Wikipedia:When sources are wrong
![]() | This is an essay. It contains the advice or opinions of one or more Wikipedia contributors. This page is not an encyclopedia article, nor is it one of Wikipedia's policies or guidelines, as it has not been thoroughly vetted by the community. Some essays represent widespread norms; others only represent minority viewpoints. |
Sources are wrong sometimes. They usually aren't; we try really hard to pick sources that will usually be right. But even those sources are sometimes wrong. Usually about little things, sometimes about big things, on rare occasions about critical things. Wikipedia:Neutral point of view gives us guidance for when sources disagree on matters of opinion, like if some sources characterize something as a language and some characterize it as a dialect. Wikipedia:No original research, and a litany of essays such as Wikipedia:The Truth, give us guidance for when we personally disagree with a source based on our own subjective views. But what about when sources disagree on basic facts, and one source is just clearly right, or where a source makes a claim that defies common sense and does not make a compelling case for its veracity? This essay outlines a few different approaches to such situations, based on actual examples, including ones encountered in writing good and featured articles.
Six approaches are given below, and can be synthesized as:
- If a source is trivially wrong, either ignore it (Approach 1), emphasize the correct answer without noting the error (2), or, if the error is particularly relevant, mention it in a footnote (3) or rarely prose (4a).
- If an error is significant but there is still a clearly right answer, mention it in a footnote (3), or in prose (4a) if it is of encyclopedic note.
- If there is a rough consensus of sources, state that in prose, while elaborating in a footnote (3) or prose (4b) based on the significance of the matter.
- If it is actually unclear what the correct answer is, but there is strong reason to think many or most sources are wrong or there is nothing approaching a consensus of sources, explain the conflict in prose (4b/4c) or do not discuss the topic at all (5).
- In rare cases, potentially include a likely incorrect statement (or even article) if there is no way to justify another approach as within editorial discretion (6).
How can we tell a source is wrong?
[edit]For this essay to apply, we need to be able to show, unambiguously, that a source is either wrong or very likely to be wrong. If you are reading this because you think the cold streak you've been having disproves climate change, this essay is not for you. This essay concerns cases where, based on an analysis of existing reliable sources for a topic (or the lack thereof), we can infer errors in a particular source. We can detect this in a few ways:
Type of error | Example |
---|---|
A secondary source conflicts with a primary source, where there is no way for the former to know better than the latter. | A reliable map says that a village called Greenacre exists, rendered in all caps. A crime happens in that area and a news publication says it happened "Near the villages of Green and Acre". |
A primary source makes a claim that is explicitly called out as wrong by reliable secondary sources. | A reliable map says that a village called Greenacre exists. Cartography Monthly calls this out as an error, saying that really it's Green and Acre. |
A non-contemporaneous source conflicts with a contemporaneous source, where there is no way for the former to know better than the latter. | A reliable newspaper quotes a victorious athlete as saying "Today is one of the greatest days in my life". Years later, an obituary in another paper renders the quote as "Today is the greatest day of my life". |
A contemporaneous source makes a claim that is explicitly called out as wrong by reliable non-contemporaneous sources. | A reliable newspaper quotes a victorious athlete as saying "Today is one of the greatest days in my life". Years later, an obituary in the same paper notes that, based on the recorded interview at the time, he was misquoted and actually said, "Today is the greatest day of my life". |
A source used terminology in a way that is obviously inaccurate, in a way that can be shown by citing a dictionary. | A non-medical publication refers to someone having a myocardial infarction as a result of blood loss, mixing that term up with cardiac arrest. |
A relatively small number of sources conflict with the overwhelming consensus of sources, where there is no reason to think that the minority is more reliable on this subject. | One hundred sources detail an actress's upbringing in Canada and continued participation in Canadian life. Three sources casually refer to her as "American actress so-and-so". |
A source makes an exceptional claim that is not directly contradicted by any source, but if it were true would be very unlikely to only appear in that one source. | Ninety-nine sources describe someone's career as a sculptor, while one source also mentions a successful poetry career, in a way that it would make no sense for the other ninety-nine to have intentionally excluded. |
A source makes an obviously wrong statement of the sort that is exempt from Wikipedia:No original research. | An anglophone newspaper mistranslates a Chinese phrase in passing. |
Often this analysis is augmented by being able to detect the plausible error using common sense. Likely the three sources thought the actress was American because she appears in Hollywood films; the sculptor had the same name as a poet from around the same time and place; etc.
