Independent analysis · English Wikipedia · 27 July 2026
The article was born as an openly partisan definition, was neutralised within ten weeks, and has flipped its framing at least six times since. Here is who did it, what they changed, and what the record does — and does not — support.
This is not a captured page. It is a decade-long, genuinely two-sided argument that neither side has permanently won, fought almost entirely in the language of Wikipedia sourcing policy rather than politics.
The dispute is narrower and stranger than “left versus right.” It is a use–mention problem: is this an article about a behaviour — conspicuously expressing moral values — or about an accusation that someone is doing so insincerely? Ten years of editing is essentially that question being re-litigated, and the lede rewritten each time the answer changes.
A handful of accounts dominate. Two pushed hard to expand the article; three pushed hard to cut it. The cutters mostly won — but the evidence suggests they won because their removals had the stronger formal policy footing, not because they outnumbered anyone.
Each segment is the framing that led the article during that period. Verified from the full wikitext of 33 archived revisions.
10 April 2016 · created by Dwanyewest
“Virtue signalling” is a pejorative term for a person expressing or promoting socially progressive views, including advocacy for women’s rights and civil rights.
That is the article’s actual first sentence, and it is both maximally pejorative and maximally politically coded — it names the political content the term is supposedly aimed at. It survived roughly ten weeks.
Current revision · 23 June 2026
Virtue signalling is a pejorative neologism for the expression of a moral viewpoint with the intent of communicating good character, frequently used to suggest that the person is more concerned with appearing virtuous than with actually supporting the cause or belief in question.
Same verdict — pejorative — arrived at from the opposite direction. The 2016 version called it a slur against progressives. The 2026 version calls it an accusation of hypocrisy, with no political target named. In between, the word “pejorative” was removed from the lede and restored at least six times.
499 editors, but the page belongs to about eight of them. Verified from consecutive revision-size deltas across all 1,256 revisions.
| Editor | Edits | Net bytes | Active window | Direction |
|---|---|---|---|---|
| M2sh22pp1l (was Bacondrum) | 97 | −18,979 | 2019–2020 | Cut |
| Yaris678 | 51 | +2,711 | 2016–2018 | Mixed |
| Netoholic | 50 | +15,489 | 2017–2020 | Expand |
| Sangdeboeuf | 35 | +4,092 | 2024–2026 | Mixed |
| Wowaconia | 33 | +31,979 | 2017 | Expand |
| Aquillion | 32 | −25,324 | 2019–2024 | Cut |
| HouseOfChange | 31 | +1,370 | 2021–2026 | Mixed |
| Grayfell | 25 | −19,189 | 2017–2024 | Cut |
The structure is unusually clean. Wowaconia (+31,979, entirely in 2017) and Netoholic (+15,489) are the expansionists. Aquillion (−25,324), Grayfell (−19,189) and M2sh22pp1l (−18,979) are the reducers. Their active windows barely overlap — Wowaconia is a 2017 phenomenon, M2sh22pp1l a 2019–2020 one, Sangdeboeuf a 2024–2026 one. The article was fought over in shifts.
The highest-edit-count account, M2sh22pp1l, is the account formerly named
Bacondrum — renamed 25 May 2021 by an administrator.
The article history shows the new name; the talk pages and the deletion debate,
which aren’t rewritten by renames, still show the old one.
Confirmed independently: a lookup of Bacondrum now returns
missing, and the account’s own edit summary describes reverting
“to revision 963014620 by Bacondrum” — i.e. to its own
earlier edit. This is a routine rename, not a sockpuppet and not concealment.
It matters only because any analysis that misses it will mis-attribute the single
most active participant. Verified
The honest answer requires separating what the record shows from what it merely permits.
What is observable: specific editors made changes that shifted the article’s framing, in both directions, repeatedly, over ten years. That is simply true and the diffs show it. Verified
What is inference: that this amounts to bias rather than disagreement. Here the evidence cuts against the obvious story. Inference
Both camps argued from policy, not ideology — original research, synthesis, source reliability. The clearest statement of the real disagreement came from an editor in 2024:
“The lead was twice changed in a way that completely changes the definition from dealing with allegations of disingenuousness to being disingenuous. That’s not quite the subject, though. The way it’s most often used implies an allegation of disingenuousness.”
Rhododendrites, talk page, 17 June 2024
Three demolitions define the article’s shape more than anything else.
The 2017 article documented where the phrase came from using the places it actually appeared: a LessWrong forum thread, a PJ Media column, and — remarkably — a reader comment posted beneath that column. Grayfell removed 9,138 bytes of it as original research. Among the deleted material:
“It’s Moral Preening, Virtue Signalling, Competitive Pearl-Clutching, Flashing Tribal Signs — call it what you will — it’s a Superior Dance beyond the Church Lady’s wildest imagination.”
A commenter named “dicentra”, cited in the article until 2017
This is the article’s foundational conflict: can a term’s colloquial origin be documented from the primary sources where it actually appeared? Grayfell’s answer — no, that is original research — ultimately prevailed. That material is absent from the article today.
A student editor on a UCLA course assignment added 19,579 bytes of signalling-theory background: Darwin, sexual selection, Zahavi’s handicap principle, contract theory. Aquillion deleted 13,139 bytes of it within the day — “none of this mentions the term” — as synthesis.
Worth stating plainly, since it would be easy to misread as activism: this was coursework, not advocacy. Verified
Over 55 minutes, M2sh22pp1l made 17 consecutive edits taking the article from 10,679 to 5,717 bytes — a 46% reduction — ending with the summary “this is all cruft really.” Further trimming took it to 4,408 bytes, its lowest point since 2017. A speedy-deletion tag followed, then a proposed deletion, then a full deletion nomination.
Another editor put the obvious accusation on the record:
“The nominator has spent the last couple weeks absolutely devastating the content of this article via removal of significant sourced portions, I’m sure in preparation for this ill-conceived AfD so he can claim that the article is ‘a dictionary entry’.”
Netoholic, deletion discussion, June 2020
An uninvolved passer-by independently checked the byte arithmetic and confirmed the scale of the cuts. So did I — the numbers are accurate. Whether the sequencing was deliberate preparation or ordinary cleanup that happened to precede a good-faith nomination is not determinable from the edit record, and I decline to assert either. What is verifiable is the ordering and the magnitude. Inference
The deletion failed. It was closed as keep within three days.
The same 3,132 bytes, added and removed four times in two days.
On 31 May 2020, five minutes apart, the article’s definition was rewritten from an academic origin story into a pejorative — the framing that, with interruptions, still leads today.
The same edit removed a block of academic citations that had been footnoted under the note “Academic usage predates pejorative.”
Stated plainly, because the gaps matter as much as the findings.
All data was pulled read-only from the public MediaWiki API: the complete revision history (1,256 revisions, 10 April 2016 to 23 June 2026), 420 talk-page revisions, both talk archives, the deletion discussion, the page’s full protection and deletion logs, and contribution profiles for 35 accounts. Byte figures are computed from consecutive revision-size deltas across the entire history rather than sampled. No edits were made and no editor was contacted.
The findings were then independently re-checked against the live API before publication: the total revision count, the first revision’s timestamp and ID, the verbatim 2016 opening sentence, the account-rename evidence, the current lede, and the page’s protection status and watcher count. All matched exactly.
Every claim above is marked Verified where it comes from fetched data, Inference where it is reasoning on top of that data, and Could not verify where the record does not settle the question.