Federative Republic of Brazil

The name on the map needs explaining, because it belongs to the state and not to the regime, and because it is still in use. República Federativa do Brasil — rendered by the map's dataset as the Federated Republic of Brazil, and in Brazil's own official English usage as the Federative Republic of Brazil — became the country's official name with the constitution of 24 January 1967, replacing the Estados Unidos do Brasil (United States of Brazil) adopted at the fall of the Empire in 1889. The constitution of 1988, written after the dictatorship, kept it: its first article still defines the Federative Republic of Brazil as a democratic state of law. The name therefore spans the last three years of military rule, the transition and every government since, and it is printed on Brazilian passports today. Nothing about it marks a break in 1985, and the entry's dates should not be read as the lifespan of a state. The period 1964–1985 is the span of a regime that governed a state which carried on, under the same name and almost the same borders, before and after it.

What makes the case worth close attention is that the regime combined high growth with routine torture, then negotiated its own exit without accounting for either. Between 1968 and 1973 the economy grew at an average of about 11 per cent a year, an expansion advertised as the milagre econômico; in the same years the political police ran interrogation centres in which prisoners were beaten, burned and shocked, and in which some were killed and their bodies hidden. The dictatorship did not fall; it withdrew, on terms it largely set, having passed an amnesty law in 1979 that covered torturers as well as their victims. The constitution of 1988 dismantled most of its legal architecture, but the amnesty survived, upheld by the Supreme Federal Court in 2010. Only in 2014, when a National Truth Commission delivered its report, did the Brazilian state formally describe what it had done.

At a glance

Why did the republic break in 1964?

João Goulart never had an easy mandate. He was vice-president when Jânio Quadros resigned in August 1961, and the military ministers tried to block his inauguration; the compromise that let him take office stripped him of most of his powers under a parliamentary system, which a plebiscite in January 1963 reversed by more than four to one. He then governed as a full president with an unstable coalition, a Congress in which his own party was divided, and an economy that had stopped growing: GDP rose 0.6 per cent in 1963 while inflation ran at about 80 per cent. His answer was a programme of reformas de base — agrarian reform, urban reform, tax reform, education, the vote for illiterates and enlisted men — announced as a break with the constraints of the 1946 constitution rather than as legislation to be negotiated through it.

The crisis of early 1964 was therefore about method as much as content. On 13 March Goulart signed decrees expropriating land beside federal highways and nationalising private oil refineries, and announced the rest of the programme to a rally in Rio de Janeiro. Six days later the March of the Family with God for Freedom brought several hundred thousand people into the streets of São Paulo, and similar marches followed elsewhere. On 25 March sailors refusing orders over pay and conditions occupied their union building in Rio, and Goulart granted them amnesty — a decision that convinced senior officers that discipline in the ranks had collapsed. On 30 March he spoke at the Automóvel Clube to a gathering of sergeants and called them his comrades. That night the army commander in Minas Gerais, Olympio Mourão Filho, began moving troops towards Rio.

The coup took three days and killed almost nobody. Mourão's column advanced; the First Army in Rio joined it on 1 April rather than fight; Goulart flew from Rio to Brasília and then to Porto Alegre, found no division willing to resist, and on 4 April crossed into Uruguay. On 2 April the president of the Senate declared the presidency vacant while Goulart was still in the country, and Congress installed the speaker of the Chamber, Ranieri Mazzilli, as interim president. The United States had been preparing for the possibility since 1963: a naval task force with an aircraft carrier and tankers sailed towards Santos and an airlift of fuel and ammunition was assembled, and both were withdrawn once it was clear the Brazilian army would not split.

Calling this a military coup is accurate but incomplete, and the document that followed said so itself. The first Institutional Act, issued on 9 April by the three service ministers acting as a "Supreme Command of the Revolution", opens by fixing "the concept of the civil and military movement" and claims that "the victorious revolution invests itself in the exercise of constituent power". Its preamble states that "the constitutional processes did not work to remove the government, which deliberately intended to bolshevise the country" — a phrase quoted by both sides ever since. The act was signed by the commanders of the three services, but it was made possible by governors, newspapers, the conservative wing of the Catholic Church, industrialists organised in the Instituto de Pesquisas e Estudos Sociais, and a middle class frightened by inflation and by the word reform. That collaboration was real and public, and it underlies the argument, now standard among Brazilian historians, that 1964 was a civil-military coup.

