Science (control)
Water
1 confirmed rewrite episode; the largest by persistence-weighted mass had a 43.5% durable-spine drop. Language openings emphasize different things.
Something to inspect — not a judgment of bias or bad faith.
Overview
Use this briefing to choose what to inspect. The tabs contain the underlying comparisons.
Open the live English Wikipedia article ↗Largest by persistence-weighted mass: 43.5% durable-spine drop. See Rewrite for exact revision receipts.
Different websites cited: 35 → 27. We do not rate those sites. See Citations.
Did editors fight over it?
The page was not dominated by back-and-forth undoing. A quiet rewrite can still matter — it just means few people fought it on the page. About 7,900 recorded revisions in the measure.
Who wrote today’s text?
Half of the wording still on the page is about 7.7 years old or newer. 15.3% was written in the last 3 years. The ten most active accounts wrote 39.1% of what is there now (authorship is more spread out; 725 different accounts overall). Snapshot data on this page runs through 2026-07-01.This is background only — a small group of writers is common on specialist pages.
Available evidence: Rewrite, Vocabulary, Citations, Framing, Facts, Versions.
1 confirmed rewrite episodeA bounded window in which established wording was substantially replaced.
Events bounded by stable revisions where long-lived wording was substantially replaced.
Snapshot corpus through 2026-07-01 (revision 1361772051).
Candidates and exact outcomes
The detector scored 49 interval rows and sent 1 candidate window to exact checking: 1 confirmed, 0 rejected. A candidate window can span multiple interval rows, so rows marked as candidates do not equal the rewrite episode count.
Persistence-weighted changeThe share of wording lost, weighted so text that survived more snapshots counts more. by interval
Each row reports established wording lost, standing wording gained, and paired change as a replacement lead. The bar shows loss for continuity with earlier reports. Below-floor anomalies remain descriptive evidence but cannot receive exact confirmation. A coverage gap is shown descriptively but excluded from detector decisions. Candidate-window labels report the exact revision-level decision for the broader window containing that interval.
How the detector reached this state
- 1 · Coarse scanA snapshot comparison that finds candidate windows for closer inspection.Candidate signal foundThe coarse scan found a deletion-heavy window worth exact checking.
- 2 · Interval scoring49 intervals scoredThe bars below show persistence-weighted wording loss for each interval.
- 3 · Exact checkA revision-level test of whether a candidate window's durable spine collapsed.1 rewrite episode confirmedRevision-level checking found the required durable wording replacement.
- 2002-07-01Loss 28.1%165 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2003-01-01Loss 46.8%1,203 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2003-07-01Loss 1.7%73 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2004-01-01Loss 6.7%520 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2004-07-01Loss 15.3%1,791 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2005-01-01Loss 87.4%13,219 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2005-07-01Loss 5.7%299 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2006-01-01Loss 9.8%897 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2006-07-01Loss 15.6%2,187 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2007-01-01Loss 25.3%5,790 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2007-07-01Loss 10.2%3,094 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2008-01-01Loss 40.4%17,570 removed · 0 added PWRConfirmed candidate window: 2007-07-01 → 2008-01-01
- 2008-07-01Loss 1.8%752 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2009-01-01Loss 2.0%1,110 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2009-07-01Loss 3.6%2,560 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2010-01-01Loss 11.9%10,323 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2010-07-01Loss 2.4%2,198 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2011-01-01Loss 3.8%4,072 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2011-07-01Loss 4.7%5,691 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2012-01-01Loss 3.6%4,859 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2012-07-01Loss 0.5%792 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2013-01-01Loss 0.7%1,117 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2013-07-01Loss 6.3%11,730 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2014-01-01Loss 1.1%2,167 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2014-07-01Loss 2.2%4,631 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2015-01-01Loss 2.6%5,887 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2015-07-01Loss 1.3%3,106 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2016-01-01Loss 0.3%895 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2016-07-01Loss 0.8%2,289 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2017-01-01Loss 7.8%23,997 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2017-07-01Loss 0.8%2,332 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2018-01-01Loss 6.4%21,243 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2018-07-01Loss 1.1%3,665 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2019-01-01Loss 2.2%7,859 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2019-07-01Loss 3.3%12,533 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2020-01-01Loss 1.0%3,870 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2020-07-01Loss 4.3%18,544 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2021-01-01Loss 0.8%3,475 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2021-07-01Loss 2.8%13,678 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2022-01-01Loss 0.1%681 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2022-07-01Loss 0.6%3,332 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2023-01-01Loss 1.6%9,696 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2023-07-01Loss 2.1%13,372 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2024-01-01Loss 0.3%1,839 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2024-07-01Loss 0.5%3,381 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2025-01-01Loss 0.4%2,725 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2025-07-01Loss 0.3%2,619 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2026-01-01Loss 3.0%24,679 removed · 0 added PWRMeasured: no anomaly threshold crossed
- 2026-07-01Loss 0.5%4,389 removed · 0 added PWRMeasured: no anomaly threshold crossed
Exact candidate decisions
Every coarse lead and its revision-level decision, including rejected candidates. RedlinesA before-and-after view showing wording removed and added. compare the full candidate window.
| Candidate window | Exact outcome | Coarse signal |
|---|---|---|
| 2007-07-01 → 2008-01-01View redline | Confirmed: 43.5% durable-spine dropExact pair: 2007-12-02T01:00:19Z → 2007-12-02T01:43:14Z · 42 minutes | coarse interval · 40% persistence-weighted loss · 17,570 PWR massThe absolute persistence-weighted amount of text lost. |
Exact-event attribution: revisions 175167658 → 175175239
3,501 tokensOne tracked unit of text, usually a word or punctuation mark. removed · 14 surviving replacement tokens · Snek01 was associated with 100.0% of removals · Snek01 was the origin author of 100.0% of surviving replacement text.
