Introspection Illusion

What it is

The 1977 review in Psychological Review is one of the most cited papers in psychology and the version that circulates is cruder than what it says.

Its finding has three parts. People are sometimes unaware of a stimulus that importantly influenced a response. Sometimes unaware of the response. And sometimes unaware that the stimulus affected the response at all.

The explanation is the useful part. When people report on the processes mediating the effect of a stimulus on a response, they are not introspecting. They draw on a priori implicit causal theories: judgements about whether a particular thing is a plausible cause of a particular behaviour.

So somebody asked why they chose a product does not retrieve the cause. They generate the most plausible candidate from a theory they already hold about what moves people.

In effect

The boundary condition is what makes this usable, and it is the half most summaries drop. Nisbett and Wilson state that accurate reports will occur where the influential stimulus is salient and is a plausible cause of the response, and will not occur where the stimulus is neither.

That converts a famous pessimism into a prediction. Asked why they bought a coat, somebody will often answer serviceably, because price, fit and colour are salient and plausible. Asked why they chose the second-cheapest wine, they will not, because position on a list is neither.

The consequence for research is not that asking people things is useless. It is that a direct question about a salient attribute is a reasonable instrument and a direct question about why one option beat another is not, where the deciding factor was layout, order, a default or a price ending. The respondent is not lying; they are doing what the paper says they do.

What it does not say

It does not say people have no self-knowledge, which the paper explicitly denies.

It is a review rather than an experiment, published in 1977, summarising a literature at a moment. Work on introspection, metacognition and confabulation has continued and this record has not surveyed it, so what is reported here is the paper's claim and its own stated boundary rather than the current consensus.

And there is no internal signal that distinguishes a retrieved reason from a constructed one, which is why the practical move is to ask what else was present at the time rather than to trust or distrust the account.

---