How did a coup become a regime?

Most of the men who took power in April 1964 intended to be gone within a year. Castelo Branco, elected by Congress on 11 April and inaugurated on 15 April, was the candidate of the army's "Sorbonne" group — officers who had served in Italy with the Brazilian Expeditionary Force, admired constitutional forms and wanted a short, legal, modernising intervention. His economic ministers, Roberto Campos and Octávio Gouvêa de Bulhões, produced the Plano de Ação Econômica do Governo: spending cuts, credit restriction, tax reform, a central bank, and a wage formula setting pay increases below expected inflation. Inflation fell from about 92 per cent in 1964 to roughly 25 per cent by 1967, and wage-earners carried the cost: the real minimum wage fell in every year of the period, and the union movement was placed under supervision. The same government created the National Housing Bank and the severance fund, the FGTS, which replaced the old stability rule that had given long-serving workers a claim on their jobs.

Castelo Branco's legalism was real but conditional, and the condition was that the left should not win elections. In October 1965 the opposition won gubernatorial elections in Guanabara and Minas Gerais, and on 27 October the government issued the second Institutional Act, abolishing all existing parties, restoring the indirect election of the president by Congress and authorising purges of the civil service. A complementary act then manufactured two parties: the National Renewal Alliance, ARENA, which supported the government, and the Brazilian Democratic Movement, MDB, licensed to oppose it. The third act, in February 1966, made governors and vice-governors indirectly elected by the state assemblies and turned the mayors of capitals and "national security" towns into appointed offices. The fourth, in December 1966, recessed Congress and reconvened it to vote on a draft constitution within a fixed timetable.

The constitution of 24 January 1967 was approved by a Congress from which the government had removed its most vocal opponents, and it did in constitutional language what the acts had done by decree. It renamed the country the Federative Republic of Brazil, kept the two-party system, gave the president a decree-law power, and wrote national security into the text. A new National Security Law, a press law and a military justice system competent to try civilians completed the structure. Castelo Branco handed over to his war minister, Artur da Costa e Silva, on 15 March 1967, leaving a regime in which the forms of constitutional government were intact and its substance was not.

Costa e Silva was expected to be the hard-liners' president, and he was, though he began by promising to "humanise" the revolution. His government faced a strike wave in 1968 — at Contagem in Minas Gerais in April, at Osasco in São Paulo in July — and a student movement that had begun with the killing of a secondary-school student, Edson Luís, by military police in Rio de Janeiro on 28 March 1968 and culminated in the March of the Hundred Thousand through the city on 26 June. The MDB did well in the municipal elections that year, and in September a deputy, Márcio Moreira Alves, urged Brazilians to boycott the independence-day parades and urged women to withhold their favours from officers. The government demanded that Congress lift his immunity. On 12 December the Chamber refused, by 216 votes to 141.

What did AI-5 change?

The fifth Institutional Act was issued the next day, 13 December 1968, and it was not an instrument of the same kind as its predecessors. Where the earlier acts had amended the constitution, AI-5 suspended it. It gave the president power to recess Congress and the state legislatures and legislate in their place; to intervene in states and municipalities without constitutional limit; to suspend political rights for ten years and annul elected mandates; to dismiss or retire judges, professors and civil servants; to confiscate property; and to impose prior censorship on the press, music, theatre and film. Its tenth article suspended habeas corpus "in cases of political crimes, crimes against national security, the economic and social order and the popular economy" — the clause that made torture practicable, because a prisoner's family could no longer ask a judge to produce him.

Congress was closed for ten months and 41 federal deputies lost their mandates, along with senators, state deputies and councillors. Purges of the universities and the judiciary, the censorship of newspapers and the arrest of writers and musicians followed within weeks. The regime had found the form it would keep until 1978: the constitution as a text, AI-5 as the switch that turned it off whenever the government needed.

