The 1966–1976 Cultural Revolution Mortality Figure Inventory: Competing Series Documented as Non-Interchangeable

Evidence inventory — research use only; not published; no advocacy; no ranking of figures. Compiled: 2026-09-24 (EDT). Coverage: the open evidence question on how many people died as a result of the Cultural Revolution (1966–1976). This is a figure-responsibility file, not a death toll: it inventories competing published mortality series with their definitions, methods, provenance, and uncertainty. No single national figure is endorsed, averaged, merged, or selected as “best.” Every headcount below appears only inside its attribution frame. This preserves the numbers-discipline established in the project’s 1959–1961 famine data record: competing series are documented as non-interchangeable; no bare numbers appear anywhere.

(a) Why no single responsibly sourced figure exists — the definitional taxonomy

Published series differ on at least four independent axes, and differences on any one axis make two series non-interchangeable:

Who is counted. At minimum five different counted populations appear: (i) armed-conflict (武斗) deaths only; (ii) all violent deaths including persecution-driven suicides; (iii) “abnormal deaths” (非正常死亡) as a statistical category in official documents and county annals — which includes deaths from persecution, execution, armed conflict, and suicides but not deaths from famine or disease that would have occurred anyway; (iv) all excess deaths during the CR period regardless of mechanism; (v) total victims of persecution (implicated/struggled, not dead). Series counting (iii) and (v) are frequently confused in secondary circulation — they must not be.
Period boundaries. The official periodization is May 1966 – October 1976 (1981 Resolution). Most violence, however, is concentrated 1966–1969 (Red Guard phase and factional armed conflict) and 1968–1971 (military-administered suppression and campaigns such as the Cleansing of the Class Ranks and the “May 16 elements” investigation). County-annals-based estimates variously cover 1966–1969, 1966–1971, or the full decade. A 1966–1971 figure and a 1966–1976 figure are not the same quantity.
Reported deaths vs. actual deaths. County gazetteers (县志), compiled by local governments and published officially, report deaths — but published gazetteers are widely documented to understate violent deaths relative to the classified (内部) versions of the same counties’ records. Andrew Walder’s team therefore treats gazetteer numbers as reported deaths and builds statistical models to infer actual deaths; the reported figure and the modeled figure from the same database are different quantities. For example, from the published county annals of Lingshan County, Guangxi, only 8 people are shown as having died, while the county’s classified document records 3,220 victims; Binyang County’s published annals show 37 victims against 3,951 in its classified document (attributed: Song Yongyi, via the Guangxi Massacre record). Any series that does not state whether it reports gazetteer counts or modeled estimates is not usable.
State secrecy. Song Yongyi (2011) states directly: for the government, the CR death toll “is a highly classified state secret,” although the state does maintain “abnormal death” statistics nationwide (attributed: Song Yongyi, “Chronology of Mass Killings during the Chinese Cultural Revolution (1966–1976),” Online Encyclopedia of Mass Violence / Sciences Po, published August 25, 2011). The absence of an official figure is structural, not an accident of research.

Consequence: the responsible output is an inventory of attributed series, each carrying its own definition, method, period, and uncertainty — never a single merged figure.

(b) The official PRC position

**The 1981 Resolution on Certain Questions in the History of Our Party assesses the Cultural Revolution as “ten years of turmoil” (十年动乱) and attributes chief responsibility to Mao, without quantifying the death toll** (attributed: the Resolution text; recorded in the project’s prior evidence). No subsequent official publication has added a number.
Deng Xiaoping, interview with Italian journalist Oriana Fallaci, 1980: “How many people really died in the Cultural Revolution — that is an astronomical figure, a figure that can never be estimated” (attributed: the interview as quoted in secondary reporting). This is the closest thing to an official acknowledgment that quantification is considered impossible — PROBABLE as an accurate quotation of Deng’s position; the full interview text is a separate primary-source question and is not asserted here.
Bottom line on the official record: no official death toll has ever been published. What circulates as “official” figures are all official-adjacent — internal documents, remarks by leaders, or books credited to Party-history organs — and they are inventoried separately below with their provenance caveats.

