What the claim's own words commit it to
A claim has to be judged on the words it uses, and this page has not previously said what those words commit it to. Doing so first prevents the commonest error in this material, which is to demote a claim for the failure of numbers it never asserted.
"Massacred" and "murdering" are qualitative terms. They are satisfied by a single episode of the killing of unarmed people in numbers, and three such episodes are established on this page's own evidence: Parkal on 2 September 1947, where police and Razakars fired on a crowd of more than 1,500 that had gathered to hoist the Indian flag, killing 22 and wounding over 150 23; Gorta in May 1948, where a lawyers' committee inspecting three sites found human skulls and bones and concluded that over 200 had been murdered and their bodies burnt 4; and Bhairanpally on 27 August 1948, where an eyewitness described villagers being lined up in rows and shot at close range 22. Whether the state-wide toll was one thousand or ten thousand does not bear on whether these were massacres. They were.
"Terrorised", "raping and looting", "under the Nizam's protection", "to keep the state out of India" and "stopped only by the Indian Army's police action" are each supported by the material set out below and in the concessions above. The claim as quoted therefore stands. Two things remain genuinely contested, and this page is about them: whether religion was the operative criterion by which the Razakars selected their targets, or whether it was political allegiance in a state whose political divisions ran largely along confessional lines; and every state-wide number, on either side, that has been attached to the claim since.
That second question is where almost all the public argument actually happens, and it is worth being precise about its relationship to the claim. The figures examined below — 880 murdered, roughly 980 dead, 200 at Gorta, and against them 27,000 to 40,000 or 2,000 Muslims killed in the reprisals — are attached downstream of the claim by its carriers and its opponents. They are not part of the proposition quoted above, and their collapse is not the claim's collapse. It is a separate finding, and it applies to both columns.
What can actually be established, and by which document
Three kinds of evidence carry the claim's core, and they are of very different weight. The strongest is judicial, and it belongs to the trial courts rather than to the Supreme Court. A Special Tribunal convicted Qasim Razvi in the Bibinagar dacoity case on 11 September 1950, sentencing him to seven years' rigorous imprisonment under section 330 of the Hyderabad Code with concurrent two-year terms on three further counts; the Hyderabad High Court affirmed the convictions on 13 April 1951 apart from the charge under section 123 9. That is testimony taken on oath, cross-examined, and reviewed on appeal. Nothing else on either side of this page has that standing.
What the Supreme Court did in January 1953 was narrower, and an earlier version of this page overstated it. Hearing Razvi's constitutional challenge to the Special Tribunal, the Court recited the case's history — describing it as one "in which, according to the prosecution, a serious raid was committed by a party of armed Razakars in village Bibinagar, about 21 miles from the city of Hyderabad, attended with robbery, looting, arson, assault and other violent acts on the afternoon of 10th January, 1948" — and recorded that after the raid "the police administration of the State of Hyderabad was at that time under the complete control of the Razakars and they tried to minimise the gravity of the occurrence as far as possible" 9. The qualifier "according to the prosecution" was dropped from that quotation on this page and has been restored. The Court was deciding preliminary points under Articles 14 and 21 and on the validity of the Tribunal, and stated that "whether the appeal will be heard further on its merits will depend upon the decision we arrive at in the present hearing" 9. It made no findings of fact about the raid. The fact-finding is the Tribunal's and the High Court's, and that is where the weight sits.
The second kind is the contemporaneous official record of an interested party. K. M. Munshi was the Government of India's Agent-General in Hyderabad from January to September 1948 — that is, the man whose job was to build Delhi's case. His memoir is therefore partisan by position and specific by access. He counted, from October 1947 to 3 April 1948, "no less than 260 incidents in which the Razakars acted with savage brutality" 3. He set out the Razakar oath: "In the name of Allah, I do hereby promise to fight to the last to maintain the supremacy of the Muslim power in the Deccan" 5. He wrote to the Hyderabad Prime Minister on 28 February 1948 that "the Razakars are a private Army operating in the State with the active aid and cooperation of the present Government" 6. His diary for 22 August 1948 reads: "The editor of the Imroze, Sheobullah [Shoebullah], has been shot down and his hands cut off. His brother-in-law is also seriously wounded. The Ittehads are liquidating their enemies in true Fascist style" 8.
The third kind is village testimony, gathered decades later and preserved locally. It is the least controllable and the most vivid. At Bhairanpally in Maddur mandal — a village in what is now Siddipet district — an eyewitness told The Hindu in 2017 that the Razakars "made the villagers stand in three, ten and even twenty in a row and fired from close range into the chest of the one standing in front", after three earlier attempts on the village had been driven off with slings 22. Testimony of this kind cannot be aggregated into a total, and this page does not try. It establishes that specific things happened in specific places; it does not establish how many.
Independent third-party observation exists for the surrounding conflict but not for the village raids. The US Embassy in New Delhi reported to Washington in July 1948 that Razakars had ambushed Indian troops crossing a strip of Hyderabad territory, killing five, and that in the British and American view "the situation has deteriorated" 12. That is a real, dated, non-Indian corroboration of Razakar armed action. Outside observation of the interior districts has not been located, but it is now known to have existed: on 20 September 1948 the same Chargé cabled Washington that General Bucher had told him "American missionaries in Bidar Hyderabad state had some harrowing stories to tell" and that Bucher "had sent two American press correspondents to Bidar to obtain story" 54. Bidar is the district containing Gorta. What those missionaries and correspondents reported has not been found, and this page's earlier statement that no comparable outside observation exists was a claim about the search rather than about the record. It is withdrawn and replaced by this: the observation existed, it has not been located, and locating it is the most promising single item in the open questions below.
The clause about the Nizam, and how far this page can test it
The claim says the militia operated "under the Nizam's protection". That clause has been conceded on this page rather than examined, and it deserves the same treatment as the rest.
