The theory that willpower operates as a limited resource.
What the evidence says
- The ego depletion model, also called the strength or resource depletion model, conceptualises self-control as ”a limited resource that becomes depleted after a period of exertion”: on its account, people who complete an initial self-control task have reduced self-control capacity on a subsequent one[1]
- Original demonstration: participants who resisted chocolate and ate radishes gave up on subsequent puzzles faster than those who ate chocolate freely — resisting food depleted a shared self-control resource[2]
- The model’s glucose account, as a review by Gailliot and Baumeister proposes it: blood glucose is ”one important part of the energy source of self-control”, acts of self-control ”deplete relatively large amounts of glucose”, and self-control failures are more likely when glucose is low[3]
- The same review proposes that ”restoring glucose to a sufficient level typically improves self-control”[3]
- The challenge to the model came from re-analysis of the existing literature first: a series of focused meta-analytic tests, designed to address limitations in earlier appraisals, found “very little evidence that the depletion effect is a real phenomenon, at least when assessed with the methods most frequently used in the laboratory”[4]
- A preregistered Registered Replication Report across 23 laboratories (total N = 2,141), running a standardised sequential-task protocol, found an effect whose confidence interval encompassed zero: d = 0.04, 95% CI [-0.07, 0.15][1]
- A larger preregistered multisite test across 36 laboratories (N = 3,531) — using two different self-control manipulations rather than a single protocol — found a nonsignificant confirmatory result (d = 0.06), with Bayesian meta-analysis making the data 4 times more likely under the null than under an informed-prior alternative[5]
- That same project found depletion was not moderated by trait self-control, willpower beliefs, or action orientation; it was larger only for participants reporting more fatigue[5]
- The same project’s exploratory analysis cuts the other way: on the full sample, ignoring exclusion criteria, the effect was statistically significant (d = 0.08), and the Bayesian analysis found the data about equally likely under the null and the informed-prior hypotheses — so the preregistered test found no evidence for depletion without establishing its absence[5]
- The question is contested rather than settled in either direction: a review weighing the arguments on both sides concludes that “despite several hundred published studies, the available evidence is inconclusive”[6]
- Decision-making depletes the same resource as self-control: participants who made choices among consumer goods or courses showed reduced subsequent self-control (less physical stamina, more procrastination, lower persistence) compared to those who deliberated without choosing — the act of committing to a choice, not just thinking, is what depletes[7]
- In a real-world test, judges’ favorable parole rulings dropped from ~65% to nearly 0% within each decision session and reset to ~65% after food breaks — the authors propose depleted mental resources, and say plainly that they cannot tell what restored them: ”we cannot unequivocally determine whether simply resting or eating restores the judges’ mental resources because each of the breaks was taken for the purpose of eating a meal”. The finding itself is contested on case ordering[8]
- That parole finding was challenged in print. Weinshall-Margel & Shapard examined the authors’ data, added 12 hearing days (n = 227) and interviewed court staff, and argue case ordering was not random: the board completes all cases from one prison before breaking, unrepresented prisoners usually go last within a session, and attorneys with shared counsel present their cases together — best first, weakest last. They also note the authors had no data on prisoners’ in-prison behaviour, a factor specified by Israel’s Parole Release Law of 2001[9]
- It is a letter offering an alternative explanation, not a replication showing the effect absent, and the original authors replied. So: the finding is contested on the grounds that case order was not random, and it has not been refuted[9]