(c) Official-adjacent acknowledgments (attributed; provenance graded)

These are the figures most often presented as “the government’s own numbers.” None is a published official toll, and their provenance varies from traceable internal compilations to unverifiable circulated quotations.

C1. The 1.728-million internal-document figure (1984) — STRONG/ATTRIBUTED, primary text not located.

Two widely-circulated claims, per a May 15, 2026 scholarly roundup (中国瞭望): “one is the internal material compiled by the CCP Central Research Office in 1984, which estimated abnormal deaths over the ten years of the Cultural Revolution at 1.728 million” (attributed: the roundup; original internal document not located — see PRIMARY-SOURCE-TARGET 1).
The same figures appear in a detailed breakdown attributed to 《建国以来历史政治运动事实》 (co-compiled by the Central Party History Research Office 中共中央党史研究室): “more than 4.2 million people were detained and investigated; more than 1,728,000 died; 135,000 were executed as counterrevolutionaries; 237,000 died in armed conflict; 7.03 million were injured or disabled; more than 71,200 families were entirely destroyed” (attributed: the compilation as quoted in secondary reporting; the compilation’s own text not independently located — see PRIMARY-SOURCE-TARGET 2).
Definition check: within this breakdown, “1,728,000 died” is the aggregate abnormal-death figure, of which armed-conflict deaths (237,000) and executions (135,000) are named subcategories. The figure therefore counts category (iii) in section (a), not all excess deaths. Do not compare it directly with series that include suicides separately or that exclude armed conflict.
Freshness: the May 2026 roundup confirms this figure is still the stable circulated form; nothing indicates the underlying internal document has been published or revised.

C2. The Party-History-Research-Institute-credited 1,490,300 figure — CONFLICTING with C1.

The Tufts University “Mass Atrocity Endings” research site (archived 2019-10-08) reports: “this tallies with official estimates of the number of non-conflict related deaths reported by a book credited to the Party History Research Institute (which estimated that 1,490,300 deaths took place in China during the Cultural Revolution)” (attributed: the Tufts site; the credited book is not named in the site text — see PRIMARY-SOURCE-TARGET 3).
Conflict with C1: 1,490,300 (non-conflict deaths, one book) vs. 1,728,000 (total abnormal deaths, 1984 internal material) are different definitions from different attributed sources; they cannot be combined or averaged. The two figures may reflect different counting scopes or different source documents; which is not established.

C3. Hu Yaobang’s 1-million figure — ATTRIBUTED, provenance via secondary reporting.

The Washington Post, July 17, 1994 (“How Many Died? New Evidence Suggests Far Higher Numbers for the Victims of Mao Zedong’s Era”): “Hu Yaobang, former communist Party chief, has been quoted as saying that 1 million people died during the period, but the figure apparently excludes the fighting between Red Guard and army factions” (attributed: the Post article). The same article notes Western scholars had “long accepted that half a million people died in the Cultural Revolution” relying on official Chinese sources, and that Brookings scholar Harry Harding believed “about 1 million died.”
Hu is separately quoted as saying 200 million people were criticized, struggled against, or implicated — “so almost every household had someone who suffered unjust treatment and persecution” (attributed: secondary reporting). Definition check: the 200-million figure counts victims/implicated (category (v) in section a), not deaths; it must not be set beside death figures.
Uncertainty: the Post article is the located source; Hu’s original statement (venue, date, exact wording) is not established in this inventory.

C4. Ye Jianying’s circulated figures (20 million dead / over 100 million persecuted) — ATTRIBUTED, unverifiable through official channels.

