> ## Documentation Index
> Fetch the complete documentation index at: https://meta.niceshare.site/llms.txt
> Use this file to discover all available pages before exploring further.
# Frequency Illusion
> Frequency Illusion is the belief that a noticed thing happens a whole lot. Zwicky named it in 2005. Dutch check: 17 of 18 often-claims under 73 percent.
**Category**: Effects
**Type**: The belief that a phenomenon happens a whole lot once you have noticed it
**Origin**: Arnold M. Zwicky, Stanford University, "Why are we so illuded?" September 2006. The name's first dictionary attestation, as Ben Zimmer reports it, is Zwicky's Language Log post of 7 August 2005
**Also known as**: Baader–Meinhof phenomenon, a reader's name from 16 October 1994 for a second printed sighting within a day
**Quick Answer** — Frequency Illusion is the belief that a phenomenon happens a whole lot once you have noticed it. Arnold Zwicky set that sentence down for language, and he named two processes behind it: selective attention, and confirmation bias. A 1994 newspaper letter had already named a sharper cousin, a second printed sighting within 24 hours. A 2022 count of Dutch usage books found that most checked claims of "often" sat under the bar the author used for that word.
## What is Frequency Illusion?
Frequency Illusion is the belief that, once you notice a phenomenon, it happens a whole lot.
> Once you notice a phenomenon, you believe it happens a whole lot.
That sentence is Arnold M. Zwicky's, in a September 2006 abstract. The noticing comes first. The belief about the rate comes second. A second meeting with a word is not, by itself, a measurement that the word has become common. [Confirmation Bias](/effects/confirmation-bias) is his name for keeping the sightings that fit the belief and setting the misses aside. [Availability Heuristic](/effects/availability-heuristic) is a different operation: judging a rate by how easily an example comes to mind.
### Frequency Illusion in 3 Depths
* **Beginner**: A new word shows up in the paper the next day, and the day feels full of it. The 1994 letter stated that as a rule about print within 24 hours. The letter did not count how often the word had been there all along.
* **Practitioner**: Before you write "often," name the other form and the years you searched. In the Dutch check, "often" was treated as at least 73 percent of the two forms together. Seventeen of 18 checked word claims sat under that bar.
* **Advanced**: Two grammar claims sat above the same bar, so a high-frequency word in a usage book is not always an illusion. A share of 65.7 percent was still called one, because 65.7 is under 73. The bar is a survey mean for one Dutch word, not a law of noticing.
## Origin
The wording on this page is from Arnold M. Zwicky, "Why are we so illuded?", Stanford University, September 2006. The abstract is about 490 words. It lists five illusions about language. Frequency Illusion is the fifth: once you notice a phenomenon, you believe it happens a whole lot. His example is a widespread belief that quotative "all" occurs "all the time" in the speech of some young people. The abstract does not print a count of that "all." This page does not add one.
He writes that Frequency Illusion is a result of two processes. Selective attention is noticing things that are salient to us and disregarding the rest. Confirmation bias, in his parenthesis, is looking for things that support our hypotheses and disregarding potential counterevidence. He also writes that hardly anyone has a panoptic view of language variation, so people think about it from personal experience. The same abstract's recency illusion is a different belief: if you have noticed something only recently, you believe that it originated recently. This page does not use that name for the frequency belief.
Ben Zimmer's Language Log post of 19 March 2019 says the Oxford English Dictionary's March 2019 update added frequency illusion, first attested in Zwicky's Language Log post of 7 August 2005, "Just Between Dr. Language and I." Zimmer quotes the dictionary gloss: a quirk of perception whereby a phenomenon to which one is newly alert suddenly seems ubiquitous. This page did not reopen the 2005 archive, and it did not open the dictionary file. The sentences used above are the ones printed in the 2006 abstract. Marten van der Meulen, in the 2022 paper below, attributes the same two illusion sentences to the 2005 post.
The older public name is not Zwicky's. The St. Paul Pioneer Press Sunday Bulletin Board, in the web reprint dated 16 October 1994 at 5:05 a.m. CDT, prints a reader letter signed Gigetto on Lincoln. The page also prints an updated time of 17 April 2021. That later time is not a second letter. Zimmer reports that Dan Kelly, the former feature editor who published the submission, identifies the pseudonym as the reader Terry Mullen. This page did not re-interview Kelly.
