A national experiment reveals where a growth mindset improves achievement

The intervention was a "growth mindset" program delivered entirely online in two 25-minute sessions. Students read scientific articles explaining that the brain forms new connections when you struggle with challenging work—like a muscle getting stronger. They then wrote a letter to a future struggling student, giving advice based on this idea (a "saying-is-believing" exercise designed to internalise the message).

The comparator was a control condition that received a parallel online session about the brain's ability to remember facts (a "memory mindset" lesson). This controlled for the effects of receiving any positive, brain-related message, the novelty of an online module, and the act of writing a letter.

The primary outcome was grade point average (GPA) in core academic subjects (math, science, English, social studies) at the end of the 9th grade (the first year of high school in the US). A secondary outcome was enrolment in advanced mathematics courses (e.g., Algebra II or higher) by the end of 10th grade.

The study included **12,490 students** from **76 public high schools** across the United States. The sample was designed to be nationally representative of 9th graders in US public schools (the first year of high school). Students were aged approximately 14–15 years. The sample was 49% female, 51% male. Racial/ethnic composition: 52% White, 22% Hispanic/Latino, 13% Black, 8% Asian, 5% other. Importantly, the sample was stratified to include schools with a wide range of achievement levels and socioeconomic compositions. Students were recruited from schools that agreed to participate; within each school, all 9th graders were invited.

**Grade point average (GPA):** Collected from school administrative records for core academic subjects (math, science, English, social studies) at the end of 9th grade. GPA was standardised within each school (mean = 0, SD = 1) to allow comparison across schools with different grading scales.

**Advanced math enrolment:** School records of whether students were enrolled in Algebra II or higher by the end of 10th grade (a binary yes/no measure).

**Peer norms:** Measured via a survey of all 9th graders in each school (not just study participants). Students rated their agreement with statements like "In this school, it's normal to be challenged by schoolwork" and "Students here respect those who work hard." These were aggregated to the school level to create a "peer norms" score.

**Pre-registered covariates:** Prior achievement (8th grade GPA), demographic variables (gender, race/ethnicity, free/reduced-price lunch status), and school-level characteristics.

**Study design:** This was a **randomised controlled trial (RCT)** with students randomly assigned within each school to either the growth mindset intervention or the control condition. Randomisation was done at the student level using a computer algorithm, and students were blind to which condition they were in (they were told they were participating in a "study of learning strategies").

**Blinding:** Students were blind to condition. Teachers and school staff were not informed which students received which condition. The researchers analysing the data were also blind to condition until after the primary analyses were pre-registered and locked.

**Duration:** The intervention itself was two 25-minute sessions, completed online during school hours, spaced about two weeks apart. The primary outcome (GPA) was measured at the end of 9th grade, approximately 6–8 months after the intervention. The secondary outcome (advanced math enrolment) was measured at the end of 10th grade, approximately 18 months after the intervention.

**Statistical approach:** The primary analysis used a multilevel model (students nested within schools) with pre-registered covariates. The key test was the interaction between the intervention and prior achievement (8th grade GPA). The authors hypothesised that the intervention would benefit lower-achieving students more than higher-achieving students. They also tested whether school-level peer norms moderated the effect. All analyses were pre-registered on the Open Science Framework before data collection was complete. A blinded Bayesian analysis was conducted by an independent team to corroborate the frequentist results.

**What this design can and cannot prove:**

**Can prove:** Because of random assignment, the study can establish that the growth mindset intervention *caused* the observed differences in GPA and math enrolment, compared to the control condition. The large, nationally representative sample means the results generalise to the US population of 9th graders in public schools.

**Cannot prove:** The design cannot tell us *why* the intervention worked (the mechanism). It cannot tell us whether the effect would persist beyond 10th grade. It cannot tell us whether the same intervention would work in other countries, with younger or older students, or in private schools. The design also cannot rule out that the control condition (memory mindset) had some unintended effect (e.g., making students feel worse about their memory, which could inflate the apparent benefit of the growth mindset condition).

**Major methodological strengths:** Pre-registration, blinding of analysts, independent Bayesian replication, large nationally representative sample, objective outcome measures (school records, not self-report), and testing of a pre-specified moderator (peer norms).

**Major methodological weaknesses:** The intervention was delivered in school, so students could have talked to each other about the content (though they were asked not to). The control condition was not a "no-treatment" control, so the effect size is relative to a specific alternative. The study relied on school records, which may have missing data (though attrition was low, <5%).

**Primary outcome (9th grade GPA):** The intervention improved GPA among lower-achieving students. For a student at the 25th percentile of prior achievement (8th grade GPA), the intervention raised 9th grade GPA by **0.11 standard deviations** (95% CI: 0.04 to 0.18, p < 0.001). This translates to roughly a 0.11 grade point increase on a 4.0 scale (e.g., from a C+ to a B-). For students at the 50th percentile, the effect was smaller (0.05 SD, p = 0.03). For students at the 75th percentile, there was no significant effect (0.01 SD, p = 0.68).

**Secondary outcome (advanced math enrolment):** The intervention increased the probability of enrolling in advanced math by **5 percentage points** (from 47% to 52%) among lower-achieving students (those below the median on prior achievement). This effect was significant (p = 0.01). For higher-achieving students, the effect was not significant (1 percentage point, p = 0.45).

