Is Willpower a Battery? – Ego Depletion and What Remains of It After the Replication Crisis

The Hook

It is 5:40 p.m. Since the morning you have fought your way through an architecture review, diplomatically moderated two difficult customer calls, and three times resisted the temptation to answer an email in the heat of the moment. Now it is time for the code documentation you have been putting off for days. Instead, you open a news site. And you have an explanation ready that feels scientific: "My willpower is used up for today."

This explanation has a name – ego depletion, the "exhaustion of the self" – and for many years it was an influential idea in social psychology. The notion is intuitive: self-control behaves like a muscle or a battery. Use it, and you have less of it left afterwards.

The problem: when the research community began testing the effect in large, preregistered studies across many laboratories, it shrank to a size that can barely be distinguished from zero. The story of ego depletion is therefore not only a story about willpower, but a lesson in how science corrects itself – and in what a "well-established" finding is actually worth.

The Core Concept: Self-Control as a Limited Resource

In 1998, Roy F. Baumeister, Ellen Bratslavsky, Mark Muraven and Dianne M. Tice published an article in the Journal of Personality and Social Psychology with a question in its title: "Ego depletion: Is the active self a limited resource?" (vol. 74, issue 5, pp. 1252–1265).

Their starting hypothesis: choice, active response, self-regulation and other forms of volition might all draw on one common inner resource. They tested this with a design that later became the standard – the sequential-task paradigm: one group first has to complete a task that requires self-control, a control group a version without that demand. Afterwards both work on a second, quite different task that also requires self-control. If the first group performs worse there, this is taken as evidence of depletion.

The four experiments of the original paper:

The authors concluded that the self's capacity for active volition is limited and that seemingly quite different acts share a common resource. It was an elegant, parsimonious theory – and that is precisely what made it so attractive. It later became known as the strength model of self-control.

Example 1: The Meta-Analysis That Seemed to Confirm the Effect

A good decade later, the theory looked solid. In 2010, Martin S. Hagger, Chantelle Wood, Chris Stiff and Nikos L. D. Chatzisarantis presented a meta-analysis of 83 studies in Psychological Bulletin (vol. 136, issue 4, pp. 495–525). Result: a significant effect of ego depletion on performance in self-control tasks. They also found effects on effort, perceived difficulty, negative affect, subjective fatigue and blood glucose levels. The later replication study by Hagger and colleagues retrospectively describes this meta-analytic effect as medium-sized.

What is remarkable is how cautiously the meta-analysis itself was worded: in its own words, its findings provided "preliminary" support for the effect. And it found support for motivation and fatigue as alternative explanations – a hint that the pure resource model need not have been the whole story even then. For example, it reported that motivational incentives improved self-control in "depleted" individuals. For a strictly physical tank model this is awkward: an empty tank does not refill because you offer the driver money. Proponents of the model, however, explained the finding with a conservation hypothesis: depleted people hold back their remaining reserve and still spend it when the incentive is large enough (Muraven & Slessareva 2003). This made the model more flexible – but also harder to refute.

Example 2: Glucose – the Resource Gets a Name

If self-control consumes a resource, which one is it? A paper published in 2007 by Matthew T. Gailliot, Roy F. Baumeister and colleagues in the Journal of Personality and Social Psychology (vol. 92, issue 2, pp. 325–336), with the programmatic title "Self-control relies on glucose as a limited energy source: Willpower is more than a metaphor", gave a physiological answer: blood sugar. According to its abstract, acts of self-control lowered blood glucose, low glucose after an initial task predicted poorer self-control on the following task, and a glucose drink eliminated the impairment.

The metaphor had become literal. But this hypothesis, too, did not withstand later scrutiny. A meta-analysis by Junhua Dang (Lund University) in Appetite (2016, vol. 107, pp. 222–230) tested exactly the three corollaries of the glucose view – a drop in the level, a link between remaining glucose and performance, restoration through glucose intake – and, in its own words, found clear and consistent evidence against the glucose view: none of the three corollaries was supported. This stands in tension with the blood-glucose effect that the 2010 meta-analysis still reported; Dang's later meta-analysis was specifically tailored to the three corollaries of the glucose hypothesis. Only merely rinsing the mouth with a glucose solution showed a significant effect, though with alarming signs of publication bias. I am of the opinion that the rinsing effect in particular – if it is real – points to motivation or reward expectation rather than to a refilled energy store, since the sugar is not swallowed when rinsing.

Example 3: The Multilab Replications

The turning point came in two steps. First, later meta-analyses questioned the size and existence of the effect and identified signs of possible bias – one of them already carries in its title the conclusion that self-control does not seem to rely on a limited resource (Carter, Kofler, Forster & McCullough, Journal of Experimental Psychology: General, 2015, vol. 144, issue 4, pp. 796–815).

Then came two large, preregistered tests across many laboratories. The first was a Registered Replication Report (RRR): in this format the study protocol is fixed and peer-reviewed in advance, and many independent labs run the same standardized experiment. Whether the study is published does not depend on whether "something comes out of it".

