A reason and a rationalisation are indistinguishable from the inside

Two different things wearing one word

A reason is a consideration that actually produced a decision. A rationalisation is a consideration that would have justified the decision, offered afterwards, whether or not it played any part. The important and uncomfortable point is that the second is not a lie. It is produced sincerely, it is often perfectly sensible, and the person producing it has no internal marker distinguishing it from the first kind.

Most discussions of this treat rationalisation as a character failing found in other people. It is better understood as the normal output of a system that was never built to report on its own operations. You have access to the conclusion and to whatever considerations are currently available to you. You do not have a log of the process.

What introspection actually delivers

The standard research line here is that self-reports about the causes of behaviour are constructed rather than retrieved. When people are asked why they chose something, they produce an answer built from plausible theories about what generally influences choices of that kind — the same theories they would use to explain a stranger. Sometimes those theories are right, which is precisely why the whole business is so hard to detect.

A memorable line of experiments demonstrates the gap by covertly swapping the option a person selected and then asking them to explain their selection. A substantial share of people accept the substituted item as their own choice and go on to give fluent reasons for preferring it. The finding has been reproduced across several materials and laboratories, though the interpretation is argued about: whether it shows that reasons are routinely confabulated, or that people are simply poor at noticing a swap under time pressure.

Why the account is usually plausible and sometimes correct

The explanation-generating machinery is good at its job. It knows what sort of things influence decisions, it knows what you value, and it produces an answer that fits the evidence available — which includes your own behaviour, now treated as a piece of data about yourself. That is why rationalisations rarely sound absurd. They sound like reasons because they are built to the same specification.

This also means the account is frequently right, and the sceptical position that all self-explanation is fiction goes considerably further than the evidence. Deliberate decisions made slowly, with the considerations written down before the choice, leave a record that later reporting can draw on. The unreliability concentrates in fast, preference-driven and socially loaded decisions, which unfortunately is most of them.

The test that is actually available

Since you cannot inspect the process, the workable check is whether the stated reason has any commitments attached to it. A genuine reason predicts other things: it applies to similar cases, it would have produced a different decision had the fact been different, and it survives having its weight questioned. A rationalisation tends to be strangely local — it explains this decision and nothing else.

So the question to ask yourself is a counterfactual one. If the consideration you are citing had pointed the other way, would you have decided differently? If the honest answer is no, then whatever produced the decision, it was not that. This does not tell you what the real cause was. It does tell you that you have not found it yet.

Where the distinction has consequences

It matters most when a reason is being used to settle a disagreement. Two people arguing about a decision are often comparing after-the-fact accounts rather than causes, which is why such arguments can be conducted with complete sincerity and no possibility of resolution. Neither party is lying and neither has access to the thing being disputed.

It matters again when you are learning from your own decisions. A record built from reconstructed reasons teaches you very little, because it records what seemed defensible afterwards rather than what actually moved you. The only fix is a record made in advance — what you expected, what you were weighing, what would change your mind — written before the outcome exists to reshape the memory.

What this does not license

It does not license dismissing anybody’s stated reasons as mere rationalisation, and doing so is a cheap rhetorical move that makes disagreement unresolvable in the same way it makes self-examination pointless. The claim here is epistemic, not moral: reports about causes are unreliable, in yourself first of all.

Nor does it imply that reasoning is decorative. Reasons given in advance, in public, with consequences attached, constrain behaviour in ways that after-the-fact accounts do not. The practical response is to shift more of your reasoning to before the decision, where it can still do work, rather than to give up on reasoning because the retrospective version is unreliable.