Where We Left Off
I am Punjabi and this is the part I am least qualified to write impartially. Chapter Nine's hidden assumption is where I have tried hardest to argue against myself, and Chapter Four is where the militancy's own killings are set out without mitigation.
Part Five recorded Punjab as both the largest beneficiary of central policy and one of its largest casualties. This part is the fifteen years in which the second overtook the first.
Five findings carry into this part.
Punjab’s rivers were allocated by notification and by an agreement that ended a live court case. Part Five’s Chapter Five: the 1976 notification left the state about a fifth of rivers flowing through its own territory, Punjab sued, and in December 1981 the Prime Minister brokered an agreement with a Chief Minister of her own party and the case was withdrawn. Part Five’s verdict — that Punjab had a weak case about the water and a strong case about the procedure — stands, and Chapter One is about what the procedure did to the politics.
The agricultural package made Punjab rich and locked it in. Part Four and Part Five both found it: seed, subsidised inputs, near-free power and a guaranteed price, producing prosperity and a crop pattern nobody has since been able to change.
A warning is received only where somebody has the authority to act, an incentive to believe, and a decision it changes. Part Six’s finding. Chapter Five applies it to a soldier who wrote the position paper he was asked for.
When a state that counts everything has no count of something, somebody decided that. Part Two’s rule, used on the Partition dead and on Hyderabad. Chapter Seven is where it does the most work in this series.
An excess is a legitimate thing done too much. Part One promised to name that word every time. Part Nine named it over the sterilisation programme. It is named again here.
One thing about who is writing this. I am Punjabi. I have written about this material before, at length, and I hold views about it that predate any of the research in this part. That is a problem for a book of this kind and there is no way to solve it except by being explicit about where the risk lies. It lies in Chapters Five to Seven, where a Punjabi account writes itself and would be wrong. Chapter Four exists partly as a corrective, and Chapter Nine’s closing box is the finding I found hardest to write.
How to read this book
The Six Boxes
Six coloured boxes run through the series, each doing one job. Here they are with examples from the material ahead.
A Word Box explains a hard word the moment it first appears.
Akal Takht: the seat of temporal authority in Sikhism, established in 1606, standing directly opposite the Harmandir Sahib in Amritsar.
The distinction matters for everything in Chapter Five. The Harmandir Sahib — the Golden Temple — is a place of worship. The Akal Takht is the institution from which binding directions to the Sikh community are issued. In a rough analogy, one is a cathedral and the other is the office of the authority that governs it.
Why it matters here: the army’s assault in June 1984 was directed principally at the Akal Takht, and it was reduced to rubble. That is why the operation is understood in Sikh terms as an attack on an institution rather than on a building.
An In Real Terms box turns a number too big to picture into something with a body.
The official count of Sikhs killed in Delhi between 31 October and 7 November 1984 is 2,733. Most of them died on 1 and 2 November.
That is roughly a thousand people a day, for two days, in the capital city of a country with a functioning police force, in an area of a few dozen square kilometres.
Of the 587 first information reports registered, 25 have produced a conviction. Twelve of those are for murder.
Forty-one years. Twelve murder convictions. Chapter Seven is about the arithmetic between those two numbers.
A How We Actually Know This box shows the physical evidence and then says what it cannot prove.
The evidence in this part is unusually uneven and it is worth mapping before starting.
Very strong: November 1984 in Delhi. Four commissions, nine committees, two special investigation teams, a documented death toll, an FIR count, a conviction count, and High Court judgments naming the failures. The Indian state has investigated this more thoroughly than almost anything else in this series.
Strong: the constitutional and political chronology — the dismissals, the notifications, the President’s Rule proclamations, the texts of resolutions and accords.
Weak: everything about the militancy’s internal decisions, and everything about the counter-insurgency’s methods between 1987 and 1993. There was no commission, the press was restricted, and the principal investigator of the disappearances was himself abducted and killed.
Which produces an asymmetry this part has to state rather than exploit: the five days in Delhi are among the best-documented atrocities in Indian history, and the following nine years in Punjab are among the worst-documented.
An Argument box appears where serious people disagree, each side at its strongest.
The framing question, and both national and Sikh accounts get it wrong in opposite directions.
The demands were made in religious language, from religious institutions, by a religious leadership. The Anandpur Sahib Resolution opens by asserting the Sikhs’ status as a distinct people. The militancy’s stated objective was a state defined by faith. The killings of Nirankaris in 1978 were about doctrine. And the violence of November 1984 was directed at people identified by their turbans and their names, which is the definition of a communal attack.
Read the operative demands rather than the preamble. Chandigarh. River waters. The transfer of Punjabi-speaking areas. Limits on central power. Those are the standard grievances of an Indian state against the centre, and versions of them were being pressed at the same time by Tamil Nadu, by West Bengal and by the north-east. What made Punjab different was not the demands but that the state’s constitutional routes were closed one after another — a notification instead of a tribunal, a dismissal instead of an election, and a charter of demands relabelled as treason.
What would settle it: comparing how the centre handled equivalent federal demands from states where the population was not religiously distinct. Part Five began that comparison and nobody has completed it.
Why people care so much: because a religious conflict has no policy answer and a federal one does, and both Indian and Sikh nationalisms have preferred the version with no answer.
A Hidden Assumption box digs out a belief sitting underneath an argument that both sides accept without discussing.
Everybody who writes about these fifteen years assumes that they were about Punjab.
The Indian version: a separatist movement in a border state, backed from outside, defeated at great cost. The Punjabi version: a community’s grievances answered with tanks. Both treat the events as belonging to one place and one people.
Look at the instruments instead of the events. An elected state government dismissed by the centre. River waters allocated by executive notification rather than by the tribunal the law provides. A live Supreme Court case ended by a political agreement. A senior officer who gave contrary professional advice passed over for the post. A state governed under emergency powers for most of a decade. Detention without trial. An official inquiry into a massacre producing almost no convictions.
Not one of those is a Punjab instrument. Every single one appears elsewhere in this series, applied to places with no religious distinctiveness, no militancy and no foreign border. Kerala was dismissed in 1959. Kashmir’s head of government was removed without a floor test in 1953. The Emergency detained a hundred and eleven thousand people. Three judges were passed over in 1973 and a fourth in 1977, and a Vice-Chief of Army Staff in 1983.
Which reframes the question. “Why did Punjab explode” has an answer that is not about Punjab. The right question is why the same instruments produced an insurgency here and produced resentment in Kerala — and it is the same answer Part Seven reached about Kashmir. Where the state’s presence is not itself in question, these instruments damage a government. Where it is, they damage the claim.
The uncomfortable corollary, which Chapter Nine returns to: Punjab’s presence in India was not in question in 1978. It became a question over the following six years, and it did so while the instruments were being applied.
And every chapter closes with a Remember This box: the chapter in the plainest words available.
Part Five’s verdict stands: Punjab had a weak case about the water and a strong case about the procedure. Chapter One is what the procedure did to the politics.
The evidence is uneven and this part says so: November 1984 in Delhi is among the best-documented atrocities in Indian history and the nine years that followed in Punjab are among the worst-documented.
Not one instrument used in Punjab was a Punjab instrument. Every one appears elsewhere in this series, in places with no militancy and no border.
1What Punjab Was in 1978
The richest state in India per head, and the one with the longest list of things it could not get decided. Both facts matter and the second one is the subject of this part.
1.1 — The prosperous half
Parts Four and Five established the first half and it should not be skipped, because a story about grievance that leaves out the prosperity is a story about somewhere else.
By 1978 Punjab was the richest Indian state per head. It had the canal system the British began in the 1880s, extended by Bhakra. It had received the agricultural package in full — the seed, the subsidised fertiliser, the near-free power for pumping, the credit and the guaranteed price. It fed the country’s buffer stock.
Part Five’s ledger recorded Punjab in the column of states that gained, and that entry stands. Whatever else this part describes, it does not describe a neglected place.
Union Territory: a part of India administered directly by the central government rather than by an elected state government.
It is meant for places that cannot sensibly be states — small islands, tiny enclaves, the national capital — and for temporary arrangements pending a decision.
Chandigarh was made one in 1966 as a temporary arrangement pending a decision about which state should have it. The temporary arrangement is now fifty-nine years old, which is longer than most Indian states have existed.
Why it matters here: Part One’s rule about emergency instruments applies to administrative ones too. A holding arrangement with no expiry becomes the settlement, and nobody ever has to defend it.
1.2 — The list
And it had a list of unresolved questions, most of which had been open since the state was reorganised in 1966.
Chandigarh. Built as Punjab’s capital after Partition took Lahore. On reorganisation it became a Union Territory serving as the capital of both Punjab and Haryana, pending a decision. The decision was promised repeatedly and not made.
The river waters. Part Five’s Chapter Five in full. The 1955 allocation gave Rajasthan the largest share of two rivers that do not flow through it, and undivided Punjab did not object — which is Punjab’s own concession and stands. After 1966 the successor states failed to agree, and in March 1976 the centre issued a notification dividing the water, leaving Punjab with roughly a fifth of the flow of rivers running through its territory.
Punjabi-speaking areas. The 1966 reorganisation left areas Punjab claimed on the Haryana side, pending a determination that kept being deferred.
Procurement terms. The state that grew the country’s grain wanted the price set higher and the input costs lower, which is the ordinary politics of every farming region anywhere.
1.3 — The routes, and the order they closed
Here is the thing this chapter exists for, and it is a claim about procedure rather than about substance.
Every item on that list had a constitutional route by which it could have been settled. Track what happened to each.
The river dispute had a tribunal. India has an Inter-State River Water Disputes Act, and its whole purpose is deciding exactly this kind of question independently of whoever is in office. Instead the matter went by central notification in 1976. Punjab filed suit in the Supreme Court in July 1979 — the second route. In December 1981 the Prime Minister brokered a tripartite agreement with a Chief Minister of her own party, and Punjab withdrew the case. Route two closed. A tribunal was eventually appointed in 1986 and never delivered a final award.
