The Roman Empire

The Roman Empire was governed for three centuries by a position that had no legal existence, and the man who invented it spent forty years making sure nobody had to say so out loud.

This course begins where the course on the Republic ends: with Octavian, renamed Augustus in January 27 BC, holding every army and claiming to have handed the state back to the Senate and people. The question here is not whether the Republic was really restored, which it was not, but what was built in its place, because the thing built in the 20s BC is what governed sixty million people until the third century broke it. If you have not read the earlier course, all you need is this: the Republic was an aristocracy of annual elected magistrates that could not control the generals its empire required, and it destroyed itself in a century of civil war that Augustus ended by winning.

Powers, not a throne

Start with what Augustus actually held, because the list is more surprising than any summary of it.

From 31 to 23 BC he was consul every year, which was legal but conspicuous, since it meant one of the two annual places was permanently occupied. In 23 BC he gave that up and took two things instead. The first was tribunicia potestas, the power of a tribune of the plebs without the office: the right to propose laws to the assembly, to veto any magistrate, and personal sacrosanctity, held for life and renewed annually so that it could be counted. He counted it for thirty-seven years, and imperial documents dated themselves by it for the next four hundred. The second was proconsular imperium, held permanently, not lapsing when he entered the city as a proconsul's normally did, and outranking that of any other governor.

To these were added the office of pontifex maximus in 12 BC, when the previous holder finally died, and the honorific pater patriae in 2 BC. He was consul thirteen times in all, and he took no office that had not existed in the Republic.

That is the whole legal basis of the Roman Empire. There is no law creating the emperorship, no coronation, no regalia, and for a long time no agreed word: princeps, meaning the leading man, was a term of respect the Republic had used of senior senators; imperator, which he turned into a first name, meant a victorious general; Augustus was a religious adjective with no political content at all, which was exactly its merit.

Example. Every element in that list existed under the Republic. What was new?

The combination, and its duration. A tribune's veto, a proconsul's command and a chief priesthood were all ordinary republican things, and the Republic's protection against them was that each was held briefly, by an elected man, alongside a colleague with equal power. Augustus removed the brevity, removed the colleague where it mattered, and kept the names. Notice also which two he chose. Tribunician power gave him a permanent legal presence inside the city without a magistracy, and proconsular imperium gave him the armies. Between them they cover the two things that had decided every political dispute since 133 BC: control of the streets and control of the legions. The genius is not in inventing anything, since he invented nothing. It is in having noticed exactly which republican instruments were load-bearing.

Now you. Why did he never take the dictatorship, which the people offered him in 22 BC during a famine, and which would have been simpler?

The Res Gestae, and how to read it

Augustus wrote his own account of his reign, the Res Gestae Divi Augusti, a document of thirty-five short chapters, and had it inscribed on bronze pillars at his mausoleum. The bronze is long gone. What survives is a copy cut into the wall of a temple in Ankara, with fragments from two other towns in the same province, which is why a text written for the city of Rome is known from central Anatolia.

It is the most useful document in Roman history and one of the least trustworthy, and both for the same reason: it is a monarch's own statement of what he wished his reign to be understood as, in his own numbers. It is full of exact figures, and they are the interesting part. He lists the cash he gave the plebs, the number of citizens in each census, the veterans he settled, the gladiators shown, the temples repaired (eighty-two in one year), the number of times the temple of Janus was closed to signify peace throughout the empire (three, against twice in all previous Roman history).

What it never mentions is a defeat, a rival, a proscription, a name of any opponent, or the words for the powers it describes. Mark Antony appears only as an unnamed faction. The reader is told the sums spent on the treasury and the army and never told that the man spending them owned Egypt.

The household that became a government

The Republic had almost no state apparatus, and Augustus did not build one either. What he did was staff the empire's new jobs from his own house.

Egypt, taken in 30 BC, was not given to a senatorial governor. It was administered by a prefect of equestrian rank, a man of the second order who could not raise a rival claim, and no senator could set foot in the province without the emperor's permission. Rome itself got a prefect of the city with three urban cohorts, a prefect of the grain supply, and in AD 6 a corps of seven cohorts of vigiles, watchmen who fought fires and patrolled at night, all of them equestrian commands answering to the emperor. The praetorian guard, nine cohorts of them, was created as a standing bodyguard under two prefects, also equestrians.

Below that level the work was done by the emperor's freedmen and slaves, running correspondence, petitions and accounts out of what was formally a private household. The imperial finances themselves were never fully separated from the man: the state treasury, the aerarium, continued under the Senate, while the emperor disposed of provincial revenues, his own vast private estates, and Egypt, which he held as personal property.

Paying the soldiers, permanently

The one genuinely new institution of the reign is the least glamorous, and the earlier course explains why it was needed: since Marius, soldiers without property had depended on their general to obtain land at discharge, which is what made armies personal.

In AD 6 Augustus set up the aerarium militare, a military treasury, endowed with 170 million sesterces of his own money and funded thereafter by two new taxes on Roman citizens: five per cent on inheritances and one per cent on goods sold at auction. Service was fixed at twenty years, and the discharge payment at 12,000 sesterces.

Example. Work out roughly what that fund had to pay each year, and say what the arithmetic implies.

Twenty-eight legions of about 5,000 men is 140,000 legionaries. With twenty-year terms, a steady state discharges about one twentieth of them annually, or 7,000 men, and at 12,000 sesterces each that is 84 million sesterces a year, before the auxiliaries are counted. The emperor's 170 million endowment covers two years of it. So the gift is a founding gesture and the taxes are the institution, which is why he was willing to be the first Roman ruler to tax Roman citizens directly since 167 BC and to take the unpopularity of doing it. The point is worth holding on to for the rest of this course: the empire's central fiscal problem, from Augustus to the fifth century, is that soldiers have to be paid on time by somebody other than their commander.

Now you. The legionary's pay was 900 sesterces a year. What does a 12,000 sesterce discharge bounty represent, and why set it that high?

What the empire bought with it

It is easy to describe the principate as a confidence trick and stop there. The two centuries that followed suggest something was actually delivered.

The civil wars stopped. Between 31 BC and AD 68 no Roman army fought another in Italy, after a century in which that had happened in every generation. Piracy stayed suppressed, the grain came in, and a common coinage made the Mediterranean a single market to a degree it would not reach again for a very long time. Provincials, who had been the property of whichever governor drew the lot, got an appeal to somebody above him.

Tacitus, no friend of the system, gives the mechanism in one sentence at the start of the Annals: Augustus seduced the soldiers with bonuses, the people with cheap grain, and everybody with the sweetness of peace. It is meant as an indictment and reads equally well as a description of a functioning government.

The one thing he could not build

Which brings us to the flaw, and it is structural rather than personal.

If the principate is not an office, it cannot be filled. There is no election for it, no law of succession, no procedure of any kind, because admitting that a procedure was needed would have admitted that the thing existed. What Augustus did instead was to build an heir, by marriage and adoption, and to have him voted the same powers in the old man's lifetime so that the transfer would look like continuity.

He tried for forty years. His nephew and son-in-law Marcellus died in 23 BC. His friend, general and son-in-law Agrippa died in 12 BC. His grandsons Lucius and Gaius, whom he had adopted as sons, died in AD 2 and AD 4. Only then, at sixty-six, did he adopt his stepson Tiberius, a man he did not much like, and require Tiberius in turn to adopt Germanicus. When Augustus died at Nola in August AD 14, aged seventy-five, Tiberius already held tribunician power and an imperium equal to his, and the succession still had to be staged in the Senate as a piece of theatre in which the heir declined and the senators insisted.

Example. Tacitus reports that at the first meeting of the Senate after the death, Tiberius asked the senators to define which part of the state's business he should undertake, and they refused to answer. What was the difficulty?

That nobody had a vocabulary for the transaction, and both sides knew it. The senators could not offer him a position that did not officially exist, and to name it would be to admit that Rome had a monarchy and that they were voting for it. Tiberius could not simply take it without confessing the same thing. So the meeting consisted of a man asking for something he could not describe from a body that could not grant it, and it ended with the powers being conferred anyway. That is what a constitution looks like when its central institution is unmentionable. It also explains why the accession of every subsequent emperor is improvised, and why the improvisations get worse.

Now you. Suetonius says that on his last day Augustus asked his friends whether he had played his part well in the comedy of life, and asked them to applaud as he left. What claim is being made?

Where this leaves the empire

Augustus left a state with a professional army, a working revenue, a governed Mediterranean and a ruler whose position was undefined and therefore untransferable by any rule.

For fifty-four years the problem was concealed by family. Tiberius, Caligula, Claudius and Nero were all descended from or married into the same house, and the fiction of continuity held because a Julio-Claudian was always available. The next lesson is about what happened when there was not one, and about the discovery that Tacitus says the year AD 69 let out of the bag: that emperors could be made somewhere other than Rome, by people other than senators.