In August 1969 Costa e Silva suffered a cerebral thrombosis and was left incapacitated. The constitution made his civilian vice-president, Pedro Aleixo, the successor, and Aleixo had voted against AI-5. The three service ministers therefore issued the twelfth Institutional Act on 31 August, declaring the president temporarily impeded and taking over as a junta: Admiral Augusto Rademaker, General Aurélio de Lira Tavares and Brigadier Márcio de Sousa e Melo. Their two months in office produced a burst of legislation. On 4 September the American ambassador, Charles Elbrick, was kidnapped by a small revolutionary group; the government released fifteen prisoners and flew them into exile, then issued the thirteenth act creating the penalty of banishment, followed by a fourteenth imposing death sentences for "revolutionary or subversive war". On 14 October the sixteenth act declared the mandates of Costa e Silva and Aleixo extinct and set a date for a new election; a complementary act reopened the Congress that had been in recess since AI-5, and on 17 October the junta promulgated Constitutional Amendment No. 1.

That amendment is usually called the constitution of 1969, and it deserves the name. It rewrote the 1967 text from top to bottom, absorbed the powers of the institutional acts into the constitutional order, broadened the definition of crimes against national security, and provided for death, perpetual imprisonment and banishment in cases of "adverse psychological warfare". It also reapportioned the Chamber in favour of the smaller, poorer and more easily controlled states. Congress elected General Emílio Garrastazu Médici on 25 October, and he took office on 30 October 1969.

What was the economic miracle?

The years from 1968 to 1973 produced the fastest sustained growth in Brazilian history. Measured in constant prices, GDP rose 9.8 per cent in 1968, 9.5 in 1969, 10.4 in 1970, 11.3 in 1971, 11.9 in 1972 and 14.0 in 1973 — an average of just over 11 per cent a year. Industrial output grew faster still, led by automobiles, electrical goods, chemicals and steel. Inflation stayed between about 16 and 25 per cent. Exports grew at around 20 per cent a year, and the country began to sell manufactured goods as well as coffee, soya and iron ore. Foreign investment flowed in, attracted by subsidies, a protected market and a labour force whose unions, in the corporatist structure inherited from Vargas's Estado Novo, had been placed under government supervision and whose wages were set by a formula the government controlled.

That formula is the hinge of the story. Wage increases were calculated on the previous year's inflation plus half the expected productivity gain, and the official inflation forecast used in the calculation was consistently lower than the inflation that followed, so every settlement delivered a real loss. The minimum wage fell by roughly 40 per cent in real terms between 1964 and 1974 on the trade union confederation's figures, and was still only about 82 per cent of its 1964 value in 1977. Income distribution worsened sharply: the Gini coefficient for the economically active population rose from about 0.50 in 1960 to 0.57 in 1970 on the estimates that became the reference point for the debate, and the share of national income going to the poorest half fell. In 1972 the economist Albert Fishlow published an analysis showing that growth had been accompanied by a regressive redistribution, and the government's economists replied that distribution had to wait for the cake to grow. Edmar Bacha's phrase for the result — Brazil as "Belíndia", a Belgium for the rich inside an India for the poor — dates from 1974 and has outlived every boom since.

The state was the third partner. The Médici government ran the largest public investment programme in Brazilian history to that date: the Trans-Amazonian Highway, announced in 1970 and driven through the forest from the north-east towards the Peruvian border; the Rio–Niterói bridge; the Itaipu hydroelectric project, agreed with Paraguay by treaty in 1973; the steel mills, the petrochemical complex, the shipyards and the telecommunications system. State enterprises borrowed abroad and hired, private firms supplied and took subsidies, and foreign capital took the technologically advanced sectors. Official propaganda sold all of it as a national achievement: "Brazil, love it or leave it", "this is a country that is going places", and the 1970 World Cup, won in Mexico City in June, were folded into one message by the president's public relations office.