Approach 1: Quietly get it right
[edit]Suppose John Doe's birthday is January 13, but a single source transposes the digits and says it is January 31. Doe is a well-known figure, and there is zero doubt in any other source as to his birthday. The erroneous source does not include any claim that it has uncovered Doe's true birthday and that all other sources are wrong; it is just a passing reference.
In this case, the correct birthday can be given in the article, without any further ado. If there is reason to think it may confuse editors (for instance, the same source is cited in the article for other claims), a hidden comment may be included along the lines of <!--Incorrectly stated as Jan 31 in Smith 2011-->
.
This approach is particularly useful for claims made in sources that are of lower quality than their peers. If breaking-news coverage of an event identifies Richard Roe as a resident of Springfield, but all subsequent coverage says he is from Shelbyville, and the misidentification is not itself encyclopedically notable, we can likely say "Shelbyville" without elaboration. If a non-specialist source says scientists have made a breakthrough in teleportation, and specialist sources say no, it was quantum teleportation, we probably don't need to humor this error. This approach is best-suited to cases where a single source errs, or where a handful of sources err out of a large pool of available sources, and where the error is unremarkable.
There are probably 100 reliable-source articles published every day that contain errors so blatant that they do not need to be mentioned in any article. To cherry-pick a few from a certain paper of record:
- A December 2017 article in The New York Times called a specific floating object a "houseboat", in summarizing a U.S. Supreme Court case that ruled the exact opposite. We do not need to qualify our description of that ruling based on a mistake made four years after the fact, especially given that the same journalist at the same paper got it right in his initial coverage.
- A May 2023 article in The New York Times reports that King Charles III "represents nearly 1,000 years of unbroken royal lineage". This is completely incorrect. Even setting aside various changes to succession rules, the British royal family was deposed for 11 years in the 1600s. We do not need to note, at the article Commonwealth of England, that according to The New York Times, the British royal family has never left power.
Approach 2: Get it right in detail, without mentioning the error
[edit]This approach is similar to #1, but a bit more detailed. Sometimes an error will be reasonably common, but also not something anyone asserts with any seriousness. Suppose that Jane Doe, a prominent resident of New Jersey, was born in neighboring New York. From time to time, sources dashing off an article about her include the words "born in New Jersey" or "lifelong New Jerseyan". We do not need to address this mistake head-on; however, in going into detail about the circumstances of Doe's birth (... was born in Manhattan, New York, in 1987. She lived in New York for the first seven years of her life, until ...) we address any confusion a reader might have, without needing to call out specific erroneous sources.
- A June 2023 article in The New York Times describes Martinique as a French territory. It is not a territory; it is a single territorial collectivity, which, despite the name, is not a kind of territory. Many other sources make the same mistake. Our article on Martinique does not need to say "sometimes mistakenly referred to as a territory", even though as august a source as the Times has done so. Instead, our explanation of Martinique's actual constitutional status suffices to correct any errors by sources.
- Many, many sources incorrectly state that The Washington Post and The New York Times both ran the Unabomber manifesto in print. This may be due to the ambiguity of the word "publish": The Times co-published it in the sense of paying for half of the cost of printing and circulating it, but the supplement only ran in the Post. This is unequivocal in contemporaneous reporting, but is frequently stated incorrectly in subsequent journalism. Our articles on the manifesto and
its author do not explicitly reference this common error, but do explicitly note that only the Post ran the work, serving to quietly debunk the mistake.
- Many sources over the years have misstated the circumstances of Geraldo Rivera's name change from Gerald Riviera. Our article on Rivera does not go into detail about this misinformation, but devotes two and a half paragraphs to the topic of Rivera's birth name and name change. In cases like this, omitting incorrect claims is as much a matter of WP:DUE as anything else.