Because the emperorship was not an office, nobody could inherit it, and every accession for three centuries was therefore an improvisation whose details reveal who actually held the power to decide.

The previous lesson ended with Augustus dead at Nola and Tiberius receiving powers that the Senate could not name. This lesson follows what happened at each of the next handful of transfers, because the sequence answers a question the constitution never addressed on paper. When an emperor died, three groups could plausibly claim a voice: the Senate, which had voted the powers; the praetorian guard, which was in the room; and the legions on the frontiers, which had the weapons. Between AD 14 and AD 193 all three tried, and the results were not close.

The family as a procedure

The first four successions were solved by blood, and by an ambiguity that suited everyone: if the ruler is not an officeholder but simply the greatest man in Rome, then his heir is a private matter, and private matters were settled in Roman law by wills and adoption.

Tiberius, adopted in AD 4, already held tribunician power and a proconsular imperium when Augustus died, so the transfer looked like a continuation. Gaius, whom the army had nicknamed Caligula, came in through Tiberius's will and the backing of the praetorian prefect. When Caligula was murdered in his palace in January 41 by officers of his own guard, the Senate met on the Capitol and debated restoring the Republic, and while it debated, the praetorians found Caligula's uncle Claudius in the palace, took him to their camp, and proclaimed him. Nero, thirteen years later, was produced at the camp gate by the prefect Burrus while Claudius's death was still unannounced and Britannicus, Claudius's own son, remained inside the palace.

Four transfers, none of them illegal, none of them following a rule, and only the first arranged in advance. Notice the direction of travel. The Senate is present at every one and decides none of them.

The price of a coronation

Claudius, who owed his position to the men who had carried him to the camp, paid them for it: Suetonius says 15,000 sesterces to each praetorian, the first time a Roman ruler had openly bought his accession from his own guard.

Example. A legionary was paid 900 sesterces a year. Work out what 15,000 sesterces represents, and what the transaction established.

Sixteen and two thirds years of a legionary's pay, handed over in one payment, to men who were already the best paid soldiers in the empire. As a bribe it is enormous; as a precedent it is worse. Two things were now true. First, a donative on accession had become the expected price, so every later emperor had to find that sum in his first week, whatever the state of the treasury. Second, and this is the part that mattered, the praetorians had learned that the death of an emperor was the most profitable event that could happen to them. An institution that profits from a vacancy will eventually create one, and within a century and a half the guard was doing precisely that. The Roman state had accidentally attached a large cash prize to regicide.

Now you. In 193, after the praetorians killed the emperor Pertinax, they announced that the position would go to whoever offered most, and Didius Julianus won by promising 25,000 sesterces a man. What does the auction actually demonstrate?

The year of four emperors

Nero was the last of the Julio-Claudians, and when he lost the loyalty of the Spanish and Gallic commands in 68 there was no relative left to produce. He killed himself in June 68. What followed is the single most instructive year in Roman political history.

Galba, governor of nearer Spain, aged seventy-two, was proclaimed by his legions in April 68 and recognised by the Senate after Nero's death. He then made the one economy his position could not afford, refusing to pay a donative, with the line that he levied his soldiers and did not buy them. It was the correct constitutional sentiment and it killed him: on 15 January 69 the praetorians cut him down in the Forum and proclaimed Otho, who had been passed over as his heir.

Otho lasted ninety-one days. The armies of the Rhine had already proclaimed their own commander, Vitellius, on 2 January, before Galba was even dead, and their advance guard beat Otho at Bedriacum in northern Italy in April. Otho killed himself. Then, on 1 July, the prefect of Egypt had his legions swear to Vespasian, who commanded the war in Judaea; the Danube armies joined him, beat Vitellius at the second battle of Bedriacum in October, and took Rome in December in fighting that burned down the temple of Jupiter on the Capitol.

Example. List where each of the four was proclaimed and say what the pattern shows.

Galba in Spain, Vitellius on the Rhine, Vespasian in Egypt with the eastern and then the Danubian armies behind him, and only Otho in Rome itself, by the guard. Three of the four were made in frontier provinces by the men they commanded, and the one made in the city lasted three months. Tacitus states the conclusion in a single clause: the secret of empire had been let out, that an emperor could be made somewhere other than Rome. It is worth being precise about what the secret is. Not that armies mattered, which everyone had known since Sulla, but that the legal machinery in Rome would ratify whatever the strongest army had already decided, and that the decision could therefore be taken on the Rhine. From 69 onwards every serious political actor in the empire knows this, and the question of who commands which frontier becomes the central question of imperial politics.

Now you. Vespasian's soldiers proclaimed him on 1 July 69, but his reign was later dated from that day rather than from the Senate's recognition in December. Why does the choice of date matter?

The nearest thing to a constitution

Vespasian's accession produced the one document that comes close to defining the position, and it survives because a bronze tablet was reused as an altar and rediscovered in the fourteenth century.

The lex de imperio Vespasiani is a law of AD 69 or 70 conferring the emperor's powers, and it works by citation: he is to have the right to make treaties as Augustus, Tiberius and Claudius had it, to convene the Senate as they had it, to extend the boundary of the city as Claudius had it, and so on through a list of precedents. The final surviving clause is the striking one. It authorises him to do whatever he judges to be in accordance with the advantage of the state, and declares valid anything done before the law was passed.

Two things follow. First, by 70 the position had become definite enough to legislate about, which it had not been in 14. Second, the definition is a list of what previous holders had done plus a clause of unlimited discretion, which is a description of custom rather than a constitution. That is what the Roman Empire had instead of a founding law for its entire existence.

Adoption, and why it looked like a principle

For eighty-four years after 96 the empire had five rulers, each of whom adopted his successor, and none of whom was succeeded by a son. Nerva, an elderly senator installed after the murder of Domitian, adopted Trajan, who commanded the army of Upper Germany, under pressure from the guard. Trajan's own adoption of Hadrian was announced by his widow as he lay dying in Cilicia and was doubted at the time. Hadrian chose Antoninus, requiring him to adopt in turn both Marcus Aurelius and Lucius Verus. Marcus and Verus then ruled jointly, the first time the position had been formally shared.

Gibbon called this the happiest period in the history of the world, and later writers have called the arrangement the adoptive principle, as though the Romans had solved succession by selecting on merit.

Example. Test that claim against the family circumstances of the five emperors.

Nerva had no children. Trajan had none. Hadrian had none. Antoninus Pius had two sons, both of whom died before he became emperor. Marcus Aurelius had a son who survived, Commodus, and he made him co-ruler at sixteen and successor at eighteen. The adoptive principle turns out to be a run of five rulers who had no surviving sons, followed immediately by one who did and who behaved exactly as a dynast. There is a real point underneath the correction: adoption was a genuine Roman legal instrument, respectable and normal, and it let the empire recruit its rulers from a wider pool than a bloodline would allow, which is why the second century looks well governed. But it was never a rule, no emperor ever declared it one, and it collapsed on first contact with a living heir. Beware of any pattern in Roman history that is visible only in a period when the alternative was unavailable.

Now you. Marcus Aurelius is often blamed for Commodus. Was the choice avoidable?

What the ritual settled

By the second century the making of an emperor had a shape: acclamation by soldiers, a donative paid to them, recognition by the Senate, and a vote of powers modelled on the tablet Vespasian's lawyers drew up. The order of those steps is the whole story, since the vote comes last and confirms rather than chooses.

That is not a small thing. A regime in which the army chooses and the civil authority ratifies is stable exactly as long as the army is unified, which for a century it broadly was: there was one civil war in 69, one in 193, and long stretches with none. It becomes lethal when several armies of comparable size disagree, and the third century will show what that looks like.

Which makes the army the institution to understand next. It decided who ruled, it consumed most of what the state collected, and it was the only Roman organisation present in every province. The next lesson is about what it actually consisted of: how many men, on what terms, standing where, and paid how.

The Roman army of the first century was the largest organised body of men in the western world, and it was not primarily a fighting force: it was a permanent, salaried, geographically fixed institution whose existence shaped the empire's taxes, its politics and its provincial society.

The previous lesson showed the legions choosing emperors in 69 and 193. That makes the composition of the army a political question rather than a military one, so this lesson asks what it actually was. Four things are worth carrying away: how many men there were, on what terms they served, where they stood, and what happened to a province that had them in it.

The order of battle

Augustus finished the civil wars with about sixty legions and reduced them to twenty-eight, which is the number the earlier course on the Republic ends with. Three of those were destroyed in the Teutoburg forest in AD 9 and never re-raised, leaving twenty-five, a figure that stayed close to constant for two centuries: Trajan had thirty, Septimius Severus thirty-three.

