C13 Growth mindset for struggling students
narrowed 1 attack: 1 narrowed · serves Q1 · last change 6 November 2026
A short online lesson teaching that intellectual ability can be developed (a growth-mindset intervention) raises the grades of lower-achieving secondary-school students by a small amount, but does little for students on average; that struggling students gain more than other students is not established across studies: in a meta-analysis of 63 studies, samples with low grades or at risk of failing gained about 0.07 standard deviations (0.09 in studies that checked the lesson changed mindsets), not significantly more than other students, and the best-designed studies were too few to check (Macnamara & Burgoyne 2023), although the largest preregistered trial did test it and found a larger gain for students in the lower half of their school's grades than for the upper half (about 0.10 against 0.01 grade points; difference P = .034; Yeager et al. 2019).
In practice
Do not expect a short lesson about the brain growing with effort to lift a whole class. If you use one, aim it at students who are struggling, and back it with a classroom culture that treats challenge as normal.
Go deeper
Where it holds
Main limits
- About short, standalone lessons, not about praise or feedback in general (C4).
- Effects in a US national trial depended on peer norms in the school (Yeager et al. 2019).
- Overall effects weak in a meta-analysis of 43 studies (Sisk et al. 2018).
- Struggling students did not gain significantly more than others in a meta-analysis of 63 studies: about 0.07 standard deviations for samples with low grades or at risk of failing, 0.05 for the rest (Macnamara & Burgoyne 2023); for adolescents specifically the pooled effect was d̄ = 0.04, and the "highest-quality" subset rests on a bar lowered to 60% of best practices because no study met every criterion.
- Within the US national trial, students in the lower half of their school's grades gained more than the upper half (about 0.10 against 0.01 grade points by the end of ninth grade; Yeager et al. 2019), so the difference is shown in one large trial but not across studies.
All limits (6)
Inside the circle: the core of the claim. Outside: every bound and carve-out in the claim file, numbered, with the attack that forced or added it where the record shows one.
forced by a narrowing attackin the original wordingwritten at creation, or added without a narrowing
Warm cards in the list were added by later attacks; plain cards were in the original draft or written at creation.
Short (about an hour or less), standalone growth-mindset interventions in secondary school, outcome grades or achievement; not praise or feedback in general (C4), not long-term programmes.
Written when the claim was created (16 October 2026); not forced by an attack
Evidence on file: Yeager et al. 2019 (Nature; preregistered national US trial; grades improved for lower-achieving students, context-dependent on peer norms; correction 2026-10-19: "lower-achieving" there means at or below the school-specific median GPA, half of all students, not low achievers in absolute terms; core-course GPA at the end of ninth grade +0.10 grade points for them (95% CI 0.04 to 0.16), +0.01 for higher achievers (P = 0.634), intervention × lower-achiever interaction B = 0.09 (95% CI 0.01 to 0.17, P = 0.034; Extended Data Table 1 legend); grades followed only to the end of ninth grade); Sisk et al. 2018 (two meta-analyses; intervention meta k = 43, N = 57,155; overall effects weak, possible benefit for low-SES or at-risk students).
Written when the claim was created (16 October 2026); not forced by an attack
Wording changed later: Added with attack #2 (corrected; it did not narrow the claim) · Yeager et al. 2019 · 19 October 2026.
Differential bound (2026-10-16; enriched 2026-11-06): Macnamara & Burgoyne 2023 (Psychological Bulletin; 63 studies, N = 97,672) coded academic challenge as a preregistered moderator: high-challenge samples d = 0.07 [0.01, 0.13], low-challenge 0.05, difference ns (Q(2) = 0.28, p = .870); high-challenge levels pool children, adolescents and adults.
Forced by attack #1 · Macnamara & Burgoyne 2023 · 16 October 2026
The secondary-school–relevant developmental cell is adolescents d̄ = 0.04, 95% CI [0.002, 0.09], p = .041 (Table 7; Q(2) across ages = 0.47, ns).
Added with attack #3 (corrected; it did not narrow the claim) · Macnamara & Burgoyne 2023 · 6 November 2026
Meta-Analysis 3's "highest-quality" set (6 studies, N = 13,571, overall d̄ = 0.02 [−0.06, 0.10]) rests on a bar lowered to ≥60% of the authors' 10 best-practice criteria, because zero mindset-shifting studies met every criterion (p. 28).