Hagger et al. 2016 (Perspectives on Psychological Science, vol. 11, issue 4, pp. 546–573): 23 labs, 2,141 participants in total, a standardized protocol based on a sequential-task paradigm by Sripada and colleagues. Result: an effect of d = 0.04 with a 95% confidence interval from −0.07 to 0.15 – an interval that includes zero.

Vohs et al. 2021 (Psychological Science, vol. 32, issue 10, pp. 1566–1581): this second test was formally not an RRR but a preregistered multisite test in which each lab implemented one of two protocols. The first author was Kathleen D. Vohs, a long-standing proponent of the resource model – an important detail, because she and Baumeister had criticized the protocol of the first replication in a commentary (Baumeister & Vohs 2016) as new and mostly untested. 36 labs, 3,531 participants. The confirmatory test was not significant (d = 0.06). A preregistered Bayesian meta-analysis with an informed prior found the data to be four times more likely under the null hypothesis than under the alternative. Only an exploratory analysis of the full sample, ignoring the exclusion criteria, yielded a significant but very small effect (d = 0.08), for which the Bayesian analysis found the null and the alternative hypothesis about equally likely. Exploratory analyses also showed that the effect was larger for people who reported more fatigue – but it was moderated neither by trait self-control nor by beliefs about willpower.

For context: a d of 0.04 to 0.08 means that the distributions of the two groups overlap almost completely. I am of the opinion that an effect of this size – even if it exists – plays practically no role in explaining your everyday behavior at 5:40 p.m.

Example 4: "All in Your Head?" – the Willpower Mindset

As early as 2010 – long before the large replications called the effect into question – a paper by Veronika Job, Carol S. Dweck and Gregory M. Walton (Psychological Science, vol. 21, issue 11, pp. 1686–1693), titled "Ego Depletion – Is It All in Your Head?", had offered an influential alternative explanation. It did not question the effect, but its cause. In four studies the authors argued that depletion depends on whether people believe that willpower is limited. Those who saw willpower as a scarce resource showed the depletion effect; those who did not, did not. A longitudinal field study found that these beliefs predicted changes in eating behavior, procrastination and goal striving under demanding circumstances.

It was a tempting message – you would only have to change your belief. But it, too, held up poorly under scrutiny. In 2023, Nicholas P. Carruth, Jairo A. Ramos and Akira Miyake published a preregistered replication of Study 1 of the original paper in PLOS ONE (original: N = 60; replication: N = 187). They found neither a significant main effect of ego depletion nor a significant moderation by willpower mindset. In an accompanying opinion article in Frontiers in Psychology (2023), Miyake and Carruth reviewed 13 independent studies from 11 papers and found only one with an unambiguous moderation effect in the predicted direction. Their conclusion, stated in the title: the existing evidence is "weak at best". This fits with the fact that Vohs et al. 2021, in exploratory analyses, likewise found no moderation by beliefs about willpower.

Framework: What We Can Learn from Ego Depletion

The story can be summarized in a pattern that is useful far beyond psychology – including for evaluating benchmarks, A/B tests and architecture decisions:

It is important to note what the evidence does not show: it does not prove that there is no fatigue from exertion at all. Everyone knows the feeling of being exhausted after a long day, and Vohs et al. found, in exploratory analyses, a larger effect among people who reported more fatigue. What the large replications show is narrower: the classic laboratory paradigm – a short self-control task measurably lowers performance on a second task – does not, in standardized and preregistered form, yield an effect reliably distinguishable from zero. Whether a different paradigm or a longer, stronger demand would show a robust effect is an open question. This is where the Duhem-Quine problem comes in: a negative result may hit the theory – or the manipulation that was supposed to test it. This is exactly how Baumeister and Vohs argued against the first replication in 2016; the second test, with Vohs as first author and protocols whose selection involved proponents of the resource model, weakened this counterargument but did not eliminate it entirely: critics object to the operationalization here as well (Englert & Bertrams 2021).

The Central Takeaway

1. Distrust the battery explanation as an excuse. By current evidence, "my willpower is used up" is not a well-supported description of what is happening inside you. I am of the opinion that it is often more productive to ask about motivation, fatigue or simply unattractive task design.
2. Don't plan around an empty tank, plan around friction. I am of the opinion that the practical question is not "How much willpower do I have left?" but "How can I design the unpleasant task so that it requires less self-overcoming?" – for example through smaller work steps or fixed time slots.
3. Weigh evidence by quality, not quantity. Two preregistered multilab studies with more than 5,600 participants combined carry more weight than many small, unregistered individual studies.
4. Check the follow-up explanations too. A mechanism that makes a finding concrete (glucose) or an additional assumption that reinterprets it (mindset) is not automatically better supported than the finding itself – both failed their own tests here.

Cross-References in the Vault

• The Marshmallow Promise: How a Candy Test Redefined Willpower for Half a Century – and Was Then Put to the Test Itself – another willpower classic that is more nuanced than its reputation.
• The Peak of Not Knowing: Why Psychology's Most Famous Effect May Be Just a Statistical Mirage – a famous effect under methodological re-examination.
• The 38 Witnesses Who Never Were: The Bystander Effect Between Laboratory Finding and Video Evidence – how a popular narrative and the evidence diverge.
• Paid for What You Already Loved: Self-Determination Theory and the Crowding Out of Motivation – motivation as an alternative to the resource explanation.
• Which One of You Was It? The Duhem-Quine Problem and the Underdetermination of Theory – why a negative replication result does not automatically refute the theory.
• Desirable Difficulties: Why Effortless Learning Deceives and Memory Lives on Effort – a counterexample: learning effects that hold up well across many replications.