A legion was nominally about 5,000 men, all Roman citizens, organised in ten cohorts, with its own artillery, surveyors, engineers, medical staff and clerks. Beside it stood the auxilia, units of non-citizens raised in the provinces, in cohorts of infantry and alae of cavalry of roughly 500 men each, and by the early second century they were about as numerous as the legions. Almost all of the empire's cavalry was auxiliary. Add the nine praetorian cohorts, the urban cohorts and the fleets at Misenum and Ravenna, and the total under Tiberius is somewhere around a quarter of a million men under arms.

Example. Set 250,000 soldiers against the empire's population and its frontier, and say what kind of army that makes it.

Against a population of about sixty million it is one soldier per 240 inhabitants, which is a small army for a state that size. Against the frontier it is smaller still. The land and river frontier from the mouth of the Rhine round to the Sahara runs on the order of 10,000 km, so an even spread gives 25 men per kilometre, and the spread was not even. That arithmetic rules out one popular picture immediately: the frontier cannot have been a defended wall, because there were nowhere near enough soldiers to hold a line of that length against a serious attack. What the army could do was garrison the crossing points, patrol, escort, collect customs, and concentrate: pull ten thousand men from three provinces into one field force and destroy whatever had come across. It was a mobile striking force with a customs service attached, which is a different thing from a defensive perimeter, and the difference matters for what happens in the fourth century.

Now you. The same arithmetic gives one soldier per 240 civilians. What does that ratio tell you about how the provinces were held?

The terms

A legionary enlisted for twenty years under Augustus, in practice often longer, with four more years under the standards as a reserve; by the second century twenty-five years was normal. He swore an oath, the sacramentum, renewed each year, and it was sworn to the emperor by name rather than to the state. He could not legally marry while serving, a ban that lasted until Septimius Severus lifted it in the 190s, and which did not stop soldiers having families in the settlements outside the fortress, merely stopped those families having legal standing.

His pay under Augustus was 900 sesterces a year, delivered in three instalments, from which the army deducted food, boots and straps, hay for the unit's animals, clothing and a contribution to the camp's festival, with the remainder credited to a savings account held with the standards. Pay records on papyrus from Egypt show these deductions itemised line by line, which is how we know that the nominal figure and the money a soldier saw were different numbers. Domitian raised the pay to 1,200 sesterces, a rise of a third, and it stayed there for a century.

Example. In AD 14, on the news of Augustus's death, the legions in Pannonia mutinied and demanded a denarius a day, discharge after sixteen years, and no further service under the standards afterwards. Test the demand against the numbers.

A denarius a day is 365 denarii a year against the 225 they were getting, an increase of 62 per cent. Their actual daily rate was about 0.62 of a denarius, so they were asking for a round figure that happened to be a large rise, which is a familiar shape for a wage demand. The second and third clauses are the revealing ones. Discharge at sixteen years instead of twenty, and no reserve service after it, are demands about time rather than money, and they say that the grievance was less about the rate than about men in their forties still being held. Tacitus, who reports all this, is hostile to mutineers and gives their case anyway, in detail, including a centurion nicknamed Bring Another for the rate at which he broke his vine staff across soldiers' backs. When a hostile source preserves the other side's argument in usable form, that argument is usually authentic, because inventing a good case for your enemy is not what propaganda does.

Now you. The mutiny was settled and then, quietly, the concessions were withdrawn. Why could the state not simply have paid?

The citizenship machine

The auxiliaries were the part of the army that changed the empire most, and they did it by leaving it.

An auxiliary served twenty-five years as a non-citizen and received Roman citizenship on discharge, for himself and, in the earlier practice, his children, together with the right of legal marriage. The grant was recorded on a pair of small bronze tablets, wired and sealed with the names of seven witnesses, copying a master list posted in Rome. These are the military diplomas, and over a thousand of them survive, because a bronze certificate of your legal status is the sort of thing a family keeps and buries with you.

Consider the flow. If about 125,000 auxiliaries serve twenty-five year terms, roughly 5,000 men reach discharge each year, and each of them turns into a citizen household somewhere in a province. Over two centuries that is on the order of a million new citizen families, mostly in frontier regions, created by the army rather than by any political decision. That is the largest single engine of Romanisation in the empire, it ran automatically, and no emperor had to argue for it in the Senate.

Where the soldiers stood

The map is the politics. Under the early empire the great concentrations were on the Rhine, on the Danube, and in Syria facing the Euphrates. Britain held three legions after the conquest, Spain one, Africa one, Egypt two and later one. Italy had none: only the praetorians, the urban cohorts and the fleets.

That last fact is the constitutional heart of the system. The emperor governed, through his own appointed legates, exactly those provinces where the legions were, while the Senate appointed the governors of the peaceful ones. An Italian senator could reach the consulship and never in his life command an armed force. Power was therefore geographically fixed, and the fixing tells you where trouble comes from.

Example. Given the distribution, predict which provinces will produce challengers, then check the prediction against the two civil wars this course has already met.

Any command holding three or more legions, so the two German armies, the Danube commands in Pannonia and Moesia, Syria, and Britain. In 69 the challengers were Galba in Spain, backed by Gaul, Vitellius on the Rhine, and Vespasian with the Judaean and then the Danubian armies. In 193 there were exactly three: Septimius Severus in Pannonia, Pescennius Niger in Syria, Clodius Albinus in Britain. Six of the seven come from the multi-legion commands, and the exception, Galba, had to be adopted by the Gallic and German armies before he was anything. The prediction holds, and emperors could do the arithmetic as easily as we can, which is why Domitian forbade two legions to share a fortress and why later emperors kept splitting the large provinces into smaller ones. The whole administrative geography of the empire is partly a device for making sure no single governor commands enough men to try.

Now you. Egypt held two legions and a prefect of equestrian rank rather than a senator. Why the special treatment?

The army as a settlement

A legionary fortress was a town. Outside its walls grew the canabae, a civilian settlement of traders, craftsmen, innkeepers and the unofficial families of serving soldiers, and many of these became real cities: Bonn, Mainz, Vienna, Budapest, Belgrade, Leon, York and Chester all began as places where a legion stood.

The army also built. Roads, bridges, aqueducts, drainage and public buildings were made by soldiers with their unit's stamp on the bricks, which is how an archaeologist can date a wall to a legion and a decade. And it spread Latin: recruits from a dozen provinces served together, were commanded, paid and disciplined in Latin, and left after twenty-five years as Latin-speaking citizens.

Recruitment itself changed under the same pressure. Italians largely vanished from the legions during the first century; by the second, units recruited locally, so the army of the Rhine was largely raised on the Rhine and the army of Syria in Syria. That made the soldiers cheaper to find and rooted them in the region they defended, and it also meant that by the third century the men who chose the emperors were provincials, commanded increasingly by provincial officers, and the political weight of Italy fell accordingly.

What it cost, and what comes next

An institution of that size on fixed money wages is an accounting problem before it is anything else. A rough figure gives the shape: 125,000 legionaries at 900 sesterces and a similar number of auxiliaries on lower rates comes to something near 190 million sesterces of basic pay, before discharge bounties of some 84 million, before the praetorians at three times legionary rates, and before food, animals, weapons and building.

That is the largest recurring expense any pre-industrial state had ever taken on, and it fell due every year whether or not there was a war. The next lesson works out where the money came from: what the empire taxed, how much it raised, and how heavy the burden actually was on the people who paid it.

An empire that pays a quarter of a million men in cash every year is a fiscal machine before it is anything else, and almost everything later emperors do can be read off the arithmetic of that payroll.

The previous lesson counted the army and ended with a rough basic payroll near 190 million sesterces, plus discharge bounties, equipment and food. This lesson asks where that money came from. Nothing resembling a Roman budget survives, so the figures here are reconstructions built from scattered numbers in Josephus, Plutarch, Suetonius, papyri and inscriptions, and their uncertainty is part of the lesson rather than a defect in it.

The money itself

Roman coinage had a fixed internal ratio: an aureus of gold was worth 25 silver denarii, and a denarius was worth 4 sesterces, the sesterce being the unit Romans quoted large sums in. Under Augustus the aureus was struck at 42 to the Roman pound of 327 grams, so about 7.8 grams, and the denarius weighed about 3.9 grams at something close to 98 per cent silver.

The state was the only mint, and it minted for a practical reason: soldiers and officials had to be paid in something, and taxes had to come back in the same thing. Coins were also the empire's most widely distributed publication. A provincial who never saw a Roman building handled the emperor's portrait every week, with a legend on the reverse announcing a victory, a virtue, a new road or the arrival of an heir.

Example. In AD 64 Nero cut the denarius from about 3.9 grams to 3.41 and its fineness from about 98 to 93.5 per cent. Work out what he gained.

Silver per coin falls from 3.82 grams to 3.19, a drop of 17 per cent, so a pound of silver that used to yield about 86 denarii now yields about 103, which is 20 per cent more coins from the same metal. The old coins keep circulating at the same official value, so nobody is robbed on the day, and the state can pay 20 per cent more wages from the same bullion. That is the attraction, and it is a tax that requires no law, no census and no collector. The cost arrives later and elsewhere: prices adjust as the new coins accumulate, and the adjustment falls on anyone holding cash rather than on the people the emperor would have had to face in the Senate. Every debasement in this course works exactly like this, and the third century will show what happens when it is done twenty times in fifty years.

