Point Zero Eight

Growth mindset is probably the most successfully exported idea in modern psychology. It is in schools, in management training, in corporate values statements and on posters. The evidence behind it has been the subject of an unusually pointed academic argument, and the part most relevant to a business is one neither side spends much time on.

Our Grades For These Claims

Applying the scheme from the first article in this series. This literature needs four separate grades.

That the overall intervention effect is small is Grade A. Every meta-analysis we located agrees on this, including those by researchers who defend the theory. The dispute is about whether the small effect is real, not about whether it is small.

That targeted effects exist for at-risk and low-socioeconomic-status students is Grade B. It is reported by the 2018 meta-analysis, by a 2023 meta-analysis, and by the national trial[2].

That growth mindset interventions produce meaningful gains in general populations is Grade D.

That any of this applies to adults in a workplace is ungraded, because we located no evidence either way. That is the most important line in this article and it has its own section.

A Note On Method

Everything here is verified to August 2026.

We obtained the published abstract of the 2018 meta-analyses[1] and portions of the 2023 systematic review including its own hosted copies[4][6].

We did not obtain the national trial paper itself and report it only as three subsequent methodological papers describe it[3][2][7].

Our sources conflict on the number of studies in the 2023 review, and we report the conflict in its own section rather than choosing.

One claim in this article concerns the financial interests of researchers. We have reported it precisely as its source states it, flagged what we could not verify, and given the opposing side its own section.

This article reviews education and psychology research. It is not training, human resources or educational advice, and nothing here is a judgment about any particular programme or provider.

What The Theory Claims

The proposition being tested, in the words of the people testing it.

Mindsets, also called implicit theories, are beliefs about the nature of human attributes such as intelligence. The theory holds that individuals with growth mindsets, being beliefs that attributes are malleable with effort, enjoy many positive outcomes including higher academic achievement, while their peers who have fixed mindsets experience negative outcomes[1].

A later paper states the applied version: students who believe their personal characteristics can change will achieve more than students who believe their characteristics are fixed, and that proponents of the theory have developed interventions to influence students' mindsets, claiming that these interventions lead to large gains in academic achievement[4].

Two observations, ours.

There are two distinct claims here and they need separating. First, that mindset correlates with achievement. Second, that changing mindset improves achievement. The 2018 paper tested them separately, and so does this article.

And note the phrase claiming that these interventions lead to large gains. That characterisation comes from critics of the theory and we flag it as such, but it is the claim the rest of the evidence is measured against.

The 2018 Meta-Analyses

The first serious aggregate test.

Sisk, Burgoyne, Sun, Butler and Macnamara published To What Extent and Under Which Circumstances Are Growth Mind-Sets Important to Academic Achievement? Two Meta-Analyses in Psychological Science[1].

The first meta-analysis examined the strength of the relationship between mindset and academic achievement and potential moderating factors, across k = 273 studies and N = 365,915 participants[1].

The reported conclusion: overall effects were weak, but some results supported specific tenets of the theory, namely, that students with low socioeconomic status or who are academically at risk might benefit from mind-set interventions[6].

On the intervention side, a source reports the overall effect size for interventions was d = 0.080 (p = .010)[2].

Two observations, ours.

The scale of the first analysis is remarkable. Three hundred and sixty-five thousand people is far beyond what most psychological questions ever get, and it makes the "weak" verdict hard to attribute to insufficient data.

And the authors reported the supportive findings alongside the unsupportive ones, which is worth noting given how the subsequent dispute developed.

What Point Zero Eight Means

Translating the number, with our own arithmetic.

A standardised effect of d = 0.080 is small enough that the word "small" understates it. Our own calculations from that figure:

A treated student ends up above the average untreated student about 53.2 percent of the time, against 50 percent by chance alone.

The two distributions overlap by roughly 96.8 percent.

Both figures are ours, computed from the reported effect size under standard normal assumptions, and appear in no paper.

Set it against the other effects this series has reported.

Brainstorming penalty, our own conversion: d = 1.39.

Feedback interventions, average: d = 0.41.

The disputed nudge headline: d = 0.43.

