In 1998, Roy Baumeister and three colleagues put people in a room that smelled like fresh-baked chocolate chip cookies. On the table sat two plates: warm cookies, and a pile of raw radishes. One group was told to eat the cookies. One group was told to eat the radishes and nothing else. A third group ate nothing at all. Then every single person was handed the same task — copy a geometric figure, following a set of rules they couldn’t possibly satisfy. There was no correct answer. The only measurable thing was how long someone kept trying before they quit. The people who had been sitting there eating radishes stopped first. Willpower, the paper argued, is a finite resource drawn from one shared pool, spent on whatever demands self-control, and eventually gone. The cookie eaters had fuel left. The radish eaters had burned theirs on resisting dessert.
That finding became one of the most quoted results in modern psychology. It turned into the muscle metaphor — willpower as a battery, draining through the day, refilled by sleep — and from there into the entire architecture of self-help. Do your hardest work in the morning. Never grocery shop while hungry. Never make big decisions at night. Whole companies reorganize their meeting schedules around a cognitive tank that has been assumed to exist for almost thirty years.
🧠 The Muscle in Your Head
The 1998 paper contained a second experiment that mattered just as much, and gets quoted far less. Participants were asked to write a short speech arguing against their own position on something they cared about. Half of them got to choose which topic to argue against. Half were assigned one. Then everyone did the frustrating puzzle. The people who had chosen their topic gave up sooner than the people who had been ordered to write the speech.
Sit with that for a second, because it runs against the obvious reading. Doing something against your beliefs should be the exhausting part. Here, the exhausting part was being the one who decided. Choice itself was pulling from the same well. That is the version of the theory with the sharpest teeth, and it is the one that explains why a day of small decisions — which shirt, which email to answer first, which of six nearly identical options — leaves you flattened by evening.
🤔 Twenty-Three Labs Said No
Here is where the story turns. In 2010, Martin Hagger and Nikos Chatzisarantis published a meta-analysis that seemed to confirm the effect. Six years later, the same two researchers ran a much harder test: twenty-three laboratories, a standardized procedure, 2,141 participants, and the whole thing pre-registered before anyone collected a single data point. The effect came out at d = 0.04, with a confidence interval running from −0.07 to 0.15. It was small enough, and uncertain enough, to contain zero. In Perspectives on Psychological Science, where that replication was published, the authors did not bury the lede.
It got worse. In 2021, thirty-six labs and 3,531 participants ran the same test again, this time with two different depleting procedures. The confirmatory analysis found d = 0.06 and did not reach significance. A Bayesian analysis put the data four times more likely under the hypothesis that the effect does not exist than under the hypothesis that it does. Letting the strictest exclusion rules go — looking at everyone, including the data they’d set aside — produced a significant d = 0.08. But even there, the Bayesian comparison came out roughly even. Two enormous, careful, pre-committed attempts, and the famous effect is either very tiny or not there.
There was a rescue story, of course. If willpower depletion is at least partly a belief — if you get depleted because you expect to — then the whole thing isn’t dead, it’s just in your head. That idea had real support behind it. And when the thirty-six-lab study looked for it directly, willpower beliefs did not moderate the effect at all. Neither did trait self-control, nor action orientation. The one thing that did track: how tired people said they felt. The strongest predictor of collapsing on the second task was the person’s own report of fatigue, which is not a resource measurement. It’s a feeling. There is even a 2010 study that induced the feeling directly, convincing people they were more or less depleted than they actually were, and their performance followed the belief. Older adults, meanwhile, don’t reliably show the effect at all; standard depletion manipulations work mainly on undergraduates, who are the population nearly every study drew from.
So what’s actually true? Something is happening in that room. It’s smaller than anyone advertised, it may depend on the paradigm being sensitive enough to catch it, and the argument now is about measurement as much as about willpower.
🔗 Who Sells You the Empty Tank
Tilt the finding toward commerce and it stops being academic. A version of ego depletion became a standard explanation for why people buy things they didn’t intend to buy: the day drains the tank, the shopper goes passive and impulsive, and the eleventh option wins by exhaustion. There’s a tidy story there about high-end options too — that people whose self-control is worn down will pay a premium for status goods because owning them restores a feeling of control.
That story is now doing a lot of work on a foundation that came back at d = 0.06. The reliable mechanics underneath a tired decision are cognitive load and choice architecture, both of which hold up under scrutiny, and neither of which requires a finite willpower reservoir. Design that reduces what a tired person has to figure out still works. It just works for duller reasons than the ones the sales deck claims.
There’s a gentler version for a long workday, too. The part of the depletion theory that survived was the felt-tiredness part, and that part is worth respecting. Whether or not there’s a tank, the sensation of an empty one changes how you perform. So does the morning-tank belief — not because it’s true about the resource, but because believing you have fuel is itself a little fuel.
🎲 This Afternoon’s Small Test
The next time you catch yourself about to quit something and you’re reaching for the explanation “I’m out of willpower,” ask a different question: am I actually depleted, or do I just feel that way?
They are not the same claim, and only one of them has held up. The other is worth keeping around anyway, because it’s the one that predicts your next hour.
Then try the small, strange finding that has always been the loveliest detail in this literature: tasting something sweet — a sip, not even swallowed — was enough to restore performance in a depleted state, which suggests the signal matters more than the sugar. The mechanism is contested. The excuse to keep chocolate nearby is not.