The miracle was real, and so was the bill. It was financed by borrowing, by suppressed wages and by postponing the redistribution that a country urbanising at that speed could not permanently avoid. Between 1970 and 1980 the urban share of the population rose from 55.9 to 67.6 per cent and the rural population fell in absolute terms for the first time; the cities absorbed millions of migrants into construction, domestic service and the informal economy, with no land reform in the countryside to hold them. The regime's own statistics showed the concentration. What they could not show was the political cost of a growth model that required an immobilised labour movement, and that is where the next decade's crisis began.

How did the state police its opponents?

Repression was organised, and it was organised early. The National Information Service, the SNI, was created by Castelo Branco in 1964 under General Golbery do Couto e Silva and became the centre of a network that included the information centres of each service, the army's Centro de Informações do Exército, the navy's CENIMAR, the air force's CISA and the federal and state police. In 1969 the government created the Operação Bandeirante in São Paulo, funded by private firms as well as the state, and in 1970 formalised it as the DOI-CODI system — Departamento de Operações de Informações, Centro de Operações de Defesa Interna — with units in every state under army command. These were not ordinary police stations. They were interrogation centres where prisoners were held without charge, beaten, burned, given electric shocks, hung from a bar in the "parrot's perch", and subjected to mock executions and threats against their families.

The record is unusually well documented for a regime of this kind, because the military courts that tried political prisoners generated files, and those files were copied and smuggled out in the 1980s for the project published as Brasil: Nunca Mais. Its analysis of the denunciations reaching military prosecutors shows the curve of violence precisely: 308 complaints of torture in the four years from 1964 to 1968, 1,027 in 1969 and 1,206 in 1970 alone. The method was to arrest a suspect, extract names under torture, arrest the names, and present the resulting confession to a military court that convicted on it. Prisoners who died were reported as having hanged themselves, jumped from windows or been shot while escaping. Vladimir Herzog, a journalist and television editor, was killed at the São Paulo DOI-CODI on 25 October 1975 and photographed hanging from a cloth strip; the army's inquiry accepted the suicide version. Manuel Fiel Filho died in the same building on 17 January 1976. Rubens Paiva, a former deputy, was arrested in Rio in January 1971 and never seen again.

The armed left the system was built to destroy was small, and it grew after the repression rather than before it. Urban guerrilla groups appeared in 1968, and their actions — bank robberies, arms seizures, the kidnapping of ambassadors — were met with a campaign that killed or captured their leaders within three years: Carlos Marighella was shot in São Paulo in November 1969, Carlos Lamarca in Bahia in September 1971. The one sustained rural campaign, the Araguaia guerrilla in the south of Pará, lasted from 1972 to 1974 and ended with the disappearance of most of its survivors and of peasants accused of helping them; the army's operation there became a case before the Inter-American Commission on Human Rights that Brazil contested for two decades. The armed struggle never controlled territory, never threatened a city and never came close to unseating the state. Its importance is that it supplied the regime's justification, and that the regime's account of it — a country on the edge of a communist takeover — became the founding myth of the dictatorship's legitimacy.

Alongside the violence ran a legal apparatus that made it look like procedure: the National Security Law of 1967, rewritten in 1969 and again in 1978; the press law of 1967, used against journalists while censors worked inside newsrooms; the censorship of music, theatre and cinema by federal police; and compulsory "moral and civic education" in schools. The regime tolerated a narrow opposition and used it. The MDB, the progressive wing of the Catholic Church, the bar association and parts of the press were allowed to denounce individual cases so long as they did not challenge the system, and the space they occupied was small enough to close whenever it widened. Compared with Argentina after 1976, where the armed forces killed perhaps 30,000 people, or with Chile after 1973, Brazil's death toll was lower. Its use of torture was not: by the standards of the 1970s the Brazilian system was as systematic as any in the hemisphere, and it reached far more people than the death toll suggests.

What did the oil shocks do to the model?