The widely repeated form: at the closing session of the December 1978 Central Work Conference (December 13, 1978), Ye Jianying said that during the Cultural Revolution the country “整了1亿人，死了2,000万人” (“struggled against 100 million people, 20 million died”) and wasted 800 billion yuan (attributed: widespread secondary circulation).
A detailed component breakdown attributed to Ye’s disclosure at the enlarged Politburo meeting after the 1st Plenum of the 12th Central Committee is reported via Yang Jisheng’s article “道路·理论·制度——我对文化大革命的思考” (《记忆》 no. 104, published November 30, 2013): (1) more than 4,300 large-scale armed-conflict incidents, more than 123,700 dead; (2) 2.5 million cadres struggled against, more than 302,700 cadres illegally detained, more than 115,500 cadres abnormal deaths; (3) in cities, 4,810,000 people labeled historical counterrevolutionaries, current counterrevolutionaries, class aliens, counterrevolutionary revisionists, or reactionary academic authorities, more than 683,000 abnormal deaths; (4) in the countryside, more than 5.2 million landlords/rich peasants (including some upper-middle peasant families) persecuted, 1.2 million abnormal deaths; (5) more than 113 million people subjected to political blows of varying degrees, more than 557,000 missing (attributed: Yang’s article as quoted in secondary reporting; Yang’s article text and Ye’s disclosure not independently verified in this inventory — see PRIMARY-SOURCE-TARGET 4).
Arithmetic note (inference, not a finding): summing the four death components (123,700 + 115,500 + 683,000 + 1,200,000) gives approximately 2.12 million — which does not reconcile with the “20 million” headline figure. The two circulated forms are internally inconsistent; neither can be treated as a coherent series. Record both; reconcile neither.
The May 2026 scholarly roundup explicitly notes: “Ye’s claim cannot be verified through official channels” (叶的说法却无从通过官方途径考究). The published text of Ye’s December 13, 1978 speech (which exists in reproduction via the 胡耀邦史料信息网, noted as excluded from the official collections 《三中全会以来》) was not accessible in this inventory; whether the death figures appear in that text is unresolved (see PRIMARY-SOURCE-TARGET 5).
Definition check: whatever Ye said, the “100 million” counts the persecuted/implicated (category v), the “20 million” purports to count deaths; and the component breakdown, if genuine, counts abnormal deaths. The circulated form has no stated methodology.

C5. The 400,000 “Beijing government announcement” claim — ATTRIBUTED, provenance unverified, LOW grade.

A Japan Foreign Policy Forum article (No. 34, Discuss Japan) states: “The Beijing government made an announcement in later years that 400,000 people had died during the Cultural Revolution. However, this number only refers to the confirmed deaths of party officials and intellectuals. The exact number of deaths is not known even today… but a figure of one million or more is most commonly cited” (attributed: the forum article). No announcement document, date, or venue is given in the article. Treat as an unverified secondary claim about an official figure; do not present it as established. PRIMARY-SOURCE-TARGET 6 records the provenance gap.

(d) Named scholar positions (attributed; methods stated; not merged)

Democide/aggregation methods:

R.J. Rummel (1991): ~7.73 million — democide analysis aggregating available estimates across Mao-era killing categories and periods (attributed: Rummel’s China’s Bloody Century figures as summarized in Song Yongyi 2011). Method: estimate-averaging across categories and repeated aggregation, the same procedure Rummel applied to other Communist states. Caveats: Rummel’s published period boundaries vary across editions (the figure is carried in Song’s encyclopedia entry as a Cultural Revolution figure); his procedure averages other analysts’ estimates rather than building from primary counts, so it inherits all of their definition differences. Treat as an attributed aggregation, not a measured count.
Ding Shu （丁抒， overseas Chinese scholar): ~2–3 million — “using diverse analyses” (attributed: Song Yongyi 2011, summarizing Ding’s disagreement with Rummel). The method detail is not stated in the secondary report; record the figure with method-unspecified. PROBABLE as an accurate report of Ding’s position; method UNRESOLVED.

Elite-historian judgment methods:

John K. Fairbank (Harvard, 1992): ~1 million — “rough estimate” (attributed: Song Yongyi 2011). Method not stated in the secondary report; a historian’s judgment figure, not a count. ATTRIBUTED.