Now you. Why would an emperor short of money debase rather than simply raise the rate of tax?

What was taxed

Two direct taxes carried the system. The tributum soli fell on land, at rates that varied by province and were assessed from a census of holdings; the tributum capitis was a head or poll tax, again varying, and in Egypt it fell on adult males with exemptions that people fought hard to prove. The assessment machinery is well documented in Egypt, where the household census ran on a fourteen year cycle and thousands of declarations survive on papyrus, listing every occupant of a house by name, age and status.

Around these sat indirect taxes: customs dues, the portoria, levied at province boundaries and ports at rates of two to two and a half per cent internally and far higher on the eastern trade; a five per cent tax on inheritances and another on manumissions, both falling on Roman citizens; and a one per cent duty on auction sales. Egypt and Africa also paid a large part of their obligation in grain rather than money, shipped directly to feed Rome.

Two absences matter. There was no income tax, because there was no way to measure income. And Italian land paid no regular tribute at all, a privilege dating from 167 BC that survived until Diocletian, so the province that supplied the emperors was the one province that did not pay for them.

Example. A papyrus in Vienna, the Muziris papyrus, records a contract over a single cargo carried from the Indian port of Muziris to Egypt in the second century. The cargo is valued at close to seven million sesterces and the customs duty on this route was a quarter of the value. What does one document tell you?

That the eastern trade was enormous and that the state took a very large cut of it. Seven million sesterces is about what it costs to pay 7,800 legionaries for a year, in one ship, and the quarter duty on it is roughly 1.75 million sesterces from a single voyage. Pliny the Elder complained that India, China and Arabia between them drained 100 million sesterces a year out of the empire, which is often quoted as evidence of a ruinous trade deficit; set against a state revenue in the region of 800 million it would be about an eighth, which is large. But read what the customs rate does to the complaint. If the goods pay 25 per cent on entry, the treasury takes a quarter of that flow back at the frontier, so the trade Pliny thought was bleeding Rome was one of the state's better revenue streams. A moralist counting bullion out and an accountant counting duty in are looking at the same ships.

Now you. How much should one contract be allowed to prove?

The budget, as far as it can be reconstructed

Modern estimates of imperial revenue in the first and second centuries cluster around 800 million sesterces a year, with serious arguments for figures a good deal higher. Estimates of what the army cost cluster around 400 to 500 million. Take the middle of both and the army absorbs somewhere over half of everything the state collected, with the grain distributions in Rome, building, the imperial household, salaries and gifts sharing the rest.

That single ratio explains a great deal of what follows in this course. A state spending more than half its revenue on soldiers has no room for a bad year, no reserve to speak of, and no way of increasing the army without a proportional increase in revenue. When later emperors need more soldiers, they will not be able to afford them, and the strain shows up as debasement, as tax in kind, and eventually as a completely rebuilt fiscal system.

How heavy was it?

Now put the revenue against the size of the economy it came from. Take sixty million people and a bare subsistence income of about 250 sesterces a head, which is the sort of figure the price of wheat and the wages in Egyptian documents support. That gives an economy of about 15 billion sesterces. Revenue of 800 million is 5.3 per cent of it.

Even doubling the revenue estimate leaves the Roman state taking a tenth of national income, against thirty to forty-five per cent in a modern developed country. By the standards of any later state the empire was very lightly taxed.

Both halves of that finding matter. It explains the durability: taxation at that level is survivable for the people paying it in most years, and the provinces had little to gain from revolt. It also explains the fragility, because a state taking five per cent has no elasticity. There is no fat to cut and no capacity to double the army in an emergency, and when the third century demands exactly that, the empire will have to reconstruct itself to get it.

Who did the collecting

Under the Republic the taxes of a province were auctioned to companies of publicani, who paid the state up front and extracted what they could, an arrangement the earlier course shows going badly wrong in Asia. The empire kept tax farming for customs but moved the direct taxes onto a different basis: the city.

Each self-governing city was assessed a total, and its own council was responsible for collecting it from the surrounding territory and delivering it. The men on the council, the decurions, were personally liable for any shortfall. In good times this was a cheap and effective arrangement, since the local rich knew who owned what and had every incentive to keep the assessment low and the peace good. In bad times it turned membership of a city council from an honour into a liability that men would go to considerable lengths to escape, which is a slow-acting poison and one of the more revealing symptoms of the later empire.

Example. Suetonius reports that Vespasian said on his accession that 40,000 million sesterces were needed to put the state on its feet. Test the figure.

Against a revenue near 800 million a year, 40 billion is fifty years of the entire income of the empire, which no ruler could contemplate raising and no state that size could conceivably spend. Compare a figure from the same author that is credible: Tiberius left 2,700 million in the treasury at his death, itself about three and a half years of revenue and remarkable enough that everyone commented on it. Vespasian's number is a hundred times the largest hoard anyone had ever heard of. The likely explanations are a corruption in the numeral, which is the commonest error in the transmission of ancient texts, or a rhetorical figure meaning an unimaginable sum. The habit worth taking from this is simple: when an ancient number is impossible, do not throw the source away, because the surrounding facts may be perfectly good. Check whether the number can be checked against another number in the same author, and say which one you trust and why.

Now you. What made Tiberius's 2,700 million sesterces possible, and why did no later emperor accumulate anything like it?

Where the money went, and what it did on the way

One more consequence deserves a paragraph, because it connects the taxes to the economy that later lessons measure.

Taxes were collected mostly in cash in the interior provinces, and spent mostly in cash on the frontiers, where the soldiers were. That flow is a one-way movement of coin out of Spain, Gaul, Asia and Africa towards the Rhine, the Danube and the Euphrates, and the only way for those interior provinces to get the coin back to pay next year's tax is to sell something to the regions where the money now sits. A tax system of that shape does not merely take money: it forces the taxed to produce a surplus for a distant market, and it thereby creates long-distance trade in ordinary goods, not just luxuries.

That is a modern reconstruction rather than a Roman theory, and it is testable, which is what makes it useful. If it is right, the archaeology should show mass movements of oil, wine, grain and pottery between provinces, peaking when the tax system worked best. It does, and a later lesson goes through the evidence.

Before that, though, the taxes had to be assessed and collected, and the previous lesson left one soldier for every 240 inhabitants. The next lesson counts the civilian side of the administration, which turns out to be smaller still.

The Roman empire was governed by fewer officials than a medium-sized modern city employs, and the way it managed that is the single most important fact about how it worked.

The previous lesson found an army of about one soldier per 240 inhabitants and a tax take of roughly five per cent of everything the empire produced. Both numbers point the same way: whatever held the provinces together, it was not the weight of the state pressing on them. This lesson counts the civilian administration and finds it smaller still, then asks what actually did the governing instead.

The arithmetic of an absent state

Under Augustus the empire's entire senior administration, senators and equestrians holding posts with real authority, numbered somewhere around 150 men. The equestrian service grew steadily, and by AD 200, with procurators, prefects and a much larger financial establishment, the figure is perhaps 350. Below them sat clerks, slaves and freedmen of the imperial household, and soldiers seconded to office work, but those are staff, not decision makers.

Example. Set 350 senior officials against sixty million people, then set the same population against the salaried officialdom of Han China, which ran to something over 130,000 men. What kind of state is Rome?

Rome gives one official per 171,000 inhabitants; Han China gives one per 460. Two empires of similar extent and similar population differ by a factor of nearly four hundred, which is not a difference of degree. The Han state reached into the county and the village, registered households, assessed individuals and moved grain by administrative order. The Roman state did almost none of that directly, and a Roman governor could not have registered a village if he wanted to, because he had nobody to send. So either the empire was not really governed, which the tax receipts and the law reports refute, or the governing was done by somebody who does not appear in the count of imperial officials. The rest of this lesson is about who. Keep the comparison rather than the figures: the Han total is disputed and the Roman one is a reconstruction, but no plausible adjustment to either closes a gap of that size.

Now you. A governor of Bithynia arrives with a staff of perhaps a few dozen, including his friends, his clerks and some seconded soldiers, to administer a province of a couple of million people. What can he actually do with his year?

The city as the unit of government

The empire was a mosaic of roughly two thousand self-governing communities, the civitates, each consisting of an urban centre and the countryside attached to it. Where cities already existed, in Greece, Asia Minor, Syria and Africa, Rome used them. Where they did not, in Gaul, Spain and Britain, Rome created them, converting tribes into civitates with a capital and a council so that the tribe would have an address, a set of magistrates and a body that could be held responsible for a tax total.