The first oil shock, in October 1973, hit Brazil harder than almost any other country. It imported about four-fifths of its oil, and the quadrupling of the price turned a manageable current-account deficit into a structural one. The government of General Ernesto Geisel, who took office on 15 March 1974, could adjust — slow growth, cut imports, accept a recession — or borrow to keep growing while the economy was restructured. It chose the second, deliberately. The Second National Development Plan, launched in 1975, poured money into import substitution in capital goods, basic inputs and energy: steel, aluminium, pulp, fertiliser, petrochemicals, nuclear power under the agreement with West Germany signed in June 1975, the Pró-Álcool programme for ethanol from sugar cane, and the Itaipu dam. The idea was to borrow abroad at negative real interest rates, build the capacity to export, and grow out of the deficit.

For a while the arithmetic worked. Growth averaged more than 7 per cent between 1974 and 1976 and about 6 per cent in 1977–79. Registered medium- and long-term external debt rose from $12.6 billion at the end of 1973 to $49.9 billion at the end of 1979, and the debt service was paid by new borrowing. Then came the second shock: oil prices doubled again in 1979, the United States raised interest rates sharply under Paul Volcker, and the banks that had recycled the petrodollars stopped lending. Brazil's interest bill rose at the same moment as its import bill, and by 1982 scheduled debt service of more than $15 billion was absorbing most of export earnings of about $18 billion. Mexico's default in August 1982 closed the market entirely, and in December 1982 Brazil went to the International Monetary Fund.

The adjustment that followed was the deepest of the postwar period. GDP fell 4.2 per cent in 1981 and 2.9 per cent in 1983, industrial output fell further, and unemployment rose in the metropolitan regions. Inflation, 15.6 per cent in 1973, reached 76.8 per cent in 1979, 110.2 per cent in 1980 and 211 per cent in 1983; by 1985 it was 235 per cent. The indexation designed to live with inflation now transmitted it. Real wages fell again, and the poorest bore the adjustment through prices rather than taxes or unemployment insurance, since neither existed at the scale required. The trade surpluses that paid the interest were extracted by compressing imports, which made the machinery and inputs the II PND had been meant to produce at home scarce and expensive.

The political consequence was decisive. The regime had legitimised itself on growth, and growth had ended. The businessmen who had financed the coup in 1964 and prospered in the boom moved towards the opposition; the middle class that had bought the apartment and the car found its salary indexed to an accelerating inflation; and the workers silenced since 1964 began, from 1978, to strike in the ABC industrial belt of São Paulo. The debt crisis did not cause the abertura — that had begun in 1974 — but it removed the regime's claim to competence, which was the only claim it had left.

How did Geisel try to open the system?

Geisel's presidency began with a strategy and a vocabulary. Distensão, the relaxation of tension, was announced in 1974; abertura, the opening, became the operative word later, and both were described by the government as "slow, gradual and safe". The author of the design was Golbery do Couto e Silva, back in the Casa Civil, who argued that the regime could not last indefinitely on AI-5 and that the army should manage a transition while it still had the strength to set the terms. The first test came in the congressional elections of 15 November 1974, when the MDB took 16 of the 22 Senate seats being contested and 160 of the 364 seats in the Chamber — 44 per cent, enough to demand parliamentary inquiries and make the government's legislative life difficult.

The reaction was immediate and it was not liberal. In July 1976 the government pushed through the "Falcão law", which reduced televised and radio electioneering to a reading of each candidate's name, party and curriculum, with a photograph; the minister of justice, Armando Falcão, explained that it would prevent "confusion in the electorate". In October 1975 the journalist Vladimir Herzog was killed in the São Paulo DOI-CODI, and the cover-up provoked a public mass in the city's cathedral and a confrontation between the Church and the army. When Fiel Filho died in the same building in January 1976, Geisel — who had been losing control of the apparatus to the hard line — dismissed the commander of the Second Army, Ednardo D'Ávila Mello, and asserted the president's authority over the military. In October 1977 he dismissed the minister of the army, Sylvio Frota, a succession candidate who had been organising against the abertura. That dismissal was the most important act of his presidency, because it settled that the president would choose his successor and that the hard line could not.

