Decision fatigue is a plausible idea resting on contested evidence

The claim and its appeal

The proposal is that making decisions consumes something, so that a person who has decided many things by mid-afternoon decides the next thing worse — with more defaulting to the easy option, more avoidance, and more reliance on whatever shortcut is nearest. It sits underneath a large amount of practical advice about scheduling important decisions early and reducing trivial ones.

The idea feels true, and that’s worth treating with suspicion rather than as support. Everybody has had an afternoon where deciding anything felt heavy. But an experience of effort isn’t evidence that a resource was consumed, and the same feeling is compatible with boredom, low blood sugar, having been in a room too long, or simply having reached the part of the day when the remaining decisions are the ones nobody wanted to make.

The theory it was built on didn’t survive

Decision fatigue arrived as an application of ego depletion, the model in which self-control draws on a limited resource depleted by exertion. That model was enormously influential, generated a very large supporting literature, and then failed to replicate when large coordinated multi-laboratory projects tested it with preregistered protocols and much bigger samples.

The failure of the parent theory does not automatically falsify the offspring, since a phenomenon can be real while its proposed mechanism is wrong. But it removes the theoretical support that made decision fatigue credible in the first place, and it means the claim now has to stand on its own direct evidence rather than on the framework it inherited.

The famous study and the boring alternative

The best-known evidence is an analysis of judicial decisions finding that favourable rulings were much more common at the start of a session than towards its end, with the pattern resetting after breaks. It is a striking result and it’s quoted constantly.

It also attracted a serious methodological objection: the order in which cases were heard was not random. Critics pointed out that scheduling practices could group certain kinds of case, and that cases where the applicant was unrepresented tended to fall later in a session, which would produce the same pattern for reasons having nothing to do with fatigue. There is a further difficulty that gets less attention — the reported swing was extremely large for a psychological effect, and an implausibly large effect is usually a sign that something other than the proposed mechanism is contributing.

What is left standing

Some closely related things are on firmer ground. Sustained attention on a demanding task does decline over time, which is well documented in vigilance research and has been for decades. Performance changes across the day in ways connected to sleep and circadian timing. And people do shift towards simpler strategies under load, which appears in work on how time pressure and cognitive demand change what information gets used.

None of that requires a depleting store of decision capacity. The observations are compatible with fatigue in the ordinary sense, with motivation declining as a session runs long, and with the entirely mundane fact that attention drifts. Those explanations are duller and make weaker promises, which is presumably why the resource metaphor travelled further.

Acting sensibly under an unresolved claim

The practical advice attached to decision fatigue happens to be defensible on other grounds, which is a fortunate accident. Scheduling consequential decisions when you’re rested is sensible whether or not deciding depletes anything. Reducing the number of trivial choices frees attention regardless of mechanism. Taking a break before a difficult judgement is cheap and has other justifications.

What should not be done is building anything that depends on the specific claim — a policy justified by the number of decisions rather than by hours worked, or an argument that a judgement should be discounted because of where it fell in a sequence. Advice that survives the failure of its stated mechanism was probably resting on something else all along, and it is worth knowing which.

The general shape of this problem

A memorable mechanism, a striking field demonstration, a large derived literature, and then a slow discovery that the mechanism does not replicate and the demonstration has a confound. This sequence has repeated often enough in behavioural research that recognising it is more valuable than any individual verdict.

The tell is usually available early. An effect that explains a great deal, applies everywhere, and is illustrated by one unusually clean real-world example is exactly the profile of a claim that will need revisiting. That is not cynicism about the field; it is the pattern the field itself has documented about its own past decade.