The Triumph of Ego Depletion

The Real Story Behind One of Social Psychology’s Most Replicated Findings

This is a guest post by Roy Baumeister, who is a social and personality psychologist who has published hundreds of articles and nearly 50 books on diverse topics. Throughout his long career, he has worked toward building a big-picture understanding of the human mind based on integrating facts and findings from large areas of research. He is ranked among the world’s most influential psychology researchers.

---

The Intellectual Origin of Ego Depletion

Caryatid Fallen Under Her Stone, Rodin

Early in my career, casting about for a theme, I decided to focus on trying to understand the human self. Lots of eminent thinkers were talking about the self, back then. Hippies were taking drugs to discover their inner selves. Researchers were also interested in the self, seeking to use scientific methods to understand it, not necessarily their own individual self, but the essence of selfhood. It sounded good to me. I signed up.

In the 1980s, some of the foremost thinkers in the field started saying that the crucial key to understanding the self was how it controls and regulates itself. There were plenty of specific self processes being studied – self-esteem, self-presentation, self-concept, self-monitoring, self-awareness — but the smartest folks (Charles Carver, Shelley Taylor) were saying that self-control was different. It was not just one more of those but the basis of the whole system. So I thought I should learn about that. I spent a year or so reading many research studies. I noticed a pattern that nobody had remarked on — the self’s ability to exert control over itself seemed to deteriorate over time with continued use. People couldn’t maintain control over their attention indefinitely, even when they were navy personnel watching their sonar screens for potentially deadly signs of an enemy submarine that might sink their ship and kill them. Even with that like-or-death incentive, they couldn’t maintain focus indefinitely. Likewise, when people were trying to do multiple self-control tasks at once, like trying to diet and quit smoking at the same time, they did a lousy job of both. (They also became unpleasantly cranky to their families or roommates, meaning they were also failing at self-control of their emotions and speech.) So maybe how the self controls itself depends on some kind of energy. And when the energy (a.k.a. willpower) runs low, self-control will fail.

This was a radical idea at the time. Psychologists in the early 1990s were besotted with the spread of personal computers, and all the fashionable theories compared the mind to a computer and analyzed how it processed information. The idea that the self contained a limited amount of energy that could occasionally run dry was wildly out of fashion. Yet multiple signs pointed in that direction. My PhD students, including Mark Muraven and Ellen Bratslavsky, tried doing some experiments to test this idea, and these worked quite well. There might really be something to this.

There was. Throughout the rest of the 1990s and 2000s, the experiments continued to turn out well. More people got interested. The first couple papers were published in social psychology’s top journal. A major research grant came through. Creative people came to work in our lab and added new directions, new insights. Not everything worked, but plenty did. Other researchers in other labs started publishing similar studies, building on the same foundation, adding new ideas, sometimes offering competing theories. I kept revising the theory, which didn’t bother me, indeed added to the excitement. For example, self-control deteriorated when people were depleted. So maybe they were out of mental fuel. But if we offered them a good incentive to perform well even when they were already depleted, they could rise to the occasion, though after this they were extra depleted. That meant it wasn’t a matter of having used up all one’s willpower, all the brain fuel (which could be a disaster). Rather, the self was conserving its remaining energy, so it wouldn’t get too low. (Eventually I learned that this happens with physical muscles too. When your muscles are tired, if someone offers you money to put forth maximum effort, you can be just as strong as you were before you got tired. Though, again, you’re extra tired afterward.)

Enter the Replication Crisis

Then, abruptly around 2015, problems arose. Social psychologists had begun to criticize their own field, saying that many findings could not be replicated in other labs. The new approach was to run multi-lab replication attempts, in which a dozen or so different laboratories would all run the same experiment and pool their results. Someone wanted to do this with ego depletion. Knowing how well it had worked for us and others, I supported this. They asked for my input. I suggested a variety of procedures that produced good results, but they said no to all of them. They wanted something simple, preferably computer-administered, easy to run in different labs on different continents, and preferably fast[1]. Fast is a problem, because ego depletion is a kind of fatigue, and if you cut short the tiring-out part, people won’t be as depleted. Waking up ten minutes early has less impact than losing four hours of sleep. I didn’t know of any procedures that suited these requirements. A colleague saw something that had worked as a computer-administered version, so I passed it along to the research group, not knowing anything about it but trying to be helpful. They embraced it. Later they would say I had recommended it. I was being set up. But I was dealing with other stuff. My precious daughter had just died of a misdiagnosed heart ailment, and I was in shock. If you haven’t lived through the unexpected premature death of the person you loved most in the world, you can’t really imagine what it’s like.

When is a Replication Not a Replication?

The replication team couldn’t even be bothered to run the full procedure they were supposedly replicating, so they used a stripped-down version. It didn’t work. It shouldn’t have been a surprise, but the vultures quickly came out and proclaimed ego depletion to be a fake. There were articles in the journals, blogs, popular press things. It was ridiculous, but there was an appetite for blood. I was bothered but still it was nothing compared to my poor dead girl.

