Growth Mindset: What the Research Really Shows About Whether Believing You Can Grow Changes Results

Does believing that your abilities can grow actually change your results? The honest answer is: possibly, a little, and in some conditions. Growth mindset research shows that the belief is linked to better grade trajectories in some student samples, that short interventions produce small and uneven effects at best, and that one critical meta-analysis finds almost no effect once study quality and bias are accounted for. The evidence does not support treating mindset as a lever that reliably improves outcomes. It supports something narrower: a modest, context-dependent tool.

This article covers what was tested, what was found, where the findings conflict, and what a sensible reader can do with the result.

What growth mindset claims, and what it does not

A growth mindset is the belief that human capacities are not fixed but can be developed over time. Mindset research examines how such beliefs influence behaviour, including challenge seeking and resilience [2].

That definition covers two different things, and they are tested differently:

• The belief itself. Do people who hold a growth view tend to do differently from people who hold a fixed view? This is usually studied by measuring beliefs and then following outcomes.
• Interventions designed to teach the belief. If a short programme teaches that abilities can be developed, do outcomes change compared with a control group? This is studied with experiments [2].

A belief can be associated with good outcomes without a short lesson in that belief producing them. Much of the confusion in the headlines comes from treating the two as one question. Your question probably covers both: does holding the belief help, and can you adopt it deliberately and get the benefit?

Early evidence: beliefs linked to grade trajectories

One of the best-known early studies followed 373 seventh graders through two years of junior high school. Students who believed intelligence is malleable (an incremental theory) showed an upward trajectory in mathematics grades. Students who believed intelligence is fixed (an entity theory) showed a flat trajectory [3]. The researchers also tested a model in which learning goals, beliefs about effort, and attributions and strategies helped explain the link [3].

A second study in the same paper tested a small intervention. Seventh graders taught an incremental theory (N=48) showed positive change in classroom motivation compared with a control group (N=43). The control group's grades continued to decline, while that decline was reversed in the group that received the intervention [3].

This is interesting, and it is also limited:

• The first study measured beliefs and then followed grades, so it shows an association over time. It cannot establish that the belief caused the improvement [3].
• Both studies involved one age group, seventh graders, and one subject, mathematics [3].
• The intervention sample was small, with 48 students in one group and 43 in the other [3].

Nothing here speaks directly to adults learning a skill or doing a job. That gap matters for the rest of this article.

The large-scale test: a short online intervention

The next step was to test whether a brief intervention could work across a whole national population of students. A study in a nationally representative sample of secondary students in the United States evaluated an online growth mindset intervention lasting less than one hour. It reported improved grades among lower-achieving students and increased overall enrolment in advanced mathematics courses [4].

Three features made this study carry more weight than many before it. The data collection and processing were independent, the analyses were pre-registered, and the results were corroborated by a blinded Bayesian analysis [4].

The study also reported something easy to miss: the effects depended on context. The intervention changed grades when peer norms aligned with the message of the intervention [4]. In other words, the lesson did not appear to work in isolation. It worked where the surrounding school culture supported the idea.

The abstract does not give an effect size for the grade improvement, so it would be a mistake to read a particular magnitude into it. What it supports is that a benefit was detected in a defined group, under identifiable conditions.

The meta-analyses: small, variable, and contested

When many studies are pooled, the picture gets less tidy. Two meta-analyses reached very different readings of broadly the same field.

The review that finds small but real effects in the right conditions

One meta-analysis included 53 independent samples testing distinct interventions, all using randomised designs. It focused on two moderators: subsamples expected to benefit most, and implementation fidelity [5]. In targeted subsamples with high fidelity, it reported an effect of d = 0.14 for academic achievement (95% CI [.06, .22]) and d = 0.32 for mental health (95% CI [.10, .54]) [5].

The same review stressed variation. Its 95% prediction intervals, which describe the range of effects to expect from future interventions, ran from -0.08 to 0.35 for academic achievement and from 0.07 to 0.57 for mental health [5]. An interval for achievement that starts below zero means a future intervention could plausibly show no benefit, or a slightly negative one, even under favourable conditions.

The review that finds almost nothing

A second meta-analysis took a more sceptical approach. It evaluated studies against best practices needed for causal conclusions and then ran three meta-analyses. Across all 63 studies (N = 97,672), it reported major shortcomings in study design, analysis and reporting, and suggestions of researcher and publication bias [6]. The overall effect was d = 0.05 (95% CI 0.02 to 0.09), which was nonsignificant after correcting for potential publication bias [6].

Narrowing the set made the estimate smaller, not larger. In 13 studies where the intervention shifted students' mindsets as intended, the effect was d = 0.04 and nonsignificant. In the highest-quality subset of 6 studies (N = 13,571), it was d = 0.02 and nonsignificant [6]. The authors concluded that apparent effects on academic achievement are likely attributable to inadequate study design, reporting flaws and bias [6].