Growth mindset interventions: d = 0.080.

Choice overload, meta-analytic mean: d = 0.02.

The mindset intervention effect sits closer to the choice overload null than to anything this series has treated as actionable.

Significant And Small Are Different Words

A point about the p-value, and it matters for reading any large meta-analysis. This section is our own.

The reported figure is d = 0.080, p = .010[2]. Those two numbers say different things and are routinely conflated.

p = .010 says the effect is probably not exactly zero.

d = 0.080 says it is not much.

Two consequences.

At the sample sizes involved here, a trivial effect will reach significance. That is arithmetic, not a criticism: significance testing asks whether an effect differs from zero, and with hundreds of thousands of observations it can detect differences far below the threshold of anyone caring.

Which means "statistically significant" is close to uninformative in a large meta-analysis, and a provider quoting significance rather than magnitude is quoting the less relevant of the two numbers.

This is the same distinction the first article in this series drew between whether an effect exists and how big it is. Here the gap between those two answers is unusually wide.

The Finding That Survived

The positive result inside the negative one.

The 2018 analysis explored moderators including developmental stage, previous failed classes, experience of situational challenges such as moving to a new school, and eligibility for free or reduced-price school lunch, alongside procedural moderators such as type of control condition and type of intervention[2].

It found that students with previous failed classes and students experiencing economic disadvantage benefitted significantly more than their peers[2].

A source notes that both Sisk and colleagues (2018) and Burnette and colleagues (2023) found that intervention effects were stronger in lower-achieving students and students from low socioeconomic status backgrounds[6].

Three observations, ours.

This is a consistent finding across independent analyses, including from the team most critical of the theory. That is meaningful corroboration.

It is also the same structural resolution this series found in the choice overload literature and the conflict literature: a near-null average concealing a real effect in an identifiable subgroup.

And it points somewhere specific. The intervention appears to do something for people who are struggling or disadvantaged, and little for people who are not. Whether that transfers to a workplace is a separate question this article takes up later.

The National Randomised Trial

The strongest single study, and it is on the supportive side.

The National Study of Learning Mindsets, reported by Yeager and colleagues in 2019, was an experiment conducted on a nationally representative sample of ninth-grade public high-school students in the United States during autumn 2015[3].

Its design: a large-scale randomized evaluation of a low-cost, nudge-like intervention, which randomised students separately within 76 schools drawn from a national probability sample of US public high schools, and which was designed to measure impact variation, both across students and across schools[7].

The intervention encouraged treated individuals to view intellectual ability not as a fixed trait, but as a muscle that can be trained through the sustained effort of seeking help when learning new academic content and trying new learning strategies[3].

The reported result: the growth mindset intervention was shown to be effective in improving the GPA of high-school students[3], and students in lower-performing schools benefitted more from the intervention, compared to an active control group[2].

We did not obtain the paper and report no effect size for it.

How To Read That Result

Our own assessment of what the trial establishes.

Four features make it strong evidence.

It is a randomised trial, not a correlation.

The sample is a national probability sample, which addresses the usual objection that psychology studies convenience samples.

The control was active rather than do-nothing, which is a considerably harder test.

And it was designed in advance to measure variation rather than to find an average, so the subgroup finding is not an afterthought.

Two features constrain what it shows.

The intervention was low-cost and nudge-like, meaning brief and computer-administered. Whatever it demonstrates is about that, not about an extended training programme.

And the benefit concentrated in lower-performing schools, consistent with the meta-analytic moderator. The trial does not establish that the intervention helps everyone; it establishes that it helped somebody, in a way the aggregate literature had predicted.

The 2023 Quality Review

The most recent and most critical analysis.

Macnamara and Burgoyne published Do growth mindset interventions impact students' academic achievement? A systematic review and meta-analysis with recommendations for best practices in Psychological Bulletin[4].

Its stated purpose distinguishes it from what came before: the earlier meta-analysis had examined the effect, however, the quality of the evidence was not systematically evaluated. So the authors conducted the first systematic and comprehensive review of the growth mindset intervention on academic achievement literature that examines both the quantity and the quality of the evidence according to a well defined set of best practices[4].