Eventually the replication group made their data publicly available to all scientists. A bright young fellow (Junhua Dang) reanalyzed the data and discovered that their replication had actually sort of worked. That is, they had mainly failed to create the state of ego depletion, so they didn’t really test the theory. To the very limited extent they succeeded in getting their research subjects into the depleted state, the results conformed to what the theory predicted. It was statistically significant, as we say. But all the publicity had already delighted in declaring ego depletion a fake.

As usual, the public and the media don’t correct themselves. Something that looks bad gets plenty of press coverage, but subsequent exonerating evidence scarcely gets reported. It’s true in crime and scandal reporting. That’s how it was with ego depletion too. Plenty of researchers were still finding the effect in their labs, but the journals didn’t want to publish the positive findings. Journal reviewers would say, I don’t believe these results because we know this effect is not real. This is an absurd and antiscientific attitude, but it was enough to sway editors, who are always looking for reasons to reject a paper. (Editors receive many more manuscripts than they can publish, so they embrace any reason to reject.)

At one point I read (and published a journal article reviewing) all the multi-lab replications done in all of social psychology, on all of its topics. Most were failures. Some would say that means that all the knowledge built up over a half century of research in social psychology was garbage, but others said, more plausibly, that the way these multi-lab studies are run is to blame, furnishing false-negative results. Crucially, late in the game we noticed that the few successes mostly involved live personal interaction rather than computer-administration, which is what we had suggested but the replication group had rejected. They wanted to do it computer-administered, which it was turning out was a broad recipe for failure across all of social psychology.

Two other groups ran multi-lab replications of ego depletion. One worked beautifully. It was published but ignored. Another got perfectly mixed results. They got a significant result if the data were analyzed one way, but nonsignificant if analyzed a different way. Standard practice is to report both analyses. But the editor didn’t want an inconclusive paper, so he told the authors to feature the failed replication and bury the successful replication in a supplementary online file that would only be found by the few people who bothered to look for it. (Editors prefer to publish failed replications, which I can understand. Failed replications are more exciting and newsworthy than successful ones. Successful replications merely confirm what we already know. But failed replications suggest revising our knowledge base, so they are newsworthy. I don’t blame the editors, though sometimes, as in this case, these policies are detrimental to the scientific goal of pursuing the truth.)

The Empirical Case for Ego Depletion

At some point I realized that if we make replication success the main criterion of quality, ego depletion is one of the most successful findings in social psychology. I published a journal article making that case, which is roughly as follows. First, there are literally hundreds of statistically significant, published findings consistent with the theory (and plenty of unpublished ones). There are close to zero in the opposite direction[2]. The successful studies included some by me and my colleagues, but plenty were published by people I’ve never met, and on multiple different continents. Some researchers do fail to get the effect, but that’s mainly because they fail to create the depleted state, not because they create it and it fails to produce its predicted results. (Thus, they do not really test the theory.) There are very few genuine failures to replicate, in the sense that the experimental procedure succeeded in creating the depleted state but the effects of the depleted state failed to appear. That is quite rare.

Sign seen at the 2017 NYC March for Science

Anyway, first criterion is abundant successful replications, including from different laboratories. Check!

Second, another new criterion is for researchers to commit themselves in advance to exactly what they are going to do, including how much data to collect and what statistical tests to use. Ego depletion has some successful pre-registered studies. Check!

Third, another criterion is getting some findings outside the lab, so it’s not just some hothouse phenomenon. Ego depletion has plenty of those too. Depleted health care professionals gradually cease to wash their hands between each patient. Depleted physicians give in to patient demands for antibiotics which are very likely useless (and thus harmful to society, which is why there are strong recommendations not to overprescribe). Accountants make more mistakes when depleted. And so on. (Again, check!)

Fourth and last, it has a pretty good record with multi-lab replications, including two complete successes, and two partial successes. (Big check for this.)

I say “two complete successes,” and that’s my occasion for writing this. There was recently published a new multi-lab replication, with several dozen researchers on different continents (none of them related to my group), all doing the same experiment with pre-registered procedures and analyses, and finding strong positive results. Their results confirmed the reality of ego depletion.

Not only did this produce a statistically significant effect – it was a pretty big one. As I said, I have carefully read all the several dozen multi-lab replications in all of social psychology, on many different topics. Very few got significant effects, but even among the few that did, most got only tiny effects, much smaller than in the original work that they were replicating. Let me add for the intelligent non-scientist, ‘significant’ means only that we can be pretty sure the result is different from zero, from absolutely no effect. It could be a tiny effect, as long as the statistics tell us that this result would not occur by chance more often than one out of twenty trials. Almost always, replications find smaller effects than in original studies. (Most multi-lab replications of various other findings have produced very small effects.) But this replication of ego depletion got bigger effects. I think it is one of the two or three biggest-sized multi-lab replication findings in all of social psychology.