The opening was not linear. On 1 April 1977, the day after the Senate rejected a judicial reform the government wanted, Geisel used AI-5 to recess Congress, and two weeks later imposed the "April package": an amendment restructuring the judiciary, a second changing the electoral rules, and six decree-laws. The package created the "bionic" senators — one third of the Senate elected indirectly by state assemblies and municipal delegates — enlarged the representation of the smaller northern and north-eastern states in the Chamber, extended the presidential term to six years and lowered the quorum for constitutional amendments. Its purpose was to guarantee that the 1978 succession and the elections that followed would not produce an opposition majority. Congress reopened on 15 April, and students who had been protesting since 1977 were met with police in the streets.

What Geisel gave back, he gave back slowly. In October 1978, with the succession settled in favour of General João Figueiredo, the government promulgated Constitutional Amendment No. 11, revoking AI-5 with effect from 1 January 1979 and restoring habeas corpus and the immunity of members of Congress for opinions expressed in office. The Institutional Acts survived on paper until the same date. The abertura, in other words, was not a surrender but a schedule, and the men who ran it intended to remain the guarantors of whatever came after. That intention explains both the amnesty of 1979 and the fact that the first civilian president after 1964 was elected by the electoral college the regime had built.

What did the 1979 amnesty settle?

General João Batista Figueiredo took office on 15 March 1979 with a mandate to continue the opening and a reputation, from his years at the SNI, for knowing where the bodies were. The mobilisation he faced had been building since 1978, when a strike wave in the ABC region and the committee movement for amnesty brought the political prisoners and the exiles into the open. The Brazilian Committees for Amnesty, founded in Rio de Janeiro in February 1978 and in São Paulo three months later, with branches in a dozen states and one in Paris, had made "ampla, geral e irrestrita" — broad, general and unrestricted — the slogan of the campaign. The bar association, the Church, the press and the MDB supported it. The government's answer was a bill of its own, sent to Congress in June 1979.

The law that emerged, No. 6,683 of 28 August 1979, granted amnesty for political crimes and crimes connected with them committed between 2 September 1961 and 15 August 1979, extended it to those whose political rights had been suspended and to civil servants and soldiers punished under institutional acts, and excluded those condemned for "terrorism, robbery, kidnapping and personal attack". The last exception was aimed at the armed left; the phrase "connected crimes" was the door through which the torturers walked, since torture had been committed in operations against political crime and could be classified as connected. The law did not name torture, did not require anyone to investigate it, and was presented as reciprocal — an amnesty for both sides — although one side had been legislating and the other had been hanging from a bar.

The parliamentary history of the law is usually misreported. The figures 206 to 201, which appear in most accounts, belong to the vote of 22 August 1979 on an amendment by the deputy Djalma Marinho that would have restricted the amnesty's scope; the government's text, in the form of the substitute drafted by Ernani Satyro, was then approved by a symbolic vote of the party leaders, without a count. The distinction matters because it shows how the law passed: not on a narrow majority of principle, but through the discipline of ARENA and the acquiescence of much of the MDB, which calculated that the return of the exiles was worth the impunity. The first prisoners were released on 28 August, though most were freed through ordinary legal remedies and pardons rather than the law itself; the last political prisoner, José Sales de Oliveira, was released on 8 October 1980.

The amnesty was followed within months by the rest of the opening. In December 1979 Law No. 6,767 dismantled the two-party system and allowed new parties, producing the PDS from ARENA, the PMDB from the MDB and, in 1980, the Workers' Party of Luiz Inácio Lula da Silva, the metalworkers' leader thrown up by the ABC strikes. In November 1980 Constitutional Amendment No. 15 restored the direct election of governors and senators, and in November 1982 the country voted for governors for the first time since 1965. The PDS won 12 of the 22 states and 235 of the 479 seats in the Chamber, but the opposition took São Paulo with Franco Montoro, Rio de Janeiro with Leonel Brizola and Minas Gerais with Tancredo Neves, and the combined opposition parties held 244 seats against the government's 235. The regime had won the election and lost the initiative.