The letter dubs it the Baader–Meinhof phenomenon, after the West German gang of terrorists, and states the phenomenon this way: the first time you learn a new word, phrase, or idea, you will see that word, phrase, or idea again in print within 24 hours. It sets topical things aside. The naming incident is a conversation with a friend about the gang, many years after the gang was in the news, and a phone call the next day in which the friend pointed to that day's newspaper. The letter does not say the reader stumbled on the second mention alone. It does not report how often the name had appeared in that paper over the previous year. The 24-hour rule is the reader's rule. It is not the 2006 definition, and it is not a probability from a sample.
## Key Points
Four cuts keep a second sighting from becoming a rate. The belief, the two processes, the denominator, and the 24-hour rule each have their own sentence.
Frequency Illusion, in the 2006 abstract, is a belief about how much a phenomenon happens. The 1994 letter is a claim that a newly learned obscure item returns in print within a day. A person can have the second meeting and still refuse the belief. A person can hold the belief with no clock set at 24 hours. [Clustering Illusion](/effects/clustering-illusion) is a clump in a small chance sample read as a cause. Two mentions in two days are not yet that clump, and they are not yet a rate.
Selective attention, as he glosses it, is noticing what is salient and disregarding the rest. Confirmation bias, as he glosses it, is looking for support and disregarding potential counterevidence. A new car in your own driveway makes that model salient. Later sightings are easier to keep. The misses, the blocks with none of those cars, do not argue back unless someone writes them down. He does not print a count of cars. This page does not invent one.
Van der Meulen treated a high-frequency word as Frequency Illusion when the targeted form was under 73 percent of the two forms together. The 73 percent is his report of the mean numerical reading of Dutch vaak, "often," in Willems, Albers, and Smeets (2020, p. 9). He says the answers varied considerably. That survey was not re-read here. He used only that mean, and he treated stronger words as at least that high. He calls the choice lenient toward the usage books. A statement of "sometimes" was left out of the test, because his reading of Zwicky is a belief that the thing happens often.
Nothing in the Dutch count tests whether a new word returns in print within 24 hours. Nothing in the 1994 letter tests whether the gang's name was under or over 73 percent of some pair. [Priming Effect](/effects/priming-effect) on this site is a faster correct decision that two strings are words when the words are common associates. A quicker word decision is not a belief about a rate in the world.
## Applications
Use the belief when someone treats a fresh sighting as a new abundance. Keep the Dutch percents inside the usage books that produced them.
A usage note that says a disliked spelling occurs often has not yet named the approved spelling or the years. In the Dutch lexical check, the comparison was the condemned form divided by the condemned form plus the proposed form, inside a window of about 30 years before the note. Copy that shape before you repeat the adverb.
The 1994 incident is a friend calling the next day about one newspaper article. A product team that reads one customer message, then a second, has the shape of that letter. The letter does not say the gang became common overnight. Ask for the count of messages that did not mention the thing.
Seventeen of 18 checked Dutch word claims fell under 73 percent. One claim, a greater-or-lesser-degree wording, sat at 76.7 percent and was not called an illusion. Two grammar claims, at 74.1 and 79.7 percent, were not called illusions either. A slide that says the books were always wrong has dropped those three.
The paper's own discussion says the statements can be genuine unsupported beliefs and can also be habits of the usage-book genre. One 1964 guide supplies 102 entries with frequency wording, about half of that decade's frequency mentions, in line with that guide's share of the decade's entries. A repeated warning is not a fresh morning of noticing.
## Case Study
Marten van der Meulen, at the Centre for Language Studies, Radboud University in Nijmegen, asked whether Dutch usage books were in the grip of Zwicky's two illusions. The paper is "Are We Indeed So Illuded? Recency and Frequency Illusions in Dutch Prescriptivism," *Languages* 7 (2022), article 42. The PDF prints the DOI 10.3390/languages7010042. It was received on 24 November 2021, accepted on 14 February 2022, and published on 22 February 2022. There was no external funding. An ethics-board statement was not applicable.