Added with attack #3 (corrected; it did not narrow the claim) · Macnamara & Burgoyne 2023 · 6 November 2026
Macnamara & Burgoyne write that Yeager et al. 2019 did not report a test of whether lower achievers' effect differed from higher achievers'; the trial's Extended Data Table 1 legend does report it (see above), so the "not established" clause rests on the meta-analysis's between-study moderator test, not on an absence of within-trial tests (corrected 2026-10-19).
Added with attack #2 (corrected; it did not narrow the claim) · Yeager et al. 2019 · 19 October 2026
1 of 6 bounds and carve-outs were forced by a specific attack.
How it changed
The ledger records 1 attack (and 2 corrections to our notes), but no earlier wording of C13 is on record, so the text cannot be replayed step by step. Only the current wording is shown.
A short online lesson teaching that intellectual ability can be developed (a growth-mindset intervention) raises the grades of lower-achieving secondary-school students by a small amount, but does little for students on average; that struggling students gain more than other students is not established across studies: in a meta-analysis of 63 studies, samples with low grades or at risk of failing gained about 0.07 standard deviations (0.09 in studies that checked the lesson changed mindsets), not significantly more than other students, and the best-designed studies were too few to check (Macnamara & Burgoyne 2023), although the largest preregistered trial did test it and found a larger gain for students in the lower half of their school's grades than for the upper half (about 0.10 against 0.01 grade points; difference P = .034; Yeager et al. 2019).
What would change this
What would kill it: A large preregistered randomised trial or meta-analysis of short growth-mindset interventions finds no grade gain for lower-achieving or at-risk secondary-school students. Added as a claim on 16 October 2026, promoted from a candidate belief in the evidence library.
Attack record
1 attack: 1 narrowed. Counts of attacks, not a score.
Full attack list (1)
A preregistered meta-analysis of 63 growth-mindset intervention studies (N = 97,672) tested whether the effect depends on students' level of academic challenge and found it does not: samples where most students had low grades or were at risk of failing gained d = 0.07 (95 % CI 0.01 to 0.13), against 0.05 for low-challenge samples, a nonsignificant difference (Q(2) = 0.28, p = .870); the same held among studies showing the lesson changed mindsets (high challenge 0.09, Q(2) = 2.98, p = .226), and the best-designed studies (6 studies, overall d = 0.02, ns) were too few to test it, so C13's contrast — a gain for lower achievers while little happens on average — is not established.
Strongest reading and verdict reasoning
Steelmans to: C13's distinguishing assertion is that short mindset lessons help lower-achieving students specifically even though the average effect is small; the largest meta-analysis to date coded exactly this moderator in advance and found no difference by challenge level, and its authors call the evidence for the struggling-students argument "limited" (p. 31).
Why: Run ordered 2026-10-16 after Jonatan supplied the full text. Full text read (APA Online First PDF, supplied by Jonatan 2026-10-16, stored privately in the source folder). Coding (p. 14): high challenge = most of the sample had low grades or was otherwise at risk of failing; moderate = situational challenge (new school, stereotype threat); low = no risk indicators; run with full samples and with available high-challenge subsamples swapped in. Meta-analysis 1, all studies (Table 7, p. 22): high d = 0.07 [0.01, 0.13] (p = .019), moderate 0.05 [−0.02, 0.12], low 0.05 [−0.01, 0.11]; Q(2) = 0.28, p = .870; unchanged with subsample replacement (2 studies). Overall d = 0.05 [0.02, 0.09], nonsignificant after correcting for publication bias (abstract). Meta-analysis 2, studies with a successful mindset manipulation check, 13 studies (Table 13, p. 27): high 0.09 [0.04, 0.14] (p < .001), moderate −0.01 [−0.10, 0.09], low 0.05 [−0.21, 0.32]; Q(2) = 2.98, p = .226. Meta-analysis 3 (6 studies, N = 13,571, best practices): d = 0.02 [−0.06, 0.10]; no moderator analyses possible. The authors note Yeager et al. 2019 reported only below-median students without testing the difference from other students (p. 11). Kill withheld: high-challenge subgroups do show a small positive pooled effect in two of the three analyses; challenge categories pool children, adolescents and adults and are not limited to secondary school; subgroup estimates are not bias-corrected. Not a restatement: C13's Bounds cite Sisk et al. 2018 for weak overall effects and a possible at-risk benefit; nothing on the claim said the lower-achiever benefit was not larger than for others. Not a preprint. Claim, Bounds and Status rewritten same turn; brief and the practical rule updated.
