomfortable replication-crisis story in this hub --- uncomfortable not because the construct is fake (it isn’t), or because the original researcher disavowed her work (she hasn’t), or because the field has reached consensus (it hasn’t). It’s uncomfortable because the honest answer is “the effect is real but a tenth the size you were told,” and “real but a tenth the size” is not a useful framing for institutional decisions that were made on the assumption that the effect was much larger.

Growth mindset is the case study where you have to decide what to do when the field is divided, the evidence is messy, and your existing strategy was built on the bigger version of the claim. It’s also the case where the strategist takeaways are the most directly applicable.

What Dweck’s Foundational Work Actually Said

The growth mindset framework grew out of decades of work by Carol Dweck on motivation, achievement, and how people respond to failure. The foundational empirical work is older than the popular framework; the framework was synthesized from earlier research starting in the early 2000s.

The most-cited empirical paper is Mueller & Dweck (1998), “Praise for intelligence can undermine children’s motivation and performance,” in the Journal of Personality and Social Psychology. The paper reported six studies of approximately 400 fifth-graders total. The finding: children praised for intelligence (“you must be smart at this”) chose easier subsequent tasks than children praised for effort (“you must have worked hard at these problems”). The intelligence-praised children also showed more negative responses to subsequent failure --- lower persistence, lower self-rated ability, lower enjoyment.

The interpretation was elegant. Praising intelligence implicitly communicates that performance reflects fixed ability. When subsequent failure occurs, the child has no good explanation other than “I’m not smart at this.” Praising effort implicitly communicates that performance reflects work. When failure occurs, the natural attribution is “I didn’t work hard enough” --- a more actionable response that supports persistence.

A second foundational paper is Blackwell, Trzesniewski & Dweck (2007), “Implicit theories of intelligence predict achievement across an adolescent transition,” in Child Development. This paper tracked seventh-grade students over two years and found that students who endorsed “growth” theories of intelligence (intelligence can be developed) showed steeper improvement in math grades than students who endorsed “fixed” theories.

These studies, combined with Dweck’s 2006 popular book Mindset, became the foundation of the growth-mindset movement. The framework that emerged was: intelligence beliefs are malleable; you can teach students to hold growth rather than fixed mindsets; doing so will improve their academic outcomes. The implication for schools, parents, and managers was enormous. If you could change the way people thought about their own abilities, you could change their achievement trajectories.

By the early 2010s, growth mindset had become one of the most influential frameworks in American education. School districts adopted growth-mindset curricula. Foundations funded growth-mindset interventions. Teacher training programs incorporated growth-mindset principles. The framework felt revolutionary because it offered a tractable, evidence-based intervention for the most stubborn problem in education --- that some students learn and others don’t, and the difference seems to be at least partly about something other than raw ability.

Then the replications started arriving.

The 2018 Meta-Analytic Verdict

In 2018, Sisk, Burgoyne, Sun, Butler & Macnamara published “To what extent and under which circumstances are growth mind-sets important to academic achievement? Two meta-analyses,” in Psychological Science. This was the most comprehensive meta-analytic test of the growth-mindset literature to that point.

The first meta-analysis examined the correlational evidence: does endorsing a growth mindset predict academic achievement? They aggregated 273 effect sizes from approximately 365,000 students. The correlation between mindset and achievement was r ≈ 0.10 --- a very small effect, accounting for roughly 1% of the variance in achievement.

The second meta-analysis examined the experimental evidence: do growth-mindset interventions improve achievement? They aggregated 43 intervention studies covering approximately 57,000 students. The average effect of growth-mindset interventions on academic achievement was d ≈ 0.08 --- a small effect that was meaningfully larger only in specific subgroups (academically high-risk students and students from low-SES backgrounds).

The Sisk meta-analysis also documented substantial publication-bias signals in the literature. The pattern of effect sizes across studies looked like a literature where successful interventions were preferentially published and unsuccessful ones were preferentially not. After publication-bias corrections, the underlying effect was likely even smaller than the unadjusted estimates.

This was, in academic terms, a serious blow to the popular framing of growth mindset. The popular framing implied that growth-mindset interventions could meaningfully change academic trajectories at scale. The Sisk evidence suggested the average effect of such interventions was very small --- probably smaller than other available interventions (better teaching, more time on task, individualized tutoring) by a substantial margin.

