The Chocolate-and-Radish Experiment That Birthed the Modern Conception of Willpower

Psychologist Roy Baumeister reflects on his groundbreaking 1998 research on self-control and shares how it became the dominant theory despite its unpopular Freudian roots.


A strong man from the late 19th century. Library of Congress

Back in 1996, Roy Baumeister conducted an experiment that was downright evil.

Together with his former Case Western Reserve University colleagues Ellen Bratslavsky, Mark Muraven, and Dianne Tice, he examined the effect of a tempting food challenge designed to deplete participants' willpower through the awful power of an unfulfilled promise of chocolate!

In the first part of the trial, Baumeister kept the 67 study participants in a room that smelled of freshly baked chocolate cookies and then teased them further by showing them the actual treats alongside other chocolate-flavored confections. While some did get to indulge their sweet tooth, the subjects in the experimental condition, whose resolves were being tested, were asked to eat radishes instead. And they weren't happy about it. As the scientists noted in their Journal of Personality and Social Psychology paper two years later (PDF), many of the radish-eaters "exhibit[ed] clear interest in the chocolates, to the point of looking longingly at the chocolate display and in a few cases even picking up the cookies to sniff at them."

After the food bait-and-switch, Baumeister's team gave the participants a second, supposedly unrelated exercise, a persistence-testing puzzle. The effect of the manipulation was immediate and undeniable. Those who ate radishes made far fewer attempts and devoted less than half the time solving the puzzle compared to the chocolate-eating participants and a control group that only joined this latter phase of the study. In other words, those who had to resist the sweets and force themselves to eat pungent vegetables could no longer find the will to fully engage in another torturous task. They were already too tired.

In the psychology world, the key finding of this seemingly silly study was a breakthrough: self-control is a general strength that's used across different sorts of tasks  -- and it could be depleted. This proved that self-regulation is not a skill to be mastered or a rote function that can be performed with little consequence. It's like using a muscle: After exercising it, it loses its strength, gets fatigued, and becomes ineffectual, at least in the short-term. Perhaps more importantly, this research would go on to serve as the foundation for at least 1,282 other studies involving everything from consumer to criminal behavior. It would, for instance, help show why people are energized by positive messages, more likely to engage in retail therapy when brokenhearted, and better off tricking themselves into eating less than willfully dieting.

In the Q&A below, Baumeister looks back on his team's influential study. He opens up about why they were initially very unhappy with the results, how some scientists continue to misuse their findings, and how this psychological experiment eventually wound up influencing the world of biology.

What was the predominant understanding about self-control back then? What was the purpose of your research?

There were three competing theories about self-control. One was that it was a kind of information processing: the mind knows what's up, figures out what to do, and does it. A second was that it was a kind of energy or strength, akin to the folk notion of willpower. And the third was that it was a skill. This last one was favored by child psychologists, who think of children growing up and acquiring skills little by little, with self-control being one of them.

One scenario in which the theories made conflicting predictions is when one exerts self-control and then gets another, different demand for self-control soon afterward. The information processing theory predicted that people should do better on the second self-control task as a result of having done the first, because the first would get the self-control module operating. This is a bit like having the program already up and running on your computer, which makes it quicker at taking on a new task. The second theory holds that your strength or energy would be depleted by the first task, so you would do worse on the second. And the skill model predicts no change from one trial to another, but gradual improvement over many occasions.

So we set up our experiment to pit these three against each other. At a lab meeting, Dianne Tice, my wife and longtime colleague, suggested confronting people with radishes and chocolates and telling them, "We really need you in the radish condition" to deplete their willpower before they solved some puzzles. Essentially, we had people engage in one self-control task and, afterward, had them do something quite different that still required self-control.


A 1919 Popular Science ad. Results may vary.

What made your study stand out?

This sort of theory was very far from the reigning models and theories in psychology when we started this work in the mid-1990s. Everyone was talking about cognitive processes and using information-processing models. Nobody talked about energy. In fact, when I first began to present this work to scientific audiences, I used to joke, "Energy models are so out of fashion that we're not even opposed to them anymore!" They were associated with obsolete Freudian theories about clashing instincts and the like. Our group was actually not very happy about these findings, because they really suggested we needed a theory based on energy, and we knew these would be tough to sell. When I was looking for a term to describe these effects, I settled on "ego depletion" as homage to Freud, because as far as I could ascertain he had been the last major theorist to say that the human self was made partly of energy and worked by energy processes.

