Growth mindset after the replication wave

Carol Dweck, a psychologist at Stanford University, spent three decades studying what she and her collaborators called implicit theories of intelligence: whether a person treats intellectual ability as something fixed at birth or as something that changes with effort and instruction. The distinction entered general circulation through her 2006 trade book Mindset, and from there into curricula, corporate training and a very large volume of commentary that had little contact with the underlying papers.

The early empirical case rested on studies with modest samples. Lisa Blackwell, Kali Trzesniewski and Dweck followed several hundred students across the transition into junior high school and reported that those endorsing a malleable view of intelligence showed a rising mathematics trajectory relative to peers who did not. A short intervention in the same paper, delivered in eight sessions, was associated with improved grades among the students receiving it.

The meta-analysis

In 2018, Victoria Sisk, Alexander Burgoyne, Jingze Sun, Jennifer Butler and Brooke Macnamara published a two-part meta-analysis in Psychological Science covering several hundred studies. The first part examined the correlation between mindset and academic achievement and found it weak. The second examined interventions and found the average effect close to zero across the full sample.

The more interesting result was where the effect was not zero. Benefits concentrated among students who were academically at risk or from low-income households — a much narrower population than the one the popular literature addressed. The authors were explicit that the broad claim, applied to students in general, was not carrying the weight placed on it.

A separate randomised trial commissioned by the Education Endowment Foundation in England, reported by Nathan Foliano and colleagues, tested a mindset intervention in primary schools and did not find a statistically significant effect on attainment.

The national study

The most methodologically serious test arrived in Nature in 2019. David Yeager, Paul Hanselman, Gregory Walton, Elizabeth Tipton, Dweck and a large collaboration ran the National Study of Learning Mindsets: a pre-registered, nationally representative trial with roughly twelve thousand ninth-grade students across the United States, using an independent research firm to handle randomisation and data collection.

It reported an effect. The effect was small, it applied to lower-achieving students rather than the whole sample, and it registered as a fraction of a grade point. It was also conditional on the school: gains appeared where the surrounding peer environment supported the framing and were absent where it did not.

Macnamara and Burgoyne have continued to argue that even this estimate is fragile and that the field's earlier enthusiasm rested on a literature shaped by selective publication. Dweck and Yeager have argued in turn that critics pool interventions of widely varying quality, treating a single poorly delivered session as equivalent to a carefully constructed one, and that heterogeneity of that kind will drive any average toward zero.

Where the record stands

Both descriptions of the dispute can be read directly. The original papers are available, the meta-analyses are available, and the disagreement between the two camps is on the record in published exchanges rather than in inference. What has not survived is the version that circulated outside the journals: a general, sizeable and reliable effect of a brief attitude adjustment on academic performance across populations.

Cairn Post marks this entry as mixed. Independent replication has recovered something in a defined subgroup under controlled conditions; it has not recovered the broader claim.