Misquotes and fact checks: the questions that matter

Quotations rarely get faked. They get worn smooth. A line passes through enough mouths and retypings to lose a clause here and gain a rhythm there. It arrives at the shape easiest to remember, rarely the shape anyone produced.

The name attached goes through the same drift, toward whichever person the audience already thinks of as wise, funny, or important.

This page is a field guide to misquotes and fact checks: the specific mechanisms that damage quotations, the symptoms each one leaves behind, and what a check on a suspect line actually consists of.

What to take away

The questions that matter

A check is only as good as the questions it asks, and these are the ones that decide an answer.

How a line goes wrong

Compression. The most common change by a wide margin. The original had a qualifying clause, a subject the sentence needed, or a second half. The circulating version does not. Compression almost always makes a line stronger and less true, because the clause that gets cut is usually the one limiting the claim.

Improvement. Repetition sands a sentence into better prose than the author wrote. Parallel structures tighten, an awkward word is replaced by a better one, a rhythm appears. Real speech is clumsy; real prose is uneven. Polish is a sign of transmission, not of genius.

Modernization. Archaic vocabulary quietly updates itself. Words that would puzzle a modern reader are swapped for words that do not, and after a few generations the quotation contains no word the credited person could plausibly have used.

Paraphrase promotion. A biographer writes "he argued, in effect, that..." as a summary. The next writer, reading quickly, puts it in quotation marks. The one after that adds a name and a date. Nothing dishonest happened at any single step. This is probably the single largest source of quotations that "sound like" someone.

Composition. Two sentences from different parts of a work, or from different works, get spliced into one memorable line. The join is usually invisible, because whoever spliced them was trying to make it invisible.

Attribution migration. Collecting has always concentrated credit in a few names, as the history of quotation collecting shows. A line credited to a moderately known person gets recredited to a very well known one in the same category: a wit to the most famous wit, a general's maxim to the most famous general, a physicist's aside to the most famous physicist. The flow runs almost entirely one way, from lesser known to better known, because the better-known name makes the line travel further. This is why the ratio of quotations credited to a handful of famous people to quotations they can be shown to have produced is so lopsided.

Character-to-author transfer. This is a separate failure from a wording error, and the words can survive it intact. A line spoken by a character in a novel, play, or film gets attributed to the person who wrote it, stripped of the fact that it is dialogue and often of the fact that the character was wrong.

Satire escaping its frame. Our material on comic lines read straight covers the version of this that does the most damage. A line written to mock a position gets quoted as though it stated one. Once the framing is gone, nothing in the sentence itself signals irony, and nothing ever puts the frame back.

Epigraph laundering. A book opens with an unsourced quotation on the dedication page. The book is reputable. The quotation now has a reputable-looking home, and later writers cite the book. The chain acquires an appearance of authority without acquiring any evidence.

Translation drift. A translated line is the translator's sentence. When several translations exist, the most quotable one wins, and the credit still goes to the original author. Sometimes the memorable phrasing has no counterpart in the original at all.

Mishearing and transcription. Spoken lines are captured by ear, and near-homophones, dropped negatives, and misplaced emphasis all survive into print. A dropped "not" inverts a sentence completely and leaves no trace of the accident.

Manufacture. Some quotations are simply made up: to sell a product, to win an argument, to fill a column, or as a joke that outlived its own punchline. Fabrications are rarer than the other mechanisms but they are the ones with a motive, and motive makes them cluster around contested subjects.

Named cases. The mechanisms are easier to see in lines that are already famous. In each of these, the wording or the name has drifted, and the earliest trace points somewhere other than the famous name.

Symptoms you can check for

You will often suspect a quotation before you can disprove it. These are the tells worth acting on: none of them is proof, and several together are a strong reason to stop using the line until you have looked.

What a check actually involves

A quotation check is a search for the earliest appearance, not the most appearances. Popularity is the thing you are trying to see past.

The corpora are the ones with dates attached: Google Books and HathiTrust for scanned books, the Internet Archive for books, periodicals, and recordings, JSTOR for journals, and the Library of Congress's Chronicling America for digitized newspapers, with subscription archives such as Newspapers.com and Gale filling in the more recent decades.

Official words live in the proceedings themselves — the Congressional Record in the United States, Hansard in Britain — and private ones in the letters, diaries, and manuscripts held by the archive that owns them.

Reference works carry the same kind of evidence and state their limits. Bartlett's Familiar Quotations, The Yale Book of Quotations, and The Oxford Dictionary of Quotations give dates, editions, and pages, and the Oxford English Dictionary's dated citations often mark where a phrase's trail begins.

For lines that live online rather than in print, the fact-checking sites are the place to start. Snopes, PolitiFact, and FactCheck.org publish checks on quotations that circulate without a source, and Quote Investigator traces a line to its earliest appearance and reports what the search found.

A negative result has limits worth stating plainly. Not finding a line proves that it is not in the corpora you searched, in the wording you searched, within the accuracy of their text recognition. That is genuinely weaker than proof of invention, and it is still far more than the zero evidence behind the caption you started with.

Saying what you found

Reference works use a small vocabulary for this and it is worth borrowing, because each term makes a different promise.

"Attributed to" is the most abused of these. In a careful reference it is a warning that the trail ran out. In a caption it usually means nothing was attempted.

When the person is alive

Extra weight applies when the credited person lives: an invented quotation is a false statement of fact about what someone said, published under their name. If unflattering, it is more than a scholarship error.

Treat any quotation attributed to a living person that you cannot open a source for as unusable, not "probably fine" or "attributed to"; drop it.

After you have already published it

Common questions

Does it matter, if the sentiment is good?

The sentiment is not what is under quotation marks. The claim is about a person. Anyone who checks and finds it hollow discounts everything else on the page, and they are right to.

Are corrected versions ever adopted?

Rarely, and this is worth understanding rather than resenting. The circulating version usually beat the original because it is shorter and better balanced. You are competing with a form that has been optimized by millions of retellings. Expect to lose, and quote the real one anyway.

Can I trust a collection that shows its sources?

More than one does not, and that is the main thing. Check what its entries carry: an edition and a page, or only a name. A collection that marks the difference between a sourced entry and an attributed one shows its own limits. That is the most useful signal a reference can send.

What counts as a primary source is what those entries should contain: the edition, the page, and the archive or collection that holds the item.