Their finding, as reported by the publisher record: examining 63 studies, N = 97,672, they found major shortcomings in study design, analysis, and reporting, and suggestions of researcher and publication bias[5].

Another source reports the same authors concluding that apparent effects of growth mindset interventions on academic achievement are likely attributable to inadequate study design, reporting flaws, and bias[8].

The Best-Practice Subset

The single most informative number in this literature, in our view.

The 2023 review reported an insignificant overall effect among studies most closely following best practices (p = 0.666)[8].

Three observations, ours.

This is a different kind of test from an ordinary meta-analysis. Rather than averaging everything, it asks what the best-designed studies found, which is the approach the first article in this series described bias-correction methods attempting by other means.

A p-value of 0.666 is not a marginal result. It is about as far from significance as a test can land.

And the structure of the finding, that effects shrink as study quality rises, is the classic signature of a literature whose apparent effect is a design artefact. It is the same shape the ego depletion case showed in the first article, where bias-corrected estimates collapsed toward zero.

We flag that we obtained this figure from a working paper describing the review rather than from the review itself, and that a reader relying on it should obtain the original.

A Finding About The Researchers

A claim we report carefully, because it is a serious one.

The publisher record for the 2023 review states, among its findings, that authors with a financial incentive to report positive findings published significantly more positive results, and our source truncates at that point[5].

Four things about how we are treating this, all ours.

It is a finding in a peer-reviewed paper in Psychological Bulletin, not an accusation from a blog, and it is reported here on that basis.

We did not obtain the analysis behind it, the criterion used to identify a financial incentive, or the magnitude of the difference. Our source is truncated mid-sentence.

It names no individual and neither do we.

And it is a claim about a pattern in a literature, which is a legitimate object of meta-scientific study, not a claim about anyone's honesty. The researcher-degrees-of-freedom material in the first article of this series describes how such patterns arise without anybody intending them.

We include it because a reader deciding whether to buy a programme should know that the question of provider incentives has been raised in the literature itself. We would not include it on weaker sourcing.

The Defence

The other side, which has one.

A source records the counter-position, attributed to Yeager and Dweck: it is concluded that large-scale studies, including preregistered replications and studies conducted by third parties such as international governmental agencies, justify confidence in growth mindset research[5].

We did not obtain that paper and report only this characterisation of its conclusion.

Two observations, ours.

The argument is methodologically serious. It does not dispute that the average is small; it points at the best-designed large studies, which is the same move the critics make and the opposite conclusion. Both sides are arguing about which subset of the literature is authoritative.

And the appeal to third-party replications by governmental agencies is a strong form of evidence if it holds, because it removes the incentive problem raised in the previous section. We did not verify those studies and cannot assess the claim.

Our own view: a reader should treat this as a live dispute between competent teams, in the same category as the incentives literature described in the fourth article of this series, and not as a question settled in either direction.

A Conflict In The Study Counts

Something our sources disagree about.

One source describes the 2023 review as examining 63 studies, N = 97,672[5], and another describes the same review as a meta-analysis on 122 studies[8]. A third refers to it as examining 63 studies of growth mindset interventions on academic performance[6].

We did not resolve this. The most plausible explanation, and it is our own conjecture rather than anything a source states, is that the two counts refer to different units, such as studies of academic achievement against total studies or total effect sizes included.

The observation this supports is the same one this series has made repeatedly: a figure reported confidently by secondary sources is frequently reported inconsistently by them, and anyone quoting a specific number should be quoting the paper.

Where We Land

Our own reading, stated so a reader can disagree with it.

Four propositions we think the evidence supports.

The average effect is small, and every party agrees on this.

There is a real and repeatedly observed effect in disadvantaged and struggling populations, found by both the critical team and the supportive one.

The general-population claim is not supported, and the quality-weighted analysis points against it.

And the dispute about the best-designed studies is genuinely unresolved, with serious researchers on both sides.

Our own summary: this looks like a real but narrow intervention that was oversold, in a way that is now being corrected at some cost to everyone involved. That is a different verdict from either "it works" or "it is a myth", and it is the one we can support.