As a researcher on self and identity, a common refrain I've heard at conferences in the late 1980s and early 1990s was, "We really don't understand the so-called agent aspect of the self." There was plenty of research on self-concepts and self-esteem, and even some on how people manage the impressions they make on others. But how the self does things was by all accounts still a deep mystery. I think that was because one needs energy theories to understand it, and, as I said, nobody in mainstream psychology was using energy theories.

As luck would have it, the 1990s saw psychological theorizing shift away from emphasizing information processing toward incorporating more biology-based ideas. So the basic, immediate skepticism that greeted our early findings gradually subsided. Nowadays, energy theories are becoming respectable again, and it's nice to think that our findings on ego depletion have contributed to their greater acceptance. By now, however, they've been replicated and extended in many different laboratories, so I am confident they are real. In fact, there was recently a meta-analysis, which I had nothing to do with, confirming the reality of ego depletion across dozens of studies.

How far has our knowledge of willpower come since, and what questions remain?

There have been several leaps forward, adding wholly new dimensions to our basic finding. For one, we found that willpower gets depleted not just by acts of self-control but also by other key things the self does: making choices and decisions, exerting initiative, perhaps planning and executing plans. This has led us to cast about for an even bigger umbrella than self-control, and it has gotten us into the interdisciplinary debates about free will. Some new work being published this year also showed that people suffer from ego depletion on an almost daily basis, outside the laboratory, and away from experimental manipulations, in the context of their everyday lives and normal activities. This has been a very welcome extension.

Another is that willpower seems tied in to the body's basic energy supply, as carried in glucose in the bloodstream. This suggests that willpower is not just a metaphor. We had been using the ideas of energy and depleted energy in a largely metaphorical fashion, but apparently it is the same basic energy that the body gets from food. It is also the same energy used for other metabolic activities. We've begun to learn about mind-body interactions -- why self-control might deteriorate during premenstrual syndrome and how it may also interact with the immune system, which sometimes uses plenty of glucose when fighting off a disease. That's why people may become cranky and irritable when they have been exposed to a cold: The immune system is using up the glucose so it's hard to be patient!

The biggest unanswered question to me is how to combine the glucose findings with the evidence that self-control can be improved by regular exercise. Quite possibly the answer is more complex, such as in how the brain allocates energy reserves.

Do you have any pet peeves about how your paper has been used by other researchers?

Some people still don't want to believe that energy has any place in psychology. They want to reduce ego depletion to an information processing sort of thing after all. There have been a handful of quite nicely designed experiments showing that the effects of ego depletion can be reduced or changed by manipulating what or how people think about what they are doing. I like the studies, but often the authors then go overboard to try to claim that ego depletion is entirely the result of their effect.

A recent set of studies, for example, showed that if you encourage people to believe they have unlimited willpower, they perform well despite ego depletion. This was a nice finding, but the authors then went wild and claimed that ego depletion is all in your head, so to speak. It's unfortunate that the field encourages scientists to overstate their findings and draw extreme conclusions. We replicated those findings in our lab, but they only occur when people are just slightly depleted. Once they are more extensively depleted, belief in unlimited willpower is no help. In fact, it makes things worse.

Could you describe this study's impact or legacy?

Well, if I say it's had a huge impact, that seems immodest, but if I try to be appropriately modest, it will sound like there's been very little impact!

Objectively though, there seems to be plenty of signs that this work has had good impact. The original papers have been cited extensively in other scientific publications, and many other researchers have borrowed our ideas and methods to use in their own work. Over the past decade, I've been fortunate to have quite a few talented people come to work in my laboratory, and the majority have wanted to do some work on these depletion and self-control processes. Beyond academic psychology, there may be some impact as well since my book Willpower spent several weeks in the bestseller lists.

Conferences on the self no longer bemoan how little understanding there is of the agent of the self. This is important to me because I was hoping to fill this important gap in psychology, since information processing and feelings aren't sufficient. We can now understand the active, executive aspect of the self, and I'd like to think that this work helped achieve a new, fuller understanding of the human self.

Presented by

Hans Villarica writes for and produces The Atlantic's Health channel. His work has appeared in TIME, People Asia, and Fast Company.

Join the Discussion

After you comment, click Post. If you’re not already logged in you will be asked to log in or register with Disqus.

Please note that The Atlantic's account system is separate from our commenting system. To log in or register with The Atlantic, use the Sign In button at the top of every page.

blog comments powered by Disqus


Cryotherapy's Dubious Appeal

James Hamblin tries a questionable medical treatment.


Confessions of Moms Around the World

In Europe, mothers get maternity leave, discounted daycare, and flexible working hours.


How Do Trees Know When It's Spring?

The science behind beautiful seasonal blooming

More in Health

Just In