Approach 3: Get it right and add a footnote
[edit]This approach is common in cases where an error is fairly widespread among sources, but still clearly wrong. Suppose Jay Doe was born in Springfield, and there are ample contemporaneous sources for this statement. But for whatever reason, midway through their life, sources started to say they were born in Shelbyville; assume, again, that none of the newer sources challenge the accuracy of the older ones. In this case, it may make sense to have a footnote explaining the source analysis, e.g. ... was born in [[Springfield]]{{efn|Many profiles of Doe published from the 1990s onward say that they were born in neighboring [[Shelbyville]].[ref][ref][ref][ref] However, a 1971 birth announcement in ''The Springfield Gazette'',[ref] a 1972 interview in the same paper with their mother,[ref] and a 1988 profile in ''The Collegetown Reliable Source''[ref] all indicate a birthplace of Springfield.}}
In other cases, one source may differ from the pack, but not in a way that can clearly be called an error. If there is a strong consensus among sources of roughly equal quality, it may make sense to assert the consensus in the encyclopedia's voice (or slightly hedged, e.g. is generally thought to have...), while noting the minority report in a footnote (Exceptionally, Adams writes that...). Or perhaps sources mostly agree but differ slightly, so we summarize their conclusions in text and go into detail in a footnote: Doe was around 6 feet tall{{efn|Height variously given as 5 feet 11 inches[ref][ref][ref][ref], 6 feet even[ref][ref], or 6 foot 1 inch[ref][ref][ref]}} See also Approach 4c below.
![A pale white arm, tattooed with the quote "The universe loves you because you are love", all lowercase and in a Minecraft-like font.](http://upload.wikimedia.org/wikipedia/commons/thumb/f/f6/End_Poem_tattoo_%28cropped%29.jpg/290px-End_Poem_tattoo_%28cropped%29.jpg)
- The Pitt Rivers Museum quotes a poem which it says refers to the
Biddenden Maids. However, the alleged poem is almost identical to Walter Scott's The Lay of the Last Minstrel, which obviously calls the poem's veracity into doubt. Our article notes both the museum's claim and the similarity to The Lay in a footnote.
- Two non-contemporaneous sources say that Kraft General Foods bought the rights to the American
Capri Sun brand in 1991. However, two contemporaneous sources says that Kraft bought the entire company Capri Sun Inc., going into detail about how the company would be integrated into Kraft, and a later court opinion is consistent with this. Because multiple sources, one of which is cited in the article for another claim, incorrectly characterize the transaction, some acknowledgment of that characterization is merited. But given that there is no reasonable dispute that the contemporaneous sources are right, and given that the distinction is relatively minor, resolving the matter by footnote is fine.
- In an essay about his End Poem, Julian Gough slightly misquotes its climactic line. There is no dispute as to what the actual line is; he even quotes it correctly when he posts the entire poem later in the essay. But saying that he quoted that line, when he in fact wrote something slightly different, could be a misrepresentation of a primary source, so a brief footnote is added about the misquotation.
New Amsterdam Theatre shows a case of sources occasionally disagreeing on small but important details. It provides the prevailing interpretations in prose and then minority reports by footnote.
The Minute Man uses footnotes to explain why certain statements were favored, for instance
The preponderance of the sources use the name The Minute Man ([4 sources]) or Minute Man ([3 sources]) for the sculpture, but Minuteman ([2 sources]) is also used.
- The [s] sound in
Sarah Ashton-Cirillo's surname is incorrectly cyrillized as a К [k] on her press pass, rather than the obviously correct С [s]; the latter is correctly used in Ukrainian-language secondary sources. Since a photo of her press pass appears in a cited source, her article mentions both cyrillizations in a footnote, but makes clear that С is correct.
- See § Tosia Altman and § Ash Street shootout below.
Approach 4: Explain the conflict in prose
[edit]This approach applies primarily in three cases.
4a: Notable error
[edit]This occurs when the error is itself of encyclopedic relevance. These are fairly straightforward and this essay will not dwell much on them.
- The article D. B. Cooper discusses how the subject of the article, who used the pseudonym Dan Cooper, was given the moniker "D. B. Cooper" due to a mixup by a journalist.
