Not in the way the popular theory says. The idea that self-control is a limited resource that drains with use, called ego depletion, was tested in 23 labs with 2,141 people and came out at an effect size of 0.04, statistically indistinguishable from zero. A second test across 36 labs and 3,531 people found the data 4 times more likely under no effect than under the theory.
What did survive is smaller and more useful. The faint effect that remained was bigger in people who said they were tired.
Where the drained battery idea came from
The strength model of self-control treats willpower like a muscle or a fuel tank. Use it on one task, and you have less for the next. In the classic experiment, people first resist something, say a plate of cookies, and then give up sooner on a hard puzzle.
An early meta-analysis pooled experiments like that and reported a robust, medium sized effect, d = 0.62. It became one of the most cited ideas in psychology, and it shaped a lot of diet advice: avoid temptation all day and you will have nothing left by evening.
A 2014 reanalysis by Carter and McCullough applied methods that estimate and correct for small-study effects such as publication bias to the data from that meta-analysis. They found very strong signals of publication bias, and a corrected depletion effect that was no different from zero. Their recommendation was to stop elaborating the model and first establish whether the basic effect exists at all.
They also framed the whole episode as a case study in how small studies, published selectively, can build a confident literature around an effect that may not be there. That framing matters for anyone who absorbed the battery idea from a diet book, because the book was usually citing the original pooled number.
What the 23 lab replication found
A registered replication report in Perspectives on Psychological Science, published in 2016, did exactly that. Twenty-three laboratories ran one standardised, preregistered depletion protocol with a combined sample of 2,141 people.
Pooled, the depletion effect was d = 0.04, with a 95 percent confidence interval from -0.07 to 0.15. That interval includes zero comfortably. For comparison, the effect the theory was built on was more than fifteen times larger.
The protocol was based on a sequential-task paradigm from Sripada and colleagues, chosen in advance so that no lab could tune its method after seeing its data. That is what preregistration buys: a result nobody could nudge toward the answer they hoped for.
The 2021 test that gave the theory every chance
Supporters argued the 2016 protocol was a weak manipulation. So a 2021 multisite project in Psychological Science, led by Kathleen Vohs, one of the theory's own proponents, ran a different design across 36 labs and 3,531 people. Each lab used one of two procedures meant to tire self-control, then measured performance on a second self-control task.
The preregistered confirmatory result was nonsignificant, d = 0.06. A Bayesian analysis using an informed prior found the data about 4 times more likely under no effect than under the depletion hypothesis. Relaxing every exclusion rule produced d = 0.08, with the evidence roughly balanced either way.
One exploratory result is worth keeping. The effect was larger for people who reported more fatigue, and it was not moderated by trait self-control, willpower beliefs, or action orientation. Tiredness predicted the effect. Believing in willpower did not.
To put d = 0.06 in everyday terms, it means the depleted and non-depleted groups differed by six hundredths of a standard deviation on the second task. That is far too small to explain why a whole evening of eating feels different from a morning.
So why does 9pm feel so much harder?
The replications do not say evening eating is imaginary. They say the reason is probably not a tank you drained by resisting a biscuit at 11am.
The fatigue finding fits better with what is already known about sleep and appetite. Short sleep shifts the hormones that govern hunger, which the sleep and appetite hormones piece covers with its own trials. And food noise is real as a pattern of intrusive food thoughts, not a moral weakness that shows up when reserves run low. Treating a tired, hungry evening as a willpower failure points you at the wrong fix.
It also changes how to plan a day. If the battery model were right, the sensible strategy would be to ration decisions and hoard self-control for the evening. With the effect this close to zero, there is little to ration. The more promising targets are the ones the fatigue result points at: sleep, not arriving at dinner very hungry, and not keeping the foods you struggle with within arm's reach.
Why OffRamp's streak forgives a missed day
OffRamp counts a logging streak, and the streak logic is deliberately forgiving: on the free tier, one missed day in a row is bridged rather than resetting the count to zero, and Pro extends that gap. An unlogged today that is still open does not count as a miss at all.
That rule comes from the same place as this article. A streak that resets on one lapse treats consistency as a willpower test you either pass or fail, which is precisely the model that did not replicate. What actually keeps weight maintenance going is continuity over weeks, and the habit research says a single missed occasion does not undo a habit being formed.
The daily check-in follows the same idea. It asks for appetite, food noise and energy on a 1 to 5 scale, not for a rating of how well you resisted anything. Staying motivated in maintenance goes further into building systems that do not lean on self-control at all. None of this is medical advice, and if evening eating feels out of your control, that is worth raising with your prescriber or a clinician.