Three things in the opened sources bear on it directly. The Special Tribunal's finding, recited by the Supreme Court, is about the state's own police and not only about the militia: after Bibinagar there was "neither any proper police investigation nor any serious attempt to arrest the culprits or bring them to trial", because the police administration "was at that time under the complete control of the Razakars" 9. Munshi's letter of 28 February 1948 puts the charge to the Hyderabad government in writing and on the record: the Razakars are "a private Army operating in the State with the active aid and cooperation of the present Government", with a volunteer force he there puts at 150,000 6. And the militia was neither disbanded nor disarmed at any point between the Standstill Agreement of October 1947 and the Army's entry in September 1948; on Munshi's own figures its rolls grew across that period from more than thirty thousand to over a hundred thousand 53. At Parkal, in September 1947, it was the Nizam's police firing alongside the Razakars 23.
Against that stands the Nizam's own account, which is quoted here with its interest stated rather than at face value. Broadcasting on 23 September 1948 — after his surrender, and while his position under the new dispensation was being settled — he said that a "quasi-military organisation" had "by duress compelled the Nawab and other trusted ministers to resign and forced the Laik Ali Ministry on me" 28. That is a defeated ruler's account of his own responsibility. It reaches this page at tier C through Wikipedia rather than from the printed source; its dating of the episode to November 1947 conflicts with the October 1947 dating usual in secondary accounts; and Munshi records having drafted and approved the Nizam's earlier surrender message. It is good evidence that the militia had captured the ministry. It is not independent evidence that the ruler was a bystander, and this page does not treat it as such.
What this page cannot do is test the Nizam's own acts. The matters usually cited on that question — the arms procurement associated with Sidney Cotton's flights, the loan to Pakistan, the appeal to the UN Security Council, and the precise terms of the Standstill Agreement — were not sourced for this page, and nothing above rests on them. They are listed in the open questions. On what was opened, "under the Nizam's protection" is established as to his government and his police, and untested as to the ruler personally.
The figures for Hindu deaths, and where each one actually comes from
Four numbers do most of the circulating, and each has thinner provenance than its repetition implies. None of them is the claim; all of them are attached to it.
Gorta, in Bidar district, is the largest single figure, and the source is better than an earlier version of this page allowed. Munshi prints in full the report of a Lawyers' Vigilance Committee which reached the village after Chincholi — on the archive scan's reading, "the 17th of May, 1948", though the OCR is ambiguous between the 11th and the 17th, and the committee's description of decomposing animal carcasses and of cattle "breathing their last for want of water and fodder" fits a delay of about a week rather than of a day. The previous assertion that the committee arrived "the day after it was burned", and the inference built on it, are withdrawn. Munshi dates the burning to 10 May; the Government of India's own commemorative account dates the Gorta massacre to 9 May, and this page cannot resolve the discrepancy 423.
The committee found the village deserted — "Nobody was there except one wounded person and three aged women" — and recorded that "Till our visit to the village no Panchnama had been prepared by the police either of (the) human bones or (the) dead bodies or the gutted houses", the sub-inspector stating "that he was unable to identify the bones and the dead bodies burnt". The figure of 200 first enters through a survivor who had returned looking for his cattle and who "appeared to be utterly frightened": "More than 200 persons have been mercilessly murdered and the dead bodies have been burnt in cow-dung cakes and heaps of fodder". An earlier version of this page stopped the quotation there, which reversed the document's weight, and the correction matters more than the original point. The passage continues: the committee went with the sub-inspector into the house "where the dead bodies had been gathered together and burnt", where "Human skulls and bones were found everywhere" and "Their panchnama was prepared in our presence"; then to heaps of kadbi in the field of one Narayan Rao Mektedar, and to two further heaps in the field of one Channappa, finding human bones and half-burnt remains at both. Its conclusion is in its own voice: "It is evident from the Inspection of these places that over 200 persons (had been) murdered. After the preparation of (the) panchanama the Sub-Inspector collected the bones and sent them for medical examination" 4.
The honest description is therefore this. Two hundred-plus at Gorta is a partisan lawyers' committee's on-site estimate, made from burnt and scattered remains at three separate locations, in the presence of a police officer, followed by the collection of bones for medical examination. It is not an enumeration of named dead, no medical report has been produced, and the committee was an interested body reporting in an interested memoir. But it is an inference from physical inspection, not merely a distraught survivor's figure relayed at second hand, and this page had it wrong.
Against it stands the village's own memorial stone, which the source describes as "engraved with the names of 23 'martyrs' that is all but hidden away behind a stairway of the Panchayat building" 24. That figure now gets the four questions this page puts to every other: who compiled it, when, from what, and what its method would omit. None of them can be answered from the source, which reports only that the stone exists and is neglected. A memorial list is not a body count. At a site whose population of 2,500 had fled, whose dead were burnt in fodder heaps, and whose remains the police could not identify, a short list of named dead is what one would expect whether the true figure is two dozen or two hundred. Gorta's toll is indeterminate between those poles. It is not a case of a number drifting upward, and the claim that it was has been removed from this page.
Bhairanpally's figures vary by about a third among the accounts that name a source, and by more if the round number is included. The Hindu's 2017 eyewitness report gives 96 for 27 August 1948, datelined Maddur (Siddipet dt.) 22; The Hindu's 2023 opinion column gives 88 and places the village in Jangaon 21; the Government of India's commemorative site gives "about 118" 23; a round 200 also circulates in popular accounts, for which no named source could be traced and which is therefore not used here. Part of that spread may not be disagreement at all: this page has not been able to establish whether the two Hindu reports refer to the same village, or whether the figures cover the same scope — the dead of 27 August alone, or the dead across the several assaults on the village since June. Until that is resolved the spread is evidence that nobody has reconciled the accounts, not evidence of inflation.
