The Frequency Illusion

What it is

The term comes from linguistics. Zwicky wrote it into a post called Just Between Dr. Language and I, in a passage about errors that people who think carefully about language make about language itself. He gave the frequency illusion as a companion to the recency illusion, the belief that a word or construction is new when it has long been in use. His account of the frequency illusion was short: once people notice a phenomenon, they register nearly every instance that crosses their path, and their estimates of how often it occurs rise as a result.

His example was the quotative use of the word all, the way some speakers use it to introduce reported speech. It was said to be constant in the speech of some of his students' friends. In recorded and transcribed conversation it turned up rarely. The gap between the felt rate and the recorded rate is the illusion in its simplest form, and it is also the reason the term has the word illusion in it: the error lies in the estimate and not in the thing being counted.

The Oxford English Dictionary entered frequency illusion in its March 2019 update, with a 2005 citation taken from Zwicky's post, and defines it as a quirk of perception in which something you have just started to notice seems to be everywhere. The term has since travelled well beyond language to words, names, objects and numbers of every kind. Popular writing also knows the same experience under another name, and the two refer to one and the same thing.

Three features define the experience. Something is noticed for the first time. It is then registered again, repeatedly and soon afterwards. And the person concludes that it has become more common. The first two are events in the person's attention. Only the third is a judgement about the world, and it is the judgement that the word illusion marks as mistaken.

In effect

Zwicky tied the frequency illusion to selective attention, the ability to favour some inputs and leave others unprocessed. The classic listening experiments are the clearest demonstration that unattended material goes largely unprocessed while material of personal importance can break through. Colin Cherry's 1953 paper set what became known as the cocktail party problem, and Neville Moray reported in 1959 that a listener's own name could break through from a message the listener had been told to ignore. A newly learned word is a plausible case of the same kind, because it becomes recognisable and so is registered each time it appears. That step, from names in a listening task to new words in everyday reading, is an inference from the attention literature and not a measured result.

A second process is expectation. Raymond Nickerson's 1998 review of confirmation bias, published in Review of General Psychology, collects experiments in which people noticed and remembered information that fitted what they already believed, often without intending to. In one study he reports, mock jurors recalled statements that fitted their chosen verdict as having been presented, including some that were never presented. A person who has begun to suspect that something is turning up everywhere is positioned like those observers: each sighting counts as support, and a day without a sighting leaves no record.

A third process is the judgement of frequency itself. Amos Tversky and Daniel Kahneman's 1973 paper on the availability heuristic reports that people judge how often something occurs by how easily instances come to mind. In their letter-position study, 152 participants judged whether each of five consonants, K, L, N, R and V, appears more often as the first or the third letter of English words. Most chose the first position, with a median ratio of two to one, although all five letters occur more often in the third position. Kahneman's Thinking, Fast and Slow adds the idea he calls WYSIATI, short for what you see is all there is: the fast system builds the best available story from whatever information is active and does not register what is missing.

Taken together, the three processes describe how a vivid recent sample can stand in for a long, unrecorded past. A record kept independently of the person's own noticing, such as a dated archive or a count begun on the day of the first sighting, separates a real rise from the illusion.

What it does not say

The frequency illusion does not say that the noticed thing is rare, or that it has not become more common. Some words and objects do enter wider use, and a record will show it when they do. The term names a gap between a felt rate and an actual rate, and the gap has to be measured before anyone can say which way it runs.

It is also a description of an experience and not a result from experiments on the illusion itself. Zwicky's account is an observation about attention, and the links to confirmation bias and to ease of recall are inferences from separate research literatures. The studies cited here measure selective attention, the noticing and remembering of belief-consistent information, and judgements of frequency from ease of recall. They do not measure how often people of any kind experience the frequency illusion, and they give no effect size for it.

Finally, it makes no claim about carelessness or fault. Selective attention is what lets a listener follow one voice among many, and judging by ease of recall is a quick way of estimating frequency that is often useful. The same processes that produce the illusion do ordinary work the rest of the time.

---