4b: Entrenched error
[edit]This occurs when the error runs deep enough that we cannot say, in the encyclopedia's voice, that it is an error. This may occur particularly when an analysis of primary sources strongly suggests that secondary sources have erred, but secondary sources make the claim confidently and it cannot be fully ruled out that they are relying on information we are not privy to.
- See § Harry R. Truman and § Tosia Altman below.
4c: Somebody's gotta be wrong
[edit]When sources disagree in small ways, this can be smoothed out by standard application of WP:NPOV. Other times, one source will sharply diverge from the pack, as discussed in Approach 3 above. However, sometimes sources will differ enough that it cannot be ignored, without any clear consensus winner. In these cases one may have to simply present the differing accounts in prose and let the readers take them as they will. Doe was born in 1950 according to the State Gazette, 1952 according to the National Tabloid, or 1953 according to the Hollywood Woodsman.
- Four different, mostly mutually exclusive narratives exist as to the fate of
Saint Vincent Beer. The article notes all four without implying any one is correct.
- Within two years of her first mention in a reliable source in 1918, Irene Ryan had had her birth surname spelled three different ways (Noblitt, Noblett, or Noblette), and the ambiguity persists to this day. Her articles gives all three spellings, with a footnote linking the earliest findable instance of each spelling.
- See § Ash Street shootout below.
Approach 5: Say nothing
[edit]Sometimes it's clear, based on editorial analysis of a source, that it is wrong, but there is no substitute source that is clearly right. Here it bears mentioning: We do not engage in original primary or secondary research, but we are creators of original tertiary analysis. That is what we do as encyclopedists. We can and do look at things and say "We know this to be wrong, even if no source says it is wrong". This of course must only be done on objective standards. We cannot disqualify sources simply because we disagree with them. But when a source makes a claim that is patently implausible and does not present a compelling case for it, we may choose to ignore it, even in the absence of an opposing source.
Construction of the World Trade Center cited a tertiary fact sheet by PBS for the claim that 60 workers died creating the building. This would be an extremely high death toll, and almost certainly would be highlighted in primary and secondary sources in greater detail, so the source is presumed unreliable for this particular claim. However, as no other source asserts any other death toll, the line was removed and the article is now silent as to death toll. A suboptimal state, maybe, but better than getting it wrong.
- What's the southernmost point in New Jersey? This ornithology paper says it's Cape May Point. This one says it's Cape May City. Ornithology papers are not where we'd expect to find an intentional challenge to the validity of a map, so we can rely on any high-quality map of New Jersey to say that the first paper's wrong: it's Cape May City. However, again WP:DUE comes into play: Since there's not much coverage of this matter at all in sources we would consider reliable with respect to geography, neither article discusses the question.
- A number of high-quality sources have asserted as fact a particularly unsavory detail regarding Biljana Plavšić's actions immediately after the
Bijeljina massacre. In an RfC, a rough consensus of editors agreed that, while the sources were of high quality, the sources they themselves relied upon were sufficiently questionable as to make the detail suspect, and agreed to exclude it.
- See § Arrests of Ulysses S. Grant below, regarding the unusual, and to date untested, but not-inconsistent-with-policy notion of deleting an article that would normally pass GNG, if all sources are wrong.
Approach 6: Get it wrong(?)
[edit]Sometimes this might be the only answer. If a large majority of sources are adamantly, confidently wrong, then we, as chroniclers of consensus, may have no option but to reflect that. This is a dicey topic. Wikipedia:Verifiability, not truth has been downgraded from policy to partial policy to essay, while at the same time words vaguely reflecting it remain in Wikipedia:Verifiability. No one likes the idea of Wikipedia getting it wrong, but it may sometimes be inevitable.
- This version of the article Priming (psychology) from late 2012 was largely consistent with the state of the art. It described priming as a real phenomenon well-supported by research. A few months later, evidence began to emerge that many of the studies demonstrating priming could not be replicated. Prior to that shift in academic attitudes, however, it was correct for our article to describe priming as real—even though we were wrong.
- See § Arrests of Ulysses S. Grant below.