Parkal, on 2 September 1947, is the most stable and the earliest: 22 killed and over 150 wounded when the police and the Razakars fired on a crowd of more than 1,500 that had come to hoist the Indian flag against a prohibitory order 23. It falls inside the 1947–48 period the claim itself states, and a previous version of this page wrongly treated it as a mismatch with the claim; that sentence has been deleted. What Parkal actually adds is that the Nizam's police were firing alongside the militia from the beginning, which supports the state-complicity clause rather than troubling it.
The two state-wide totals in circulation are the weakest material here, and their provenance is now traceable. "880 murdered and about 1,125 women raped before August 1948" is routinely presented as a contemporaneous tally. Its traceable source is an opinion column in The Hindu of 15 September 2023, which says only: "According to one report, they were responsible for the murder of 880 people and the rape of about 1,125 women before August 1948" 21. The report is not named, and no earlier appearance of the pairing could be found. The figure of roughly 980 deaths, likewise presented as derived from the 1948 White Paper, is an explicit reconstruction published in The Wire in October 2023: its author writes that the White Paper "describes 53 events of Razakar atrocities in four districts of Marathwada", that "as per author's assessment, in the 53 events around 200 deaths occurred, probably 20 rapes were committed... and around 50 incidents of loot happened", and that "[i]f one extends the same proportion to the 260 events, probable deaths would be around 980" 20. That is honest arithmetic, honestly labelled, resting on the untested assumption that the 53 detailed incidents were representative. It is not a count and its author does not say it is. Both of those 2023 pieces were written in explicit opposition to the then-forthcoming film, one of them titled "How the Movie 'Razakar' Sets Out to Distort History", and their purpose is disclosed here because it is relevant to how they are read. What they establish is provenance, and provenance is checkable independently of motive.
One number on this side of the ledger has been getting a pass, including from this page, and the same rule has to reach it. Munshi's "260 incidents in which the Razakars acted with savage brutality" is a bare assertion in the memoir of an interested official 3. He publishes no schedule, states no criterion for what counts as an incident, and does not say who compiled it or how. It should not be described as corroborated: the White Paper's schedule of 53 detailed Marathwada incidents corroborates that a schedule existed for four districts, not that the state-wide figure was 260 20. On this page's own standard it is a total without a method, and the consequence runs downstream — Ravibabu's 980 takes 260 as its denominator and inherits the defect entire.
Finally, the Government of India's own commemorative account gives a figure rarely quoted beside the others. Produced by the National Council of Science Museums, an autonomous body under the Ministry of Culture, with the Salar Jung Museum, it states: "Altogether, over 2000 people were killed and several hundred women were raped by the Razakars and Nizam's Police during this period" 23. Three things must be said before it is used for anything. It is an open-ended floor — "over 2000" — and not a point estimate; it covers the Razakars and the Nizam's police together, not the militia alone; and "this period" is nowhere defined. It carries no source of any kind. It cannot therefore be set against the higher circulating totals as a ceiling, and this page withdraws the suggestion that it "sits below" them. What it shows is only that the state's own commemorative narrative has not settled on a number either.
The figures for Muslim deaths, held to exactly the same standard
The rule this page follows is that a figure is worth what its method is worth, and the rule cannot change when the victims change. What follows applies to the reprisal killings of September–October 1948 exactly the tests applied above: is the method disclosed, is the author interested, is the provenance traceable, does the quotation survive being read to the end. Applying the same standard to both columns is a statement about evidence. It is not a claim that the two columns are equal in size, in kind, or in culpability, and nothing below should be read as one.
The Sunderlal delegation is the fullest document and is regularly described as a commission of inquiry. It expressly denied being one: "We clarified our position, whenever opportunity presented itself saying that ours was not a Commission of investigation or Inquiry into events proceeding or following the police action and that ours was merely a goodwill mission charged with the task of restoring better communal relations" 1. Its field-work was substantial for a three-week tour — nine of sixteen districts, seven district headquarters, 21 towns, 23 villages, over 500 people interviewed from a further 109 villages not visited, 31 public meetings and 27 private gatherings 1. Its central sentence is: "We can say at a very conservative estimate that in the whole state at least 27 thousand to 40 thousand people lost their lives during and after the police action" 1. Three things must be said about it. The sentence says "people", not "Muslims"; the surrounding text makes plain that Muslims were meant — "The sufferers were Muslims who formed a hopeless minority in rural areas" — but the number as written is not confessionally specified 1. The report discloses no method for the range whatever: no enumeration, no district returns, no basis, and its district sub-totals ("not less, if not more than 18,000" in four districts, "at least 5 thousand" in four more) are asserted rather than shown 1. And the mission's own interest has to be read the way the White Paper's interest is read below: it was a goodwill mission whose declared purpose was communal pacification and the reassurance of Muslims, a purpose that would select for accounts of Muslim suffering and against systematic enumeration, and one whose commissioning Sardar Patel afterwards denied 2. Interest analysis is a rule here, not a tool used on one archive.