He searched the Normative Database of Dutch. Section 2.1 says the database holds 5,678 entries from prescriptive publications aimed at mother-tongue speakers of Dutch in the Netherlands, for 1910–2018. The abstract dates the same project 1900–2018. The decade table begins in the 1910s. This page does not collapse those two start years.
Frequency wording was common. 1,249 entries contain at least one mention of frequency, and those entries hold 1,786 frequency statements. Grammar accounts for 1,064 of the statements, which he prints as 59.5 percent. Lexis accounts for 36.8 percent. Another 67 statements sit at other levels, such as spelling or pronunciation. High-degree terms are printed as 59 percent, 1,054 tokens. Middle-degree terms are 331, 18.5 percent. Low-degree terms are 402, 22.5 percent. The leading word was vaak, "often," with 317 mentions. The illusion test used high-degree wording only.
For words, he drew two checkable items in each decade he could, skipping the 1920s and the 1950s, for 18 items. The share is the condemned form divided by both forms, in a window of about 30 years before the usage note. The first corpus was C-CLAMP, about 200 million words of magazines and newspapers from 1837 to 1999. If the two forms together produced fewer than 100 hits, he used Delpher. He warns that Delpher's character recognition is weak in some periods and that its representativeness is open to question. He still uses the relative share.
Seventeen of the 18 word claims were under 73 percent, so he calls them Frequency Illusions. In seven of those cases the targeted form was under 5 percent of the pair. The worked example is duistering, "darkness," which Moortgat in 1925 called the usual form against duisternis. In C-CLAMP for 1896–1925 he finds 8 of the condemned form and 909 of the approved form. That is 0.9 percent of 917. A different item is compounds ending in -toename, "increase," the form Grauls attacked in 1957 in favor of -toeneming. In the 1928–1957 slice of the same corpus, that attacked form is 50 hits and 47.2 percent, and he still calls that an illusion. A later, separate count found 64 hits for -toeneming, the ending Grauls had preferred, and 8,767 for -toename. Those two windows are not one result. The near miss is ons inziens, "in our opinion," at 7,791 hits and 65.7 percent in Delpher for 1946–1975. It was the majority of that pair, and it was still under 73, so it still counts as an illusion in his rule. The only lexical exception is meer of mindere mate, "a greater or lesser degree," at 1,457 hits and 76.7 percent in Delpher for 1986–2005. That one is not an illusion.
For grammar he did not draw the same random pairs. He used counts already cleaned for nine morphosyntactic variables. The methods date that usage material 1995–2004 and the frequency notes roughly 1990–2015. The appendix says the reported grammar counts combine the Corpus of Spoken Dutch with the 1995–1999 section of C-CLAMP. The unpublished source of those counts was not re-read. Seven of the nine variables had high-frequency notes in that window. Two of them sat above 73 percent, so their three notes are not illusions: u heeft, a "you have" form, at 827 hits and 74.1 percent in a 2009 note, and hele mooie auto, "very nice car," at 2,482 hits and 79.7 percent in notes from 1993 and 2011. The other 21 notes, on five variants, sat under the bar. One of those is a neuter noun followed by wat rather than dat, at 339 hits and 7.4 percent across the combined spoken and written material.
A stronger adverb failed on its own terms. Houët wrote in 2000 that in speech this wat was used almost exclusively. The spoken slice was 35.8 percent. That is higher than 7.4 percent, and it is not exclusive. It was not even the majority form. Heidbuchel wrote in 1962 that omdat, "because," was used almost exclusively against doordat. In C-CLAMP for 1932–1962, doordat occurs 832 times, 8.8 percent of the variants. Weverink wrote in 2012 that almost everybody used ze rather than hen or hun for a spoken object. The spoken corpus held over 2,300 of hen and hun in those roles. The paper does not print the size of the ze side of that last comparison. "Over 2,300" stays a floor, not a percent.
The lesson is a mismatch inside usage books, under one stated bar, for high words such as "often." It is not a count of how often a new car appears on a commute. Seventeen of 18 is the lexical sample he checked. It is not 17 of every 18 speakers.
## Boundaries and Failure Modes
Frequency Illusion is the wrong summary when a second sighting is treated as a measured surge, when one Dutch bar is treated as the meaning of "often" in every language, or when the books' misses are treated as a rate for shoppers and roads.