Chandigarh had a political route — a decision by the government of India, promised in successive settlements and repeatedly postponed.
And electoral politics was the route for everything else. An Akali government took office in Punjab in 1977, after the Emergency. It was dismissed in 1980, when Congress returned to power at the centre and used Article 356 across a set of states.
Set the dates in order. 1976: the water goes by notification. 1980: the elected government goes. 1981: the court case goes. 1982: the Akalis launch an agitation, having run out of the other three.
That sequence is the argument of this part, and everything from Chapter Three onwards happens inside it.
Imagine a dispute with your neighbour about a boundary.
You go to the tribunal that exists for boundary disputes. Before it rules, the council issues a notice deciding the matter administratively. You go to court. While the case is pending, the mayor arranges a meeting with your solicitor — who is a member of the mayor’s party — and your case is withdrawn. You vote for a new council. The mayor dissolves it.
You have now exhausted every method that exists. Nothing that happens next is justified by that. But if you want to know why somebody in your position eventually stops using methods, the answer is on the record with dates against it.
All of which rests on the plainest kind of evidence available to this series.
Underneath all four items on that list sits one concrete question that could have been settled at any point in fifty-nine years.
It is the smallest item on Chapter One’s list and the one that has outlasted everything else.
It was built as Punjab’s capital, on land taken from Punjabi villages, to replace Lahore. On reorganisation it was made a Union Territory as a temporary arrangement pending a decision, and the decision was promised in successive settlements including the accord of 1985, which set a date. The date passed. Fifty-nine years is not a pending decision; it is a decision, taken by never taking it, and every Punjabi government of every party has said so.
Haryana was created without a capital and has had no other since. Transferring Chandigarh means either leaving Haryana without one or building a new one at enormous cost, and every proposed package — a transfer plus compensating territory — has foundered because the compensating territory was itself disputed. Successive central governments have not refused; they have failed to find terms both states would accept, which is a different thing and is what federations look like when two members want the same object.
What would settle it: a decision. There is no evidential gap here at all, which distinguishes it from everything else in this part.
Why people care so much: because it is the one item on the list that requires no inquiry, no money and no reinterpretation of history — and it is still open.
All of which rests on the plainest kind of evidence available to this series.
The chronology in section 1.3 is entirely composed of dated public acts: a gazetted notification, a filed suit, a signed agreement, a withdrawn case, and a proclamation under Article 356 laid before Parliament.
None of it requires anybody’s account of anybody’s motives, which is why this chapter can be categorical where the rest of the part cannot.
What it cannot show: whether the closing of those routes was intended to have the effect it had. Each act had its own justification at the time, and there is no document connecting them. This part argues that the sequence mattered regardless of intent, which is a weaker claim and the only one the evidence supports.
One thing this chapter does not claim. It does not say that closing those routes caused what followed, or that anybody closed them in order to cause it. Each act had its own justification and there is no document connecting them. What it says is that by 1982 a state with an ordinary list of federal grievances had no ordinary way left to press any of them, and that this is a fact about a procedure rather than an opinion about a motive.
By 1978 Punjab was the richest state in India per head, and Part Five’s ledger records it as a state that gained. A story that leaves that out is about somewhere else.
It also had a list open since 1966: Chandigarh, the river waters, the Punjabi-speaking areas, and procurement terms.
Every item had a route. 1976 the water went by notification, 1980 the elected government went, 1981 the court case went. The agitation of 1982 came after the other three.
2The Resolution
It is described in Indian politics as a secessionist document. Very few people who describe it that way have read it, and the text did not change between the two descriptions.
2.1 — What it is
In December 1972 the working committee of the Shiromani Akali Dal appointed a twelve-member sub-committee to draft a statement of policy. Its members included Surjit Singh Barnala, who would later be Chief Minister, Gurcharan Singh Tohra, Balwant Singh and a retired Major-General.
The committee’s report was adopted at Anandpur Sahib on 17 October 1973. It was endorsed with modifications at the party’s conference at Ludhiana in October 1978.
It is a party’s charter of demands — the sort of document any political party produces, comparable in form to a manifesto or a white paper.
2.2 — What it actually demands
Set out plainly, because almost nobody does.
On the centre and the states: that the constitution be recast on genuinely federal lines, with the central government’s role limited to defence, foreign relations, currency and communications, and all remaining powers vested in the states.
On specific Punjab questions: Chandigarh to Punjab; the transfer of Punjabi-speaking areas; an equitable share of river waters; state control over irrigation, agriculture and industry.
On religion: recognition of Sikhism as a religion distinct from Hinduism, and protection of Sikh identity.
On the economy: redistribution, poverty eradication, action against monopolies — which is to say, roughly the economic language every Indian party used in 1973.
Now look at that first demand again, because it is the one that matters and it has a history in this series.
Defence, foreign relations, communications. Part Seven’s Chapter Three set out what the Instrument of Accession transferred to the Union in 1947: defence, external affairs and communications.
The Anandpur Sahib Resolution’s central constitutional demand was, in substance, that every Indian state be given what Jammu and Kashmir was actually given in 1947 — and which, as Part Seven documented, was taken away from it by executive orders between 1954 and 1965.
The resolution is a published political document and its text can be read. That should make this chapter simple and it does not, for one reason.
It circulated in several versions and translations. The original was in Punjabi. English renderings differ, sometimes materially, and the 1978 Ludhiana endorsement modified it. Through the 1980s the government habitually quoted the most maximalist available rendering and the Akali Dal the most moderate — and both were quoting real text from real versions.
This is not a minor complication. A great deal of the national argument about whether the document was secessionist has been conducted by two sides reading different translations at each other, in good faith, for forty years.
What this chapter therefore does: it states the operative demands, which are consistent across versions, and states the preamble’s nationalist language, which is also present across versions. Where the versions diverge is mostly in emphasis and tone, and no claim here depends on a disputed passage.
What it cannot show: which version any given official had read when describing it. That is unknowable and it matters more than it should.
2.3 — What is genuinely awkward in it
A Punjabi account stops there. This one cannot.
The resolution’s preamble is not a federalism document. It asserts that the Sikhs have constituted a distinct political people since 1699, and speaks of an autonomous region in northern India in which Sikh interests would be the fundamental principle of state policy.
That is more than devolution. A territorial unit whose declared basis is the interests of one religious community is not the same proposition as a state with more powers, and pretending otherwise is not honest. It is the strongest thing the document’s critics have and this part concedes it.
There is a further complication. The resolution circulated in several versions and translations, and the government habitually cited the most maximalist rendering while the Akalis cited the most moderate. Both were quoting real text.
2.4 — What changed
The document was adopted in 1973. It was endorsed in 1978. It was referred, in substance, to the commission on centre-state relations that the Indian government itself appointed in 1983 to consider exactly the questions it raised.
Between 1978 and 1982 it was the platform of a party that had governed Punjab in coalition and had partners in Delhi. By 1983 it was being described in national politics as a charter for secession.
Not a word of it had changed.
Federalism: a system in which powers are divided between a central government and constituent units, with each having authority the other cannot simply take away.
The Indian constitution is federal in structure and heavily weighted to the centre: the Union List is long, residual powers sit with the centre, and Article 356 lets it dismiss a state government. Part Three called it a federation with a hole in it, and Part Five showed the hole in operation.
Demanding more federalism is therefore an ordinary constitutional position in India, held at various times by parties in Tamil Nadu, West Bengal, Andhra Pradesh and the north-east.
Why it matters here: the Anandpur Sahib Resolution’s core demand is a federalism demand, and it was not treated as one.
The way to test whether that demand was extreme is to hear it in another voice.
Try the central demand on a different state and hear how it sounds.
Suppose Tamil Nadu asked that the Union government be responsible for defence, foreign affairs, currency and communications, and that everything else — police, land, education, industry, water — sit with the state.
That is a radical constitutional proposal. It is also a recognisable one, of a kind argued for by regional parties in several Indian states across the 1970s and 1980s, and it is what the Sarkaria Commission was appointed in 1983 to consider.
The same words, from a party representing a religious minority in a border state, were read as treason. The difference was not in the sentence.
Which is the argument this chapter cannot avoid.
This is the founding dispute of the whole period and both readings are held by serious people.
Read the preamble rather than the shopping list. A document asserting that a religious community has been a distinct political nation since 1699, and calling for a territory where that community’s interests are the basis of state policy, is not asking for better centre-state relations. Limiting the Union to four subjects would leave a state with its own everything else — which is a confederation, not a federation. And whatever the drafters meant, the movement that carried the document by 1983 was demanding a separate country, which is the only test of a political text that matters.
The operative demands are a federalism charter and the Indian government treated them as one for a decade: it negotiated on them, and in 1983 it appointed a commission on centre-state relations to examine precisely these questions. Governments do not refer secession demands to expert commissions. As for the movement of 1983, that is an argument about 1983 and not about the document — and Chapter Four shows what happened in between.
What would settle it: the drafting record of the 1972–73 sub-committee, and a definitive comparison of the circulating versions. Neither exists in accessible form.
Why people care so much: because if it was a federalism demand, the state answered a constitutional argument with an army, and if it was a secession demand, the state answered a secession movement the way states do.
Notice what both of those positions are doing. They are both reading, and both believe the reading settles something.
Both sides of that argument assume that the question is what the document said.
Its defenders quote the demands. Its critics quote the preamble. Both are conducting a reading, and both believe that establishing the correct reading would establish something about the conflict.
A political document’s meaning is not fixed by its text. It is fixed by who is holding it up and what their opponents need it to mean, and both of those changed completely between 1973 and 1984 while the words sat still.
Follow the same sentences through. In 1973 they are an opposition party’s manifesto in a state the party does not govern, and nobody outside Punjab reads them. In 1977 they belong to a governing party in coalition with parties that are simultaneously in office in Delhi, and the demands are negotiable. In 1980 the government holding them is dismissed. In 1982 they are the platform of an agitation. By 1983 they are being cited in Parliament as evidence of treason, and by 1984 they are being cited by men with rifles as authority for something the document does not ask for.