Against it stands an official estimate an order of magnitude lower, and the interest attaching to it is the largest on this page. Major-General J. N. Chaudhuri was the Military Governor of Hyderabad — that is, the commander of the army which the Sunderlal delegation found had participated in looting and worse, estimating the number of dead produced under his own administration. He is recorded as having admitted that "[i]mmediately after the military occupation…a large majority of Hindus in all districts of Hyderabad thought that Hindu Raj had come into being" and that "[a]s a result of this retaliatory action by Hindus, some Muslims in certain districts suffered", and as having concluded: "It is estimated that about 2,000 Muslims may have been massacred" 19. Three disclosures are owed. He discloses no method, exactly as the Sunderlal delegation discloses none. The document is now identified: Chaudhuri's report to the Ministry of States of November 1948, quoted at length by Purushotham, whose fuller text the page previously lacked — "The districts particularly affected were Bidar, Osmanabad, Gulbarga, Nalgonda & Warangal. Though accurate figures cannot be obtained yet of what the casualties were, it is estimated that about 2,000 Muslims may have been massacred. Almost all of this happened in the smaller villages off the main communications" 40. The Scroll.in intermediary is no longer load-bearing for it. And the same witness has to be weighed once, not twice: Chaudhuri is also the source of the finding, six months later, that the pre-invasion Razakar atrocity figures were "deliberately inflated for propaganda purposes or were unreliable" 39. This page will not take the second and discount the first, or the reverse. What distinguishes them is direction of interest: the inflation finding ran against his position as the officer who had to justify the annexation, and the casualty estimate ran with his position as the officer commanding the forces implicated. That is a reason to weigh the first more heavily than the second. It is not a reason to treat either as a measurement. Purushotham also records, from the same files, that "Confidentially, however, there was an official consensus that the fact of widespread violence was irrefutable" 40.
Chaudhuri's private formula for the whole episode — "in political physics, Razakar action and Hindu reaction have been almost equal and opposite" — was the working assumption on which the new government designed its amnesty policy 10. It is quoted here as what it is: the perpetrating administration's own claim of proportionality, made by the officer commanding. It is not this page's assumption and nothing in this page's argument rests on it. Nehru, for his part, wrote in a note of 14 November 1948 of "the massacre of possibly some thousands of Muslims by Hindus", adding: "This information is contrary to what I had believed and I should like it to be verified" 2.
A secret UK High Commission report of late November 1948, seen by the US Embassy, recorded that "[i]nitial entry Indian troops was followed by pillage, arson and murder by Indian forces, especially Sikhs", and that "[d]isciplinary action soon brought troops under control" 11. That is valuable and hostile-to-interest evidence about the conduct of the entry, and it is kept for that. It is not evidence about scale, and this page previously used it as though it were. The same telegram states the report was "based on personal observation and conversations with top Indian and Hyderabad officials and UK businessmen", reports that the military government "is popular with public" and praises Chaudhuri's impartial handling of disruptive elements. It is a November assessment of order and administration drawn substantially from the incoming administration, not a survey of village reprisals across nine districts over the preceding weeks. Its silence about the districts is not a bound on the district toll, and the inference has been withdrawn.
The higher figures need the same care, in both directions. Wilfred Cantwell Smith's 1950 article reported the 50,000 as rumour and labelled it so: "It is widely held that the figure mentioned therein for the number of Muslims massacred is 50,000. Other estimates by responsible observers run as high as 200,000" 2. That much is honest hearsay about an unpublished document, and it has since hardened into a citation. But Smith did not say only that, and this page previously implied he did. Having visited Hyderabad in 1949 he offered his own assessment — "Thousands upon thousands were slaughtered; many hundreds of thousands uprooted" — and a proportional estimate of his own: "Of the total Muslim community in Hyderabad, it would seem that somewhere between one in ten and one in five of the adult males may have lost their lives in those few days" 2. The criticism that actually holds is not that Smith trafficked in rumour, but that he discloses no basis for the proportion, and that Margrit Pernau's near-identical figure — that the reprisals "cost the lives of one-tenth to one-fifth of the male Muslim population" — may not be independent of him 2.
There is one body of official counting that bears on scale and that this page had omitted while citing the same article for Chaudhuri's figure. The first of these is not an enumeration at all, and the page's own standard requires the working to be shown. Purushotham gives the underlying arithmetic: "The Hyderabad government gave out a fixed quota of 4,500 'widow pensions to the women whose husbands had been killed.' Of these, 'about 75 percent' were given to Muslims 'who suffered immediately after the Police Action on account of the reaction.' According to government figures, then, at least 3,375 Muslims were killed during and after the Police Action" 40. The primary is a fixed pension quota of 4,500, not a count of widows; 3,375 is a derived floor bounded by that quota, which makes it weaker as a measurement and stronger as a floor than the page has been saying. The second is a genuine administrative count: the government-appointed Rehabilitation Committee reported in 1951 that in 47 villages of Osmanabad district alone there were "approximately 2,500" widows and 6,000 orphans, and framed its recommendation on "approximately 10,000 widows and 25,000" orphans for that district 19. and the government-appointed Rehabilitation Committee's finding in 1951 that in 47 villages of Osmanabad district alone there were "approximately 2,500" widows and 6,000 orphans, with its recommendation framed on an estimate of "approximately 10,000 widows and 25,000" orphans for that district 19. These are not death tolls and cannot honestly be converted into one. But they are administrative counts with registers behind them and a bureaucratic purpose, compiled by the government that succeeded the violence — which is the kind of methodical counting this page had said never happened. They bracket the scale from below in a way that neither circulating total does. The correction strengthens the page's thesis rather than weakening it: the material that could have tested the range was collected for other purposes and never aggregated.
So the same test returns the same answer in both columns, and that answer is not a symmetry of events. For Razakar victims: many specific, dated, located incidents, judicially tested in at least one case, and no defensible state-wide total. For Muslim victims of the reprisals: two official estimates twenty-fold apart, neither disclosing a method, one of them the estimate of the officer commanding the forces implicated and reaching this page only at second hand, alongside widow and orphan registers that indicate a large event without measuring it. Sherman, who has worked the files, states the position flatly: "It is difficult to elaborate the scale and nature of the violence that occurred. The information available to the historian is incomplete" 16. In the same article this page cites eight times she also states a figure, which this page had not quoted: "Conservative estimates suggest that 50,000 Muslims were killed. Others claim several hundred thousand died" 46. Her own footnotes cut both ways and are quoted here for that reason: the 50,000 traces to W. C. Smith, p. 46, "a report commissioned but never published by the Indian Government", and against the higher claim she records that "the version of their report cited in this article does not calculate or estimate the total number killed" — a doubt now resolved by Purushotham, who read the original and the fourteen pages of Confidential Notes at the Nehru Memorial Museum and Library in 2010 and concluded that "it is impossible to either confirm or refute the figures presented in the Sunderlal Report" 41. Quoting a source for its agnosticism while omitting its estimate is the kind of selection this page audits in others. Anyone offering a single comparative number for the two halves of 1948 in Hyderabad is not working from the record.