The authors also reported that researchers with a financial incentive to report positive findings published significantly larger effects than those without such an incentive [6]. That is a finding of this one review, and it is a claim about a pattern in the published record. It is not a demonstration that any particular study was distorted.

Reading the two side by side

The numbers are closer than the conclusions. Both reviews report small averages: 0.14 in the targeted, high-fidelity subsamples of the first, and 0.05 overall in the second [5][6]. The disagreement is about what those numbers mean.

• The first review treats the variation as the story. Effects are small on average but differ by who is studied and how well the intervention is delivered [5].
• The second treats the small average, its fragility under bias correction, and the weak design of many studies as the story [6].

The two also draw their subsets differently. The first selected subsamples expected to benefit and high-fidelity implementations. The second selected by study quality and by whether mindsets actually changed [5][6]. Reasonable readers can disagree about which filter is more informative, and the disagreement is not resolved by the sources here.

Context matters: who benefits and where

Taken together, the studies point to heterogeneity rather than a single effect.

• Who. The national study found improved grades among lower-achieving students, not across the board [4].
• Where. Effects on grades appeared when peer norms aligned with the intervention's message [4].
• How well it was delivered. The first meta-analysis found that the effects it reported depended on targeted subsamples and implementation fidelity, and it reported wide prediction intervals around them [5].

There is also a single exploratory study worth mentioning, mainly to show how far the field reaches. In 135 undergraduates who completed a modified laboratory social stress task, a higher growth mindset of behaviour was associated with lower cortisol at 25 minutes after the task began in men, but not in women [1]. This is one small, early study of an association. It should not be read as a reliable result, and it offers no guidance about managing anxiety or stress.

What a sensible reader can take from this

The research does not show that believing you can improve is a proven driver of results. It also does not show that the belief is worthless. What it supports is more modest: the belief is linked to better trajectories in some samples, brief interventions show small and uneven effects, and context seems to decide a lot [4][5][6]. None of it replaces strategy, practice and an environment that supports learning.

The suggestions below are practical suggestions, not findings. The sources do not test these personal practices.

• Pair the belief with a concrete strategy. "I can get better at this" is vague. "I will rewrite the first paragraph twice and compare" is something you can do. If you keep one, keep the second.
• Notice the environment you are in. The national study suggests that surrounding norms shaped whether the message mattered [4]. It can help to ask whether the people around you treat effort and mistakes as normal parts of learning, and to spend more time with those who do.
• Judge by your own results over weeks. A belief that feels convincing is not evidence that it is working. Look at what changed in your output.

If you want help turning a vague aim into something trackable, How to Set Goals You Will Actually Track covers the mechanics, and A 20-Minute Weekly Review gives a simple way to look back at results.

A two-week personal test

This is self-observation, not a measure of mindset, and it will not tell you whether growth mindset works in general. It can tell you something about your own practice.

1. Pick one skill you currently avoid. Choose something specific, such as giving short presentations, writing a first draft, or a particular spreadsheet task.
2. On day one, write down what you believe about your ability to improve it. One or two sentences. Include doubts.
3. Practise one small, specific strategy for ten minutes on five days. Name the strategy in advance, for example "outline in three bullets before writing" or "rehearse the opening aloud once".
4. After each session, record three things: what you tried, what changed, and what did not.
5. At the end of two weeks, reread your notes and ask: did the practice or the belief seem to matter more? Did anything shift in your belief, and did that follow the practice or come first?

Treat the answer as a hint rather than a conclusion. Two weeks is short, and many things vary at once. If the practice helped regardless of your belief, that is useful. If your belief changed after you saw yourself improve, that is useful too.

What we still do not know

Several questions remain open, and the sources are clear about them.

• Effects vary widely. The prediction intervals in the first meta-analysis show that future interventions could land well above or below the average, and the review itself emphasises extensive variation [5].
• Study quality is disputed. One review finds major design and reporting problems and suggestions of bias, and concludes that apparent benefits are likely attributable to those problems [6]. Another finds small positive effects in targeted, high-fidelity subsamples [5]. These are not reconciled.
• Some outcomes are barely studied. The first review describes the literature on social functioning as too nascent to test moderators, and its average effect there came with a prediction interval from -0.50 to 1.22 [5].
• The evidence centres on students and young people. The studies discussed here involve seventh graders, secondary students and undergraduates [1][3][4]. Whether and how the findings carry over to adults at work is not established by this research.
• There is more to learn. Researchers who have worked in this field for decades describe much more to learn, even while describing the studies so far as generative [2].

So the position that holds up is the cautious one. Believing that abilities can develop may help a little, for some people, in some settings, and the size of that help is contested. Use it as one small input alongside concrete strategy, regular practice and an environment that supports both, and judge it by what changes in your own work.