Five meanings, one text, eleven years.
This is why the reading argument is unresolvable and why both sides can quote accurately for ever. And it generalises: a charter of demands is a container, and what it holds depends on who is carrying it. The useful question is never what a political document says. It is who gained by which reading, in which year, and what they did with it.
Carry one date forward. In 1983 the Indian government appointed a commission on centre–state relations to examine the questions this resolution raised. In the same year the resolution was being described in national politics as a charter for secession. Both of those were the position of the same government in the same twelve months.
Adopted 17 October 1973. Its central constitutional demand was that the centre be limited to defence, foreign relations, currency and communications — very nearly the list the Instrument of Accession gave the Union in 1947.
Its preamble is a different document: a distinct political people since 1699, and a region where Sikh interests would be the basis of state policy. That is more than devolution and this part concedes it.
Between 1978 and 1983 it went from a coalition partner’s platform to a charter for treason. Not a word of it changed.
3The Man Who Was Built
The claim that Congress built Bhindranwale to split the Akali vote is the most repeated allegation in Punjabi politics. This chapter examines it, finds the evidence thinner than the claim, and argues the question is the wrong one anyway.
3.1 — The claim
Jarnail Singh Bhindranwale became head of the Damdami Taksal, a Sikh seminary, in 1977. Over the following six years he became the most consequential figure in Punjab.
The claim, made repeatedly in Punjab and by a number of Congress figures afterwards, is that he was promoted by Congress in the late 1970s as a religious counterweight to the Akali Dal — a way of splitting the Sikh vote and undermining the party that had beaten Congress in 1977.
Several senior Congress figures have said versions of this publicly. It is the standard explanation in Punjab and it is repeated as though established.
Vote bank: a bloc of voters expected to vote together, which a party tries to secure whole rather than persuading one at a time.
It is ordinary electoral arithmetic everywhere and it is not sinister in itself. What makes it consequential is the standard method for attacking somebody else’s: you do not try to win the bloc over, you find something that will split it, and the cheapest splitter is usually a rival identity or a purer version of the same one.
The risk is equally standard. A splitter recruited to divide a bloc has his own objectives, and once he has a following he no longer needs the party that promoted him.
Why it matters here: this is the mechanism the allegation in this chapter describes, and it is the reason the allegation is plausible without being proved.
3.2 — What the evidence supports
Grading it properly is the job of this section.
Solid: that Congress and the Akali Dal were direct electoral competitors for the Sikh vote, and that anything dividing the Akali base was in Congress’s electoral interest. That is not an allegation, it is arithmetic.
Documented: that he was allowed to operate with a freedom the law did not require. He was arrested in 1981 in connection with the killing of a newspaper editor and released. Weapons were carried openly in and around the temple complex for a long period. Police action was repeatedly not taken.
Alleged: that Congress deliberately promoted him as a strategy. This rests on statements made by participants afterwards, in a period when everybody involved had reasons to attribute the disaster to somebody else. There is no document.
So the honest position is that the electoral interest is obvious, the permissiveness is documented, and the deliberate strategy is asserted rather than proved.
A political vacuum is not an abstraction. It is a specific and countable thing: the number of people in a place who have a grievance and no remaining institution through which to press it.
In Punjab by 1982 that group included farmers who wanted a better procurement price, a party that had won an election and been dismissed, lawyers whose case had been withdrawn over their heads, and young men with education and no work in a state whose prosperity had stopped rising.
None of those people wanted a separate country in 1978. Most of them never did.
What a vacuum does is not convert people. It removes the alternatives, so that whoever is offering something becomes the only person offering anything — and he does not have to be persuasive, he only has to be present.
3.3 — Why the question is the wrong one
And now the reason this chapter is short.
Suppose the strongest version of the claim is true and Congress did build him. It still explains very little, because it does not explain why the building worked.
Political parties promote convenient religious figures constantly, in every country, and almost none of them becomes what he became. The variable is not the promotion. It is whether there is anything for the promoted figure to occupy.
Chapter One listed what was available in Punjab by 1982. A river dispute settled by notification. A court case ended by political agreement. An elected government dismissed. A charter of demands relabelled as treason. Every constitutional route to a set of ordinary state grievances closed, in sequence, within six years.
A political space of that size does not stay empty. If it had not been him it would have been somebody, and the somebody would have had the same material to work with.
The most repeated allegation in Punjabi politics deserves a proper hearing on both sides.
The electoral logic is unarguable: Congress had lost Punjab to the Akalis in 1977 and needed the Sikh vote split. He was given platforms, was permitted to campaign against Akali candidates in 1980, was arrested for a killing and released within weeks, and was allowed to arm and fortify a religious complex over months while the police did not act. Senior Congress figures have subsequently acknowledged versions of this. A pattern of official permissiveness that consistent is not accident.
Permissiveness is not sponsorship. Governments avoid confrontations inside religious sites everywhere, for obvious reasons, and the reluctance to enter the Golden Temple was shared by every party and by the police themselves. Congress state governments were also being killed by the movement he led, which is a strange outcome for a strategy. And every account of deliberate promotion was given after 1984 by people with a catastrophe to explain and an interest in placing it elsewhere.
What would settle it: nothing that exists. There will be no document.
Why people care so much: because whoever produced him is understood to own everything that followed, which is why both sides need the answer and why the box below argues the question is the wrong one.
What both sides of that argument share, though, is an assumption about how political figures come to matter at all.
Everybody arguing about Bhindranwale assumes that he was created.
The Punjabi version: Congress built him, therefore what followed is Delhi’s responsibility. The Indian version: he arose from Sikh fundamentalism, therefore what followed is Punjab’s. Both are arguments about origin, and both parties want the origin because whoever produced the man owns the consequences.
People are not created. Spaces are — and a space is created by closing things, which is a matter of record rather than of allegation.
This is why the origin argument has run for forty years without moving. It cannot be settled, because the evidence for deliberate promotion is memoir and the evidence against is absence, and neither side will ever produce the document. Meanwhile the thing that can be established — which routes closed, in which year, by whose act — is not seriously disputed by anybody and is barely discussed.
The general rule matters beyond Punjab. When a political figure appears who seems to come from nowhere, the productive question is never who sponsored him. It is what was closed, and when — because sponsorship is cheap and available to everybody, and vacancy is not.
And it cuts against the account I grew up with. “They built him” is a comfortable sentence for a Punjabi to say, because it puts the agency entirely elsewhere. The version in this chapter is less comfortable and I think more defensible: the closing of the routes is Delhi’s, and what was done with the space that opened was not only Delhi’s, and Chapter Four is about that.
Which is why this chapter is the shortest in the part. The origin question absorbs enormous energy in Punjab and settles nothing, and the question it displaces — what was closed, and when — is answerable from a gazette.
The claim that Congress built him is the standard explanation in Punjab. The electoral interest is obvious, the permissiveness is documented, and the deliberate strategy is asserted rather than proved.
Even at its strongest the claim explains little, because it does not explain why the building worked. Parties promote convenient religious figures constantly and almost none become this.
People are not created. Spaces are — and a space is created by closing things, which is a matter of record rather than allegation.
4The Slide, 1978 to 1984
This chapter exists to prevent the rest of this part being dishonest. What was done in Punjab in these six years, by people from Punjab, is set out here without mitigation.
4.1 — April 1978
On 13 April 1978, Baisakhi, a group from the Damdami Taksal and the Akhand Kirtani Jatha marched to protest against a convention of the Sant Nirankari Mission in Amritsar. There was a clash. Thirteen Sikhs were killed.
Note who died. The first deaths of this period were Sikhs, killed in a dispute between Sikhs and a sect they regarded as heretical, and the prosecution that followed produced acquittals which were bitterly received.
That is where the cycle starts, and it starts inside Punjab.
4.2 — What followed
The list is not short and it should be read in full rather than summarised.
April 1980: the head of the Nirankari sect was assassinated in Delhi.
9 September 1981: Lala Jagat Narain, the founder and editor of a major Punjabi newspaper group and a persistent critic, was shot dead. Bhindranwale was named in the investigation, arrested in September and released in October.
25 April 1983: A.S. Atwal, a Deputy Inspector General of Police, was shot dead on the steps of the Golden Temple as he was leaving it. His body lay there for hours because the police would not enter.
5 October 1983: at Dhilwan, a bus was stopped, passengers were separated by religion, and six Hindu passengers were shot dead. The following day the state government was dismissed and Punjab was placed under President’s Rule.
December 1983: Bhindranwale moved into the Akal Takht itself and fortified it. Shabeg Singh, a retired Major-General who had trained the Bengali resistance in 1971 and had been dismissed from the army, organised the defences.
And running through all of it, continuously: killings of Hindus in Punjab, of Nirankaris, of Congress workers, of police officers, of journalists, and of Sikhs who disagreed. By 1984 the largest single category of people being killed by the militancy was Sikhs.
A Deputy Inspector General of Police is a very senior officer — the rank that commands a range of districts.
He was shot on the steps of the Golden Temple, in daylight, in April 1983, as he walked out of it. His body lay where it fell for hours, because the force he commanded would not go in to collect it.
Whatever else is true about this period, a state whose senior police officers can be killed on the steps of a building its own force will not enter is not exercising authority in that place. That is the situation the Indian government faced in 1983, and no argument in the later chapters of this part is available to anybody who has not accepted it.
4.3 — What this means for the rest of this part
Three things follow and all three constrain what the later chapters can argue.
The state’s problem was real. A police officer shot on the steps of a shrine his force would not enter, passengers pulled off a bus and killed by religion, and an armed group fortifying a religious complex are not grievances. They are a collapse of public order, and no government anywhere would have left them alone.
Punjab had agency. Part Five’s Chapter Eight made this concession about Kashmir and it applies here with more force. Nobody in Delhi ordered the Dhilwan killings. The decision to fortify the Akal Takht was taken in Amritsar.
The victims were mostly Sikh. Any account of this period that presents the militancy as a Sikh community’s defence against the Indian state has to explain the Sikhs it killed, and mostly does not.