Displacement: the evidence category this page did not open
This page argues that no state-wide magnitude is recoverable, and it reaches that conclusion by examining enumerations of the dead. That is narrower than the thesis sounds, and the narrowing should be visible rather than buried.
There is a second category of evidence that bears on the reach and intensity of the village raids without depending on counting corpses: the movement of people out of the Hyderabad districts during 1947–48 into Sholapur, Gadwal, Chanda and the other bordering Indian districts, and the Hyderabad State Congress camps that received them. Camp registers, relief expenditure, and border-district administrative correspondence are the ordinary sources for that, and they are not testimonial and not compiled by either side's partisans for the purpose of establishing a toll. None of them was opened for this page.
What is visible in the sources that were opened is only trace evidence. The Sunderlal delegation records abduction of women out of the state "to Indian towns such as Sholapur and Nagpur", and armed men entering from Sholapur and other Indian towns during the reprisals — traffic in both directions across the same border 1. The Gorta committee found the village empty because survivors had taken shelter in adjoining villages 4. That is not a displacement record and cannot be made into one here.
The correct statement of this page's finding is therefore: no death toll for either community is recoverable from the material that has been examined. That is not the same as saying nothing about scale is recoverable at all, and the difference is a research programme rather than a rhetorical hedge. Until the camp and border-district records are examined, the thesis is a thesis about death tolls.
Whom the Razakars killed, and what that does and does not settle
The claim's word is "Hindus". The record shows a target set defined by political allegiance that was overwhelmingly, but not exclusively, Hindu — and the two facts have to be held together rather than traded off.
The clearest single test is Shoebullah Khan, editor of the Urdu daily Imroze, who attacked Razvi in print and was attacked in Hyderabad on the night of 22 August 1948; his hand was severed and he was shot 8. Sherman's account is that the cases the new administration chose to bring to court were designed to minimise ethnic tensions; that Razvi and four associates were prosecuted for the murder of "a fellow Muslim"; and that the Bibinagar dacoity case was preferred alongside it because such cases were believed likelier to "inspire in the public feelings of pure abhorrence or deep nationalism, rather than enmity between Hindus and Muslims" 10. Those are two observations, and an earlier version of this page welded them into a single causal claim about the Shoebullah prosecution that Sherman does not make; the rationale attaches most directly to Bibinagar and at most to the choice of cases generally. The Special Tribunal convicted in the Shoebullah case and the High Court acquitted 10. Sherman's summary of the summer of 1948 stands on its own: the Razakars "targeted not only Hindus, but Muslims whose loyalty was in doubt" 10.
The village pattern points the same way without dissolving the communal fact. Sundarayya, the participant historian of the Telangana insurgency the Razakars were deployed against, insists that resistance was not organised on religious lines: "just as in the earlier period, the people both Hindus and Muslims, had fought in the villages against landlords unitedly, so also they fought shoulder to shoulder against the Razakars without falling prey to religious fanaticism" 13. Serish Nanisetti reports two data to the same effect: he cites the historian Inukonda Thirumali's list of 34 well-known peasant struggles between 1944 and 1946, of which only two were against Muslims, and quotes Amit Kumar Gupta that both the Razakar raids on behalf of the mostly Hindu landlords and the Sangham resistance to them were "surprisingly free from all religious articulations" 21.
Both of those have to be bounded before they are used, and this page did not bound them. Thirumali's list counts whom peasant struggles targeted — landlords — and cannot speak to whom the Razakars attacked; it is a datum about the insurgency, not about the militia's target set. Gupta's window, 1944 to 1946, closes before the campaign the claim is about: Razvi's rise and the Razakar expansion postdate it, and Munshi locates the worst of the violence in "the intensive stage which began about the end of February" 1948 3. Neither Thirumali nor Gupta was consulted directly for this page; both reach it through an opinion column written against a film. The reframing they support is real but its evidentiary base here is thinner than the page previously implied.
And the reframing can be pushed too far, which this page declines to do. The oath Munshi records is confessional in terms 5. The recruiting slogans Sundarayya records are confessional in terms, and he is a hostile witness on that point 13. Sundarayya also concedes that most urban Muslims backed the organisation 14. The dead at Gorta, Parkal and Bhairanpally were Hindu cultivators. To say the violence was agrarian and counter-insurgent is not to say the victims were not Hindus; to say the victims were Hindus is not by itself to establish that religion was the operative criterion of selection. The claim collapses that distinction. The historiography has not, and the disagreement is live rather than settled. The Government of India's own pamphlet, for what it is worth, counted the Communists as having murdered 2,000 people between 15 August 1947 and 13 September 1948 — the same order of magnitude the state's commemorative site now gives for the Razakars, and equally unsourced 10.
What the claim leaves out, and what the sources it leans on actually say
The document most often deployed to complicate the claim is the Sunderlal delegation's report, and the document most often deployed to support it is Munshi. Both are more even-handed than their partisans allow, and quoting them to the end costs neither side anything — which is the lesson of the Gorta passage above and of the Smith passage before it.