Case studies
[edit]Harry R. Truman
[edit] Harry R. Truman managed to pass GAN while falsely asserting a middle name of Randall, the result of an unsourced edit in 2007. (Truman's actual middle name is thought lost to time.) The error was corrected by the time the article became featured, but it was too late for several reliable sources, which appear to have copied the incorrect name from the article. We cannot definitively say that the sources are wrong, but we are also very skeptical that they are right, so instead we say
Some noncontemporaneous sources have given his middle name as Randall,[1][2] but Truman stated that he did not know his middle name, just the initial R.[3][4]
We thus avoid saying in the encyclopedia's voice that this was his middle name, but also avoid saying in the encyclopedia's voice that the more recent sources are wrong, which would also be original research since there is some (slim) chance they know something we don't and some (maybe a bit less slim) chance that they are coincidentally right.
Sources
|
---|
|
Tosia Altman
[edit]Tosia Altman is a fairly straightforward case, but useful to juxtapose to Truman. Like Truman, alternate names (Tova and Taube) appear for her in non-contemporaneous sources, so Approach 4 might be the right call here. However, these names never overtook Tosia as the primary way to refer to her, and are not otherwise important to the article, so Approach 3 is taken, with the situation explained via footnote.
Arrests of Ulysses S. Grant
[edit]The question of whether Ulysses S. Grant was arrested in 1872 posed a dilemma that multiple editors said they had never encountered before. A large number of reliable sources asserted the claim as fact. A small number questioned it but then still tended to conclude it was fact. Most editors (including this essayist, the article's principal author) agreed that the claim was suspect, but disagreed as to whether it was within editorial discretion to say that, or whether Approach 6 was the only way forward. Eventually, two options were considered:
- Determine that, due to inadequate skepticism about the event, all cited sources should be considered unreliable as to the facts of the matter, and take the article to AfD for failing GNG in a rather roundabout way. This can be viewed as an extreme case of Approach 5.
- Reframe the article to be entirely about the claim that Grant was arrested, while remaining adamantly silent as to the veracity of that claim. This would be an extreme case of Approach 4.
The second option was settled on at least provisionally, and then, coincidentally, a new reliable source came out that did question the arrest, a deus ex machina that thankfully resolved the dispute but annoyingly prevented the possibility of a full case study into the extreme edge case of an article where 100% of sources may be wrong. As of July 2023, the article frames the arrest as an alleged event of disputed historicity.
Ash Street shootout
[edit]The Ash Street shootout is an extreme case of Approach 3, and to a lesser extent 4c. While the broad outline of the event is not in dispute, almost every detail is. If one were to summarize it without taking any sides at all, one would have to say that some number of Army Rangers, reinforced by some other number of Rangers, either harassed or were harassed by some number of people who might have been dealing drugs and may or may not have been Crips, due to a surveillance camera—which was installed either shortly before or some time before, and may have had its presence revealed in a newspaper article but may not have—and then either fired upon them or were fired upon, resulting in a gunfight of some number of minutes, in which some number of bullets were fired, which struck either zero, one, or more than one people, leading to either some or all of the Rangers' guns being confiscated.
Little of what appears in the sources can clearly be called wrong, but almost every important detail is a case of "somebody's gotta be wrong". The article uses short summaries of differing accounts in prose, and then elaborates by footnotes, with a dedicated section for unclear numbers and separate explanatory notes for the questions of whether the alleged drug dealers were Crips, who fired first, and what the story was with the camera; the conflicting casualty figures are handled in prose.
Powder House Island
[edit] Powder House Island had a large amount of dynamite on it, which exploded on June 27, 1906. All sources at the time agreed on this, ranging from multiple newspapers to court documents related to the incident (not disputed by the plaintiff, defendant, or judiciary). A single article in the Detroit Free Press, written the day after, said the explosion happened on nearby Fox Island (previously used as a dynamite storage area by the same dredging company); 110 years later, a local newspaper writing about the history of the islands said that the explosion happened in 1907. Since these are obvious errors that contradict otherwise-unanimous agreement among a great many sources, the article uses Approach 3.