The idea that a good deed licenses a bad one has thinner support than its fame

The claim, stated fairly

Moral licensing is the proposal that behaving well in one instance makes a person more willing to behave badly in the next, because the earlier action has established a credential that the later one can be charged against. The typical experiment has participants do something creditable — recall a good deed, express an admirable preference, choose a virtuous option — and then measures behaviour on an unrelated task where a small transgression is available.

The idea has obvious appeal. It has a mechanism, it explains a pattern people believe they have observed in themselves, and it produces a satisfying reversal of the ordinary expectation that good behaviour begets more of the same. Those are the properties of a claim that travels quickly, and none of them is evidence.

What the accumulated evidence looks like

A meta-analysis pooling the published studies found an average effect that was small, and reported indications of publication bias in the underlying literature — the pattern in which the smaller studies show larger effects, which is what you would expect if null results had failed to reach print. Since then, several preregistered replications of specific licensing demonstrations have returned nothing, or effects too small to distinguish from noise.

This isn’t a debunking, and it should not be reported as one. It is the familiar picture of a claim whose supporting literature was produced under conditions known to inflate effects, and which has not yet been tested cleanly at sufficient scale to say whether something real is underneath. The honest description is unresolved, leaning weak.

The rival account nobody finds as interesting

A competing prediction is that behaviour is consistent rather than compensatory: having done something creditable, a person is more likely to do so again, because the action has supplied evidence about what kind of person they are. There is experimental work supporting this direction too, which is awkward, since the two accounts predict opposite outcomes from the same setup.

When a literature contains support for both directions, the most likely explanation is that the effect depends on conditions that nobody has yet pinned down, and the second most likely is that neither effect is reliably present. Proposals about which conditions matter exist — whether the first act was framed as progress towards a goal or as evidence of a trait, whether the acts are seen as belonging to the same domain — but these are hypotheses generated to reconcile a mixed literature rather than findings established in advance.

Why the idea persists anyway

It survives because it’s easy to confirm from memory. Everybody can recall an occasion when a virtuous act was followed by an indulgence, and the recollection feels like data. It is not: the occasions when good behaviour was followed by more of the same are unmemorable precisely because nothing surprising happened, so the sample available to introspection is filtered by the very pattern it’s being used to demonstrate.

It also has the shape of a clever counterintuitive insight, which is a genre that circulates on its own merits. Notice how many popular claims about behaviour take the form of a reversal — the thing you would expect turns out to be the opposite — and how much better that structure travels than a finding that confirms the obvious. Reversals are more interesting, and interestingness isn’t correlated with truth.

What to do with an idea in this state

The reasonable treatment is neither belief nor dismissal. Hold it as a possibility worth watching for in your own conduct, since checking costs nothing, and refuse to build anything on it — no policy, no incentive scheme, no argument that depends on it being true.

This is a general skill and moral licensing is a good case for practising it. A great many behavioural claims sit in exactly this state: originally demonstrated, widely repeated, meta-analysed to a small and possibly artefactual effect, and never conclusively resolved either way. Being able to hold such a claim without either endorsing or rejecting it is more useful than any particular verdict, because the state is common and permanent verdicts are rare.

The version that is on firmer ground

There is a narrower and better-supported observation nearby, which is that people are strongly motivated to maintain a self-image as decent and reasonable, and will reinterpret their own behaviour to preserve it. That much shows up across many lines of research, and it does not require any licensing mechanism.

From that starting point, the interesting question is not whether one good act permits one bad one, but how a sequence of small reinterpretations lets a person hold a stable self-image through behaviour that has drifted a long way. That process is slower, less quotable and much harder to demonstrate in a single session, which is presumably why the tidier claim became the famous one.