Decision fatigue

Why does a person who chose carefully all morning start taking whatever is put in front of them by evening?

By six in the evening the same person who spent the morning weighing options carefully will take the first thing offered, agree to the default, or put the decision off entirely. The obvious explanation is that something ran out. Deciding costs, you did a lot of it, and the tank is empty.

That story is so intuitive that it named itself and got into the language before anyone checked it properly. It is worth asking what would have to be true for it to be right — and what the evening's behaviour looks like if it is wrong.

Take the fuel model seriously first, because it makes real predictions. If self-control draws on a consumable resource, then any act of control should impair the next unrelated one; the impairment should not be reversible by paying people more; and topping the resource up should restore performance.

The first prediction is the classic finding: people made to resist a plate of biscuits later gave up sooner on an unsolvable puzzle. Hundreds of studies followed that shape. The second and third are where it came apart.

Consider incentives. A muscle genuinely out of fuel does not care what you offer it. But the depletion decrement can be abolished by offering money, by telling participants their effort will help someone, or simply by telling them that exerting willpower is energising rather than draining. A resource whose availability depends on what you believe about it is not a resource. It is a decision about allocation.

Then the fuel itself. The glucose account — that self-control burns blood sugar, and sweet drinks restore it — was the model's most physical claim and the most fragile. The brain's metabolic rate barely moves with cognitive effort of this kind, the arithmetic never worked, and merely rinsing a sweet solution in the mouth and spitting it out produces the same restoration, which fuel delivery cannot explain.

And then the replications. A large pre-registered, multi-laboratory replication reported in 2016 found an effect indistinguishable from zero; a second, differently designed multi-lab registered attempt a few years later found effects that were at best negligible. This is the crucial point for anyone quoting the literature: the flagship result of ego depletion did not survive pre-registration.

So what does survive? Notice that we have not denied the evening behaviour — we have only denied the fuel. An alternative reads effort as a cost accounting problem. Mental effort is aversive and is felt as such; the longer you spend on effortful choosing, the more attractive the alternatives to it become, and attention drifts from what you must do toward what you would rather do. Nothing is consumed. The exchange rate changes. That account explains everything the fuel account explained and also explains the incentive result, which the fuel account cannot.

What about the famous field evidence — the judges whose favourable parole rulings declined steeply across each session and jumped after a food break, from roughly two-thirds of cases early on to near zero just before a break? Those figures are recalled and should be treated as approximate, but the more important caveat is structural: the case ordering was not random. Prisoners without legal representation were systematically heard later, and a denial takes less time to write than a grant, so a judge managing a session rationally produces the same curve without any fatigue at all. The effect size was also implausibly large for a psychological process. It is the single most-cited demonstration of decision fatigue and it is not clean evidence of it.

The correction here is narrower than "decision fatigue is a myth", and the narrowness matters.

What is unsupported is the strong resource claim: a general, finite, glucose-backed pool of willpower that any act of control drains for all the others. What remains well supported is smaller and less dramatic. Effortful choosing is subjectively costly and people avoid it. Over a long run of decisions, the balance tips toward whichever option requires least justification — the default, the status quo, the recommendation already on the table. And the decision environment determines the size of that shift far more than any inner reserve does: whether a default exists, whether the choice must be defended, whether anything else is competing for the same attention.

Two honesty notes. Field studies that show later-in-session decisions getting worse — prescribing, screening, ordering vaccinations — all share the parole study's confound, because what is scheduled late is rarely a random sample of what is scheduled early; several are suggestive, none is decisive. And "decision fatigue" as commonly used slides between three different things: being tired, having lost interest, and having genuinely less capacity. Only the third would be depletion, and it is the one with the weakest support.

• Why subjective mental effort is aversive at all, and what the sensation is tracking.
• How pre-registration changed the published effect sizes across social psychology more broadly.

Every term the collection defines is gathered in the glossary.