Each had a council, usually of about a hundred decurions, drawn from men who met a property qualification, at Pliny's home town of Comum set at 100,000 sesterces. Each had annually elected magistrates, usually two senior ones with judicial power and two responsible for streets, markets and public buildings. It is the constitution of the Roman Republic in miniature, reproduced in two thousand copies, and that similarity is not an accident: it was the only model of self-government Rome had.

Divide sixty million by two thousand and the average community holds 30,000 people. The average is misleading in the usual way, because Rome itself held around a million, Alexandria and Antioch several hundred thousand, and the median civitas was a small town of a few thousand with farmland around it. But the order of magnitude is right, and it is the useful figure: the empire delegated the government of a few tens of thousands of people to a hundred local landowners, and then dealt only with them.

What a charter actually says

The best evidence for how much autonomy this meant is a set of bronze tablets found near Seville in 1981, the charter of the small town of Irni, granted under the Flavians around AD 91. About two thirds of the text survives, and it is the fullest Roman municipal constitution we have.

It reads like a manual: how magistrates are elected and in what order the voting groups are called, how the council is summoned and what quorum it needs, how public contracts are let, how the accounts are audited, how a freed slave's status is recorded, how much a magistrate may spend on games. Two provisions are worth more than the rest.

The first is that the local courts may hear civil cases up to a value of 1,000 sesterces, and nothing above. That is a little over a year's pay for a legionary, so it covers the ordinary business of a small town and stops precisely where property of any real size is involved. Above the line, the case goes to the governor. Roman rule is not the absence of interference: it is a threshold.

The second is that a man who holds a magistracy at Irni becomes a Roman citizen when he leaves office, along with his parents, his wife and his children. The town is a Latin community, not a Roman one, and the charter turns its annual elections into a citizenship machine. That provision is where this lesson connects to the next one, and it explains a great deal about why local men competed for offices that cost them money.

Why the local rich paid for the privilege

They did pay, and heavily. A man entering a magistracy or priesthood owed a summa honoraria, an entry fee to the treasury of his own town, recorded at a couple of thousand sesterces in modest places and tens of thousands in the wealthy cities of Africa. On top of that he was expected to give: to pave a street, roof a portico, endow a festival, distribute cash at his son's coming of age, or build something with his name across the front of it. The system has a modern name, euergetism, and the inscriptions on which it rests number in the tens of thousands, because the whole point of the gift was that it was recorded.

Pliny the Younger, whose letters make him the best-documented benefactor in the empire, gave his home town a library with 100,000 sesterces for its upkeep, offered to pay a third of the cost of hiring teachers if the parents found the rest, and endowed a fund for the maintenance of local children. The endowment is worth doing the arithmetic on, because he explains his reasoning.

Example. Pliny conveyed an estate worth more than 500,000 sesterces to an agent of the town, then took the land back charged with a permanent rent of 30,000 sesterces a year. Why not simply hand over the cash, and what return did he set?

The return is 30,000 on 500,000, or six per cent, which is close to the ordinary yield on Italian land. He explains the structure himself: a gift of cash would be spent, and an estate given outright would be farmed badly by whoever the town appointed. So the town gets neither the capital nor the management, only an income charged on land that stays in the hands of somebody with every reason to keep it productive. Notice what that tells you about ancient public finance. There is no bank, no bond and no institution that can hold and invest money on a town's behalf, so an endowment meant to last has to be a charge on land, which is why so many benefactions take the strange-looking form of a rent or an obligation attached to a farm rather than a sum in a treasury.

Now you. Trajan's alimenta scheme worked the same way on a state scale. At Veleia in northern Italy an inscribed bronze table records loans of 1,044,000 sesterces secured on local estates at five per cent, funding 245 legitimate boys at 16 sesterces a month, 34 legitimate girls at 12, one illegitimate boy at 12 and one illegitimate girl at 10. Do the two sides balance?

The governor and the leaks in the system

Delegated government works while the delegates are willing and solvent. Pliny's posting to Bithynia and Pontus around AD 110 is the best surviving picture of what happened when they were not, because he wrote to Trajan about almost everything and the correspondence survives, 121 letters in the tenth book of his collection, most with the emperor's reply attached.

The province's cities had been spending. Nicomedia had put 3,318,000 sesterces into an aqueduct, abandoned it, demolished it, and started another. Nicaea had spent more than ten million sesterces on a theatre that was cracking, and had a gymnasium burnt down and rebuilt to no plan anybody could follow. Claudiopolis was excavating an enormous bath at the foot of a hill, financed by the entry fees of newly enrolled councillors. Pliny is not investigating corruption in any of these cases. He is reporting waste, on a scale that is easy to feel: the Nicomedian aqueduct is the annual pay of nearly 3,700 legionaries, poured into a hole and then knocked down.

Example. Trajan sent Pliny to Bithynia with a special commission rather than letting the province take its turn under an ordinary proconsul. What does the appointment itself tell you, before a single letter is read?

That the normal system had no way of noticing a problem until it was large. Bithynia's cities had been mismanaging their finances for years and two of its recent proconsuls had been prosecuted for extortion, and nothing in the machinery brought either fact to Rome's attention early. There was no audit, no inspectorate and no reporting line from a city council to anybody. Correction came as an exception, a trusted man with an enlarged brief, and it came after the money was gone. Rome paid that price knowingly, because the alternative was a salaried provincial civil service it could neither afford nor staff. Notice too what 121 letters implies about the volume of business reaching an emperor who governed forty-odd provinces: it was small enough that Trajan could read it, which is only possible because almost nothing came up.

Now you. Trajan repeatedly refuses Pliny's requests, most famously by rejecting a fire brigade of 150 men for Nicomedia after the city burned. Why refuse something so cheap and so obviously useful?

What the cities got out of it

It is easy to read all this as extraction: Rome takes the tax, the local rich pay the costs of government and buy the honours, and the imperial power supplies almost nothing. That undersells the exchange.

A city under Roman rule was safe from its neighbours in a way no Greek city had ever been, and the wars that had periodically wrecked Asia Minor stopped. Its elite kept the local standing they already had and gained a route out of it, through equestrian rank, imperial service and eventually the Senate. Its ambitions could now be pursued in a currency other than violence, which is why the cities of the east put such extraordinary energy into competing over titles, temples and games. That competition looks trivial and was not: it was what political life became once the alternative had been closed off, and the buildings it produced still stand.

The price is worth naming now because it becomes lethal later. The men who ran the cities were personally liable for the tax total, and the honours were expensive. While being a decurion was worth more than it cost, the system financed itself and needed no enforcement. When the third century made the burden heavier and the honour cheaper, it stopped working, and the state's response was to make the office compulsory and hereditary, which is a good working definition of an institution that has died.

For now the exchange held, because the reward was real. The next lesson takes up what that reward was: membership itself, and how a status that began as the privilege of a few Italian towns spread until an emperor could grant it to every free inhabitant of the empire at once.

Rome's most unusual institution was not its army or its law but its willingness to turn conquered people into Romans, and no other ancient state did it on anything like the same scale.

The previous lesson ended with local notables paying for the privilege of governing their own towns on Rome's behalf. This lesson asks what they were buying. The short answer is membership, in a status that could be inherited, earned, bought or given, and that spread until it covered everyone. Along the way the ruling class of the empire stopped being Italian, and the emperors themselves stopped being Italian, without anybody deciding that this should happen.

What the status actually did

Roman citizenship was a bundle of legal capacities rather than a nationality. A citizen could contract a marriage recognised in Roman law, so his children inherited from him under that law; he could own property in the Roman form, make a valid will, and sue in a Roman court. He was exempt from the poll tax that provincials paid. He could not lawfully be flogged or tortured by a magistrate without trial, and he had the right of appeal, which by the imperial period meant appeal to the emperor.

Those protections were real and they were also patchy. Governors flogged citizens when it suited them and were occasionally prosecuted for it, and the status did more for a rich citizen than a poor one, because enforcing it required somebody to complain and be listened to. But the exemptions had cash value and the legal capacities had commercial value, and the poll tax exemption alone was worth having in a province like Egypt.

The one thing citizenship did not do is what a modern reader expects: it carried no vote worth using. The assemblies at Rome had stopped legislating under Tiberius, and a citizen in Gaul was never going to travel to Rome anyway. It was a legal status and a claim on protection, not a share of political power. That is precisely why it could be given away so widely.

Counting the Romans

Augustus recorded three censuses in the Res Gestae: 4,063,000 citizens in 28 BC, 4,233,000 in 8 BC and 4,937,000 in AD 14. Tacitus gives a further figure from Claudius's census of AD 47: 5,984,072. Four numbers across seventy-five years, from two independent sources, and they let you watch the status spread.

Example. Between AD 14 and AD 47, the citizen body grows from 4,937,000 to 5,984,072. Can natural increase account for that?