What Chapter One’s sequence explains is why there was a space. It does not explain, excuse or diminish what was done in it, and this part will not be read as though it does.
The events in section 4.2 are among the best-documented in this part, for an unhappy reason: they happened in public, to public figures, in front of a press that was still operating freely.
The killing of a newspaper editor, of a serving Deputy Inspector General on the steps of the Golden Temple, and of bus passengers on a highway are all matters of contemporaneous reporting, police record and subsequent prosecution.
Compare that with what is available for the same state between 1987 and 1993, which Chapter Eight describes, and the difference is the whole of the evidential problem in this part.
What it cannot show: who ordered what. The chain of command inside the militancy was not documented, its records do not exist, and the people who would know were mostly killed. Attribution of individual killings to individual decisions is largely unavailable, which is why this chapter lists acts rather than assigning them.
Which raises the question of how much of that was already unavoidable, and from when.
Chapter One argued that routes closed. Whether that made what followed unavoidable is a different question.
Once an armed group is fortifying the holiest site of a religion, killing police officers on its steps and selecting bus passengers by faith, the space for negotiation has been destroyed — and destroyed by the people occupying it, not by anybody in Delhi. Governments cannot negotiate with a movement whose participants are killing the moderates of their own community, because there is nobody left with the standing to make an agreement stick. The Akali leadership by 1983 could not deliver what it might have signed.
Everything in that description is true and dated 1983. The routes closed in 1976, 1980 and 1981. There is no version of this in which a settlement was unavailable in 1978 or 1979, when the demands were Chandigarh, water and a boundary, and when the Akali Dal was in office and could deliver. The state did not face an armed movement in 1979 and it did by 1983, and the intervening years are the ones with the decisions in them.
What would settle it: the record of the Akali–centre negotiations of 1982 and 1983 — what was on the table and how close it came. Accounts differ and the files are not open.
Why people care so much: because the answer determines whether June 1984 was a tragedy with no alternative or the end of a road somebody chose to walk down.
I have put this chapter before the chapters about the state’s conduct rather than after them, deliberately. Placed afterwards it would read as mitigation offered late. Placed here it does what it is for, which is to establish that everything in Chapters Five to Eight happened in a situation that Punjabis had a substantial hand in creating.
13 April 1978: thirteen Sikhs killed in a clash with a sect. The cycle starts inside Punjab, between Sikhs.
An editor shot in 1981. A Deputy Inspector General shot on the temple steps in April 1983. Six bus passengers separated by religion and killed at Dhilwan in October 1983. The Akal Takht fortified in December 1983.
The state’s problem was real, Punjab had agency, and the militancy’s largest category of victim was Sikhs. Chapter One explains why there was a space. It does not excuse what was done in it.
5June 1984
The Vice-Chief of the Army Staff was asked to plan it, advised against it, proposed an alternative, and was passed over for the post that would have let him decline to carry it out.
5.1 — The officer who was asked
Lieutenant General S.K. Sinha was Vice-Chief of the Army Staff and next in line to be Chief. He was asked to prepare a position paper on an assault on the Golden Temple complex.
He advised against it. His grounds were that an armed assault on the holiest site of a religion would be understood as sacrilege and would have consequences far outlasting any military result, and he proposed an alternative approach.
He was not made Chief. General A.S. Vaidya was appointed instead, with Lieutenant General K. Sundarji as Vice-Chief, and those two planned and executed the operation.
Part Six established the test: a warning is received only where somebody has the authority to act, an incentive to believe, and a decision it changes. Here the warning came from the officer whose professional judgment the government had specifically commissioned, and the response was to appoint somebody else.
This is the third supersession in this series and they should be read together. Three judges passed over in April 1973 for deciding against the government. One judge passed over in January 1977 for dissenting. And a Vice-Chief of Army Staff passed over in 1983 for advising against an operation.
5.2 — What was done
From 3 June 1984 Punjab was sealed. A thirty-six hour curfew was imposed. Rail, road and air links were cut. Telephone and telegraph lines were severed. Electricity was interrupted. Foreign journalists were removed from the state and complete censorship was imposed.
3 June was also the martyrdom anniversary of Guru Arjan Dev, one of the days on which the complex is fullest. Large numbers of pilgrims were inside when it was sealed.
The assault began on the night of 5 June. Commandos of 1 Para entered the complex and met prepared, disciplined fire from positions Shabeg Singh had sited on rooftops, in the clock tower and in the drains. They were pinned down with heavy casualties.
By the early hours of 6 June the infantry could not take the Akal Takht. Armoured personnel carriers were brought in; one was disabled by an anti-tank rocket. Permission was sought and given for tanks to use their main armament, and at around half past seven in the morning Vijayanta tanks fired high-explosive shells into the Akal Takht.
The complex was under control by the morning of 7 June. Bhindranwale, Shabeg Singh and Amrik Singh were killed. Around three dozen other gurdwaras were entered by the army simultaneously.
5.3 — What was destroyed
The Akal Takht was reduced to rubble.
The Sikh Reference Library — manuscripts, handwritten volumes, correspondence, the documentary record of a religion — burned. What was in it is not recoverable and there is no full catalogue of what was lost.
Casualty figures are disputed and this part will not pretend otherwise. The government’s own White Paper gave figures in the region of five hundred civilians and militants killed and eighty-three soldiers. Independent estimates, including from people who were in Amritsar, run considerably higher, and the count of pilgrims present when the complex was sealed has never been established.
5.4 — The professional objection
The most damaging criticism of the operation came not from Punjab but from inside the Indian military establishment.
A retired Major-General who was a contemporary of the operation’s commander made the point most simply. If the police assessment was that the militants held two or three hundred weapons, most of them shotguns and bolt-action rifles, why was the army needed at all?
Lieutenant General Jagjit Singh Aurora — who had accepted the Pakistani surrender at Dhaka in 1971 — was among the operation’s most outspoken critics.
And the military consequence arrived within days. Sikh soldiers deserted or mutinied in several units across India. That had not happened to the Indian Army before and has not since.
Siege: surrounding a fortified position and waiting, rather than assaulting it.
A siege is slow, politically uncomfortable and looks like weakness. It is also, against a fixed position with finite supplies and no relief, the method that costs fewest lives on both sides — and it leaves the buildings standing.
Its disadvantage is time, and time in this case meant a religious complex under encirclement for weeks in full view of the world, with the government’s authority visibly in question every day.
Why it matters here: a siege was among the alternatives available, it is what the officer in section 5.1 is understood to have proposed, and the reason it was rejected was political rather than military.
Before weighing whether any of that was necessary, it is worth being precise about which parts of it are established and which are not.
The operation is documented from three directions of very different quality.
Solid: the sealing of the state, the dates, the units involved, the use of tank main armament, the deaths of the three leaders, and the destruction of the Akal Takht. These are agreed by the army’s own accounts, by the government’s White Paper and by every critical source.
Documented: that Sinha advised against and was passed over. It appears in his own account and in standard military histories and is disputed by nobody, including the officers who conducted the operation.
Contested: the casualty figures and the number of pilgrims present. The state was sealed from 3 June and no independent observer was inside.
What none of it can show: what the pilgrim count was. The complex was sealed on a day of pilgrimage with people inside, and nobody counted them going in — which means the denominator for every casualty claim, on every side, is missing.
Which leaves the question that has been argued in India for forty-one years.
This is the most argued-about military decision in independent India and the government’s case is stronger than most Punjabi accounts allow.
By June 1984 an armed group with military expertise had fortified a complex in a border state, was killing police, journalists, Hindus and dissenting Sikhs, and was being resupplied. The state had no authority in Amritsar. Every month of delay strengthened the defences and the argument that the government could not govern. A siege of a religious site would have been a spectacle lasting weeks, would have brought crowds that could not be controlled, and would very likely have ended in an assault anyway, later, against better defences.
Nobody serious argues the state should have done nothing. The objection is to the form: a full military assault, with armour, on the holiest site of a religion, on a day of pilgrimage, with the state sealed and the press excluded, having passed over the officer who advised otherwise. Each of those was a separate choice. The date could have been different. The armour could have been withheld. The Reference Library could have been secured. And a government that had already decided the political cost was acceptable did not test whether it was, because it had removed the one professional voice telling it the cost.
What would settle it: Sinha’s position paper and whatever alternative it proposed. It has never been released.
Why people care so much: because if the assault was the only option, the Akal Takht is a casualty of an insurgency, and if it was one option among several, it is a casualty of a decision.
Both of those positions, though, are conducting an argument about soldiering, and the decisions that mattered were not made by soldiers.
Everybody arguing about June 1984 assumes that it was a military operation.
It is discussed as generalship. Was the plan sound, was the intelligence right, should tanks have been used, could the commandos have done it differently. Its defenders praise the execution and its critics attack it, and both are grading soldiers.
The decisive choices were not military and were not made by soldiers. The date was a political choice. The sealing of the state and the exclusion of the press was a political choice. The decision to assault rather than besiege was a political choice about how long a government could be seen to wait. And the choice of who would command was made by passing over the officer whose professional advice was inconvenient.
The military content of this operation was the part that got overruled — which is exactly the finding Part Six reached about the Forward Policy of 1961, in the same words, twenty-three years earlier.
The distinction changes where the failure sits. If Blue Star was bad generalship, the fix is better planning, and the Indian Army has spent forty years absorbing its tactical lessons. If it was a political decision wearing a uniform, then better planning fixes nothing, because the planners were not consulted on the question that mattered.
Watch which version each side needs. The Indian account needs the military framing because it makes the operation a regrettable necessity executed under fire. The Punjabi account needs it too, because a military assault on a shrine is a cleaner grievance than a political calculation about the optics of waiting. Neither account wants the version in which the decision was about how a government would look.
One item on the list of what was destroyed deserves separating from the rest, because it is the only one that cannot be repaired.
A library is not a building. It is the only copy of things.
The Sikh Reference Library held handwritten volumes, correspondence, and manuscript material assembled over generations — the documentary memory of a religion four centuries old.