County-annals quantitative methods (the Walder research program — the methodologically strongest published series, but period-bounded and definition-specific):

Andrew Walder and Su Yang （苏阳）: 750,000–1.5 million killed in rural China, “with roughly equal numbers permanently injured” — based on statistics drawn from 1,500 Chinese county annals (attributed: Song Yongyi 2011). Definition check: this counts killed in rural areas — a narrower category than “abnormal deaths” — and is rural-only, not a national total.
Su Yang and Andrew Walder (later round): 492,000–1,970,000 rural deaths — based on 1,530 county annals (attributed: the May 2026 中国瞭望 roundup). This is a different estimation round of the same research program with a different county-annals base; it is not interchangeable with the 750,000–1.5 million figure even though both come from the Walder/Su Yang program.
Andrew Walder (2,213 county/city annals): 1.1 million–1.6 million “abnormal deaths” in 1966–1971, and 22 million–30 million people politically persecuted (attributed: the May 2026 roundup; the 1966–1971 finding is also described in a 2016 CUHK article, “政治运动中的集体暴力”). Definition check: this is abnormal deaths (category iii) for 1966–1971 only — it does not cover 1972–1976 and does not claim to be a full-decade total. The 22–30 million is persecuted/implicated (category v), not deaths.
The database’s reported (not modeled) deaths: from the county-annals database itself, the recorded “abnormal deaths” total 232,799, of which 24,315 (10.44%) were recorded suicides and 31,282 (13.44%) recorded armed-conflict deaths (attributed: the 2016 CUHK article). The article stresses these are reported deaths, not actual deaths — the database’s modeling (the 1.1–1.6 million range) exists precisely because reported deaths understate the real number. Do not cite 232,799 as a death toll.
**Andrew Walder, China Under Mao (2015): the Cleansing of the Class Ranks campaign (1968–1970) alone killed 600,000–800,000** (attributed: Pacific Affairs review of the book, citing p. 277). The review further notes Walder’s case that “more than half of the deaths caused by the Cultural Revolution occurred amidst military-administered demobilization and campaigns such as the Cleansing of the Class Ranks.” Definition check: this is one campaign’s deaths — a component, not a total — and it reframes the CR death toll as concentrated in 1968–1971 state-administered campaigns rather than in 1966–1968 Red Guard factional violence.
**Su Yang （苏阳）, Collective Killings in Rural China during the Cultural Revolution (Cambridge University Press, 2011):** based on 1,530 county annals plus fieldwork and interviews across Guangxi, Guangdong, and Hubei. Findings on mechanism rather than a national total: among 65 accessible official Guangxi county documents, 43 counties report local massacres, 15 record death tolls over 1,000, and the average death toll was 526 among all counties reporting massacres (attributed: the book as summarized in the Guangxi Massacre record). Su proposes a “community model” of collective killings, challenging genocide/mass-killing frameworks. Provincial finding: in Guangxi and Guangdong, more than half of all counties experienced “collective killings,” and the county-annals timing data indicate the great majority of deaths resulted from killings rather than from struggle sessions or armed conflict (attributed: the May 2026 roundup). Provincial death ranking from the database: Guangxi, Guangdong, Inner Mongolia, and Shanghai show the highest violence; Shanghai has the highest per-county average (1,053 abnormal deaths) — e.g., in Baoshan County’s 1968 Cleansing of the Class Ranks, 334 of 2,339 investigated died abnormally; in Shanghai County, 385 of more than 5,600 investigated (attributed: the 2016 CUHK article).

Documentary/dissident-scholar methods:

Song Yongyi （宋永毅）, research librarian, Dickinson College (formerly Cal State LA): the most prolific documenter of CR mass killings. His 2011 “Chronology of Mass Killings during the Chinese Cultural Revolution (1966–1976)” (Online Encyclopedia of Mass Violence / Sciences Po, ISSN 1961-9898) states the CR “claimed the lives of several million people” and reports that “estimates by various scholars range from one-half to eight million” — presenting the field’s range, not his own single figure (attributed: the encyclopedia entry). His book 《文革大屠杀》 (Massacres during the Cultural Revolution) argues the CR massacres were actions of “state apparatuses” — “the direct slaughter by the regime towards its citizens” (attributed: the book as summarized in the Guangxi Massacre record). He edited the Chinese Cultural Revolution Database (Chinese University of Hong Kong Press) and, with Tan Hecheng, published classified Dao County massacre archives (《湖南道县及周边地区文革大屠杀机密档案》, 2017). Method note: Song’s distinctive contribution is documentary — classified investigation files and county-level massacre chronologies — rather than a single national estimate; his “several million” is a characterization of the field’s evidence, not a computed series. He also documents the understatement gap between published and classified county records (Lingshan: 8 published vs. 3,220 classified; Binyang: 37 published vs. 3,951 classified — see section a).
**Tan Hecheng （谭合成）, The Killing Wind (Oxford University Press, 2017): Dao County, Hunan, 1967: 4,519 people recorded as killed (4,193) or driven to suicide (326); across all counties and cities involved in the Dao County massacre region, 7,696 killed and 1,397 forced to commit suicide; 2,146 permanently injured and disabled (attributed: the book as summarized in the Daoxian Massacre record; Song Yongyi’s 2009 Sciences Po account of the Dao County massacre corroborates the event). Definition check:** one county cluster, 1967 — a documented component, not a national series.
Guangxi provincial totals: “official records show between 100,000 and 150,000 people were killed” in the Guangxi Cultural Revolution (attributed: secondary reporting citing official records). Song Yongyi documents that independent researchers in Guangxi counted 421 people eaten in cannibalism cases, with reports across 27 counties (attributed: Song as quoted in secondary reporting). Provincial figure, 1967–1968 — a component, not a national series.

Journalistic/reference positions:

Encyclopaedia Britannica: 500,000–2,000,000 — a reference-range characterization, not a computed series (attributed: Britannica’s Cultural Revolution overview).
**Evan Osnos, The New Yorker, May 6, 2016: “up to one and a half million had been executed or driven to suicide” (attributed: the article). Definition check:** executed-or-suicide — narrower than abnormal deaths; not interchangeable with the county-annals abnormal-death ranges.
Adrian Brown, Al Jazeera, May 13, 2016: “some historians say the number of dead could be as high as two million” (attributed: the report as quoted in Facts and Details).
Tufts “Mass Atrocity Endings” research site (archived 2019): working figure of 1.5 million fatalities; notes “estimates of fatalities during the Cultural Revolution vary wildly — ranging between one million and 20 million (though the former is probably closer to the true number than the latter)” (attributed: the site). The 1–20 million range reflects the full span from Hu Yaobang-adjacent figures to Ye Jianying-adjacent figures; the site’s own judgment favors the low end — PROBABLE as an accurate report of the site’s position, which is itself an editorial judgment, not evidence.
Facts and Details (aggregator): “No one knows exactly how many died, but estimates range from hundreds of thousands to 20 million” (attributed: the site). An aggregator’s range, not a series.
The Washington Post, July 17, 1994, additionally notes that the CR’s “purification of the class ranks” (1968–70) and “May 16 elements” (1968–69) campaigns — “least known in the West and least studied by scholars” at the time — were “now revealed to be among the bloodiest,” supporting a higher toll than the then-accepted half-million (attributed: the article). This anticipates Walder’s later quantification (600,000–800,000 for the Cleansing of the Class Ranks alone).

Not verified in this inventory — do not use:

