Willpower doesn't run down like a battery: the ego-depletion effect was close to zero in two large preregistered multi-lab replications
Interpretation
The original limited-resource findings (Baumeister 1998) and a supportive 83-study meta-analysis (Hagger 2010) did not survive bias-focused reanalysis (Carter 2015, disputed by Dang 2018) or two preregistered replications: 23 labs, N = 2,141, d = 0.04 (Hagger 2016) and 36 labs, N = 3,531, d = 0.06 (Vohs 2021). Graded moderate rather than strong because both replications used students and general adults on short computer tasks, not athletes or training. Related sport-science evidence on mental exertion before exercise is mixed: 73 studies, g = -0.38 (Brown 2020), but a bias-corrected estimate near zero (Holgado 2020). Evidence that consistent people rely more on habits than on effortful restraint (de Ridder 2012, 102 studies; Galla and Duckworth 2015, N = 2,274) is correlational and limited.
Practical implication
Don't plan training around rationing discipline. Reduce how much each session depends on in-the-moment willpower: fixed times and places, kit prepared, if-then plans for known obstacles, and hard sessions where they're least likely to be disrupted. A possible small cost of mentally demanding days can't be ruled out. All extrapolated from non-athlete lab research; untested in triathletes.
Sources
10 sources, independently verified.
Original ego-depletion experiments: people who ate radishes instead of tempting chocolates quit sooner on unsolvable puzzles; making a meaningful choice or suppressing emotion caused similar drops on a following task. Authors proposed a single limited resource for self-control. Small lab studies; the effect later failed large preregistered replications.
Meta-analysis of 83 studies found a significant ego-depletion effect on self-control performance and on effort, perceived difficulty, negative affect and subjective fatigue; described as medium-sized in later work. Motivational incentives reduced the effect. Did not correct for publication bias; later bias-focused analyses and replications suggest the pooled effect was inflated.
Focused meta-analyses addressing publication bias and small-study effects found very little evidence that the depletion effect is real with the laboratory methods most often used. Its bias-correction approach was disputed by Dang 2018.
Registered Replication Report of a standardised sequential-task ego-depletion protocol across 23 labs. Pooled effect d = 0.04 (95% CI -0.07 to 0.15), indistinguishable from zero, against a medium-sized effect in the earlier meta-analysis. Critics noted it tested a single protocol.
Preregistered multi-lab project using two depletion procedures. Confirmatory effect d = 0.06, non-significant; Bayesian analysis with an expected effect of 0.30 found the data about 4 times more likely under the null. Exploratory full-sample effect d = 0.08 (significant, Bayes roughly equal); effect somewhat larger in people reporting more fatigue; not moderated by trait self-control or willpower beliefs.
People who believed willpower is not limited showed no depletion effect after a demanding task; manipulating beliefs reproduced this; in a field study, willpower beliefs predicted eating, procrastination and goal striving in demanding periods. Not supported as a moderator in the 2021 multi-lab project.
Prior cognitive exertion reduced physical performance, g = -0.38 (95% CI -0.46 to -0.31); aerobic g = -0.26, isometric and motor -0.57, dynamic resistance -0.51, maximal anaerobic trivial (0.10, ns). Between-subject designs g = -0.65 vs within-subject -0.28. Mostly small lab studies in non-athletes.
After the authors' 2021 correction (doi:10.5334/joc.178, miscoded effect-size signs fixed, two RPE measures added): pooled effect of mental fatigue on endurance performance dz = -0.53 (95% CI -0.76 to -0.28; originally -0.50) and on RPE dz = -0.32. Egger's test and a three-parameter selection model indicated publication or reporting bias; bias-corrected performance effect -0.14 (95% CI -0.46 to 0.16), non-significant. Conclusions unchanged: evidence does not conclusively show that mental fatigue harms exercise performance.
Trait self-control had a small-to-medium relationship with behaviour; the association was significantly stronger for automatic than for controlled behaviour, and for imagined than actual behaviour. General-population and student samples, mostly correlational.
Across 6 studies, habits for exercise, healthy snacking, sleep, study and meditation partly mediated the link between self-control and goal outcomes, and with greater automaticity and less reported effortful inhibition. Mostly correlational and prospective designs in young people.