The 2019 National Study That Complicated the Picture

If Sisk 2018 had been the only major test, the story would be relatively clean: growth mindset effects are small, smaller than popular versions claim, and not a particularly important lever for academic achievement. But the picture got more complicated.

In 2019, Yeager, Hanselman, Walton, et al. published “A national experiment reveals where a growth mindset improves achievement,” in Nature. This was the National Study of Learning Mindsets, an unusually large preregistered randomized controlled trial: 12,490 ninth-graders across 65 schools, with a brief online growth-mindset intervention.

The headline finding was that the intervention produced an average GPA improvement of about 0.10 grade points. That sounds small --- and it is --- but it varied substantially across subgroups. Lower-achieving students showed larger effects (around 0.10 GPA points). Higher-achieving students showed roughly null effects. The effect was also larger in schools with norms more supportive of academic achievement.

The Yeager study was the strongest evidence to date that growth-mindset interventions, when carefully designed and deployed in specific conditions, could produce measurable academic effects. It was also evidence that the average effect was small (consistent with Sisk’s meta-analysis) and that the population-level intervention is much less impactful than the popular version suggests.

How to interpret Yeager 2019 depends on how you frame the question. If the question is “does growth-mindset intervention work at all?” --- the answer is yes, in specific subgroups, at small magnitude. If the question is “is growth-mindset intervention a meaningful lever for closing achievement gaps?” --- the answer is more equivocal. A 0.10 GPA improvement in lower-achieving students is real, but it’s smaller than the effects of many other interventions, and it would not by itself close meaningful achievement gaps.

The 2019 UK Trial That Found Nothing

Meanwhile, a different kind of growth-mindset evidence was emerging from outside the United States. The UK Education Endowment Foundation funded several large-scale randomized trials of growth-mindset interventions in British schools.

Foliano, Rolfe, Buzzeo, Runge & Wilkinson (2019) reported on the “Changing Mindsets” trial: approximately 5,000 Year 5 pupils across 101 schools, randomly assigned to receive growth-mindset training, with attainment outcomes measured at the end of Year 6. The trial measured impact on math and reading achievement.

The result: no significant impact on either outcome. The intervention, as deployed in a large UK sample, did not improve the achievement measures it was designed to improve.

This was an important counterweight to the Yeager 2019 result. A large preregistered RCT in a different national context, using a careful growth-mindset intervention, found null effects on the central outcome. The honest interpretation: growth-mindset interventions are highly conditional on context, implementation, and population. Sometimes they produce small effects in specific subgroups (Yeager 2019). Sometimes they produce null effects on average (Foliano 2019). Whether they will work in any given context is hard to predict in advance.

A 2023 update from Macnamara & Burgoyne in Psychological Bulletin synthesized the post-2018 literature and reinforced the basic Sisk verdict: aggregate effects remain small (d typically 0.05 to 0.10), substantial publication bias is present in the literature, and there is no robust general benefit of growth-mindset interventions across populations.

Why the Original Looked Bigger

The growth-mindset story is the cleanest available example of how a research program can become a cultural movement at scale before the field has aggregated rigorous evidence on the underlying effect sizes.

The original effect sizes were inflated. Mueller & Dweck 1998 reported substantial differences between intelligence-praise and effort-praise conditions. Subsequent larger studies have generally found smaller effects when the same paradigm is tested at scale. This is the standard winner’s curse pattern --- original small-sample studies tend to detect inflated effects.

The original was a praise-manipulation study; the cultural framework was much broader. Mueller & Dweck 1998 specifically studied how praise affects subsequent task choice and persistence in elementary-school children. The popular growth-mindset framework extends this finding to claims about beliefs, abilities, learning across the lifespan, and institutional culture. The extension was always greater than the evidence supported. Even if the original praise effect is real and replicable, the broader claims about “fixed vs growth mindset” as a stable trait that predicts life outcomes are much weaker.

The framework filled a real institutional need. American schools had been searching for tractable, scalable interventions to address persistent achievement gaps. Growth mindset offered something that looked deployable --- a brief workshop, a curricular change, a poster on the wall. The institutional appetite for the framework was enormous. Foundations funded growth-mindset research and interventions. School districts adopted growth-mindset curricula. The institutional momentum produced enormous cultural amplification, much of which got ahead of the evidence.