**Jung Chang and Jon Halliday, Mao: The Unknown Story (2005): the book attributes “well over 70 million deaths in peacetime” to Mao’s rule overall (attributed: the book’s publicity and reviews). A CR-specific figure of “3 million” is sometimes circulated as theirs, but no CR-specific figure was verified against the book in this inventory** — UNSUPPORTED as a Chang/Halliday CR figure. Do not attribute the 3-million figure to them.
“Song Guangxun” as a CR-mortality scholar: correction recorded. Web and Chinese-language searches locate only Song Guangxun (宋广训, b. 1930, Baxian, Hebei), a woodcut print artist (attributed: Ashmolean Museum Eastern Art blog; chineseposters.net artist index). No mortality series by this name exists. The task brief’s “Song Guangxun” cannot be resolved to a CR-death researcher; the relevant “Song” in this field is Song Yongyi (宋永毅). Do not cite “Song Guangxun” as a mortality source.
Wang Nianyi’s 《大动乱的年代》 (1988, Henan People’s Press): a secondary critique (hosted at ccrhm.org) contains garbled OCR suggesting the book mentions 1,772 deaths in a single sentence, but the passage cannot be reliably parsed to determine which book or what definition is meant. Unresolvable from this source — do not cite a figure.
**MacFarquhar & Schoenhals, Mao’s Last Revolution (2006): suggested as a source in the prior ; no CR-specific mortality figure was independently verified against the book in this inventory. Lookup remains open — do not attribute a figure.**

(e) Contemporaneous dissident/reform voices (1966–1976)

No contemporaneous (1966–1976) dissident or reform voice located in this inventory advanced a quantified CR death toll; the democratic/reform responses of the period (Yu Luoke’s bloodline-theory critique, the Li Yizhe big-character poster, the April 5, 1976 Tiananmen movement, Deng Xiaoping’s 1975 rectification — all recorded in the project’s prior evidence) addressed the CR’s political mechanisms, not its mortality. The first attributed quantifications come post-1978 (Ye Jianying’s circulated 1978 remarks; Hu Yaobang’s quoted 1 million; the 1984 internal document). Record accordingly: do not back-project later figures as contemporaneous positions.

(f) Civil-society and international-institution positions

United Nations / FAO: No contemporaneous UN or FAO assessment of CR mortality has been located. The structural reason is the same as for the 1959–1961 famine (documented in the famine data record): the PRC did not hold China’s UN seat until October 1971, so no UN-body position from 1966–1971 is possible, and no retrospective UN quantification of CR deaths is known. Status: no documented UN/FAO position located — EVIDENCE GAP, not a finding of absence of concern.
International civil society (contemporaneous): no documented Amnesty International, Human Rights Watch, or comparable organizational position on CR mortality located; China was effectively closed to outside investigation during the CR years. Retrospective documentation is scholarly and journalistic (see section d).
Record accordingly: the international record on CR mortality is almost entirely retrospective scholarship, not contemporaneous institutional response. Do not present later scholarly or journalistic work as if it were a contemporaneous institutional position.

(g) Later-outcome evidence

The CR’s end (arrest of the “Gang of Four,” October 1976), the 1978–1981 reassessment (Ye Jianying’s December 1978 speech; the 1981 Resolution), and the post-1983 “遗留问题” (leftover problems) investigations — including the classified Guangxi investigation of the 1980s that Walder later used — produced the documentary base for both the official-adjacent figures (section c) and the county-annals research (section d). The investigations generated figures; they did not publish a national toll.
The 1981 Resolution’s “total negation” verdict and the rehabilitation of victims are documented facts (recorded in the project’s prior evidence); they carry no quantification.
Do not assert that any later figure (e.g., the 1984 internal 1.728 million) was produced by or endorsed through these campaigns unless a source states so.

Source conflicts

C-D2a — The two official-adjacent abnormal-death figures.

SOURCE A: 1984 Central Research Office internal material — 1.728 million abnormal deaths, ten years (attributed: May 2026 roundup).
SOURCE B: Book credited to the Party History Research Institute — 1,490,300 non-conflict deaths (attributed: Tufts site).
CONFLICT: Different totals, different stated scopes (all abnormal deaths vs. non-conflict deaths), different attributed documents. BEST AVAILABLE INTERPRETATION: two separate attributed claims; do not merge or average. CONFIDENCE: CONFLICTING.
WHAT WOULD RESOLVE IT: PRIMARY-SOURCE-TARGETs 1–3 (the underlying documents).

C-D2b — Ye Jianying’s 20 million vs. the field’s measured series.