The rise is 1,047,072, or 21.2 per cent in thirty-three years, which is an annual growth rate of 0.58 per cent. That is far above what a pre-industrial population sustains. Ancient populations grew, when they grew at all, at something nearer 0.1 per cent a year, and at that rate the citizen body of AD 14 reaches only about 5,103,000 by AD 47. So roughly 880,000 citizens have to come from somewhere other than births.

The sources are known and they add up. Every slave formally freed by a Roman owner became a citizen, which is a Roman peculiarity and a large flow in a society with millions of slaves. Every auxiliary discharged after twenty-five years became one, and at around 5,000 discharges a year over thirty-three years, counting the children included in the earlier grants, that alone is several hundred thousand. Add grants to individuals, to whole towns, and to the magistrates of Latin communities year by year. The honest qualification is that we do not know for certain what the census counted, whether adult males alone or men, women and children, and the answer changes the absolute size of the citizen body by a factor of three. It does not change the growth rate, because the same basis was presumably used each time, and the growth rate is what the argument rests on.

Now you. The last well-attested Republican census, in 70 or 69 BC, returned 910,000. Augustus returns 4,063,000 in 28 BC. What happened, and what should make you cautious?

The ways in

Birth to two citizen parents was the ordinary route. Beyond it, four doors stood open.

Manumission was the widest. A slave freed by proper procedure became a citizen, with some restrictions on himself and none on his children, so the descendants of slaves were full citizens within a generation. Military service was the second: the auxiliary diploma of the previous lesson, arriving five thousand times a year. The third was office. In a community holding Latin rights, the annual magistrates and often their whole families became citizens on leaving office, so an entire local elite converted itself over a couple of decades. Vespasian granted Latin rights to the whole of Spain in the early 70s, which put several hundred communities on that escalator at once.

The fourth door was the emperor's personal grant, and it could be bought.

Example. In the Acts of the Apostles the tribune commanding the garrison in Jerusalem is about to have Paul flogged, discovers he is a citizen, and says that he acquired the status himself for a large sum. Paul replies that he was born to it. Read the exchange as evidence.

It gives you three things at once. First, the protection was operative: the tribune stops the flogging immediately and is alarmed, which means the rule was known to bind even a soldier in a distant province. Second, the status was purchasable, presumably through officials at court who arranged grants for a fee, and the tribune's name in the text, Claudius Lysias, tells you which emperor's reign he bought it in, because a new citizen took the name of the man who made him one. Third, and most interesting, birth outranks purchase in the exchange. Paul's reply is a small piece of social one-upmanship, and it only works if everyone present understands that a bought citizenship is second best. That is the ordinary fate of any status that can be acquired: it keeps its legal content and loses its social edge, and the edge is what people are actually competing for.

Now you. New citizens took the family name of the emperor or patron who enfranchised them. What does that convention allow a historian to do?

Claudius, the Senate, and two versions of one speech

In AD 48 the emperor Claudius proposed that leading men from Gallia Comata, the parts of Gaul conquered by Caesar less than a century earlier, should be eligible for the Senate. We have his speech twice: as Tacitus wrote it up in the Annals, and as it was actually delivered, because the Lyon assembly had it inscribed on bronze and a large part of the tablet survives.

Compare them and you get a rare controlled experiment in how ancient historians work. Tacitus gives a tight, well-organised argument: Rome has always taken in outsiders, from Sabine kings onwards; Sparta and Athens fell because they held their conquered at arm's length; what is now old was once new. It is lucid, forceful and about a third the length of the real thing.

The bronze shows Claudius wandering. He digresses into the ancestry of the Etruscan kings, interrupts himself to acknowledge that he is going too far back, worries about what people will say, and defends his own antiquarian learning. The argument Tacitus reports is genuinely in there, and so is a great deal that Tacitus cut.

Example. Tacitus did not invent the speech and did not reproduce it either. What is the right conclusion about his method, and how far can it be generalised?

That he worked from the substance and rewrote the form. The historical content, the precedents, the appeal to Rome's history of absorption, the answer to the objection that Italians should be enough, is all present in the original; what Tacitus supplied is compression and shape. So for this one speech we can say that he is reliable about what was argued and unreliable about how it sounded, which is roughly the standard an ancient reader expected, since a historian was supposed to give a speaker the words appropriate to the occasion.

The generalisation has to be careful, and in two directions. Favourably, Tacitus had access to the acta senatus and evidently used them, so his senatorial debates rest on record rather than imagination. Unfavourably, this is one speech, one that happens to have been inscribed because provincials were proud of it, and it is exactly the sort of case where a historian is likely to have been careful. It does not license trusting the speeches in Tacitus's account of Tiberius's court, where no record was public and the whole point of the scene is often what could not be said aloud. One verified instance raises the odds; it does not settle the class.

Now you. The Senate resisted. Within a century the resistance had become irrelevant. Why?

The edict of 212

In 212 Caracalla granted Roman citizenship to all the free inhabitants of the empire. The text survives, badly damaged, on a papyrus from Egypt, in a Greek version whose opening gives a pious reason: the emperor wishes to bring all his people to the gods in thanksgiving for his own preservation. There is an exclusion clause naming the dediticii, a category we cannot securely identify, and it evidently mattered to very few.

Cassius Dio, writing a generation later and hostile, says the point was money: certain taxes, notably the five per cent on inheritances, fell on citizens only, and Caracalla had just doubled that tax to ten per cent. Universal citizenship makes a citizens-only tax into a universal one. The motive is plausible, the arithmetic works, and it is not the whole story, because the edict also removed the poll tax basis in some provinces and Dio wanted Caracalla to look grasping. The safest reading is that a fiscal motive and a religious one are not in competition: the grant was cheap to make, useful to the treasury, and it stated a fact that a century of drift had almost brought about anyway.

What is genuinely striking is how little seems to have changed. There was no celebration recorded, no crisis, and no rush of new privileges. The best evidence that anything happened at all is the naming habit: from 212 the papyri of Egypt fill with people called Aurelius, Caracalla's family name, appearing suddenly in front of names that had been Greek or Egyptian the year before. Whole villages of Aurelii, in documents about donkeys and rent.

What was left to be unequal about

The reason so little changed is that legal distinction did not disappear; it moved. Across the second century Roman law had been developing a division between honestiores, the respectable, meaning senators, equestrians, soldiers and town councillors, and humiliores, everyone else. The two were punished differently for the same offence: the honestior was fined, exiled or beheaded, the humilior was beaten, sent to the mines, thrown to the beasts or crucified. By the time everyone was a citizen, the protections citizenship had once carried against flogging and degrading execution belonged to rank instead.

That is the thing to carry forward. A society does not become equal by extending a status to everyone; it re-founds the inequality on a distinction the extension does not touch. The empire ended up with universal citizenship and a sharper legal hierarchy than it had begun with, and the second development is what shapes the late Roman world.

Membership on those terms was worth having as long as the empire itself was worth belonging to, and for two centuries it plainly was. The next lesson asks what paid for it, and finds that the answer changed early: after Augustus, the conquests stopped, and an empire that had always been funded by the next war had to learn to live on what peace produced.

Every Roman institution before Augustus assumed that the empire would keep getting bigger, and from the first century onwards it essentially stopped, which changed what the empire had to live on.

The previous lessons have described a state whose largest expense was a standing army, whose revenue came from taxing provinces, and whose citizenship spread outward through those provinces. All of that describes a machine running on a fixed input. Under the Republic the input had not been fixed: conquest brought indemnities, plunder, slaves and new tribute, and the earlier course on the Republic shows the wars of the second century BC paying for themselves several times over. This lesson asks when that ended, and why.

The year it stopped

In September of AD 9 the governor of Germany, Publius Quinctilius Varus, marched three legions and their auxiliaries through wooded country towards a reported rising, guided by a Roman-trained Cheruscan officer named Arminius who had arranged the report. Over three or four days the column was destroyed. Varus killed himself. The legions were the Seventeenth, Eighteenth and Nineteenth, and their numbers were never used again by the Roman army, which is as close to an official admission of disaster as Rome ever made.

The site was found in 1987 near Kalkriese, north of Osnabrück, by a British army officer with a metal detector, and the excavation has produced the shape of the ambush: a turf and timber wall built along a sand ridge, with the boggy ground of a moor on the other side, funnelling the column into a gap a few hundred metres wide. Among the finds are coins, and the coins date the site, because none of the many hundreds recovered is later than the issues Varus himself countermarked. Archaeology rarely confirms a literary account this cleanly.

Suetonius says Augustus beat his head against a door and called on Varus to give back the legions. Tacitus reports something more consequential: in the papers left at his death, Augustus advised that the empire be kept within its existing frontiers. Whether that was strategic wisdom or the rationalisation of an old man who had just lost an eighth of his army, his successors largely followed it.