There is no full catalogue of what was in it, which means there is no list of what is gone. A historian working on Sikh history today does not know what he cannot consult.
The Akal Takht was rebuilt within a few years. That is what a building is. The library is the part that cannot be.
Four months later, two of the men guarding the Prime Minister who ordered it shot her in her own garden. That is Chapter Six, and this part does not present it as justification for anything.
Lt Gen S.K. Sinha, Vice-Chief of Army Staff, was asked for a position paper, advised against an assault, proposed an alternative, and was passed over for Chief. The third supersession in this series.
Punjab was sealed on 3 June — curfew, communications cut, press excluded — on a day of pilgrimage. Tanks fired into the Akal Takht on the morning of 6 June. The Sikh Reference Library burned.
The decisive choices were political, not military — the date, the sealing, assault rather than siege, and who would command. The military content was the part that was overruled.
6The Thirty-First of October
The mobs had voter lists. That single fact settles what kind of event this was, and it is the fact most consistently left out of the word used to describe it.
6.1 — The morning
On the morning of 31 October 1984, Indira Gandhi was shot in the garden of her residence by two of her own bodyguards, Satwant Singh and Beant Singh. Beant Singh was killed immediately; Satwant Singh was tried and executed.
She died that day. Rajiv Gandhi was sworn in as Prime Minister that evening.
What followed in Delhi began within hours.
6.2 — The five days
Mobs formed and moved through the city. The worst of it was on 1 and 2 November, and it continued to the 7th.
Sikh men were dragged from homes, shops, buses and trains and killed, very often by being burned alive with kerosene and a tyre. Houses were looted and burned. Gurdwaras were destroyed. Women were assaulted.
It was not confined to Delhi. Killings occurred in Kanpur, in Bokaro and in other cities.
The army was available and was not deployed effectively for days.
6.3 — What made it organised
Three things distinguish this from a spontaneous outbreak, and every one is documented in the inquiries the Indian government itself appointed.
The mobs carried lists. Voter rolls and ration records were used to identify which houses were Sikh, in neighbourhoods where the mobs did not live. Somebody supplied those.
They were equipped. Kerosene and iron rods arrived in quantity, distributed rather than gathered.
The police did not act. Not everywhere and not universally — some officers acted and saved lives, and they should be named wherever the record allows. But across large parts of the city the police stood by, refused to register complaints, and in some documented instances disarmed Sikhs who were defending themselves.
Names of local politicians appear across the inquiries as having led or directed mobs. Chapter Seven follows what happened to those names.
Every element of section 6.3 comes from inquiries the Indian government itself appointed, principally the Misra and Nanavati Commissions and the committees between them.
The use of electoral rolls and ration records, the distribution of kerosene, and the failure of the police to register complaints or intervene are not allegations by victims’ organisations. They are findings.
That matters because the standard rebuttal to any account of November 1984 is that it comes from an aggrieved community. It does not. It comes from commissions appointed by successive Indian governments, several of them by governments of the party being examined.
What it cannot show: the chain of supply. Somebody obtained the rolls, copied them and distributed them. The inquiries established that this happened and did not establish who did it at each point, which is why Chapter Seven’s conviction record looks as it does.
6.4 — The sentence
Asked about the killings, the new Prime Minister said that when a great tree falls, the earth shakes.
Read as a description it is a metaphor about grief. Read as a response by the head of a government whose capital city had just produced two thousand seven hundred corpses in five days, it is something else: an explanation of the killings as a natural consequence, offered by the one person with the authority to stop them.
Riot: a violent public disturbance by a crowd, generally spontaneous, generally without prior organisation, and typically involving more than one group.
The word is used almost universally for what happened in Delhi in November 1984, including in the titles of the official inquiries.
It fits none of the facts in section 6.3. Crowds that arrive with electoral rolls identifying households by religion, carrying distributed kerosene, in areas where they do not live, while the police do not intervene, are not rioting. There is a different word for an organised, one-sided attack on a defined group with official passivity, and the inquiries themselves describe the elements of it while continuing to use the other word.
Part One promised to name the concealing word every time it appeared. Part Nine named “excesses”. This is the same operation performed by a different word.
One of those three facts does more work than the other two, and it is worth slowing down on.
An electoral roll is an ordinary administrative document. It lists, street by street and house by house, who lives where — and in India it lists their names, and Sikh names are identifiable.
A crowd forming spontaneously in anger does not have one. It attacks what it can see.
To have one, somebody must obtain it from a government office, copy it, and distribute it to people who will use it. That takes hours at minimum, and it requires access.
Which is why the lists are the single most important fact about November 1984. Everything else could be a crowd. The lists could not.
What is not in dispute is that it was organised. Who organised it, and at what level, is the question every inquiry has circled without closing.
That it was organised is established. By whom, and how far up, is the live question and the one every inquiry has circled.
What the inquiries actually established is local: named councillors and constituency-level figures leading mobs in their own areas, using rolls their own workers could obtain, in wards they knew. That is a pattern of many local actors behaving similarly in the same hours, which is what happens when a signal is understood rather than when an instruction is issued. No inquiry in forty-one years has found a document, an order or a meeting at the top, and several have looked.
Grant that no order was written. A capital city’s police force does not stand down simultaneously across dozens of stations for three days by coincidence, and the army was available and not deployed effectively. Whatever was or was not said, the operative decisions — that the police would not act and the army would not be used — could only be taken at the top, and they are what converted local mobs into three days of killing. And the head of government’s public characterisation of the deaths was that the earth shakes when a tree falls.
What would settle it: the Delhi Police control room records and deployment logs for 31 October to 3 November 1984. Their non-production has been raised repeatedly before the commissions.
Why people care so much: because “organised locally” and “directed centrally” carry entirely different consequences for a national party, and the difference between them is one set of missing records.
Whichever answer is right, two things this chapter deliberately does not do, and they should be stated before the summary. It does not present the assassination as anything other than a murder, and it does not present the killings that followed as a response to it in any sense that mitigates them. Two men killed a Prime Minister. What was then done in Delhi was done to people who had no connection to either of them beyond a name and a turban.
She was killed on the morning of 31 October 1984 by two of her bodyguards. The killing in Delhi began within hours and was worst on 1 and 2 November.
Three things make it organised, all documented by the government’s own inquiries: the mobs carried voter lists and ration records, kerosene was distributed, and the police largely did not act.
“Riot” fits none of those facts. A crowd forming in anger does not arrive with an electoral roll.
7The Counting
Two thousand seven hundred and thirty-three deaths. Five hundred and eighty-seven police reports. Twelve murder convictions in forty-one years. Part Two’s rule has never had a clearer application.
7.1 — The number
A committee under a retired official, appointed on the recommendation of an earlier commission, was given the specific task of establishing how many Sikhs were killed in Delhi between 31 October and 7 November 1984.
It reported in August 1987 and put the figure at 2,733. It also gave a national figure of about 3,325.
The later Nanavati Commission accepted 2,733, noting that Sikh representative bodies disputed it and claimed more than three thousand, and that in the absence of definite evidence the figure had to be treated as more or less correct.
Independent human rights organisations have put the Delhi figure nearer four thousand. Higher estimates circulate and are not established.
This part uses 2,733 throughout, for the reason Part One gave: when the state’s own figure is the lowest defensible one, use it, because a charge built on the state’s own number cannot be answered by disputing the number.
7.2 — The machinery
Over the following four decades the Indian state investigated this more than almost anything else in this series.
| Body | When | What it did |
|---|---|---|
| Marwah Commission | 1984 | Appointed to examine the police role. Wound up before completing its work; its material was passed on. |
| Misra Commission | 1985–86 | The first full inquiry. Recommended the further committees below. |
| Jain–Banerjee, Poti–Rosha, Jain–Aggarwal | 1987–1993 | A succession of committees on registration of cases. Their recommendations were repeatedly obstructed, including by writ petitions from those they named. |
| Kapoor–Mittal | 1987–1990 | On police conduct. Its two members could not agree on method. |
| Ahuja Committee | 1987 | Established the death toll: 2,733 in Delhi. |
| Narula Committee | 1993–94 | Recommended registration of cases against named politicians. |
| Nanavati Commission | 2000–2005 | The most comprehensive. Named individuals and found the police response gravely deficient. |
| Mathur Committee, SITs | 2015, 2018 | Re-investigation of closed cases. The 2018 team was set up under Supreme Court supervision to re-examine 186 of them. |
Four commissions, nine committees and two special investigation teams.
Untraced: a police classification closing an investigation on the ground that the offenders could not be identified.
It is a routine and often legitimate disposal. A burglary with no witnesses and no evidence goes untraced, and nobody is at fault.
It carries a different weight when applied to a killing in daylight, in a residential street, in front of neighbours, by a crowd that came from somewhere and dispersed to somewhere. Those cases are not hard to trace. They are hard to trace if nobody records the names on the day.
Why it matters here: about 240 of the 587 reports from November 1984 were closed this way, and the 2015 and 2018 investigation teams reopened files marked untraced and secured convictions from them.
7.3 — The arithmetic
Now set that machinery against what it produced.
2,733 deaths.
587 first information reports. That is one police report for roughly every five deaths, in a jurisdiction where the ordinary practice is one report per incident.
Around 240 of those reports were closed by the police as “untraced.”
About 250 produced acquittals.
25 cases have produced a conviction. Twelve of them are murder cases.
| Stage | Number | What it means |
|---|---|---|
| People killed in Delhi | 2,733 | The state’s own figure, established by a committee appointed for the purpose. |
| First information reports registered | 587 | One report for roughly every five deaths, where ordinary practice is one per incident. |
| Closed by police as untraced | about 240 | Investigation abandoned before any court saw the file. |
| Ending in acquittal | about 250 | Tried and not proved. |
| Producing a conviction | 25 | Of these, 12 are murder cases. |
| Individuals convicted | about 440 | Some on multiple charges arising from the same case. |
Read the second row against the first. The gap between 2,733 deaths and 587 reports is not a court’s failure and it is not a delay. It is a decision taken in police stations in the first week of November 1984, and everything below it in that table is a consequence.