SOURCE A: Ye’s circulated 1978 remarks — 20 million dead, 100+ million persecuted (attributed; unverifiable through official channels).
SOURCE B: Walder’s modeled abnormal deaths, 1966–1971 — 1.1–1.6 million (attributed; method stated).
CONFLICT: An order-of-magnitude-plus gap. BEST AVAILABLE INTERPRETATION: the two figures have incompatible definitions, periods, and provenance — Ye’s has no stated method or verifiable text; Walder’s is a modeled abnormal-death estimate for 1966–1971 only. They are non-interchangeable; neither is “the” figure. CONFIDENCE: CONFLICTING.
WHAT WOULD RESOLVE IT: Nothing on Ye’s side short of a verifiable primary text (PRIMARY-SOURCE-TARGET 5); on the scholarly side, only national-archive access.

C-D2c — Rural-killed vs. abnormal-death series within the Walder program.

SOURCE A: Walder & Su Yang — 750,000–1.5 million killed in rural China (1,500 annals).
SOURCE B: Walder — 1.1–1.6 million abnormal deaths, 1966–1971 (2,213 annals).
CONFLICT (apparent): overlapping but not identical ranges from the same research program. BEST AVAILABLE INTERPRETATION: different counted populations (killed vs. abnormal deaths), different annals bases, different period statements. Both are publishable as separate attributed series; neither supersedes the other. CONFIDENCE: PROBABLE.
WHAT WOULD RESOLVE IT: the program’s successive publications stating their own scope (obtainable from the authors’ bibliographies).

C-D2d — The armed-conflict death sub-figures.

SOURCE A: 《建国以来历史政治运动事实》 breakdown — 237,000 armed-conflict deaths (attributed).
SOURCE B: Yang Jisheng’s Ye-attributed breakdown — 123,700+ dead in 4,300+ large-scale armed-conflict incidents (attributed).
CONFLICT: Both claim official-adjacent provenance; they differ by roughly half. BEST AVAILABLE INTERPRETATION: two separate attributed claims; the Yang/Ye chain’s provenance is weaker (article → circulated disclosure → no verifiable text). CONFIDENCE: CONFLICTING.
WHAT WOULD RESOLVE IT: PRIMARY-SOURCE-TARGETs 2, 4, 5.

Freshness check (recorded 2026-09-24)

The May 15, 2026 scholarly roundup (中国瞭望, 万维读者网) re-circulates the established series — the 1984 internal 1.728-million figure, Ye Jianying’s 20-million claim, Walder’s 2,213-annals model (1.1–1.6 million, 1966–1971; 22–30 million persecuted), and Song Yongyi’s documentary findings — without material revision. Song Yongyi’s 2011 encyclopedia entry remains the stable published summary of the field’s 0.5–8 million range. No source located in this inventory has changed its published series since prior project research (prior research had assembled no series; the famine-file discipline was the only prior state). No previously researched figure required correction for staleness. The one freshness-relevant note: Walder’s program has continued to publish (e.g., Civil War in Guangxi, Stanford University Press, 2023) — the provincial component findings are active research, not a settled series.

Corrections caught in this inventory

“Song Guangxun” is not a CR-mortality researcher — the name resolves to a woodcut artist (宋广训, b. 1930). Recorded above; do not cite as a mortality source. The brief’s intended “Song” is Song Yongyi (宋永毅).
The Jung Chang / Jon Halliday “3 million” CR figure is unverified — the book’s verified claim is the 70-million-plus total for Mao’s rule; no CR-specific figure was located against the book. Do not attribute.
Wang Nianyi’s 1,772 figure is unresolvable — garbled secondary OCR; no usable definition or attribution. Do not cite.
The 1989 death toll is excluded by design — the prior explicitly excluded it; MUSE15 owns the 1989 archive. No 1989 figure appears in this file.
Ye Jianying’s component breakdown (summing to ~2.12 million) does not reconcile with the circulated “20 million” headline — recorded as an internal inconsistency, not resolved.