The Problem Nobody Mentions

The most important section of this article for a business reader, and it is our own analysis.

Read back through everything above and notice the population.

Ninth-grade students. Academic achievement. Grade point average. Free and reduced-price school lunch eligibility. Previously failed classes. Transition to high school.

Every study in this article concerns children or adolescents in an educational setting.

Three consequences.

A corporate growth mindset programme for adult professionals is an extrapolation, and it is a substantial one. Different population, different age, different outcome measure, different setting, different duration.

The moderator that most consistently survives, being that effects concentrate in struggling and economically disadvantaged participants, does not obviously map onto a professional workforce, and if it maps at all it would point at your least experienced or most struggling staff rather than at everyone.

And the intervention that was actually tested in the strongest study was brief and computer-administered, which is not what is generally sold to businesses.

We searched and located no evidence on adult workplace applications in the material we obtained. We are not saying it does not exist. We are saying we did not find it, and that a provider citing this literature is citing studies of schoolchildren.

What A Business Is Actually Buying

Our own reasoning about the decision, rather than about the science.

Three questions to put to any provider.

Which population is your evidence from? If the answer is students, ask what makes it transfer.

What effect size are you claiming, and from which analysis? The credible answer is small; a claim of large gains is the claim the 2023 review was written to test.

What would count as this not having worked? A programme whose success is measured by attitudinal survey responses is measuring the thing that is easiest to move.

Two observations.

A cheap intervention with a small effect can still be worth doing. The arithmetic is straightforward and the seventh and twelfth articles in this series both made the same point about cost-to-evidence matching.

But an expensive intervention with a small effect in a different population is a different proposition, and the case for it has to be made on grounds other than this literature.

What Does Transfer

The constructive residue, and it is ours.

Strip away the intervention question and something useful remains that requires no evidence base at all.

The distinction between treating ability as fixed and treating it as developable is a real distinction in how a manager behaves, independent of whether a training session can install it.

Three places it shows up in ordinary management practice.

In hiring, whether you screen for what someone can currently do or for what they could learn.

In feedback, and the seventh article in this series is directly relevant: feedback about a person is the costly kind, and "you're not a numbers person" is about as fixed-attribute as a sentence gets.

And in who gets the difficult work, which is where a belief about fixed ability becomes a resource allocation decision with compounding consequences.

None of that depends on the intervention literature. It is a question about what you believe and how you act, and a business can examine its own promotion and assignment records for evidence of which model it is actually operating.

What To Do

Ask which population the evidence comes from. Everything in this article concerns students in schools, and a workplace programme is an extrapolation.

Treat "large gains" as a red flag. Every meta-analysis we located, including supportive ones, reports the average effect as small.

Distinguish significant from meaningful. At 365,915 participants a trivial effect reaches significance, so magnitude is the number that matters.

Note where the effect concentrates. Struggling and economically disadvantaged participants benefited more, consistently, across independent analyses.

Ask about study quality, not just study count. The 2023 review found the effect insignificant among the studies best following best practices.

Ask providers about their own incentives. The question of financial incentives in this literature was raised in the literature itself, in a peer-reviewed journal.

Match spend to evidence. A cheap intervention with a small effect can be worth running; an expensive one in an untested population needs a different justification.

Look at your own records instead. Whether your firm treats ability as fixed is visible in who gets difficult assignments and who gets promoted, and that is data you already have.

Watch the language in feedback. Statements about what someone is, rather than what they did, are the fixed-attribute version and are also the kind the feedback literature identifies as costly.

