← Blog

Focus, explained

Does willpower run out? What the replications found

By late afternoon you can't make yourself start the hard thing, and the explanation feels obvious: you used up your willpower. For years psychology largely agreed. Then the largest tests of the idea came back close to zero. Yet mental fatigue is still something you can feel. Those two facts need separating.

By ComunAI · 2 November 2026 · ~6 min read

The idea: self-control as a fuel tank

In 1998 Roy Baumeister and colleagues published the experiments that named the effect. In the best known, people who had to resist fresh cookies and eat radishes instead gave up sooner on a puzzle afterwards. They called it ego depletion: self-control draws on a limited resource, and using it leaves less for whatever comes next.¹

By 2010, a meta-analysis by Martin Hagger and colleagues could pool 83 studies, and it reported a medium-sized effect (d = 0.62).^2,3

Then the large tests arrived

In 2014 Evan Carter and Michael McCullough reanalysed that literature, correcting for the tendency of small studies with positive results to get published. They found strong signs of publication bias, and an effect that, after correction, was no different from zero.³ Preregistered projects across many laboratories followed, with the analyses fixed in advance:

The last of these was led by Kathleen Vohs and Brandon Schmeichel, both long-standing contributors to this research, and its preregistered analyses found no evidence for depletion; an exploratory analysis of the full sample reached d = 0.08. Some effect may survive. A 2019 review by Malte Friese and colleagues calls the evidence inconclusive after several hundred published studies.⁷ But the debate has shifted: it is now about whether the effect is tiny or absent, not about a medium-sized law of the mind. Evidence: Contested

The famous judges

The most quoted real-world example is a 2011 study of parole decisions. Shai Danziger and colleagues reported that favourable rulings fell from about 65% to nearly zero within each session and jumped back after a food break.⁸

The order of cases may not have been random, though. Keren Weinshall-Margel and John Shapard pointed to factors the analysis overlooked, including how cases were scheduled.⁹ Andreas Glöckner's simulations then showed that part of the pattern could arise from sensible time management alone, so the size of the effect was overestimated.¹⁰ This study cannot carry the claim. Evidence: Contested

Mental fatigue is still real

The replications tested a narrow prediction: one short self-control task impairs the next. They did not test a long day. Decline over sustained attention is covered in the post on session length, and the glucose version of the fuel idea, which Robert Kurzban argued is unlikely on physiological grounds,¹¹ in food, water and focus.

Three accounts try to explain the tired feeling:

• Motivation and attention shift. Michael Inzlicht and Schmeichel proposed that effort at one moment moves motivation and attention away from further control, rather than draining anything.¹² Evidence: Early
• A cost–benefit readout. Maarten Boksem and Mattie Tops proposed that fatigue arrives when perceived costs outweigh expected rewards.¹³ Kurzban and colleagues' opportunity-cost model is a close relative, treating effort as a signal about what else you could be doing.¹⁴ Evidence: Early as an explanation.
• A metabolic cost after all. Antonius Wiehler and colleagues had two groups do high-demand or low-demand tasks across roughly a workday. Only the high-demand group shifted toward short-delay, low-effort choices, and by evening it had higher glutamate concentration in the lateral prefrontal cortex.¹⁵ Evidence: Early — one study, and it measured choices after a whole day, not after one radish.

Does believing in it make it true?

Veronika Job, Carol Dweck and Gregory Walton found that people who saw willpower as unlimited showed no depletion in the lab, and that the belief predicted better goal striving in demanding periods.¹⁶ A follow-up linked it to less procrastination and better grades among students with heavy course loads.¹⁷ Krishna Savani and Job later reported reverse depletion among Indian participants.¹⁸

In the 36-lab project, exploratory analyses found that willpower beliefs did not moderate the effect, though the effect looked larger in people reporting more fatigue.⁶ That the belief switches depletion off is unproven. Evidence: Contested

What the evidence says helps

• Reduce temptation rather than out-resisting it. Marina Milyavskaya and Inzlicht followed students across a semester with experience sampling. Merely experiencing temptations went with feeling depleted and with worse goal attainment, while effortful resistance was unrelated to it.¹⁹ Evidence: Early — one correlational study of students; the advice is our inference from it.
• Don't read tiredness as a character flaw. Our reading of the replications. Try a smaller next step, or a real break.
• Be wary of big decisions after a long, demanding day. Our inference from Wiehler's study, and no more than that.¹⁵

Where the evidence stands

The calibrated bottom line

Willpower as a fuel tank that one hard task can drain has not held up in the largest tests, though a very small effect may remain. Mental fatigue after long, demanding work is separate and better supported, with its mechanism still open. The upshot is modest: spend less effort fighting temptation, and treat a tired afternoon as information, not a verdict on your character.