Misquotes and fact checks source check: the report note

Almost every quotation check ends with nothing found. That is a normal outcome for a misquotes and fact checks source check, not a failed attempt, and it can be published. Its value depends on saying which strings, archives, years, and reading quality the search actually covered.

When a match appears on the page

A positive result is close to conclusive: the passage appears on the page in an identified edition. Contrary argument cannot undo that observation. It establishes wording in that edition, not every version. Our misquotes and fact checks approach describes confidence as search coverage, not as a verdict.

What a null result can establish

A null result supports only the claim that a string does not appear in the material searched, as indexed. It cannot show that the sentence was never written: no corpus is complete, and no index is perfect. The logic is not symmetrical with a positive match.

That limitation does not make a negative worthless. A carefully bounded search is stronger than a caption without evidence because it records the actual reach of the inquiry. A reader can judge the gap instead of mistaking the result for a claim about all writing.

The danger of leaving the search unnamed

A vague negative is useless and slightly dishonest because readers cannot tell what was tested. A bounded one is a real contribution, and it may save the next researcher the same afternoon. State scope before the conclusion, rather than treating a search as proof about the world.

Set out the search envelope

A usable report specifies three boundaries: the exact strings, named corpora with years covered, and reading quality. Without every boundary, a reader cannot identify the material the search left untouched.

The strings field names each exact wording tested. A quotation circulates in variants, so testing one wording means testing only that wording. Search at least the current popular version, the longest version you have seen, and a short, distinctive fragment from its middle.

The most informative line names each corpus and its date range. The Library of Congress describes Chronicling America as historic newspaper pages through 1963. It contains selected U.S. newspapers published from 1777 through 1963, not every paper or year. Readers can see important gaps immediately.

For quotation reference points, consult The Yale Book of Quotations, The Oxford Dictionary of Quotations, and Quote Investigator, then verify candidate wording against the cited source. An overview of why a search returns nothing explains what different kinds of corpus cover.

The reading quality matters because searchable text may not match the page image. Newspaper text on Chronicling America is produced by Optical Character Recognition and left uncorrected, so exact-phrase searching can miss material that is physically present. A null result on a scanned corpus is weaker than one on a born-digital corpus, and the note should say which was searched.

One bounded case from Casablanca

The familiar wording "Play it again, Sam" is attributed to Casablanca, but it is not spoken in the scene at issue. Test it alongside "Play it, Sam" and the film's line, "Play it, Sam. Play 'As Time Goes By.'" The check concerns that scene only.

A useful report also names evidence that could change its result. If you cannot identify any possible discovery or source that would revise the conclusion, you have not considered where the search remains incomplete. Keep the limit visible rather than letting a scene-level check stand in for a claim about first use.

Retry predictable search failures

Before declaring a phrase absent, try ordinary explanations for a failed search. A wording may vary slightly, or an exact-match tool may miss a phrase that remains visible in the underlying material. The six retries shown below should come before the conclusion.

Python's standard-library difflib.SequenceMatcher compares strings approximately and can flag near-matches in OCR text available to you. It is a comparison tool, not an archive search engine. When only exact matching is possible, shortening the string is a cheap substitute.

The Quotes by Era guide describes how collection practices varied by period, helping explain thin decades. For the wider checking process, see how a quote gets checked. These references put gaps in context without turning absence into proof about authorship.

Publish a search claim that can be checked

Weak wording says no evidence has been found that a line was ever said. Better wording limits the claim: this wording does not appear in these named corpora before this date, and the earliest located appearance is here, crediting nobody. The second statement reports a search, not the world.

That formulation can survive a later discovery of an earlier appearance because it describes the search that was made. Its boundary remains true even if the record changes, while a claim about what was ever said would not. Name the evidence and the search limit so readers can assess the finding.

For factual claims rather than quotation attribution, consult specific published checks from organizations such as FactCheck.org, PolitiFact, and Snopes. Record the exact claim and the evidence cited; an organization's name alone is not evidence for a quotation's wording.

If an unsourced quotation is attributed to someone who is alive, it is unusable regardless of how strong the negative search is. Keep the sentiment, if useful, but present it without attribution; see using an unsourced line.

Decide when the record is ready

Stop when further attempts stop changing the result. For most lines, three wordings across two corpora is a defensible stopping point; search longer when the answer matters more. Record the date as well, so a later reader can judge how current the negative remains.