The Limits Of This Analysis

Several caveats matter. This article reviews education and psychology research. It is not training, human resources or educational advice, and nothing in it is a judgment about any particular programme or provider. Everything is verified to August 2026. We obtained the published abstract of the 2018 meta-analyses and portions of the 2023 review, and did not obtain either in full. We did not obtain the national randomised trial at all and report it solely as three subsequent methodological papers describe it, stating no effect size for it. The d = 0.080 figure reaches us through a working paper rather than from the meta-analysis itself, as does the p = 0.666 best-practice figure. Our sources conflict on whether the 2023 review covered 63 or 122 studies, and our explanation for the discrepancy is our own conjecture that no source states. Our translations of d = 0.080 into overlap and superiority percentages are our own calculations under standard normal assumptions and appear in no paper. The finding regarding authors with a financial incentive is reported from a truncated source; we did not obtain the analysis behind it, the criterion used, or the magnitude, and it names no individual. We did not obtain the defending paper and report only a characterisation of its conclusion, and we did not verify the third-party governmental replications it invokes. We located no evidence, positive or negative, on adult workplace applications; that is an absence in our search rather than a demonstrated absence in the literature. Several sources are working papers and are identified. The population critique, the questions for providers and the section on what transfers are our own reasoning, not findings.

Related Reading

No. 14 Ninety-Nine Three field experiments found nine endings increased demand every time. They worked better on new items and worse alongside sale signs, and the reason... Behavioural Finance A Third Of It Made Things Worse Why statements about what a person is are the costly kind of feedback. Behavioural Finance Four Meta-Analyses, Two Answers The other live dispute where competent teams reached opposite conclusions.

Frequently Asked Questions

Does growth mindset work?

The average intervention effect is small, and every party agrees on that. There is a repeatedly observed effect in struggling and economically disadvantaged students, found by both critical and supportive teams. The general-population claim is not well supported, and the dispute over the best-designed studies is unresolved.

How small is d = 0.080?

On our own calculation, a treated student ends up above the average untreated student about 53 percent of the time against 50 percent by chance, with roughly 97 percent overlap between the distributions. For comparison, this series has reported feedback interventions at 0.41 and a brainstorming penalty around 1.39.

But it was statistically significant. Does that not settle it?

No. Significance asks whether an effect differs from zero, and with 365,915 participants a trivial effect will clear that bar. The p-value tells you the effect is probably not zero; the effect size tells you it is not much. A provider quoting significance is quoting the less relevant number.

What did the 2023 review find?

It examined evidence quality as well as quantity, reported major shortcomings in study design, analysis and reporting along with suggestions of researcher and publication bias, and found an insignificant overall effect among the studies most closely following best practices, at p = 0.666.

Should we run growth mindset training for our team?

We cannot answer that, and neither can this literature. Every study here concerns students in schools, and we located no evidence either way on adult workplace applications. A provider citing this research is citing studies of schoolchildren, and the extrapolation needs its own justification.

Is there anything useful here regardless?

Yes, and it needs no evidence base. Whether a manager treats ability as fixed or developable shows up in hiring criteria, in feedback language, and in who gets the difficult assignments. That is visible in your own promotion and assignment records, which is better evidence about your firm than any published study.