The most senior conviction came in December 2018, thirty-four years afterwards, when the Delhi High Court reversed an acquittal and sentenced a former Member of Parliament to life imprisonment. A second life sentence followed in 2025.
7.4 — What the court said
The High Court in 2018 did not confine itself to the facts of the case. It said that the majority of those responsible had enjoyed political patronage, that they had been assisted by an indifferent law enforcement agency, and that they had escaped prosecution and punishment for more than two decades.
That is the Delhi High Court, in a judgment, in 2018, describing the conduct of the police and of politicians in 1984.
Twelve murder convictions for 2,733 deaths is one conviction for every 228 people killed.
Put it the other way. If a man was murdered in Delhi in the first week of November 1984, the chance that anybody has ever been convicted of a murder in connection with his death is under half a per cent.
And the first senior conviction arrived when the survivors giving evidence were in their sixties and seventies, and a great many of the witnesses, the accused and the police officers concerned were dead.
None of which is this book’s characterisation. Every element of it comes from the state’s own machinery.
Every figure in this chapter is the Indian state’s own, produced by bodies the Indian state appointed.
The death toll is a government committee’s. The FIR count is the Nanavati Commission’s. The disposal figures — untraced, acquitted, convicted — are court and police records. The characterisation of political patronage and indifferent policing is a High Court judgment.
That is the strongest evidential position available anywhere in this series, and it is why this chapter makes no estimates of its own and uses the lowest defensible number throughout.
What it cannot show: the true death toll. The claim that it is materially higher than 2,733 is credible — bodies were removed and in some accounts disposed of in the river, and a state that could not register more than 587 reports was not counting carefully. But the higher figures are estimates and this part does not need them.
All of which is usually explained by pointing at the legal system, and that explanation does not survive the arithmetic.
Everybody who discusses the impunity of 1984 assumes that justice failed in the courts.
It is the standard framing. Justice delayed. A slow legal system. Cases dragging for decades. Even sympathetic accounts describe an overburdened judiciary failing survivors, and the remedy proposed is always faster courts and special tribunals.
The courts were handed almost nothing to work with. Look again at the arithmetic. 587 reports for 2,733 deaths, and roughly 240 of them closed by the police as untraced before any judge saw them.
A criminal case is built at the beginning, not the end. It is built by an officer who records a complaint on the day, who takes statements while witnesses remember, who collects what is at the scene, and who names the accused. Everything a court can eventually do is limited by what that officer did in the first week.
And the officers who would have done that in November 1984 are the same force that Chapter Six describes standing by while it happened, refusing to register complaints, and in documented instances disarming the victims. The investigators were the people who had already declined to intervene.
Which is why thirty-four years of commissions produced twelve murder convictions. Not because judges were slow, but because there was nothing to try. The 2015 and 2018 teams were reduced to reopening files marked untraced and re-examining witnesses in their seventies, and one of them produced a conviction — which proves that the cases had been available and had not been built.
The general rule this series keeps meeting: where a state fails to record something, the failure is complete at the moment of not recording it, and everything afterwards is an attempt to reconstruct from an absence. Part Two applied it to the Partition dead. Part Nine applied it to detentions never entered in a register. This is the same finding at its most consequential, because here the thing not recorded was murder.
Part Two’s rule was that when a state which counts everything has no count of something, somebody decided that. This chapter is the harder version. Here the state did count — twice, carefully, by dedicated committee — and then did not investigate what it had counted. The number is not the missing thing. The cases are.
2,733 killed in Delhi, 3,325 nationally — the state’s own figures, which this part uses because a charge built on them cannot be answered by disputing them.
Four commissions, nine committees, two investigation teams. 587 police reports for 2,733 deaths. 240 closed as untraced. 25 convictions, 12 for murder. The Delhi High Court in 2018 described political patronage and indifferent policing.
The courts were handed almost nothing. A case is built in the first week, by the same force that had stood by — and where a state fails to record something, the failure is complete at that moment.
8The Decade After
The insurgency was defeated. What was done to defeat it has never been investigated, except in one district, by one man, who was abducted and killed for doing it.
8.1 — The accord, and what happened to it
On 24 July 1985, Rajiv Gandhi and the Akali leader Harchand Singh Longowal signed an accord. It promised the transfer of Chandigarh to Punjab by January 1986, a commission on the disputed Hindi-speaking areas, and a tribunal to reassess the river waters.
It was a genuine settlement of Chapter One’s list and it should be counted as a credit item. A Prime Minister who had been in office nine months negotiated directly on the substance and signed.
Longowal was assassinated on 20 August 1985, less than a month later, by militants who regarded the accord as a betrayal.
Elections followed in September and the Akali Dal won. Surjit Singh Barnala became Chief Minister.
Chandigarh was not transferred in January 1986. The date passed, then another date, then the question was referred to a commission. It has still not been transferred, forty years later.
The tribunal on the river waters was constituted in 1986 and, as Part Five recorded, issued an interim report and never a final award.
In May 1987 the Barnala government was dismissed and Punjab was placed under President’s Rule, where it remained for about five years.
So the accord delivered an election and none of the substance, and the elected government it produced was dismissed within two years.
8.2 — The special legislation
The counter-insurgency ran on a statute passed in 1985 and re-enacted in 1987: the Terrorist and Disruptive Activities (Prevention) Act, generally called TADA.
Its provisions departed from ordinary criminal law in specific ways. Confessions made to police officers were admissible in evidence, which ordinary Indian law does not permit for good reason. Detention without charge was permitted for extended periods. Trials could be held in camera and witness identities protected. In defined circumstances the burden shifted towards the accused.
Tens of thousands of people were detained under it across India. Its conviction rate was in the low single figures — around one or two per cent of those arrested.
Read those two facts together. A law under which very large numbers were detained and almost nobody was convicted was not principally functioning as a prosecution statute. It was functioning as a detention statute, which is Part Two’s finding arriving under a new name in a new decade.
Enforced disappearance: the detention of a person by agents of the state, followed by a refusal to acknowledge the detention or to say where the person is.
The distinguishing feature is not the killing but the denial of information. A family with a body has a death. A family without one has nothing: no inquest, no certificate, no case to bring, no inheritance, no widowhood, and no end.
It is treated in international law as a distinct offence for that reason, and as a continuing one — the offence does not finish when the person dies, because the denial persists.
Why it matters here: this is what Jaswant Singh Khalra was documenting, and the reason his method was cremation records is that the state had produced no other record at all.
8.3 — The man who counted
Jaswant Singh Khalra was a bank employee and human rights worker in Amritsar. In the early 1990s he did something nobody else had done: he went to the municipal records of cremation grounds.
Cremations require firewood, and firewood is purchased, and purchases are recorded. Working from those records in three crematoria in Amritsar district, he documented large numbers of bodies cremated by police as unidentified, and published his findings.
On 6 September 1995 he was abducted from outside his home. He was killed. In 2005 six police officials were convicted of his abduction and murder, and the sentences were enhanced on appeal in 2007.
The Supreme Court directed the Central Bureau of Investigation to examine his allegations. Its report confirmed approximately 2,097 cremations in Amritsar district, of which a minority were fully identified, a further group partially identified, and well over a thousand not identified at all.
Two things about that figure and they are both essential.
It is one district. Punjab has around twenty. No equivalent examination was made anywhere else, and none has been made since.
It exists because one man looked at firewood receipts. There was no commission of inquiry into the Punjab counter-insurgency. There has never been one.
This chapter has the weakest evidential base in the entire series, and the reason is the finding.
Solid: the accord and its terms, Longowal’s assassination, the election results, the dismissal of 1987, the text of TADA, the fact that Chandigarh was not transferred, the Khalra convictions and the CBI’s figure for Amritsar district. All of these are documents, judgments or things anybody can check.
Estimated: total deaths across the period. The figure in general use is above twenty thousand and it is compiled from press reporting, police statements and human rights work rather than from any official count. Nobody made one.
Unknown: essentially everything about method. How many were detained under TADA in Punjab, for how long, and with what outcome. How many died in custody. How many disappeared. How many of the dead described as militants were.
What this cannot show, and the reason: there was no commission. Every other episode of comparable scale in this series produced an official inquiry whose findings can be quoted and argued with. This one produced nothing, and the one private investigation that existed ended with its author’s murder. An absence of evidence here is not neutral — it is the direct product of a decision not to look.
8.4 — How it ended, and the ledger
The insurgency was defeated between 1991 and 1993, by a police campaign under a Director General appointed for the purpose, using the powers described above. Elections were held in February 1992 with a turnout of around a fifth of the electorate, boycotted by the Akalis. Beant Singh became Chief Minister and was himself assassinated in 1995.
Total deaths across the whole period, from all causes and all sides, are generally estimated at over twenty thousand. The great majority were Sikhs.
| Where | What was done |
|---|---|
| Punjab | River waters by notification 1976; elected government dismissed 1980; Supreme Court case ended by agreement 1981; military assault on the Akal Takht June 1984; Chandigarh promised in 1985 and never transferred; elected government dismissed 1987; President’s Rule for about five years; over twenty thousand dead. |
| Delhi | 2,733 killed in five days. 587 reports, 25 convictions. |
| Kanpur, Bokaro and elsewhere | Killings in the same week. Investigated far less than Delhi and largely absent from the national account. |
| Haryana | Received 3.5 MAF of Ravi–Beas water by the 1976 notification and has built its share of the canal. Its own case is real and is set out in Part Five. |
Two thousand and ninety-seven cremations, in one district, of bodies the police recorded as unidentified.
Punjab has around twenty districts. If Amritsar were representative — and nobody knows whether it was, because nobody looked — the figure for the state would be a number with a different shape entirely.
What is certain is narrower and stranger. The only reason anybody knows about the two thousand is that one man reasoned that cremations need firewood, that firewood is bought, and that purchases are written down.
He was abducted and killed for that reasoning, by policemen who were convicted of it ten years later.
