Willpower as a finite resource is one of the most exported ideas psychology has produced. The laboratory version is the : do one self-control task, then do worse on a second. An early meta-analysis put the effect at medium size, and later ones questioned both the size and the existence of it, and flagged possible bias. So 23 labs ran one preregistered protocol. What was left?
Almost nothing was left. Across 23 laboratories and 2,141 participants running a standardized sequential-task protocol, the meta-analytic estimate came out at d = 0.04, with a 95 percent from -0.07 to 0.15. The interval includes zero, which means the data are compatible with no effect at all. The resource depletion model of self-control lost the experimental finding it was built on.
Good self-control tracks better health, steadier relationships, success at work and school, and less trouble with crime and addiction, so understanding how it works has real stakes. The dominant account has been the strength or resource depletion model: self-control draws on a limited supply that runs down with use.
The experimental engine of that model is the sequential-task paradigm. People complete an initial self-control task and then perform worse on a subsequent one, a state named . An early meta-analysis of these experiments reported a medium-sized effect, and the idea spread far beyond the laboratory on the strength of it.
Then later meta-analyses questioned both the size and the existence of the effect, and identified instances of possible bias. Those analyses are what prompted a : 23 laboratories, 2,141 participants in total, all running one standardized protocol based on a sequential-task design agreed in advance.
The pooled result was d = 0.04, with a 95 percent confidence interval running from -0.07 to 0.15. An interval containing zero means the data are compatible with there being no effect, and an interval this narrow also means any real effect is small. Read the scope precisely, though. One standardized protocol was tested, not every operationalization of the idea, and critics have argued the chosen task pairing was insensitive. What the report rules out is a medium-sized ego-depletion effect of the kind the model needed.
wat. Twenty-three laboratories ran the willpower-drain experiment the same way and landed on a number indistinguishable from nothing.
wat wat
A large and preregistered estimate of nearly zero is how a field corrects itself. It also leaves the practical advice stranded, since budgeting your willpower across a day assumed a budget that this did not find.
How do we know?
Hagger and colleagues coordinated a Registered Replication Report in which 23 laboratories, with 2,141 participants in total, ran one standardized ego-depletion protocol and the results were meta-analyzed.
What is still unknown
This tests one standardized protocol, not every version of the idea. Defenders argue the chosen task was a poor instrument, and a confidence interval around zero rules out a large effect rather than proving the absence of a small one.
Learn something?
Where to next
What was left of the marshmallow test when they ran it on ten times as many children?self-control
If you hold a pen in your teeth, do cartoons get funnier?replication
What happened when the Dunning-Kruger effect was run on random data?replication