References

1. Sisk, V. F., Burgoyne, A. P., Sun, J., Butler, J. L., & Macnamara, B. N. (2018). To What Extent and Under Which Circumstances Are Growth Mind-Sets Important to Academic Achievement? Two Meta-Analyses. Psychological Science, published abstract, on mindsets or implicit theories being beliefs about the nature of human attributes such as intelligence; on the theory holding that individuals with growth mindsets, being beliefs that attributes are malleable with effort, enjoy many positive outcomes including higher academic achievement while peers with fixed mindsets experience negative outcomes; on interventions having been implemented in schools around the world; and on the first meta-analysis examining the strength of the relationship between mindset and academic achievement and potential moderating factors across k = 273 studies and N = 365,915 participants. Note: we obtained the published abstract only. pubmed.ncbi.nlm.nih.gov
2. Working paper on machine learning approaches to mindset interventions, on a recent meta-analysis finding overall effects on academic outcomes to be weak, with an overall effect size for interventions of d = 0.080 (p = .010), citing Sisk and colleagues (2018); on the moderators explored including developmental stage, previous failed classes, experience of situational challenges, eligibility for free or reduced-price school lunch, type of control condition, type of intervention and timing of outcome measure; on students with previous failed classes and students experiencing economic disadvantage benefitting significantly more than their peers; and on the national trial finding that students in lower-performing schools benefitted more compared to an active control group. Note: a working paper; our source for the d = 0.080 figure, which we did not obtain from the meta-analysis itself. arxiv.org
3. Working paper on treatment effect estimation, on the National Study of Learning Mindsets (Yeager et al., 2019) being an experiment conducted on a nationally representative sample of ninth-grade public high-school students in the United States during Fall 2015; on the intervention encouraging treated individuals to view intellectual ability not as a fixed trait but as a muscle that can be trained through sustained effort of seeking help and trying new learning strategies; and on the intervention having been shown to be effective in improving the GPA of high-school students. Note: a working paper. We did not obtain the trial itself and state no effect size for it. arxiv.org
4. Macnamara, B. N., & Burgoyne, A. P. (2023). Do growth mindset interventions impact students' academic achievement? A systematic review and meta-analysis with recommendations for best practices. Psychological Bulletin, author-hosted copy, on mindset theory holding that students who believe their personal characteristics can change will achieve more than those who believe them fixed; on proponents having developed interventions claiming large gains in academic achievement; on the evidence not having been systematically evaluated considering both quantity and quality; on Sisk and colleagues (2018) having previously meta-analysed intervention effects without systematically evaluating evidence quality; and on the authors conducting the first systematic and comprehensive review examining both the quantity and quality of the evidence against a defined set of best practices. Note: an author-hosted copy; we obtained portions. artscimedia.case.edu
5. Semantic Scholar record for Macnamara and Burgoyne, reporting that when examining all studies (63 studies, N = 97,672) the authors found major shortcomings in study design, analysis, and reporting, and suggestions of researcher and publication bias, with authors having a financial incentive to report positive findings publishing significantly more positive results; and separately recording the conclusion attributed to Yeager and Dweck that large-scale studies, including preregistered replications and studies conducted by third parties such as international governmental agencies, justify confidence in growth mindset research. Note: a bibliographic record. The financial incentive passage is truncated mid-sentence in our source; we did not obtain the analysis behind it, the criterion used to identify a financial incentive, or the magnitude of the difference. We did not obtain the defending paper. semanticscholar.org
6. ResearchGate record for Sisk and colleagues (2018) with third-party commentary, on Macnamara and Burgoyne (2023) having examined 63 studies of growth mindset interventions on academic performance and concluded effects were non-significant; on both Sisk and colleagues (2018) and Burnette and colleagues (2023) having found intervention effects stronger in lower-achieving students and students from low socioeconomic status backgrounds; and on overall effects having been weak while some results supported specific tenets of the theory, namely that students with low socioeconomic status or who are academically at risk might benefit. Note: a publisher record with third-party commentary, not peer-reviewed. researchgate.net
7. Working paper on assessing treatment effect variation, on the National Study of Learning Mindsets being a large-scale randomized evaluation of a low-cost nudge-like intervention designed to instil a growth mindset; on it randomising students separately within 76 schools drawn from a national probability sample of US public high schools; and on the study having been designed to measure impact variation both across students and across schools. Note: a working paper, cited for the trial's design; we did not obtain the trial. arxiv.org
8. Working paper on adaptive procedures for boundary false discovery rate control, on Macnamara and Burgoyne (2023) having conducted a meta-analysis on 122 studies and concluded that apparent effects of growth mindset interventions on academic achievement are likely attributable to inadequate study design, reporting flaws, and bias; and on their finding an insignificant overall effect among studies most closely following best practices (p = 0.666). Note: a working paper; our source for the p = 0.666 figure. Its study count of 122 conflicts with the count of 63 at references 5 and 6, and we did not resolve the discrepancy. arxiv.org

This article reviews education and psychology research and is not training, human resources or educational advice. Nothing in it is a judgment about any particular programme or provider. No paper discussed was obtained in full; the national randomised trial was not obtained at all. Two key figures reach this article through working papers rather than the studies reporting them. Sources conflict on the study count of the 2023 review. The claim regarding researcher financial incentives is reported from a truncated source and names no individual. No evidence on adult workplace applications was located in either direction.