**Moderation by peer norms:** The effect on GPA was concentrated in schools where peer norms were supportive of challenge-seeking. In schools with peer norms one standard deviation above the mean, the intervention effect for lower-achieving students was **0.17 SD** (p < 0.001). In schools with peer norms one standard deviation below the mean, the effect was **0.05 SD** (p = 0.12, not significant). This interaction was pre-registered and significant (p = 0.02).

**Bayesian replication:** The independent Bayesian analysis found a 99% probability that the intervention had a positive effect on GPA for lower-achieving students, and a 95% probability that the effect was moderated by peer norms.

**No harm:** The intervention did not harm higher-achieving students (no negative effects on GPA or math enrolment).

The effect on GPA for a lower-achieving student (0.11 SD) is small by conventional standards. To put it in context: it is roughly equivalent to the difference between a student who earns a C+ (2.3 GPA) and a B- (2.7 GPA) in a single course. Over a full year of core classes, this is a modest but meaningful improvement—about one-third of the typical gap between students from low-income and high-income families.

The effect on advanced math enrolment (5 percentage points) is more substantial. In a school of 500 students, this would mean about 25 more students taking advanced math than would have otherwise.

The effect was highly context-dependent: in schools with strong peer norms for challenge-seeking, the effect was about 50% larger (0.17 SD). In schools with weak norms, the effect was negligible. This means the intervention is not a "magic bullet"—it works best when the environment already supports the message.

**Acknowledged by authors:**

The study cannot identify the exact mechanism (e.g., did students study more, seek help more, or just feel more motivated?).

The peer norms measure was a school-level aggregate, not a measure of each student's immediate social network.

The intervention was delivered in a specific context (US public high schools) and may not generalise to other settings.

The control condition (memory mindset) may have had unintended negative effects, though the authors argue this is unlikely given that the memory mindset lesson was also positive and brain-focused.

**Critical reader notes:**

The effect size is small (0.11 SD) and only significant for lower-achieving students. For a self-experimenter, this means you might need a large sample or a very sensitive measure to detect an effect.

The study was funded by the Raikes Foundation, the Bezos Family Foundation, and the National Science Foundation—all organisations with an interest in promoting growth mindset interventions. However, the pre-registration and independent Bayesian analysis reduce concerns about bias.

The study only measured academic outcomes, not psychological outcomes (e.g., self-efficacy, anxiety, enjoyment of learning). It's possible the intervention changed how students felt without changing their grades.

The 18-month follow-up for math enrolment is impressive, but we don't know if the effect persists into college or career.

The intervention was delivered in school, with teacher support. A self-experimenter trying to replicate this at home might not get the same effect without the school context.

For someone running their own n=1 experiment:

**What to test:** A growth mindset intervention. Specifically: read a short article (5–10 minutes) explaining that the brain grows when you struggle with challenging material, then write a letter to a future self or a friend explaining this idea (the "saying-is-believing" exercise). Repeat once two weeks later. The key is to *internalise* the message, not just read it.

**Minimum meaningful duration:** At least 6 months to see an effect on grades or skill acquisition. The study saw effects at 6–8 months. For a self-experiment, track a specific learning goal (e.g., learning a new language, improving at a musical instrument, or mastering a coding skill) over at least 3–6 months.

**What to measure:**

- **Primary metric:** Objective performance on a standardised test or graded assignment (e.g., weekly quiz scores, practice exam scores, or a pre-post test of skill). Avoid self-reported "effort" or "confidence"—these are unreliable.

- **Secondary metric:** Persistence behaviour (e.g., hours spent practising, number of attempts at a difficult problem, or whether you voluntarily seek out harder challenges).

- **Context metric:** Your own "peer norms"—do the people around you (friends, family, colleagues) value challenge-seeking or do they avoid difficulty? This might moderate your results.

**Key confounds to control for:**

- **Expectancy effects:** You know you're doing the intervention, which could bias your effort. To minimise this, commit to a specific measurement protocol before starting (e.g., "I will take a standardised test every 4 weeks, regardless of how I feel").

- **Time of year:** Academic performance naturally varies with exam cycles, holidays, and stress. Run the experiment over a consistent period (e.g., one semester) and compare to a baseline from the previous semester.

- **Other interventions:** Don't start a new study routine, diet, or sleep schedule at the same time. Isolate the mindset intervention.

- **Prior achievement:** The study found the intervention only helped lower-achieving students. If you're already high-performing, you might see no effect. Consider testing on a skill you're genuinely struggling with.

**What a positive result would look like:** A 0.1–0.2 standard deviation improvement in your performance metric. For example, if your baseline quiz scores average 70% with a standard deviation of 10%, a positive result would be an average of 72–74% after the intervention. Or, if you're tracking persistence, you might see a 5–10% increase in the number of practice problems you attempt per week. The effect is small but real—don't expect a dramatic transformation.

**Additional tip:** The study found that peer norms matter. If you're doing this alone, consider sharing the mindset message with a friend or study partner and doing the "saying-is-believing" exercise together. The social reinforcement might amplify the effect.