The delegation's account of who did the killing after 13 September names three agents, not one: local non-Muslims who had suffered under the Razakars, infiltrators from across the border, and — the clause usually dropped — "a number of armed and trained men belonging to a well known Hindu communal organisation from Sholapur and other Indian towns as also some local and outside communists participated in these riots and in some cases actually led the rioters" 1. Its finding on the Indian Army is graver than the paraphrases suggest — "we had absolutely unimpeachable evidence to the effect that there were instances in which men belonging to the Indian Army and also to the local police took part in looting and even other crimes" — and it is immediately qualified: "Lest we might be understood to imply a slur on the Indian army we hasten to record our considered opinion that the Indian Army and its officers in Hyderabad generally maintained a high standard of discipline and sense of duty" 1. It also records Hindus dying to protect Muslim neighbours, and Hindu weavers defending Muslim weavers at heavy cost 1.
Sundarayya, from the other end of the political spectrum, records the same protective behaviour in the Telangana districts, and his sentence carries more than shelter: "The Hindu people in those villages rescued such ordinary Muslim people to the extent possible, gave shelter to them in their houses and rescued thousands of Muslim families from the campaign of rape and murder indulged in by the Union armies" 15. An earlier version of this page trimmed the final clause with an ellipsis while quoting the Sunderlal delegation's milder finding complete with its exculpatory qualification — an asymmetry on the page's most sensitive point, and it is corrected here. A Communist participant-historian, hostile to the Nizam and to the Razakars, attributes rape and murder to the Indian Army directly and in stronger terms than the delegation did. That statement sits alongside the delegation's finding rather than inside it, and both belong on the page.
The suppression story also needs correcting. Sardar Patel wrote to Qazi Abdul Ghaffar on 4 January 1949 disputing that any such mission had been sent — "At least I am not aware of any such mission having been entrusted to you by the Government of India" — and complaining that "there is nothing in it about the extent and consequences of Razakar atrocities" 2. The report was never officially released. But it was not sealed for sixty-five years: Omar Khalidi printed fragments in 1988, which Noorani judged unreliable; Noorani obtained the full text from the Nehru Memorial Museum and Library and Frontline published the covering report verbatim on 3 March 2001; the appended documentation followed with Noorani's book in 2013–14 21.
Nor is it true that nobody answered for the reprisals. Sherman's archival account records that alongside over 13,000 Muslims detained as associated with the Razakars and the Nizam's irregulars, "[a]nother several thousand Hindus were jailed after having been implicated in the post-police action reprisals against Muslims" — and that in the end the political settlement released almost everyone on both sides, so that "only a handful of symbolic Razakars were punished with formal imprisonment" 10. Razvi himself served until 1957. The honest statement is that the judicial response was thin in both directions, for the same administrative and political reasons, and that this is a large part of why no evidentiary record of the violence was ever produced.
How the claim travels, and what rides on the word chosen
The claim is contemporaneous with the events, not retrospective, and its earliest systematic form is a brief for intervention. The Government of India's White Paper of August 1948 assembled the case against the Nizam's government before the Army moved, and it has now been read 26. It publishes no statewide count of the dead. It carries two schedules, which the circulating accounts have run together: Appendix IV, 'Details of Atrocities' (pp. 53-55), the Marathwada tour list annexed to the letter of J. V. Joshi, a minister in the Nizam's own government who resigned over what he found; and Appendix V (pp. 58-74), a 'Catalogue of border incidents in regard to which protests have been lodged with the Nizam's Government' organised by the Bombay, Central Provinces and Madras border areas 313638. A 2018 news account attributes to the document "more than 300 incidents of atrocities allegedly committed since India's Independence by the Razakars", on the authority of a historian rather than a page 25; the figure still cannot be confirmed or refuted, because Appendix V's numbered entries are communications rather than incidents and each carries between one and ten sub-entries. What can now be said is that an atrocity death-rate taken from Appendix IV cannot honestly be scaled by a denominator taken from Appendix V, and that Ravibabu's roughly 980 does exactly that 20. What is independently attested is the schedule of 53 detailed Marathwada incidents, described by an author who worked from the document 20; Munshi's separate count of 260 is his own assertion and is treated above as such 3. That the compilation was made to justify an annexation does not falsify its contents; it explains what was collected and what was not. Nothing comparable was ever collected about the weeks after 17 September, which is the structural reason the two halves of the ledger look so different.
The embellishment is the move from enumerated incidents to unenumerated totals, and it is recent. The 880/1,125 pairing and the roughly 980 extrapolation both entered wide circulation in 2023, from a newspaper column citing an unnamed report and from a signed reconstruction that labelled itself an assessment 2120. The second move is a reclassification: "genocide", a word that does not appear in the 1948 documentation examined for this page and that none of the historians surveyed here applies to the Razakar violence, entered mass circulation with the title of the 2024 Telugu feature Razakar: The Silent Genocide of Hyderabad. Whether the term has been applied by others to the post-September 1948 killings of Muslims is a separate question this page has not surveyed and does not decide. Litigation over the film's certification settled nothing about 1948 — a bench of Chief Justice Alok Aradhe and Justice Anil Kumar Jukanti declined to halt the release on the narrow ground that "[t]he petitioner in this PIL has not challenged the Certificate issued by the Central Board of Film Certification" (reported 15 March 2024; the date of the order itself could not be established from the accessible text) 29 — and it is noted here only because the film is where the word entered wide use.
The date now carries formal commemorative status under two names. The Ministry of Home Affairs notified that "Hyderabad Liberation Day" will be marked every year on 17 September, reported on 13 March 2024 as having been notified in an official notification "released yesterday" 30; the Telangana government marks the same date as Praja Palana Day. The state-sponsored commemorative account of the period is also, as noted, the source of the lowest circulating figure for Razakar killings 23. Afsar Mohammad's oral-history work records what the same date carries at the other end: an interviewee in Karimnagar who told him that after the Police Action "every Muslim in the Hyderabad state had suddenly become an enemy of the people" 17. Both memories are of things that happened. Neither is a number.