Corrections to our notes
These fix a number or a fact in our notes. They are not attacks and do not count toward the floor of 3 survivals.
- — The preregistered national US trial that C13's lower-achiever clause leans on did test whether lower achievers gained more than other students and found they did (higher achievers B = 0.01 grade points, P = 0.634; intervention × lower-achiever interaction B = 0.09, 95% CI 0.01 to 0.17, P = 0.034, n = 12,490), so the statement relayed from Macnamara & Burgoyne 2023 that this trial never ran the test is wrong, and…
Details
What was checked: fact behind the claim text's differential clause ("that struggling students gain more than other students is not established"), not the kill-condition (a no-gain finding for lower achievers), which this result does not touch.
Why: Nightly-attacks 2026-10-19 Europe/Stockholm (a lead from the study-reading notes, re-checked tonight). Full text read (Nature published PDF, stored privately as ). Extended Data Table 1 legend: "higher-achieving students demonstrated no significant treatment effects on core course GPAs, B = 0.01 grade points (95% confidence interval = −0.03–0.06), s.e. = 0.02, n = 6,170, k = 65, t = 0.480, P = 0.634... resulting in a significant intervention × lower-achiever interaction B = 0.09 grade points (95% confidence interval = 0.01–0.17), s.e. = 0.04, n = 12,490, k = 65, t = 2.179, P = 0.034)". Main text: lower achievers (at or below the school-specific median GPA) B = 0.10 grade points, effect size 0.11. Macnamara & Burgoyne 2023 (stored full text): Yeager et al. "did not report a test of whether the treatment effect for lower-achieving students differed from the treatment effect for higher-achieving students" — contradicted by the legend. The 2026-10-16 ledger entry relayed that statement ("the authors note Yeager et al. 2019 reported only below-median students without testing the difference"); that entry is left as written and corrected here. Not a weaken or a survival: the meta-analysis's own pooled moderator test (Q(2) = 0.28, p = .870) still stands, so "not established across studies" remains accurate; the fix adds the trial's within-study test and the definition of "lower-achieving" (the lower half of each school, not low achievers in absolute terms); scope, direction and kill-condition unchanged; Status stays weakened. Not a duplicate (first ledger entry from Yeager et al. 2019). Not a preprint. Claim, Bounds, brief and the practical rule limits rewritten same turn.
- — Checked against the full text of the study: precision of the Macnamara differential and best-evidence lines.
Details
What was checked: precision of the Macnamara differential and best-evidence lines; no kill-condition test.
Why: Nightly-attacks 2026-11-06 Europe/Stockholm (the targeting rules queue). Full text Table 7: adolescents d̄ = 0.04, 95% CI [0.002, 0.09], p = .041; Q(2) across developmental stages = 0.47, p = .791. Meta-Analysis 3 (pp. 28–29): no study met all 10 best-practice criteria; threshold lowered to ≥60% of criteria to reach six studies (N = 13,571), overall d̄ = 0.02 [−0.06, 0.10]. Not a duplicate (Bounds already carried high-challenge 0.07, Q(2) = 0.28, overall 0.05, best-evidence 0.02, and the Yeager interaction correction, but not the adolescents cell or the 60% bar). Kill withheld: high-challenge d = 0.07 is a small positive effect, not a null for lower achievers; C13's "not established across studies" clause already rests on the ns moderator. Claim text, Status and kill line unchanged. Not a survival; does not count toward the floor; does not reset C13's fruitless streak. C13 returns from rest under the targeting rules(a) on this verified full-text 3g flag.
Evidence
8 linked sources: 2 supports, 5 limits, 1 challenges (first-pass draft, pending human review). supports: Yeager et al. 2019, Burnette et al. 2023 · limits: Sisk et al. 2018, Ganimian 2020, Gazmuri 2025, Outes-León et al. 2020, Foliano et al. 2019 · challenges: Macnamara & Burgoyne 2023