Why did Diretas Já fail?

Because the electoral college, not the street, chose the president, and the government still had the votes in it. The campaign for direct presidential elections began in 1983, when the PMDB deputy Dante de Oliveira introduced a constitutional amendment to replace the college with a popular vote; by the time it was voted on it had been endorsed by the governors of the three largest states, by the Church, the unions and the press, and by demonstrations that were, by the standards of any country at the time, enormous. The rally in São Paulo's Praça da Sé on 25 January 1984 drew more than 300,000 people; Belo Horizonte drew 400,000 in February; the Candelária in Rio drew more than a million on 10 April; and the closing rally in the Vale do Anhangabaú on 16 April was estimated at between 1.5 and 2 million, the largest political gathering in Brazilian history.

The amendment needed two thirds of the Chamber, 320 votes of 479, and on 25 April 1984 it received 298, with 65 against and 3 abstentions. The government's majority held, but only just, and the result exposed the arithmetic on which the regime depended. The PDS caucus was split: 54 of its deputies voted for the amendment. What saved the government was not conviction but the structure of representation the April package had built — the over-representation of the small northern and north-eastern states, where its clientelist networks were strongest — and the discipline of the party machines that controlled the college. The campaign had changed public opinion and had not changed the votes, and the two-thirds rule meant that the second fact decided the question.

The sequel showed what the campaign had achieved. With the amendment dead, the opposition's route to power ran through the college it had just denounced, and the PMDB chose Tancredo Neves, governor of Minas Gerais, with the dissident PDS senator José Sarney as his running mate; a faction of the PDS, the Liberal Front, broke with the government and joined them. When the college met on 15 January 1985, Tancredo won 480 votes to 180 for Paulo Maluf, with 26 abstentions: the same college, used against the party that had built it. The campaign had discredited the PDS, split it and made a negotiated transition the government's only remaining option. It had not failed so much as it had been outvoted by the institution it was trying to abolish.

How did the regime end, and what survived it?

Tancredo Neves never took office. He was hospitalised on 14 March 1985, the day before the inauguration, for what was diagnosed as appendicitis and turned out to be a series of infections; José Sarney was sworn in as acting president on 15 March, and Tancredo died on 21 April without having governed a day. Sarney, a former president of the PDS and a man the left had spent years denouncing, therefore became the first civilian president since 1964, presiding over a government that called itself the New Republic. The irony shaped the transition: the first civilian government needed to prove its democratic credentials precisely because of who led it.

The legal dismantling was rapid. Constitutional Amendment No. 25 of 15 May 1985 opened the franchise to illiterates, who had been barred from registering, restored direct elections for the mayors of state capitals and "national security" municipalities, and set the timetable for a constituent assembly. In November 1986 the country elected that assembly, with the PMDB dominant after the Cruzado Plan's brief success against inflation, and on 5 October 1988 it promulgated a constitution that remains in force. The text is long, detailed and social-democratic in its commitments: it defines health and education as rights, sets out an extensive list of labour and social protections, decentralises revenue to states and municipalities, and makes the armed forces the guarantors of constitutional order rather than of the government. It also dismantled the censorship apparatus and ended the military courts' jurisdiction over civilians — though the National Security Law of 1983 survived until 2021 and the press law of 1967 until 2009.

What the constitution did not do was deal with the past. The amnesty law survived the constituent assembly, and when the Supreme Federal Court ruled on it in April 2010 — in a case brought by the bar association — it upheld the law by seven votes to two, holding that it had been a political pact ratified by the 1988 constitution. Seven months later the Inter-American Court of Human Rights, in Gomes Lund v. Brazil, decided that the law was incompatible with the American Convention and ordered Brazil to investigate the disappearance of the Araguaia guerrillas; the state has complied only partially, and no one has been prosecuted for torture. Reparation moved faster than justice: a commission created in 1995 recognised the dead and disappeared and compensated their families, and an Amnesty Commission created in 2001 has processed tens of thousands of claims for reinstatement and indemnity.