Dweck and her collaborators were excellent communicators. Dweck’s book is well-written. Her TED talks are compelling. Her framing --- the distinction between “fixed” and “growth” mindset --- is memorable and immediately actionable. Compelling communication is not evidence, but it does produce cultural reach, and cultural reach can sustain a construct’s influence regardless of the evidentiary base.

The replication evidence took fifteen years to arrive. Mueller & Dweck 1998 → Blackwell, Trzesniewski & Dweck 2007 → Sisk 2018. Twenty years from the foundational paper to the first definitive large-scale meta-analysis. During those twenty years, the cultural and institutional infrastructure around growth mindset grew enormously. When the meta-analytic evidence finally arrived, it had to overcome a fully formed movement.

The Honest Verdict Today

Three layers, in order from most-supported to least-supported.

Layer 1: Some growth-mindset effects are real but small. The aggregate meta-analytic correlation between mindset and achievement is around r = 0.10. The aggregate effect of mindset interventions on achievement is around d = 0.08. These are real, replicable effects. They are also small enough that they will not produce dramatic visible changes in most individual cases, and they account for a tiny fraction of the variance in academic outcomes.

Layer 2: Effects are larger in specific subgroups. Yeager 2019 and other follow-up studies suggest the effects of growth-mindset interventions are larger in academically lower-achieving students and in supportive school environments. The average effect masks meaningful heterogeneity. If you are designing an intervention for high-risk students in a supportive school, the expected effect is larger than the population average. If you are designing a population-level intervention, the expected effect is the small population average.

Layer 3: The strong popular claims are not supported. The claim that growth-mindset interventions can dramatically improve achievement at scale, close meaningful achievement gaps, or transform school cultures, is not supported by the current evidence. The strong claims that drove the institutional adoption of growth-mindset curricula in the 2010s overstated what the underlying evidence supported.

This is a genuinely uncomfortable place for the field to be. The construct isn’t fake. The original researcher hasn’t disavowed. Some effects are real. But the popular and institutional versions claimed much more than the evidence supports, and a generation of educational policy was built on the bigger version.

What This Means If You’re a Strategist

Growth mindset is the most directly applicable case in this hub because the strategist takeaways are about what to do when an intervention is real but smaller than promised. This is the modal situation in behavioral science applied to organizations.

1. “Real but small” is the most common honest finding in applied behavioral science. Most behavioral-science interventions, when tested rigorously at scale, produce effects in the d = 0.05 to d = 0.20 range. These effects are real, but they’re often smaller than the next-best intervention available, and smaller than the effects suggested by single-study original findings. Building strategy around the assumption that interventions will produce effects of d = 0.50 or larger --- which is what the popular versions of most behavioral-science findings claim --- is systematically miscalibrated.

For organizational decisions: when you are evaluating whether to deploy a behavioral-science intervention (a new training program, a new culture initiative, a new performance-management framework), the realistic expected effect is small. The question is not “will this work?” --- the answer is usually “yes, a little.” The question is whether the expected small effect is the best available use of your time, attention, and budget. Often, larger but less glamorous interventions (better hiring, better tooling, better incentives) will produce larger effects than behavioral-science-branded interventions.

2. Subgroup effects are real but often not generalizable to your specific subgroups. Yeager 2019 found that growth-mindset interventions produced larger effects in academically lower-achieving students. This is true for the specific population they studied. Whether it will be true for the specific lower-achieving students in your organization, school district, or training cohort is much less certain. Subgroup effects often don’t replicate across contexts.

The practical implication: if you read about a behavioral-science intervention that “really works for a specific subgroup,” do not assume it will work for your apparently-similar subgroup. The conditions that produced the original subgroup effect may not be present in your environment. Pilot before scaling.

3. When the field is divided, the honest move is to communicate the uncertainty. Growth mindset is unusual among the entries in this hub because the field is genuinely divided. Reasonable researchers disagree about effect sizes, mechanisms, and policy implications. This is not a case where the verdict is settled.

When you are advising a client, a board, or a team on a contested behavioral-science framework, the temptation is to take a side (“growth mindset works” or “growth mindset doesn’t work”). The more honest move is to communicate the actual state of the evidence: real but small effects, conditional on subgroup and context, contested in the field, smaller than popular versions claim. This is harder to communicate than a clean verdict. It is also more useful for decision-making, because it correctly calibrates expectations for the intervention’s likely impact.

