Concept
Growth Mindset and Resilience-Training Replication Crisis
replicability-crisis education psychology
Bruce Daisley, in Fortitude, investigates why grit, growth mindset, and resilience training — despite being near-ubiquitous in schools (including the UK's Northern Ireland national curriculum), corporate hiring (Google explicitly hires "for growth mindset"), and Angela Duckworth's TED talk (27M+ views on TED.com, 10M+ on YouTube, the platform's 22nd most-viewed talk) — largely fail to replicate.
Growth mindset (Carol Dweck) couldn't be independently reproduced. Dweck's core finding — that reframing praise from ability ("you're smart") to effort ("you worked hard") transforms children's response to failure — showed a striking pattern once other researchers tried to replicate it: results only reliably appeared when Dweck herself was directly involved in running the study; independent replications failed. Dweck's own explanation was that others weren't following her methodology precisely enough. Timothy Bates (University of Edinburgh) ran multiple growth-mindset replications with 12-year-olds (and a second cohort of 200 12-year-olds in China): no significant effect on cognitive performance, and no relationship between children's stated mindset and their IQ, grades, or grade changes across the year. A 2018 meta-analysis of 365,000 participants found only a weak overall correlation between growth mindset and academic achievement — directly contradicting Dweck's own claimed effect sizes (that 67% of effort-praised children chose harder subsequent problems, versus just 8% of intelligence-praised children; no independent study has replicated this gap).
Grit (Angela Duckworth) is real but statistically overstated. Marcus Creed (Iowa State University) reviewed 90 studies underlying Duckworth's Grit. Duckworth's headline West Point statistic claimed grittier cadets were "98% more likely" to complete Beast Barracks summer training — but the overall completion rate across all cadets was already 94%, making the true incremental lift from grit only about 4 percentage points, not the dramatic-sounding "98%." See Duckworth Grit West-Point Restriction-of-Range Case for a separate, distinct critique of the same underlying statistic (restriction of range, via Alex Edmans).
Resilience training itself doesn't work, and may be a form of victim-blaming. Professor Alex Haslam recounts West Dorset NHS Trust's early-2000s response to a public-health warning about its toxic work environment: rather than address root causes, it sent staff on resilience training. A 2013 Military Psychology study of US cadets found voluntary resilience training attracted zero sign-ups (personnel didn't believe it would help), and personnel serving in Afghanistan who had completed resilience training showed declining resilience and morale over the training period. Daisley argues resilience training implicitly reframes systemic problems (toxic workplaces, unmanageable workloads) as individual deficiencies — a framing he connects to real-world "just be more resilient" responses to residents of Flint, Michigan (years of contaminated tap water) and to overworked NHS staff.
Resilience may be a state, not a trait. Researchers at the University of Queensland and Harvard argue resilience isn't a fixed, trainable personal capacity but a state anyone can embody given the right circumstances — the Ukrainian public wasn't described as "resilient" before Russia's invasion, but exhibited it once circumstances demanded it. Control appears to be the operative variable, not willpower or training.
Two direct animal studies linking control to resilience/stress, cited by Daisley from Martin Seligman's earlier (pre-"positive psychology") research: rats injected with a 50%-lethal cancerous tumor, then placed in a cage delivering 15 daily electric shocks — half the rats had a touchpad that could stop the shock, half had no such control. Roughly 3/4 of rats without control died; roughly 1/4 of rats with control died (see Seligman Rat Tumor-Control Study). Separately, rats immobilized via Botox injection (removing physical control generally) were given a piece of wood to chew as a partial substitute for control — those able to chew showed no rise above baseline stress; those denied it showed stress levels 7x baseline (see Botox Rat Chewing-Control Stress Study).
Practical implication: since control (not innate toughness) predicts resilience, the useful intervention for someone struggling — burnout, an unhappy relationship — is asking what control they actually have, not urging them to "toughen up."
Discussed in
- Episode 98 — Is growth mindset training a waste of time