Which produces the hardest argument in this part, because both sides of it are true.
It worked. Whether that settles anything is the question, and both positions are held by people who lived through it.
By 1990 the militancy was killing thousands a year, overwhelmingly Sikhs, including anybody in Punjab who opposed it. Ordinary policing had failed for a decade. The campaign of 1991 to 1993 broke it, elections resumed, and the state has had three decades of peace and normal politics since. Those are not small things and no alternative method was demonstrated by anybody. The people who criticise the methods most freely are living in the peace they produced.
The result is not in dispute and the objection is not to the result. It is that a decade of counter-insurgency in an Indian state has never been examined by any commission of any kind. The one systematic count that exists covers a single district, was compiled privately from cremation receipts, and its author was abducted and murdered by policemen who were subsequently convicted of it. A state confident that its methods were lawful does not leave that record uninvestigated for thirty years.
What would settle it: a commission of inquiry with the powers the Shah Commission had. None has ever been appointed and none has been seriously proposed.
Why people care so much: because the peace is genuinely valuable and nobody in Punjab wants it disturbed, which is precisely the argument that has prevented the question being asked.
One measure of how completely this decade has been left alone. Every other episode in this series with a comparable death toll has a commission attached to it whose report can be named, quoted and argued with — Partition, Hyderabad, 1962, the Emergency, November 1984. For Punjab between 1987 and 1993 there is no document to argue with, because none was ever produced.
The 1985 accord promised Chandigarh by January 1986. Longowal was assassinated within a month. Chandigarh has still not been transferred. The Barnala government was dismissed in 1987 and Punjab spent about five years under President’s Rule.
TADA: confessions to police admissible, extended detention without charge, in camera trials. Tens of thousands detained and a conviction rate of one or two per cent — a detention statute wearing a prosecution statute’s name.
Jaswant Singh Khalra counted cremations from firewood receipts in one district and was abducted and killed. The CBI confirmed about 2,097. Punjab has twenty districts, and there has never been a commission of inquiry into the counter-insurgency.
9What Would Have Prevented It
The defence at full strength, four moments, and the thing a Punjabi account gets wrong — which is the part of this book I found hardest to write.
9.1 — Four arguments for the defence
These are the strongest arguments available to a defender of the Indian state’s conduct. Two of them are conceded outright.
One. By 1983 the state faced a genuine armed insurrection. Chapter Four establishes it and this part does not qualify it. A government that left the Akal Takht fortified would have been abandoning a state.
Two. The 1985 accord was a real attempt at settlement. A Prime Minister negotiated on the substance and signed within a year of the worst events of the period. That is not the conduct of a government uninterested in a political solution, and the man who signed it for Punjab was killed by Punjabis for signing.
Three. The insurgency was defeated and the peace has held for three decades. Conceded. Punjab has had normal politics, contested elections and changes of government since 1997. Very few insurgencies anywhere end like that.
Four. Most of the killing was not done by the state. Over twenty thousand died across the period, and the militancy killed a very large share of them — including most of the Sikhs who died. Any account presenting this as a state’s campaign against a community has to account for that and usually does not.
9.2 — Four moments
Part One requires an alternative that was available at the time. There were four, and the striking thing is how cheap the first three were.
1976: a tribunal instead of a notification. The Inter-State River Water Disputes Act existed. Using it would have taken longer and produced an outcome Punjab might well have disliked — and it would have been an adjudication rather than an executive act. Cost: delay.
1981: letting the case run. Punjab’s suit was pending in the Supreme Court. Nothing required the Prime Minister to broker its withdrawal. Cost: the risk of losing.
1980 and 1987: not dismissing the elected governments. Neither dismissal was compelled. The 1980 one was part of a general use of Article 356 after a change at the centre; the 1987 one removed the government produced by the accord. Cost: living with opponents in office.
1983: the position paper. Chapter Five’s finding. The officer asked for professional advice gave it, and was passed over. Cost: taking advice.
Not one of those required a concession on the substance of any Punjabi demand. All four were procedural.
Price the four alternatives in section 9.2.
A tribunal instead of a notification in 1976 would have cost the central government several years of delay and the possibility of an award it disliked.
Letting the case run in 1981 would have cost it the risk of losing a lawsuit.
Not dismissing the elected governments in 1980 and 1987 would have cost it opponents in office in one state.
Taking the advice in 1983 would have cost it a slower and more embarrassing operation, and the appointment of a Chief of Army Staff who had disagreed with it.
Set that list against what followed: an assault on the Akal Takht, a Prime Minister assassinated, 2,733 people killed in Delhi in five days, and over twenty thousand dead across fifteen years.
Nothing about that comparison establishes causation and this part does not claim it does. What it establishes is the price of the alternatives, which is the only side of the ledger anybody can actually price.
9.3 — What I would have to retract
Three things would break this part.
If Lieutenant General Sinha’s position paper, released, showed that he proposed no workable alternative and merely recorded objections, Chapter Five’s central charge weakens to a complaint about tone.
If the record of the 1982 and 1983 negotiations showed that the Akali leadership was offered the substance of Chapter One’s list and refused it, Chapter Four’s verdict tips and much of Chapter One goes with it.
If a systematic study of the militancy’s recruitment showed that it drew on people with no connection to the grievances of Chapter One, then the space argument in Chapter Three fails.
The first two depend on files the Indian government holds. That is the sixth part in a row ending on that sentence.
The four alternatives in section 9.2 are unusual in this series because none of them is a counterfactual about outcomes. Each is a claim about a procedure that existed and was not used, and procedures leave records.
The Inter-State River Water Disputes Act was on the statute book in 1976 and a tribunal was eventually appointed under it in 1986, which establishes that it was available. The Supreme Court suit was live in 1981; its withdrawal is on the file. The dismissals of 1980 and 1987 were proclamations laid before Parliament. And the position paper of 1983 was commissioned by the government itself.
That is why Chapter Nine can name alternatives with more confidence than most parts of this series. Availability is a matter of record rather than of modelling.
What it cannot show: whether taking any of them would have changed the outcome. A tribunal in 1976 might have produced the same allocation and the same anger. The claim is that the alternative existed and was cheap, not that it was a guarantee.
Which leaves the thing this part has been circling since the front matter, and it is not a question about Punjab.
Everybody who writes about these fifteen years — in Delhi, in Chandigarh and in the Sikh diaspora — assumes that this was a Sikh problem.
The Indian version: a religious separatism in a border state. The Punjabi version: a community’s persecution by a state that saw it as suspect. Both treat the conflict as belonging to a people, and both are organised around what was done to or by Sikhs as Sikhs.
Two facts sit badly with that on either side.
The first: not one instrument used here was designed for Punjab. The dismissal of elected state governments, the settlement of a dispute by notification instead of adjudication, the ending of litigation by political agreement, the passing over of a professional who advised against, detention without trial, President’s Rule for years at a time — every one of them appears elsewhere in this series, in Kerala, in Kashmir, in the Emergency, applied to people with no religious distinctiveness at all. Punjab did not receive special treatment. It received the ordinary treatment, at full intensity, in a state where it happened to land on a community with a distinct identity to organise around.
The second, and this is the one I did not want to write: the majority of the Sikhs who died in these fifteen years were killed by Sikhs. The militancy killed police officers, politicians, journalists, moderates, accord signatories and villagers, and most of them were Sikh. An account of this period built around Sikh martyrdom at Indian hands erases them, and it is the account I grew up inside.
Which produces a finding neither side wants. The causes were federal and generic. The form was religious and local. And the dead were overwhelmingly from one community, killed by both the state and by men claiming to act for that community — which is the ordinary arithmetic of insurgency everywhere and is described that way almost nowhere in Punjab.
I have written about this material before with the community framing, and I now think it was wrong — not in its facts but in what it made those facts mean. The Indian version denies Punjab a legitimate constitutional grievance. The Punjabi version denies Punjab its own dead. Both are more comfortable than the arithmetic.
That is the case against this part. Two of the four defences I think simply win, and I have said so rather than arranging the chapter so they lose. What none of the four reaches is section 9.2, because every alternative there was procedural, available and cheap, and none of them required conceding a single Punjabi demand.
Four defences: a genuine armed insurrection by 1983; the 1985 accord was a real attempt; the insurgency was defeated and the peace has held; and most of the killing was not done by the state. Two are conceded outright.
Four moments, all procedural, none requiring any concession on substance: a tribunal in 1976, letting the case run in 1981, not dismissing the governments in 1980 and 1987, and taking the advice in 1983.
The causes were federal and generic. The form was religious and local. And the majority of the Sikhs who died were killed by Sikhs — which the account I grew up inside does not say.
10An Honest List of What We Do Not Know
Five days in Delhi have been investigated fourteen times. Nine years in Punjab have been investigated once, privately, by a man who was killed for it.
10.1 — Genuinely unknown
Six things in this part cannot be established, and their distribution is the most important thing about them.
What Lieutenant General Sinha’s position paper proposed. Chapter Five’s charge rests on there having been a workable alternative, professionally set out, before the operation.
Why it is unknown: it has never been released. This is the single document that would most change this part.
How many died in Operation Blue Star. The government’s White Paper gave figures in the region of five hundred civilians and militants and eighty-three soldiers. Independent estimates run higher.
Why it is unknown: the state was sealed, communications cut and the press excluded from 3 June. Nobody independent was present. And the number of pilgrims inside the complex when it was sealed was never established, which means the denominator is missing.
The true toll of November 1984. The state’s figure is 2,733 and this part uses it.
Why it is unknown: bodies were removed, 587 reports were registered for 2,733 recorded deaths, and about 240 of those were closed as untraced. A count that thin cannot be assumed to be complete.
What was done in Punjab between 1987 and 1993. One district was examined, privately, from cremation records.
Why it is unknown: there has never been a commission of inquiry. This is not a closed file; it is an investigation that was never held, which is a different and worse category.
Whether Congress deliberately promoted Bhindranwale. Chapter Three grades this Alleged.