The discipline of communicating evidentiary uncertainty --- without collapsing into either confident endorsement or confident dismissal --- is one of the highest-value habits for anyone whose work involves translating behavioral science into business or institutional decisions.

Sources

This article is part of an ongoing series on famous behavioral-science studies. Other entries cover the Stanford Prison Experiment, power posing, the marshmallow test, ego depletion, the facial feedback hypothesis, Bargh elderly priming, the bystander effect, and the Mozart Effect. The full hub lives at /replication-crisis/.

If you’ve deployed growth-mindset programs in your organization or curriculum and want a careful evidence review of what you can realistically expect, book a consultation.

FAQ

Is growth mindset “real”? Yes, in the sense that some growth-mindset effects are real and replicable. No, in the sense that the popular version of the framework overstates the size and reliability of those effects by approximately an order of magnitude. The honest current verdict is that effects are real but small (d ≈ 0.08 average) and conditional on subgroup and context.

Should schools still teach growth-mindset curricula? This is a policy question that depends on the alternatives and the costs. Growth-mindset programs are relatively inexpensive (a brief workshop, a poster on the wall). At that cost level, even a small effect may be worth the investment, particularly for lower-achieving students who appear to show larger effects. But if growth-mindset programs are crowding out other interventions with larger effect sizes (more tutoring, better instructional materials, smaller class sizes), the opportunity cost may be larger than the benefit.

Has Carol Dweck responded to the meta-analyses? Dweck has engaged with the replication evidence in several ways. She has acknowledged that average effects are small. She has emphasized that the effects are larger in specific subgroups and conditions, which is consistent with the Yeager 2019 finding. She has continued to defend the broader framework while acknowledging methodological refinements are needed. Her position is more nuanced than either “the original was right” or “the original was wrong.”

Is “fixed vs growth mindset” a real psychological distinction? The construct --- that people differ in whether they think intelligence is malleable or fixed --- appears to have some validity as a measurable individual difference. Whether it’s a particularly important individual difference, and whether interventions can meaningfully change it at scale, are separate questions. The current evidence suggests the construct is real but less consequential than the popular version implies.

What other interventions actually work for academic achievement? The largest effect sizes in education research are for things like one-on-one tutoring (d = 0.50 or higher), classroom feedback practices, mastery learning, and certain forms of phonics instruction for early reading. The behavioral-science interventions that get media attention (mindset, grit, character education) tend to have much smaller effect sizes. If you are choosing where to invest educational resources, the boring high-effect interventions are usually a better bet than the glamorous behavioral-science ones.

Does this mean Carol Dweck’s research is “bad”? No. Dweck’s research is methodologically careful and theoretically interesting. The construct she developed is real. The issue is that the popular and institutional adoption of her framework went far beyond what the underlying research supported. This is not the original researcher’s fault --- it’s a function of how cultural amplification works. The lesson is about how we as readers and decision-makers should calibrate our confidence in popular versions of behavioral-science claims, not about whether the original research is worth doing.

replication-crisis behavioral-science education developmental-psychology evidence-evaluation

Free Tool

Built for Experimentation Teams

GrowthLayer is the experimentation platform I built for CRO teams --- test management, AI-powered insights, and pattern recognition across your entire program.

Explore GrowthLayer → (opens in new tab)

· Start Free →

Share this article

LinkedIn (opens in new tab) X / Twitter (opens in new tab)

Copy link

Go deeper

Methodology The PRISM Method Case Studies $30M+ in Results Work Together Services & Mentoring

Experimentation and growth leader. Builds AI-powered tools, runs conversion programs, and writes about economics, behavioral science, and shipping faster.

About LinkedIn Newsletter

← Previous

Loss Aversion: What Survives of Behavioral Economics’ Most Famous Idea

Next →

The Facial Feedback Hypothesis: A Pen in Your Teeth, a Camera in the Room, and a Detective Story About a Methodological Moderator

replication-crisis behavioral-science education developmental-psychology evidence-evaluation
Share this article
LinkedIn (opens in new tab) X / Twitter (opens in new tab)
Atticus Li

Experimentation and growth leader. CXL-certified CRO practitioner, Mindworx-certified behavioral economist (1 of ~1,000 worldwide). 200+ A/B tests across energy, SaaS, fintech, e-commerce, and marketplace verticals.

About LinkedIn Newsletter