The White Paper, read directly
The founding document of this claim has now been opened, and it changes the arithmetic section rather than merely footnoting it. The Government of India's White Paper on Hyderabad of August 1948 survives as page images inside National Archives of India file 1(15)-71/48, and the pages relevant here have been read directly rather than through the journalism that has been standing in for them.
Four things in it bear on the claim's core, and all four strengthen it. First, the Razakar oath appears in a government document of August 1948, nine years before Munshi's memoir and in wording that is close but not identical: "A Razakar, on enrolment, pledges his life to the Ittehad, to Hyderabad and to his Leader and vows 'to fight to the last to maintain the supremacy of the Muslim Power in the Deccan'" 32. Munshi gives "Muslim power"; the White Paper gives "Muslim Power", and neither is a transcription of the other. Two independent attestations of a confessional oath are better than one memoir, and the page previously rested this on the memoir alone. Second, the same page reconciles the enrolment figures that open question #7 called irreconcilable: "Seventy thousand men, women and children have undergone training; 1,50,000 members are on the rolls, and recruitment is going on briskly to reach the target figure of five lakhs of Razakars" 32. Munshi's 150,000 is the roll, his 'over a hundred thousand' the trained core, his five lakhs the target. Third, Gorta is attested contemporaneously, with the government's own hedge intact and its date confirmed against the commemorative site's 9 May: "This village with all its inmates, both human and animal, was burnt down to a heap of ashes on May 10, 1948. ... The Committee was led to believe that about 200 persons were burnt to death in the village" 34. Fourth, Appendix IV — the 53-incident Marathwada schedule that has reached this page only through The Wire — turns out to be annexed to the letter of Mr J. V. Joshi, a serving minister in the Nizam's own cabinet, who resigned over it: "Unfortunately I have noticed every Hindu complaining of inhuman atrocities, arson, murder and loot perpetrated by Goonda elements. Even innocent Hindu women are reported to have been molested. These Goonda elements, in my sincere opinion, have entirely gone out of the control of Government" 37. That is testimony from inside the accused government, and it is the best-provenanced item in the whole atrocity record on this side of the ledger.
Against that, the document's own numbers are lower and less coherent than the totals now circulating under its name, and this must be said with the same plainness. Its statewide summary, under the heading "Collapse of Law and Order in the State", offers no count at all: "There is wide-spread panic and disorder throughout the State. Life in the villages has become so insecure that thousands of persons have fled to the Indian Provinces for safety. A number of villages have been burnt down in several places. Hundreds of persons have been killed, wounded or beaten" 33. That sentence lumps killed with wounded and beaten and is rhetoric, not a tally. Two paragraphs later the document gives its worst-hit district a figure — "In Bidar district, for example they have devastated 129 villages, burnt about 1,000 houses, looted property worth crores, killed 70 to 100 men and women and committed rape on about 50 women" — and two paragraphs after that gives about 200 dead in Gorta, which is a village in Bidar district 34. Either the district figure was compiled before the 10 May burning, or the White Paper contradicts itself threefold on a single page. It dates neither, and this page cannot resolve it.
The consequence is narrower than it first appears, and the temptation to overread it should be resisted. The White Paper cannot be used as a ceiling on Razakar killings: it is internally inconsistent on its own worst district, it closes in August 1948 and therefore cannot cover Bhairanpally on 27 August or the final weeks before the Army moved, and its Warangal passage claims displacement on a scale no death toll approaches — "Lakhs of inhabitants of this district have been forced to seek refuge in the neighbouring Province of Madras. Four separate investigations were made by different sets of people, all of which tally in all material particulars" 34. What it does establish is that the government making the case for annexation, at the moment of making it, published no statewide count of the dead. Every statewide total in circulation — 880, roughly 980, 'over 2000' — was therefore produced after the fact and by somebody else. That is a finding about the totals, which is where this page already located the weakness, and not about the massacres.
One further correction of the page's own framing. Appendix IV and Appendix V are different schedules and have been run together in the circulating accounts. Appendix IV, 'Details of Atrocities' (pp. 53–55), is Joshi's Marathwada tour list across Aurangabad, Jalna, Parbhani and Nanded — Ravibabu's '53 events'. Appendix V (pp. 58–74) is a 'Catalogue of border incidents in regard to which protests have been lodged with the Nizam's Government', organised by the Bombay, Central Provinces and Madras border areas 31. They are not the same thing, and an atrocity death-rate taken from the first cannot honestly be scaled by a denominator taken from the second. Whether Appendix V contains 'more than 300 incidents' still cannot be settled: its numbered entries are communications, not incidents, and each carries between one and ten lettered sub-incidents — the Bombay sequence reaches at least no. 30 by p. 65, and a second sequence runs to no. 17 for the Central Provinces before the Madras area opens at no. 18 on p. 69 and reaches no. 32 by p. 74 38. Five of its seventeen pages have been read here. What can be said is that the individual incidents plausibly number in the low hundreds once sub-entries are counted, and that they are not, as has been suggested, mostly trivial: on the pages opened they include an armed band of about 200 Razakars and State Police raiding a village, a man abducted by Razakars and 'done to death', and a cowherd boy 'abducted by the Razakars, taken to their camp and shot dead' 38.