Why it is unknown: it rests on statements by participants made afterwards, in a period when everybody had reason to place the responsibility elsewhere.
What was actually offered in the negotiations of 1982 and 1983. Chapter Four’s argument box turns on how close a settlement came.
Why it is unknown: accounts differ sharply and the files are not open.
10.2 — Solid
The chronology of closure. The notification of March 1976, the dismissal of 1980, the suit of July 1979 and its withdrawal after the agreement of December 1981. All dated public acts.
The text of the Anandpur Sahib Resolution, including both its operative demands and its preamble, and the fact that it did not change between 1973 and 1984.
The killings of 1978 to 1983. The Nirankari clash of 13 April 1978, the assassination of a newspaper editor in September 1981, the killing of a Deputy Inspector General on the temple steps in April 1983, and the Dhilwan bus killings of October 1983. All contemporaneous public record.
That Sinha advised against the operation and was passed over. Not disputed by anybody, including by the officers who conducted it.
The conduct of Operation Blue Star. The sealing of the state from 3 June, the assault of 5 and 6 June, the use of tank main armament against the Akal Takht, the destruction of the Akal Takht and the burning of the Sikh Reference Library.
Everything in Chapter Seven. 2,733 and 3,325; 587 reports; roughly 240 untraced; 25 convictions of which 12 for murder; four commissions, nine committees, two teams; and the Delhi High Court’s findings of political patronage and indifferent policing.
The accord of July 1985, Longowal’s assassination in August 1985, and the non-transfer of Chandigarh — which can be verified by looking at a map today.
The Khalra case. His documentation, his abduction on 6 September 1995, the conviction of six police officials in 2005 and the enhancement of sentences in 2007, and the CBI’s confirmation of approximately 2,097 cremations in one district.
10.3 — The asymmetry
Set the two halves of section 10.1 against each other.
For five days in Delhi: four commissions, nine committees, two investigation teams, a death toll established by a dedicated committee, an FIR count, a conviction count, and High Court findings. Insufficient, slow, and obstructed — and it exists.
For nine years in Punjab: nothing. No commission. No committee. No official count of the dead, the detained or the disappeared. One private inquiry into one district, whose author was murdered by policemen who were later convicted of murdering him.
Both sets of events were carried out substantially by agents of the same state, in the same decade, against substantially the same community.
I do not think there is a single explanation for that and I am not going to invent one. What I will say is that the Delhi killings were done by a party’s supporters and could therefore be investigated by its successors, and the Punjab counter-insurgency was done by the police and the army, which no Indian government of any party has ever shown an appetite for examining.
Part Nine ended by noting that three national catastrophes had produced three official investigations that are harder to obtain in India than abroad. This part adds a fourth case and it is worse, because here there was nothing to withhold.
Genuinely unknown: what Sinha’s paper proposed, how many died in Blue Star, the true toll of November 1984, what was done in Punjab from 1987 to 1993, whether Bhindranwale was deliberately promoted, and what was offered in 1982–83.
Solid: the chronology of closure, the text of the resolution, the killings of 1978–83, that Sinha was passed over, the conduct of Blue Star, everything in Chapter Seven, the 1985 accord, and the Khalra case.
Five days in Delhi produced fourteen official inquiries. Nine years in Punjab produced none. Same state, same decade, substantially the same community.
Sources & further reading — Part 10
Timeline, 1973–1995
Dates only. The arguments are in the chapters.
| Date | What happened |
|---|---|
| 17 Oct 1973 | The Anandpur Sahib Resolution is adopted by the Akali Dal working committee. |
| Mar 1976 | Central notification divides the Ravi–Beas waters, leaving Punjab about a fifth of the flow. Part Five’s Chapter Five. |
| 1977 | An Akali government takes office in Punjab after the Emergency. |
| 13 Apr 1978 | Clash with the Nirankari convention at Amritsar. Thirteen Sikhs killed. |
| Oct 1978 | The resolution is endorsed at the Akali conference at Ludhiana. |
| 11 Jul 1979 | Punjab files suit in the Supreme Court challenging the 1976 notification. |
| Feb 1980 | Congress returns at the centre. The Akali government in Punjab is dismissed. |
| Apr 1980 | The head of the Nirankari sect is assassinated in Delhi. |
| 9 Sep 1981 | Lala Jagat Narain, newspaper editor, is assassinated. Bhindranwale is arrested in September and released in October. |
| 31 Dec 1981 | Tripartite water agreement brokered by the Prime Minister. Punjab withdraws its Supreme Court case. |
| Aug 1982 | The Akali Dal launches the Dharam Yudh Morcha. |
| 25 Apr 1983 | DIG A.S. Atwal is shot dead on the steps of the Golden Temple. |
| 5 Oct 1983 | Six Hindu passengers are separated by religion and killed at Dhilwan. |
| 6 Oct 1983 | The Punjab government is dismissed. President’s Rule imposed. |
| 1983 | Lt Gen S.K. Sinha, Vice-Chief of Army Staff, advises against an assault on the complex and is passed over for Chief. Gen A.S. Vaidya is appointed. |
| Dec 1983 | Bhindranwale moves into the Akal Takht. Shabeg Singh organises the defences. |
| 3 Jun 1984 | Punjab is sealed: curfew, communications cut, press excluded. It is the martyrdom anniversary of Guru Arjan Dev and the complex is full. |
| 5–7 Jun 1984 | Operation Blue Star. Tanks fire into the Akal Takht on the morning of the 6th. Bhindranwale, Shabeg Singh and Amrik Singh are killed. The Akal Takht is destroyed and the Sikh Reference Library burns. |
| Jun 1984 | Sikh soldiers desert or mutiny in several units across India. |
| 31 Oct 1984 | Indira Gandhi is assassinated by two of her bodyguards. Killing begins in Delhi within hours. |
| 1–2 Nov 1984 | The worst of it. Mobs use voter lists and ration records. 2,733 are killed in Delhi by 7 November; about 3,325 nationally. |
| 24 Jul 1985 | The Rajiv–Longowal accord. Chandigarh to be transferred by January 1986; a tribunal on the waters; a commission on the disputed areas. |
| 20 Aug 1985 | Longowal is assassinated, less than a month after signing. |
| Sep 1985 | Elections. The Akali Dal wins. Barnala becomes Chief Minister. |
| Jan 1986 | Chandigarh is not transferred. It has still not been transferred. |
| 1985, 1987 | TADA is enacted and re-enacted. Confessions to police become admissible; detention without charge is extended. |
| May 1987 | The Barnala government is dismissed. President’s Rule for about five years. |
| 1987 | The Ahuja Committee reports the Delhi death toll at 2,733. |
| 1991–93 | The police campaign that defeats the insurgency. Elections held in February 1992 with about a fifth of the electorate voting. |
| 1995 | Jaswant Singh Khalra publishes his findings on secret cremations. He is abducted on 6 September and killed. Chief Minister Beant Singh is assassinated in August. |
| 2005 | Six police officials are convicted of Khalra’s abduction and murder; sentences enhanced in 2007. The Nanavati Commission reports on 1984. |
| 2018, 2025 | A former Member of Parliament is convicted and sentenced to life over November 1984, and again in a second case. Twenty-five cases in total have produced convictions. |
Part Eleven starts here. It takes November 1984 on its own — not as the end of this part but as the beginning of a question about how India handles organised communal violence, and what has changed since.
Glossary
Every hard word used in this part, in plain English.
| Word | What it means |
|---|---|
| Akal Takht | The seat of temporal authority in Sikhism, established 1606, facing the Harmandir Sahib. Directions binding on the community issue from it. Destroyed in June 1984. |
| Anandpur Sahib Resolution | The Akali Dal’s charter of demands, adopted October 1973. Its central constitutional demand was that the Union be limited to defence, foreign relations, currency and communications. |
| Dharam Yudh Morcha | The agitation launched by the Akali Dal in August 1982 to press the resolution’s demands. Literally a campaign of righteous struggle. |
| Federalism | Division of powers between a centre and constituent units, with each holding authority the other cannot simply remove. The Indian version is heavily weighted to the centre. |
| FIR (first information report) | The document a police station records when a cognisable offence is reported. Everything a court can later do is limited by what it contains. 587 were registered for 2,733 deaths. |
| Harmandir Sahib | The Golden Temple: the central place of Sikh worship, standing in the same complex as the Akal Takht and distinct from it. |
| Nirankari | A religious movement regarded by orthodox Sikhs as heterodox. The clash of April 1978 was with one of its conventions. |
| Riot | A spontaneous violent disturbance by a crowd, generally unorganised and usually involving more than one group. Used for November 1984 and fitting none of its documented features. |
| Siege | Encircling a fortified position and waiting rather than assaulting it. Slower, politically uncomfortable, and cheaper in lives and buildings. |
| TADA | The Terrorist and Disruptive Activities (Prevention) Act, 1985 and 1987. Confessions to police admissible, extended detention without charge, in camera trials. Conviction rate one to two per cent. |
| Untraced (case closure) | A police classification closing an investigation on the ground that the offenders cannot be identified. About 240 of the 587 reports from November 1984 were closed this way. |
What Comes Next
Part Eleven — November 1984
Part Eleven takes five days on their own. Chapter Seven of this part established the arithmetic; Part Eleven asks what it means. It asks:
- What is the actual documentary record of who directed the mobs — not what is alleged, but what has been found by a commission or a court, and against whom?
- The Delhi Police were the investigators of a crime in which the Nanavati Commission found them gravely deficient. Was that ever addressed, and by whom?
- Compensation was paid. To how many families, on what basis, and how does it compare with what has been paid after other organised violence in India?
- Part Two’s rule about counting has now been applied four times. What does the pattern across Partition, Hyderabad, the Emergency and 1984 actually establish about the Indian state?
- India has had repeated episodes of organised communal violence since. What, if anything, changed procedurally after 1984, and what is the evidence that it worked?
- The word “riot” is still used officially. Who benefits from it, and what would change if it were not?
Questions, not topics. If any of them is answered badly, you will be able to tell.