What the advice meant in practice

Not that nothing moved. Claudius invaded Britain in 43, in part because he needed a victory to establish himself. Trajan took Dacia in two wars between 101 and 106, and then invaded Parthia, reaching the Persian Gulf in 116. But look at what happened next: Hadrian abandoned the whole Mesopotamian conquest on his accession in 117, and no emperor after Trajan added a province that lasted. Set against four centuries in which Rome had absorbed Italy, Spain, Gaul, Greece, Asia, Syria, Africa and Egypt, the imperial record is one significant permanent addition, Britain and Dacia, and Dacia was abandoned in the 270s.

The question is why. The usual answer, that Rome had reached its natural limits, is not much of an explanation, since the empire stopped at the Rhine rather than the Elbe, and neither is a natural line. The better answer is arithmetic.

The last conquest that paid

Dacia was rich in gold. The mines of the western Carpathians were worked by Rome immediately after the conquest, with imported miners from Dalmatia, and the wax tablets recording their contracts survive. The plunder from the war was enormous, and one late source, John the Lydian, quoting Trajan's own doctor Criton, gives a figure.

Example. Criton's figure as transmitted is 5,000,000 Roman pounds of gold and 10,000,000 of silver. Test it against the size of the Roman economy.

A Roman pound is 327 grams and the aureus was struck at 42 to the pound. Five million pounds of gold is therefore 210 million aurei, which at 25 denarii and 4 sesterces to the denarius is 21 billion sesterces. Against an imperial revenue near 800 million a year, that is twenty-six years of the entire income of the empire carried off from one province, and 1,635 tonnes of gold, which is a substantial share of all the gold mined in the ancient world. The figure is impossible.

Now apply the standard correction. Ancient numerals are corrupted in transmission more often than any other kind of text, and dividing by ten is the commonest repair, giving 500,000 pounds of gold and a million of silver. That is 2.1 billion sesterces, or about two and a half years of imperial revenue, and 163 tonnes of gold. Still spectacular, still the largest windfall of the imperial period, and now within the bounds of the possible. Notice how the two versions are treated differently: the impossible one is evidence that a big number was remembered, the emended one is a usable estimate, and neither should be quoted without saying which it is.

Now you. Trajan's triumph was followed by 123 days of games with a reported 10,000 gladiators and 11,000 animals, and the building of a new forum in Rome. What is the connection to the arithmetic above, and what did it fail to buy?

Why the next province was not worth taking

The clearest ancient statement of the calculation is Strabo's, written under Augustus, about Britain. Some of the British chiefs, he says, have obtained friendship with Rome and the trade already pays customs duties on the crossing in both directions; to garrison the island would need at least one legion and some cavalry, the cost of which would have to come out of the tribute, so the revenue would be reduced by the expense of the occupation and the risks would increase. Appian, writing after the conquest had happened, says the Romans hold the better part of Britain and have no need of the rest, since even what they hold is not profitable.

Example. Turn Strabo's remark into a general rule and use it to explain the shape of the empire.

The rule is that a province is worth taking when the tax it will yield exceeds the permanent cost of the troops needed to hold it, and that both terms depend on what is already there. Tax yield depends on a settled agricultural surplus, cities that can assess and collect it, and a population used to paying somebody. Garrison cost depends on terrain and on how organised the resistance is.

Run the rule around the frontier and it predicts the map. Egypt, Syria and Asia had cities, records and a habit of taxation, and Rome took them and held them with very few troops. Germany beyond the Rhine had no cities, no coinage of its own, no scribes and a dispersed population in difficult country, so the tax yield is small and the garrison cost is permanent and high. Dacia was the exception that proves the rule, because gold in the ground substitutes for a taxable population. Britain sits at the margin and, on Appian's testimony, on the wrong side of it. The empire's boundary is not a natural line and not a strategic doctrine: it is roughly the contour where the expected revenue met the cost of the legions.

Now you. If the calculation was that clear, why did Claudius invade Britain in 43 anyway?

What the frontier actually was

With expansion over, the edges had to be marked, and from Hadrian onwards they were, in different forms. In Britain, a wall of stone running 117.5 km from Wallsend to Bowness, begun in 122, with a milecastle roughly every Roman mile and two turrets between each pair. In Upper Germany and Raetia, a palisade and later a wall and ditch running some 550 km with around 900 watchtowers behind it. Elsewhere the frontier was a river, the Rhine, the Danube and the Euphrates, patrolled by fleets.

Example. The garrison on and immediately behind Hadrian's Wall was on the order of 10,000 men. Work out what the wall could and could not do.

Ten thousand men over 117.5 km is about 85 per kilometre, and most of them were in forts rather than standing on it. That is enough to watch, to challenge, to stop small parties and to raise an alarm, and nothing like enough to hold a line against a determined force that concentrated at a chosen point. The same arithmetic that made the whole army too small for a 10,000 km perimeter applies locally.

What the physical remains suggest instead is control of movement. The wall has gates at every milecastle, roughly every 1.5 km, which is a strange feature in a defensive barrier and an obvious one in a customs post. It has a ditch and a road behind it, useful for moving men sideways along the frontier rather than for facing outward. And it is visible from a long way off, which is a message as much as an obstacle. Read it as a boundary that regulated and taxed traffic, marked where Roman jurisdiction began, and told everyone on both sides where they stood.

Now you. The Antonine Wall was built 160 km further north in the 140s, held for about twenty years, and then abandoned in favour of the older line. What does that sequence show?

Living without windfalls

A state that has stopped conquering loses three things at once, and it is worth being precise about them.

It loses the irregular income. Indemnities, plunder and confiscated royal treasuries had funded the Republic's wars and much of Augustus's building, and after Dacia there is no more of it. From then on, everything must come from taxation, which the earlier lesson showed was running at about five per cent of national income with no obvious way to increase it quickly.

It loses the main supply of slaves. Mass enslavement of defeated populations had stocked Italian agriculture, and with the wars over, the supply had to come from breeding, from the frontier trade and from exposure of infants, which is one reason slave prices in Egypt look higher in the second century than the first, and one reason tenancy gradually displaced slavery on Italian estates.

And it loses the standard proof that an emperor deserved the job. Roman political culture had measured a leader by victory for four hundred years, and an emperor who could not win a war of conquest had to manufacture the appearance of one, or find a different basis. Hadrian, who abandoned territory and built walls, was disliked in the Senate for exactly this reason, and the coins and inscriptions of the second century work hard to present the emperor as the provider of peace, security and prosperity rather than as a conqueror.

That last shift is the one that matters most for the century that follows. If the empire is no longer paying for itself with conquest, then it has to be worth having for what it produces in peacetime. The next lesson asks whether it was, and how anyone can possibly know, by counting broken pots, sunken ships and lead in the ice of Greenland.

Nobody in the Roman empire ever counted its output, so any claim about whether the imperial peace made people better off has to be built out of evidence that was never meant to be economic evidence at all.

The previous lesson left the empire with a fixed frontier, no more windfalls and a revenue of about five per cent of what its inhabitants produced. Whether that was a good bargain for the inhabitants depends on what the peace produced. This lesson is as much about method as about Rome: four independent proxies, each with a clear result and a clear defect, and the case for a genuinely larger economy rests on the fact that they fail in different directions and agree anyway.

A hill made of jars

Behind the old river port of Rome stands a mound about 35 metres high covering some 22,000 square metres. It is made entirely of broken pottery: an estimated 53 million amphorae, deposited between roughly AD 30 and 260, and about eighty-five per cent of them are Dressel 20, the fat globular jar used to carry olive oil from the Guadalquivir valley in southern Spain.

It is a hill rather than a scatter for a specific reason. Oil soaks into a porous ceramic wall and turns rancid, so an oil amphora cannot be reused or ground up for building material the way a wine amphora can. It had to be emptied, smashed, stacked with lime to control the smell, and left. The result is an accidental archive, and a very good one, because these jars were labelled. Painted on the necks are the weight of the empty vessel, the weight of the oil, the shipper's name, the estate, the district, and a control note from the customs official, often with the year given by the consuls.

Example. Take 53 million amphorae over 230 years, at about 70 litres each. What flow does that describe, and against Rome's population, what does it come to per person?

Fifty-three million times seventy is 3.71 billion litres. Spread over 250 years that is 14.8 million litres a year, or about 212,000 jars a year, which is 580 jars broken every day for two and a half centuries. Against a city of a million people it is roughly 15 litres of olive oil per head per year.

Now notice what that comparison does. Fifteen litres a head is a believable figure for a Mediterranean diet, in which oil is fuel for lamps, soap, medicine and most of the fat in the diet, and estimates from other evidence run around 20 litres. The agreement is worth something, but the direction of the discrepancy is worth more: Monte Testaccio is one dump beside one port, receiving mainly the Spanish oil that passed through official channels, and it excludes African oil, Italian oil, oil that arrived in skins and everything consumed anywhere other than Rome. So the figure it yields should be too low, and it is. A proxy that lands slightly under a plausible total, for a reason you can state in advance, is behaving exactly as it should.