The truth came last. A National Truth Commission was created by law in November 2011, installed in May 2012 and reported on 10 December 2014, after two and a half years of work, more than 1,100 interviews and the collection of documents from the armed forces and the intelligence services. Its three volumes run past 4,000 pages. The commission identified 434 people killed or disappeared for political reasons between 1946 and 1988 — 191 killed and 243 disappeared, of whom 33 have since been identified — and named 377 public agents as responsible for grave violations. It described the repression as systematic and generalised, classified the crimes as crimes against humanity, and held that the amnesty could not be opposed to the state's duty to investigate. It also estimated that at least 8,350 indigenous people died in the period covered by its mandate through state action or omission, a figure drawn from studies of only ten of the country's more than 300 peoples and, as the commission said, certainly an undercount.

How do Brazilians and historians describe the regime now?

The word matters in Brazil, and it has moved. The regime called itself a revolution, then a "military regime"; its opponents called it a dictatorship, and from the 1980s ditadura militar became standard in the press and the schools. Since the 2000s a generation of historians has argued for ditadura civil-militar — civil-military dictatorship — on the ground that businessmen, landowners, the press, the Church and the political class were not spectators but participants, and that civilian support was essential to the regime's durability. The argument is not merely semantic. It changes where responsibility lies for the torture, the censorship and the economic policy, and it changes what a democracy inherited.

The historiography has gone through three broad phases. The first, contemporary with the events, was built around the confrontation between the left and the state and treated the regime as the instrument of a class or of a Cold War alliance. The second, in the 1980s, produced the canonical narrative histories — Thomas Skidmore's The Politics of Military Rule in Brazil (1988) remains the indispensable English account — and the first archive-based studies of the security apparatus. The third, from the 2000s, has been dominated by Elio Gaspari's five volumes, by the opening of military archives under the 2011 access-to-information law, and by the truth commission's research, which for the first time documented the chain of command behind the killings and the part played by civilian firms in financing the repression.

The disputes that remain are specific. The figure of 434 dead and disappeared is a documented minimum, not a total, and the commission's report says so. The figure of 8,350 indigenous deaths covers the whole 1946–1988 period, not only the dictatorship, and rests on a sample; it is often quoted as a precise count, which it is not. The character of the 1964 coup is still argued: the American role is now documented in Brazilian and US archives, and it is smaller than the Cold War account implied and larger than the nationalist account allowed. Whether the abertura was a strategy of the regime or a victory of the opposition is a question every history of the period answers differently, and the truth is that it was a negotiation in which the regime set the calendar and the opposition set the price.

The politics of memory has not settled. The 2014 report was received by a president who had herself been imprisoned and tortured, and its recommendations — repeal of the amnesty law, criminalisation of denial, a permanent memory body — were largely ignored. The governments of 2019–2022 encouraged the commemoration of 31 March 1964 as a "democratic revolution" and downgraded the memory institutions; the practice ended with the change of government in 2023, and files have continued to be released in fragments. Brazil has thus documented its dictatorship more thoroughly than most of its neighbours and punished it less than any of them, and the entry's dates describe a regime that ended while the argument about it did not.

Common myths about the Federative Republic of Brazil

Timeline of the Federative Republic of Brazil

What the map shows next

At 1967 on the map the map draws a country that has just acquired the name it still bears, with a Congress sitting under a constitution written for it by the army and an economy stabilised by holding down wages. The Amazon is still empty on this map, Brasília is seven years old, and the frontier states the regime would later carve out of Mato Grosso and southern Pará do not yet exist.

At 1974 on the map the first oil shock has arrived and the government has chosen to borrow rather than adjust; the MDB has just won 16 of the 22 Senate seats, and Geisel is beginning the abertura with the hard line still in control of the police. By 1982 on the map the debt crisis has forced Brazil to the IMF and direct elections have returned for governors, with the opposition governing the three largest states. At 1986 on the map the map shows the country the constituent assembly was elected to refound — the same borders, the same name, and a government that had just frozen prices to stop an inflation the dictatorship left behind.