Who generated the record, and what the Military Governor found when he audited it
This page has applied interest analysis to Munshi, to the Sunderlal delegation and to Chaudhuri. It has not applied it to the contemporaneous press, and the press is where a great deal of the atrocity record originated. Sunil Purushotham, working the Ministry of States files at the National Archives, finds that the reporting was largely manufactured by an interested party: "much of the reporting on Hyderabad in many of the major national and international newspapers and on Indian government radio was derived from press releases put out by the 'publicity offices' established by the Hyderabad State Congress Committee of Action in Bezwada, Sholapur, and elsewhere." Officials said so at the time — a district magistrate reported on 27 September 1947 that "press accounts of atrocities and panic seem to be a result of the part of the [Congress] programme" — and a Congress delegation's own January 1948 finding was that newspaper reports of "goondas from Nizam's State playing havoc on the borders of Barsi taluka of Sholapur District" were "far from correct", and that it was in fact "goondas from the Indian Union" extorting money inside Hyderabad in the State Congress's name 39.
The strongest single document against the claim's downstream numbers comes from the officer with the least reason to produce it. After the annexation, having investigated, the Military Governor concluded that "the figures of Razakar atrocities investigated so far do not tally with the figures released . . . by the Hyderabad State Congress to the world outside by way of propaganda before Police Action", which were "deliberately inflated for propaganda purposes or were unreliable" 39. The source is Chaudhuri to Vellodi, 31 May 1949, NAI Ministry of States f.1(50)-H/49 — a real and locatable file, independently cited several times by Taylor Sherman 47. This is an admission against interest by the commander of the occupying force, which is the evidentiary form that carries most weight, and it is the reason the inflation of these numbers cannot be dated to a newspaper column in 2023. It was already being found in 1949.
The same standard has to reach the same witness twice, and this page will not take Chaudhuri's inflation finding while discounting his casualty figure, or the reverse. Chaudhuri is also the source of the estimate of about 2,000 Muslims massacred in the reprisals, and he is the officer commanding the forces implicated in them. Whoever discounts one of his statements for interest owes the other the same discount. What survives that symmetry is direction: on the atrocity figures his interest ran against the concession he made, and on the reprisal figures it ran towards the estimate he gave. That is not a reason to believe either number. It is a reason to weigh the first more heavily than the second, and the page should say so in those words rather than lean on whichever suits.
The symmetrical caution is owed on the other side, and Purushotham supplies it. The earliest loud reporting of the reprisals was also a press campaign by an interested state during an information war: Inqilab first claimed widespread violence on 24 September 1948, Zamindar the next day ("Aurangabad is totally destroyed. . . . It is a city of the dead"), Ehsan on 17 October that "Muslims are being butchered 'en bloc'" 44. And the gravest allegations in the Sunderlal material — the ones against the Indian Army — are, in the judgment of the historian who read the original report and the Confidential Notes at the Nehru Memorial Museum and Library, "difficult to substantiate" 42. Neither column's press record is independent evidence, and this page now says so of both.
Apportionment: five armed agents, one column of numbers
The page says that the classification question is live but under-evidences it, and the material that would evidence it has been sitting inside a source cited here three times. Between 1946 and 1951 the rural Deccan had at least five sets of armed men killing people in the same districts: the Razakars, the Nizam's police and military, landlords' retainers, Communist guerrillas, and after September 1948 the Indian Army and Union police. No statewide 'Razakar toll' has ever separated them.
The only village-level named enumeration of rural deaths that exists is the Communist Party's, and it is dominated by killings after the Police Action. Sundarayya's Appendix I, 'Martyrs of the Struggle', lists the dead by district, taluka, name and village, frequently with dates: "Killed on 27.12.1948", "Killed on 26.12.1948", "Killed in Narsimhapuram, 8.1.1949", "Killed on 20.11.1949", and one entry reading "Joined the squad as student at the age of 16 and was shot dead at the age of 17 by the Indian army" 50. Dhanagare, in a peer-reviewed article, counts the list and states the apportionment problem exactly: "Sundarayya has produced a complete list of 2,517 'martyrs of the struggle'. However, not all of them were killed by the Indian Army; some were killed by the Razakars" 53. Sundarayya's own preface attributes the aggregate to agents that do not include the Razakars at all — "about 4,000 were killed by police, military and landlord terror during 1946-51" — and disclaims verification of his own figures in terms this page would apply to Munshi: "some accounts appear to be exaggerated and appear unreal. The exact number of killed and wounded ... cannot be verified now" 51.
The insurgency was also killing at a comparable order of magnitude. Sundarayya records that "Five hundred landlords and their agents were killed by July 1949", and quotes a Congress opponent's higher figure of "957 landlords and their agents ... killed by Communist guerrillas" 52. The Government of India's own pamphlet credited the Communists with 2,000 murders between 15 August 1947 and 13 September 1948, itemised to the rupee and the rifle — "attacked 22 police outposts, destroyed village records, manhandled 141 village officials, seized 230 guns, eight revolvers and one rifle, looted or destroyed paddy worth Rs70,000, robbed cash and jewellery worth Rs10,43,668, and destroyed 20 customs outposts" 49. The granularity is the point: the state counted Communist violence to the rupee and never produced anything comparable for Razakar killings. That is a fact about what the state chose to count, and it is the structural reason this page keeps arriving at the same answer.
And the conflation this page has treated as an open question is stated as a finding in a university-press monograph: "Even more duplicitously, Razakar attacks on the communist-led peasant revolutionaries of Telangana were reported as attacks on Hindus as Hindus" 39. The government's own brief supports the political-allegiance reading from the other side: the White Paper's border-incidents chapter says the raids "began in order to terrorise those border villages which offended the Razakars and the State under the leadership of the Hyderabad State Congress", that the incursions were "ostensibly in search of 'undesirable characters'", and that "the motive for these raids was in most cases extortion of money. In these raids, even villagers from the Hyderabad side joined with guns, swords and lathis" 35. None of this makes the dead at Gorta and Bhairanpally other than Hindu cultivators, and the page's refusal to trade one fact for the other stands. What changes is that the reframing can now be cited to an archival monograph and to the accusing government's own document, instead of to a historian and a sociologist quoted at second hand in an opinion column.