Terminal state

COMPLETE-EVIDENCE-FAMILY. The inventory is complete as a figure-responsibility file: every located published series is recorded with definition, method, provenance, and uncertainty; no series is merged, averaged, or selected. PRIMARY-SOURCE-TARGETs below record the missing primary texts; they do not block the inventory’s completeness as an attribution map. Unresolved items: (i) the underlying texts for all official-adjacent figures (PST 1–6); (ii) a verified CR-specific figure in MacFarquhar & Schoenhals (open lookup); (iii) the aisixiang reproduction of Ye Jianying’s December 13, 1978 speech was inaccessible in this pass (fetch failed; do not treat as absent).

Source list (as located)

Song Yongyi, “Chronology of Mass Killings during the Chinese Cultural Revolution (1966–1976),” Online Encyclopedia of Mass Violence / Sciences Po, August 25, 2011 (ISSN 1961-9898; PDF mirror at netadvisor.org).
Andrew Walder, China Under Mao: A Revolution Derailed (Harvard University Press, 2015) — via Pacific Affairs review (Matthew D. Johnson).
Su Yang, Collective Killings in Rural China during the Cultural Revolution (Cambridge University Press, 2011) — via frontmatter and secondary summaries.
“政治运动中的集体暴力” (2016), CUHK 《二十一世纪》 (cuhk.edu.hk PDF) — Walder county-annals database findings: 2,213 annals, 1966–1971, 1.1–1.6 million abnormal deaths, 22–30 million persecuted, 232,799 reported deaths.
学界详解被掩盖的文革真相—”集体杀戮” (中国瞭望, 万维读者网, May 15, 2026) — Walder/Su Yang rounds (1,530 annals: 492,000–1,970,000 rural deaths), 1984 internal 1.728M figure, Ye Jianying 20M claim.
The Washington Post, “How Many Died? New Evidence Suggests Far Higher Numbers for the Victims of Mao Zedong’s Era,” July 17, 1994 (archived) — Hu Yaobang 1 million; Harry Harding ~1 million; half-million scholarly consensus then; bloodiest-campaigns note.
Tufts University “Mass Atrocity Endings: China — the Cultural Revolution” (archived October 8, 2019) — 1.5M working figure; 1–20M range; 1,490,300 Party-History-credited figure; phase breakdowns.
Facts and Details, “Cultural Revolution: Death Toll, Fighting and Mass Killing” — Song-sourced field range 0.5–8M; Rummel 7.73M; Fairbank ~1M; Ding Shu 2–3M; Walder & Su Yang 750,000–1.5M.
Evan Osnos, The New Yorker, May 6, 2016 — up to 1.5M executed or driven to suicide.
Al Jazeera (Adrian Brown), May 13, 2016, via Facts and Details — up to 2M per some historians.
Japan Foreign Policy Forum No. 34 (Discuss Japan) — 400,000 Beijing-announcement claim (unverified provenance).
Tan Hecheng, The Killing Wind (Oxford University Press, 2017) — Dao County figures, via the Daoxian Massacre record; Song Yongyi, “The Dao County Massacre of 1967” (Sciences Po, 2009).
杨继绳, “道路·理论·制度——我对文化大革命的思考,” 《记忆》 no. 104, November 30, 2013 — Ye-attributed component figures (via secondary reporting).
Ye Jianying, speech at the closing session of the Central Work Conference, December 13, 1978 — reproduced text noted at aisixiang.com/data/23237.html (via 胡耀邦史料信息网编者按); fetch failed in this pass; 20M/100M figures’ presence in the text UNRESOLVED.
Encyclopaedia Britannica, Cultural Revolution overview — 500,000–2,000,000 range.
Guangxi Massacre record (academic-studies section) — Su Yang’s 65-county-document findings; Song Yongyi’s published-vs-classified annals comparison; Guangxi 100,000–150,000 provincial total.
GETTR post (social media, low grade) — circulating Ye-attributed component list (420万 detained, 172.8万 suicides, etc.); recorded only as evidence of circulation, not as a source.

PRIMARY-SOURCE-TARGETs (piped to the shared register; see claims log)