Illusory correlation

The run of good results started around the time she changed her pre-game routine. She noticed. A few more good sessions followed, and the connection became harder to ignore.

The routine felt meaningful. She began protecting it. Missing one element felt risky - not because she believed it rationally, but because the coincidence had repeated often enough to feel like evidence.

But the results were there before the routine too. And the routine failed to protect against bad results plenty of times. She was tracking the hits and not the misses. The pattern that felt real had been constructed by selective attention, not by anything that was actually happening between the routine and the performance.

This is illusory correlation.

What it is

Illusory correlation is the perception of a relationship between two variables when no such relationship exists, or when it is substantially weaker than it appears. People notice and remember co-occurrences that feel salient or meaningful - and infer a link that may not be there.

The pattern does not have to be invented entirely. Sometimes there is a real but weak relationship, and the perceiver inflates it into something stronger. Sometimes there is no relationship at all, and the mind builds one from coincidence. In either case, the perceived connection exceeds what the evidence actually supports.

The mechanism

Two processes drive illusory correlation.

Salience and the availability heuristic. Unusual or emotionally meaningful pairings catch attention more readily than ordinary co-occurrences. A distinctive event happening alongside another distinctive event - both memorable, both available to recall - gets registered as a pairing. Non-occurrences, and pairings between ordinary events, get much less cognitive weight. The result is a sample that the mind assembles without being representative.

Confirmation bias. Once a perceived correlation forms, attention shifts toward confirming examples and away from disconfirming ones. The lucky shirt works -> registered. The lucky shirt does not work -> explained away, forgotten, or attributed to other factors. The perceived pattern accumulates evidence while its failures remain invisible.

The mystery and the reveal

Here is the mystery: why does a perceived correlation feel so stable even when it is pointed out to be false?

Here is the reveal: because the cognitive processes that build the correlation are the same processes that maintain it. The evidence that would break it - the misses, the non-occurrences, the baseline rate - is exactly the kind of evidence the mind is least likely to retrieve. What is remembered is what was salient. What was salient was what confirmed the pattern. The loop is self-sustaining.

Where it shows up

Superstitions. Lucky objects, rituals, and routines accumulate their perceived power through illusory correlation. The shirt worn before a win; the route taken before a good day; the specific conditions that preceded a successful outcome. The exceptions are plentiful but forgettable.

Stereotypes. Early experiments showed that participants overestimated the co-occurrence of rare behaviours with atypical group members. This is one of the documented ways illusory correlation contributes to the formation and maintenance of social stereotypes: minority groups and unusual behaviours are both distinctive, and their conjunction - even when statistically no more common than it is for the majority group - gets disproportionate cognitive weight.

Everyday judgments. Coincidental sequences generate causal stories: the presentation went well the day I wore blue, so I wore blue again. The client meeting was successful after a coffee from that cafe, so I returned before the next one. None of these are implausible causes - they are just unexamined ones, treated as signal when they may be noise.

What it is not

It is not purely a logical mistake. Illusory correlation is a perceptual and memory phenomenon, not just a reasoning error. The distorted sample is assembled before deliberate reasoning begins - through attention and recall that are not evenly distributed across occurrences and non-occurrences. Correcting the inference requires correcting the sample, not just the logic applied to it.

It does not only happen with rare events. While distinctive or rare events amplify the effect, illusory correlations form with everyday events too whenever attentional or memorial biases skew the sample that is being perceived.

It is not the same as confirmation bias. Confirmation bias is the active seeking and weighting of confirming evidence. Illusory correlation is the upstream process: the construction of a perceived pattern that confirmation bias then maintains. The two work in sequence, not in parallel.