Fixed and Growth Mindset

What it is

The prediction that follows from the distinction is behavioural rather than descriptive. Entity theorists avoid difficulty because failure would be diagnostic of their limit; incremental theorists approach it because failure is information. That much is intuitive and it is well known.

Credit needs care. Adrian Furnham writes that Dweck popularised the modern use of the distinction, wording which correctly implies she did not originate it, and he names no predecessor. This vault has not established the predecessor either and does not invent one. What it records is that the sentence points at a gap the source then leaves empty.

The two claims have to be separated before anything else can be said. That people differ in what they believe ability is, and that a short intervention can shift that belief and thereby raise attainment at scale, are different propositions with different evidence, and almost all the popular writing runs them together.

In effect

The critical side of the evidence, in order.

Victoria Sisk, Alexander Burgoyne and colleagues published two meta-analyses in Psychological Science in 2018. The correlational one covered 273 effect sizes across more than 365,000 people and found an average correlation of about one tenth, roughly one per cent of the variance, with 58 per cent of the effect sizes not significant and the relation near zero in adults. The intervention meta-analysis covered 43 effect sizes across more than 57,000 people and found a standardised effect of about 0.08, with 86 per cent of effect sizes not significant. Academic risk was not a significant moderator. Socio-economic status was: the effect in low-income households was about 0.34 across seven effect sizes, against about 0.03 and not significant in middle class and above.

Brooke Macnamara and Burgoyne published the larger analysis in Psychological Bulletin in 2023. Across 63 studies and more than 97,000 participants the effect was about 0.05 and not significant after correcting for publication bias. Restricted to studies that demonstrably shifted mindsets, it was about 0.04 and not significant. Restricted to the highest-quality evidence, six studies, it was about 0.02 and not significant. They report that 94 per cent of interventions contained confounds, and a significant interaction between publication status and author financial incentive.

The large-scale trials cut both ways and are given both ways here.

David Yeager and colleagues published the National Study of Learning Mindsets in Nature in 2019. It was preregistered, nationally representative and independently evaluated: 12,490 ninth-graders across 65 schools, given a two-session online intervention lasting under an hour. The primary preregistered result, on 6,320 lower-achieving students, was a gain of about a tenth of a grade point, a standardised effect of about 0.11, with larger benefits where peer norms supported taking on challenge. The paper's own benchmark is worth carrying: the median of 67 federal evaluations of secondary-school interventions is about 0.03 standard deviations, so a small effect here is not nothing.

The British trials point the other way, and the better-powered one is a clean null. The Education Endowment Foundation's first Changing Mindsets trial showed two months' progress on a small sample and was not statistically significant. The second, by Francesca Foliano, Heather Rolfe, Johnny Runge and colleagues in 2019, covered 101 schools and 5,018 Year 6 pupils with national tests as the outcome. It found no effect in maths, reading or grammar, and none among pupils eligible for free school meals, at a high security rating. The evaluators flag contamination, most control teachers already being familiar with growth mindset.

The replies are at full strength and belong here. Jeni Burnette and colleagues, in the same 2023 issue of Psychological Bulletin, found a standardised effect of about 0.14 on academic achievement across 53 samples, with a prediction interval running from below zero to about 0.35. Elizabeth Tipton, Christopher Bryan, Yeager and colleagues, in the same issue, argue that averaging across a heterogeneous literature asks the wrong question and that effects should be modelled as varying by procedure and context. Yeager is an author of that reply; Dweck is not. Macnamara and Burgoyne published a rejoinder, and the exchange is unresolved.

Dweck's own answer to this line of criticism belongs beside the criticism rather than underneath it. Yeager and Dweck, in American Psychologist in 2020, concede that early claims were oversimplified and that fidelity of implementation and school context matter far more than the popularisation implied. That is a concession by the originator.

What it does not say

It does not say the distinction is not real. People do differ in what they think ability is, and that difference is measurable.

It does not support the popular intervention claim. Teaching children that intelligence is malleable does not substantially raise attainment at scale, on the two largest meta-analyses available.

It does not rule out a targeted effect. A small, real effect in lower-achieving and low-income students in supportive school contexts is still live, on the strength of one large preregistered American trial. It did not appear in the best-powered British trial.

It is not a debunking. The honest word is a narrowing, and anything that uses mindset as a causal step in a longer chain is asking it to carry more than it can.

It is not the same thing as willpower mindset. That is a separate and much narrower claim about beliefs concerning willpower, with its own replication record. Dweck is the shared name, not the shared claim, and collapsing them lets one inherit the other's credibility or its obituary depending on who is writing.

---