Thus not only did they produce a clear and significant replication of ego depletion – they produced a much bigger effect than almost all of the other similar projects in all of social psychology. Again, this fits the conclusion that ego depletion is extremely well replicated, better than almost anything else in social psychology.

By the way, whenever I say that ego depletion may well be the best replicated finding in all of social psychology, I always invite disagreement, as I do here also. If you disagree, please send me a nomination of any other social psychology finding that has a better replication record. I’ve been doing this for several years, and thus far I have received only one nomination. (That had a pretty good case also. I thought it a close call. Even so, the least optimistic conclusion at present is that ego depletion is the second or third best replicated finding in social psychology.) You can use the criteria I used, as listed above, or make an argument for different criteria. Again, out of all the thousands of findings in social psychology, just name one or two that have a better replication record than ego depletion.

Ego depletion continues to be attacked. My impression is that this mainly comes from people who can’t get their own theories to work and so try to build themselves up by tearing down the honest hard work of other people. (Ego depletion is not the only target of such people.) Perhaps my judgment here is uncharitable. But the effect has been replicated hundreds of times, with different methods, by many different laboratories and research teams on different continents and with no connection to each other. There are very few genuine failures to replicate, by which I mean researchers who successfully created the depleted state but found it did not produce its usual consequences. (There will always be a few, for any phenomenon. That’s basic statistics. Moreover, social psychology has no effects that always occur.)

I don’t mind scientists who question ego depletion. Questioning everything is what a good scientist does. But I cannot imagine any competent, honest scientist looking at all that evidence and saying “no, there’s nothing there.” Anyone who seriously believes that ego depletion is not real is dismissing an enormous amount of research.

With the new, recently published multi-lab replication of ego depletion, the case should be settled. The effect is real and, if properly done, is something that goes beyond a small effect to a medium or large one. Plus, remember, what can be done in the lab is often much reduced as compared to what occurs in daily life. For example, much has been written about military incompetence, and with justification. But think what it must be like for an officer to make moment-by-moment decisions during a campaign in which he only gets a couple hours sleep (which increases vulnerability to depletion) and has already made multiple difficult decisions that will send some soldiers to their death. Nothing we can do in the lab has that kind of impact.

My research career is nearly over, and the future is in other people’s hands. But if they refuse to believe ego depletion, they will never be able to get an accurate understanding of self-control — which is one of the most important traits contributing to success or failure in life, and, as I said earlier, one of the most important keys to understanding the self.

Rodin, The Thinker at the Gates of Hell

References for intrepid scientists:

Baumeister, R.F., & Tice, D.M. (2022). Ego depletion is the best replicated finding in all of social psychology. Scholarly Journal of Psychology and Behavioral Sciences, 6. MS ID 000234. Doi: 10.32474/SJPBS.2021.06.000234.

Baumeister, R.F., Tice, D.M., & Bushman, B.J. (2023). A review of multi-site replication projects in social psychology: Is it viable to sustain any confidence in social psychology’s knowledge base? Perspectives on Psychological Science, 18, 912-935. Doi: 10.1177/17456916221121815.

The latest multi-lab replication:

Dang, Xiao et al. … (2025). Revisiting ego depletion: evidence from multi-lab collaborations. Journal of Pacific Rim Psychology, 19, 1-10. DOI: 10.1177/18344909251386084

For the uninitiated:

Baumeister, R. F., & Tierney, J. (2011). Willpower: Rediscovering the greatest human strength. New York: Penguin Press.

Baumeister, R.F., & Vohs, K.D. (2016). Strength model of self-regulation as limited resource: Assessment, controversies, update. Advances in Experimental Social Psychology, 54, 67-127. Doi: 10.1016/bs.aesp.2016.04.001

Footnotes

---

[1] The reason for speed is that research subjects are compensated based on time, usually in half-hour units. If it takes longer than half an hour, they have to pay for an hour, so that doubles the cost of conducting the experiment. But the half hour usually has to include welcome (some are late), instructions, the manipulation to deplete willpower, and then another task to measure the consequences. To get this done in half an hour, one has to keep things brief.

[2] Some skeptics have claimed, absurdly, that the hundreds of findings in support of ego depletion must just have been due to random luck, like if someone were to flip a coin ten times and get heads every time. But if this were all about luck, there would be an equal number of lucky findings in the opposite direction, showing that people perform better when depleted, rather than worse.

---

Commenting

Before commenting, please review my commenting guidelines. They will prevent your comments from being deleted. Here are the core ideas:

1. Don’t attack or insult the author or other commenters.

2. Stay relevant to the post.

3. Keep it short.

4. Do not dominate a comment thread.

5. Do not mindread, its a loser’s game.

6. Don’t tell me how to run Unsafe Science or what to post. (Guest essays are welcome and inquiries about doing one should be submitted by email).