The first boundary is the sample. The 1,786 statements are lines in a genre that argues about disputed forms. The abstract says the illusions seem due partly to unsupported beliefs and partly to genre habit and rhetorical choice. The discussion says the result should not be read as proof that every prescriptivist, or every language user, is always wrong about recency or frequency. The same paper's recency check is a separate test: a form first attested more than 30 years before the note. Of 23 recency statements he checked, 20, or 87 percent, were illusions on that rule, and 3 were not. Those years are not the frequency shares. A nineteenth-century contrast he cites, about Matthijs Siegenbeek, went the other way: that author's frequency terms were broadly in line with usage. That study was not re-read here. No cell from it is on this page.
The second boundary is the 73 percent. It is one survey mean for Dutch vaak, reported second-hand, with variability he does not print as a standard deviation. He writes that similar frequency words need not mean the same thing in another language, citing Willems and colleagues at page 11. A team that treats 73 percent as the English word "often," or as the point where a noticed car has truly become common, has left his method. Low and middle words were not eligible for the illusion label. A majority share could still fail, as 65.7 percent did. A share just above the bar could pass, as 74.1, 76.7, and 79.7 did. Passing the bar does not make the usage advice wise. It only means the high word was not, on this rule, an overstatement of the share.
The third boundary is the export, and it is the common misuse. The 0.9 percent, the 47.2 percent, the 35.8 percent, and the 17 of 18 belong to Dutch forms in the corpora named above. They are not a street, a clinic, a feed, or a household. The 1994 letter predicts a second printed meeting within a day for an obscure new item. It was not tested against a log of first meetings. Quotative "all" remains an example with no rate on this page. [Illusory Truth Effect](/effects/illusory-truth-effect) is a higher truth rating after a plausible statement is heard again. A truth mark is not a belief that the statement's topic has become more frequent.
## Common Misconceptions
Three retellings change the letter, the abstract, or the Dutch table. One puts the research on the gang. One turns a next-day phone call into a proof of abundance. One treats every "often" in the books as false.
The 1994 letter names the phenomenon after a West German group because that was the obscure name in the story. The letter is a reader's rule about print, plus one phone call the next day. Zwicky's 2006 abstract names the frequency belief and the two processes. It is not an experiment with a sample size. The counted comparison on this page is van der Meulen's 2022 check of usage books. None of those three sources is a study of the gang's own perception.
The letter's rule is "again in print within 24 hours," for an obscure new item, topical news set aside. The abstract's belief is "a whole lot," illustrated by "all the time" as what some people say about quotative "all." The Dutch bar for a high word was 73 percent of a named pair. A friend pointing to one article meets the letter's story. It does not meet the bar, because the bar was never computed for that name.
False as a summary of his tables. Seventeen of 18 lexical claims were under 73 percent. One lexical claim was at 76.7 percent. Three grammar notes, on two variants, were at 74.1 and 79.7 percent and were not illusions. Twenty-one other grammar notes were under the bar. "Sometimes" was not in the test. The author also leaves room for rhetorical habit, so a failed line is not automatically a fresh error of noticing.
## Related Concepts
The frequency belief sits next to a preference for fitting evidence, a judgment from easy examples, a clump read as a cause, a faster word decision, and a truth rating after repetition.
Zwicky's second process: looking for what supports a hypothesis and disregarding potential counterevidence. The site's page is the broader family of favoring a belief one already holds. It does not inherit the Dutch shares.
A rate judged by how easily an example comes to mind. Frequency Illusion, in the 2006 wording, starts when a phenomenon has become salient and then seems to happen a whole lot. Ease of recall is a different lever.
A streak or a pile in a small chance sample, read as proof the process is not chance. Two printed mentions after a first learning are the 1994 story. They are not a grid of counts.
A faster correct decision that two letter strings are words, when the words are common associates. The 1971 gap is in milliseconds. It is not a belief about how often a word occurs in a newspaper.
A higher truth rating after a plausible statement is heard again. The rating can move while the topic's frequency in the world is untouched. Repetition here is not the Dutch share of a condemned spelling.
## One-Line Takeaway
When a newly noticed thing seems to be everywhere, write the other form, the years, and the misses before you trust the word often.
This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.