Attribution describes public revision-history associations and origin authorship. It does not establish bias, motive, or misconduct.
Which words grew or shrank?
Words that became more common or less common between the compared versions. Handy for noticing a shift in topic or tone — not a score of bias.
Comparing rev 175167658 -> rev 175175239 (exact confirmation). Rough word count: 5,964 → 5,020. Overall: small vocabulary change.
Used more afterward
- affected1 → 2
- politics2 → 3
- human17 → 16
- land15 → 14
- very15 → 14
- well11 → 10
- only10 → 9
- potable9 → 8
- system9 → 8
- water510 → 449
- enough8 → 7
- irrigation8 → 7
Used less afterward
- jordan16 → 0
- israel13 → 0
- cubic6 → 0
- region6 → 0
- bank5 → 0
- meters5 → 0
- oecd5 → 0
- privatization5 → 0
- arab4 → 0
- dam4 → 0
- palestinians4 → 0
- suez4 → 0
These are the standout words, not a full dictionary of the article. Boring markup words are filtered out. Treat this as a hint, not a conclusion.
How the footnotes changed
How the article's own footnotes changed across the rewrite: which websites and books were cited more or less. We only show the mix — we do not rate sources as good or bad.
Looking at rev 175167658 → rev 175175239 (exact confirmation, 2007-12-02 → 2007-12-02): 0 websites were cited more (or newly), 8 less (or dropped). Number of footnotes: 54 → 35. Different websites: 35 → 27.
Mix of citation types: news 49%, book 46%, journal 3%, web 3% → book 76%, news 14%, journal 5%, web 5%.
| cited more often | before → after |
|---|---|
| none in top list | |
| cited less often | before → after |
|---|---|
| news.bbc.co.uk | 10 → 0 |
| mondediplo.com | 2 → 0 |
| monde-diplomatique.fr | 2 → 0 |
| hdr.undp.org | 1 → 0 |
| columbia.edu | 1 → 0 |
| mercurynews.com | 1 → 0 |
| environmentalindicators.com | 1 → 0 |
| worldwater.org | 1 → 0 |
Counted from the article’s own footnotes. We list domains only — we do not call any source trustworthy or untrustworthy.
Cross-language lead comparison
How the opening of the article presents the topic in different languages. Disagreement is a reason to read carefully, not a final answer.
In this narrow time window, the English lead text remained textually identical before and after (only revision metadata differs), so no genuine English-side change occurred. The Arabic and Tamil leads also remained essentially unchanged in substance, with Arabic only adding a trailing image caption ('الماء'). Since English did not change, no temporal divergence (moving away, converging, or parallel change) can be established; the only persistent difference is the pre-existing gap in detail and figures between English (highly quantitative, ~1,460 Tt, 71% coverage, aquifer/vapor percentages) versus the much shorter Arabic and Tamil descriptions (basic definition, 70% coverage, no numeric detail), which predates and persists through this window.
Compared matched historical revisions around the confirmed rewrite from 2007-07-01 to 2008-01-01.
Languages compared: en, ta, ar.
Version receipts: en: en before / en after · ta: ta before / ta after · ar: ar before / ar after
| topic | how they compare | English change | other language(s) change | languages |
|---|---|---|---|---|
| English lead unchanged while AR/TA leads unchanged too; no substantive divergence introduced in this window | difference persisted | Before: Water is a common chemical substance that is essential to all known forms of life... About 1,460 teratonnes (Tt) of water covers 71% of the Earth's surface “About 1,460 teratonnes (Tt) of water covers 71% of the Earth's surface” After: Water is a common chemical substance that is essential to all known forms of life... About 1,460 teratonnes (Tt) of water covers 71% of the Earth's surface “About 1,460 teratonnes (Tt) of water covers 71% of the Earth's surface” | Before: الماء سائل شفاف دون طعم أو رائحة أو لون... كما أن 70% من سطح الأرض مغطى بالماء “كما أن 70% من سطح الأرض مغطى بالماء” After: الماء سائل شفاف دون طعم أو رائحة أو لون... كما أن 70% من سطح الأرض مغطى بالماء، تصغير|الماء “no quotation” | AR, TA |
| AR added a thumbnail image caption 'الماء' with no textual content change | unclear change | Before: No equivalent image caption addition present. “no quotation” After: No equivalent image caption addition present. “no quotation” | Before: No trailing image caption present in the lead. “no quotation” After: تصغير|الماء added at the end of the lead. “تصغير|الماء” | AR |
Differences are invitations to read both sides — not a score of who is right.
Versions we used
These are the exact public Wikipedia versions behind the checks on this page. Open any link to read the original. Wikidata item: Q283.
| language | comparison point | article title | version | when | text used (chars) |
|---|---|---|---|---|---|
| en | before | Water | open version 175167658 | 2007-12-02T01:00:19Z | 2,555 |
| en | after | Water | open version 175175239 | 2007-12-02T01:43:14Z | 2,555 |
| ta | before | நீர் | open version 190161 | 2007-11-30T02:21:36Z | 363 |
| ta | after | நீர் | open version 199808 | 2008-01-03T07:52:39Z | 363 |
| ar | before | ماء | open version 1157634 | 2007-11-24T10:28:55Z | 338 |
| ar | after | ماء | open version 1180557 | 2007-12-02T22:32:57Z | 350 |