Food for Thought

Which explanation for your own behavior – "I'm a morning person", "I need pressure", "my willpower is used up" – have you never checked, even though you use it every day? And what would a small, honest experiment look like that could test it?

Fact Check

67 claims were checked (15 hard facts, 7 quotes, 16 source attributions, 29 synthesis claims). 9 claims were corrected, weakened or removed (1 false, 7 to be weakened, 1 not verifiable); numbers, years, journals and page ranges were correct throughout – the errors lay in the synthesis.

• The draft presented glucose and the willpower mindset as after-the-fact "rescue hypotheses" for a wavering finding. That was false: both emerged in 2007 and 2010 respectively, when the effect was still considered well established; corrected to "follow-up explanations" and "alternative explanation".
• The study by Vohs et al. 2021 was described as a Registered Replication Report with "exactly the same experiment" in all labs. In fact it was a preregistered multisite test in which each lab implemented one of two protocols.
• The argument "incentives cannot refill an empty tank" was weakened, because proponents of the model explained the finding with a conservation hypothesis (Muraven & Slessareva 2003).

Automated AI check against primary sources; no human expert review.

Sources and Further Reading

1. Baumeister, R. F., Bratslavsky, E., Muraven, M., Tice, D. M. (1998): Ego depletion: Is the active self a limited resource? Journal of Personality and Social Psychology 74 (5), 1252–1265. https://doi.org/10.1037/0022-3514.74.5.1252
2. Hagger, M. S., Wood, C., Stiff, C., Chatzisarantis, N. L. D. (2010): Ego depletion and the strength model of self-control: A meta-analysis. Psychological Bulletin 136 (4), 495–525. https://doi.org/10.1037/a0019486
3. Gailliot, M. T., Baumeister, R. F., DeWall, C. N., Maner, J. K., Plant, E. A., Tice, D. M., Brewer, L. E., Schmeichel, B. J. (2007): Self-control relies on glucose as a limited energy source: Willpower is more than a metaphor. Journal of Personality and Social Psychology 92 (2), 325–336. https://doi.org/10.1037/0022-3514.92.2.325
4. Dang, J. (2016): Testing the role of glucose in self-control: A meta-analysis. Appetite 107, 222–230. https://doi.org/10.1016/j.appet.2016.07.021
5. Muraven, M., Slessareva, E. (2003): Mechanisms of self-control failure: Motivation and limited resources. Personality and Social Psychology Bulletin 29 (7), 894–906. https://doi.org/10.1177/0146167203029007008
6. Carter, E. C., Kofler, L. M., Forster, D. E., McCullough, M. E. (2015): A series of meta-analytic tests of the depletion effect: Self-control does not seem to rely on a limited resource. Journal of Experimental Psychology: General 144 (4), 796–815. https://doi.org/10.1037/xge0000083
7. Hagger, M. S., Chatzisarantis, N. L. D., et al. (2016): A Multilab Preregistered Replication of the Ego-Depletion Effect. Perspectives on Psychological Science 11 (4), 546–573. https://doi.org/10.1177/1745691616652873
8. Baumeister, R. F., Vohs, K. D. (2016): Misguided effort with elusive implications. Perspectives on Psychological Science 11 (4), 574–575. https://doi.org/10.1177/1745691616652878
9. Vohs, K. D., Schmeichel, B. J., Lohmann, S., Gronau, Q. F., Finley, A. J., et al. (2021): A Multisite Preregistered Paradigmatic Test of the Ego-Depletion Effect. Psychological Science 32 (10), 1566–1581. https://doi.org/10.1177/0956797621989733
10. Englert, C., Bertrams, A. (2021): Again, no evidence for or against the existence of ego depletion: Opinion on "A multi-site preregistered paradigmatic test of the ego depletion effect". Frontiers in Human Neuroscience 15, 658890. https://doi.org/10.3389/fnhum.2021.658890
11. Job, V., Dweck, C. S., Walton, G. M. (2010): Ego Depletion – Is It All in Your Head? Implicit Theories About Willpower Affect Self-Regulation. Psychological Science 21 (11), 1686–1693. https://doi.org/10.1177/0956797610384745
12. Carruth, N. P., Ramos, J. A., Miyake, A. (2023): Does willpower mindset really moderate the ego-depletion effect? A preregistered replication of Job, Dweck, and Walton (2010). PLOS ONE 18 (6), e0287911. https://doi.org/10.1371/journal.pone.0287911
13. Miyake, A., Carruth, N. P. (2023): On the robustness and replicability of the moderating influence of willpower mindset on the ego-depletion effect: existing evidence is weak at best. Frontiers in Psychology 14, 1208299. https://doi.org/10.3389/fpsyg.2023.1208299

← All articles