Now you. The painted labels give the shipper's name and the year. What can be done with that which cannot be done with the volume?

Counting the wrecks

The second proxy is ships that failed to arrive. Something over 1,200 datable Mediterranean shipwrecks have been catalogued, and when they are plotted by century the shape is dramatic: a steep rise from about 200 BC, a peak running from roughly 100 BC to AD 200, and a steep fall afterwards, with the level of the later empire well below that of the late Republic and not recovering for many centuries.

Example. Read the curve at face value, then attack it as hard as you can.

At face value it says that seaborne trade in the Mediterranean grew enormously in the late Republic, stayed high through the first two centuries of the empire, and collapsed thereafter, which is a striking result because nobody was trying to measure trade and the ships sank without regard to anyone's argument.

The attacks are serious. Wrecks are found because their cargo survives, and amphorae survive superbly while barrels, sacks and bulk grain leave almost nothing; if the empire shifted from amphorae towards barrels for wine, which the evidence from Gaul suggests it partly did, the later decline is partly a change in packaging. Wrecks are found where people dive, which means near coasts, near resorts and near modern harbour works. Datable wrecks are dated by their amphorae, and amphora typologies are much finer for some periods than others, so a century with a well-studied jar sequence produces more dated wrecks than one without. And a wreck count measures voyages and losses together, so better navigation would depress it while trade rose.

Where that leaves the curve is this: its exact height is not evidence of anything, and its overall shape is too large to be produced by these biases alone. Every objection above is a reason to distrust the fall after AD 200 more than the rise before it, and the rise is a factor of several over a few centuries. So the honest summary is a strong claim about growth, a weak claim about the decline, and no claim at all about the numbers on the axis.

Now you. The same catalogue puts the peak of the curve in the last century BC and the first century AD, which is to say that most of the rise happens before the imperial peace exists. Does that undercut the claim being made here?

Lead in the ice

The third proxy comes from Greenland. Ice sheets accumulate annual layers of snow that trap whatever was in the air, and cores drilled through them can be counted back like tree rings. Cores dated in this way show atmospheric lead rising from around 600 BC, peaking near the beginning of our era at roughly four times the natural background, falling away thereafter, and not returning to that level until the later middle ages.

The connection to Rome runs through silver. Silver was extracted from lead ore by cupellation: the ore is smelted and the lead is oxidised off, and the process releases lead vapour in quantities far greater than the silver recovered. Fine particles circulate through the northern hemisphere and settle. So a core drilled in the Arctic is a record of the output of mines in Spain and Britain, laid down by people who had no idea they were leaving one.

Its virtues are that it is quantitative, continuously dated and wholly independent of what archaeologists choose to dig. Its limits are that it measures one industry, that richer ore would yield the same silver with less pollution, and that attributing northern-hemisphere lead to Rome specifically requires an argument rather than an assumption. Recent high-resolution work has found dips that line up with known events, including the years after the plague of the 160s, which is either a striking confirmation or a warning about how easily a wiggle can be matched to a date already in mind. Take it as strong evidence about the scale of Roman metal production and weak evidence about any particular year.

How many people lived in towns

The fourth proxy is settlement. Rome held somewhere near a million people, a figure supported from several directions: the grain distributions, which had around 200,000 adult male recipients, the area within the walls, and the capacity of the aqueducts. Alexandria was probably half that, Antioch, Carthage and Ephesus a few hundred thousand between them, and a long tail of provincial towns of five to twenty thousand ran to something like two thousand communities.

Any city of that size is an economic statement, because it has to be fed from somewhere else.

Example. Reckon a person's grain at 40 modii a year, a modius at about 6.55 kg, and a qualifying grain ship under Claudius's shipping incentives at 10,000 modii. What does feeding Rome require?

A million people need 40 million modii, which is 262,000 tonnes of wheat, and at 10,000 modii a ship that is 4,000 shiploads. Sailing was concentrated in roughly eight months of the year, so if the arrivals were spread across a 240-day season, Rome's ports had to receive about seventeen grain ships a day, every day, for eight months, on top of everything else that came in.

That single number carries most of the argument of this lesson. Seventeen ships a day requires harbours, warehouses, lighters to move cargo up the Tiber, a market in shipping, credit to finance voyages, insurance in some form, officials to check quantities, and a legal system that will enforce a contract between an Egyptian shipper and a Roman merchant. None of that is optional, all of it is expensive, and none of it exists unless the aggregate flows are very large. Rome was not fed by heroic effort; it was fed by an ordinary working system, and the system is the finding.

Now you. Skeletal evidence from Roman Italy shows average adult height flat or slightly declining across the imperial period. Does that overturn everything above?

What the four say together

None of these is a measurement of the Roman economy. Together they constrain it, and they do so from four directions that share no assumptions: a rubbish dump, a shipwreck catalogue, an ice sheet and a set of city sizes.

They agree on a long expansion beginning in the late Republic, a peak that lasts through the first and second centuries, and a decline afterwards. They agree that the scale involved was not matched again in Europe for a very long time, on any of the measures: no western European city reached Rome's size again until London around 1800, and northern-hemisphere lead pollution did not return to its early imperial level until the later middle ages.

Why the state was the cause

What produced it is not mysterious, and the earlier lessons have already supplied most of the causes. Piracy was suppressed and the whole Mediterranean was under one authority, so the risk premium on a voyage fell. One coinage and one commercial law covered the entire basin. Roads and harbours were built by an army that was being paid anyway. Taxes were collected in the interior and spent on the frontiers, which forced interior provinces to sell in order to get the coin back. And two thousand cities of local landowners were competing to build, which turned rural surplus into urban demand.

Every one of those is an achievement of the state, which is why the third century is going to hurt so much. This economy was not independent of the empire that produced it, and it will not survive the empire's difficulties as a separate thing.

Before those difficulties arrive, one more feature of this world needs describing, because it will turn out to matter more than anything else in this course. The empire that traded oil and counted amphorae also sacrificed, in every city and at every public act, and its relationship with the gods was a civic transaction. The next lesson looks at what happened when a group appeared that would not perform it.

Roman religion was a set of things you did rather than a set of things you thought, and almost everything strange about the empire's treatment of Christians follows from that one fact.

The previous lessons have described a state that governed through cities and asked very little of its subjects. Religion is where it asked something, and the request was small: take part. This lesson works out what taking part meant, why the request was so easy for almost everyone to meet, and why the group that could not meet it ended up producing the only universal order the empire ever issued to its inhabitants before the edict of 212.

What the sacrifice was for

The Latin for the relationship is pax deorum, the peace of the gods, and it was understood as a bargain of maintenance rather than devotion. The community performed the rites correctly, at the right times, with the right words and the right animals, and the gods in return did not send plague, defeat, flood or famine. If the rites were done wrongly they were done again from the beginning, sometimes several times, because the fault lay in the performance.

Two consequences follow. The first is that there was no creed, no requirement of belief, and no religious authority to define one. Priesthoods at Rome were offices held by magistrates and senators, not a separate class, and the emperor from 12 BC held the chief of them. A Roman could think what he liked about the gods, and educated Romans wrote sceptical books about them while holding priesthoods, without any sense of contradiction.

The second is that the system was infinitely expandable. A new god was not a rival claim but an addition, and the empire absorbed Isis, Mithras, Cybele, Jupiter Dolichenus and dozens of local deities without difficulty. Where the state did intervene, as against the rites of Bacchus in 186 BC or against magicians and astrologers repeatedly, the worry was never doctrinal. It was the same worry the previous lessons found about fire brigades: organised bodies of people meeting privately.

The emperor among the gods

Into this framework the imperial cult fitted with almost no friction. In the Greek east, where honours to rulers were an old habit, cities built temples to Rome and Augustus within his lifetime and competed for the right. At Rome the etiquette was stricter: a living emperor received no cult, but a dead one could be voted divus by the Senate, and Augustus, Claudius, Vespasian and Titus were. Vespasian is said to have remarked on his deathbed that he thought he was becoming a god, which is a joke that only works if everyone understood the procedure and nobody quite believed the claim.

What mattered practically was not theology but the oath and the gesture. Swearing by the emperor's genius, burning incense before his image, and pouring a libation were the ordinary acts of a loyal subject, performed at the opening of a court, at a public dinner, on an accession. They were closer to a national anthem than to a communion, and refusing them looked like what refusing an anthem looks like.

The exemption that had a price

One group had a standing exception. Jews did not sacrifice to Roman gods, and Rome accepted this on the grounds of ancestral custom, which was a category Romans respected. Instead the Temple in Jerusalem offered sacrifices on behalf of the emperor, which satisfied the requirement of participation without breaching Jewish law, and Jewish communities across the empire were excused the imperial cult, allowed to send the Temple tax, and exempted from military service.

The revolt of 66 to 73 ended that accommodation without ending the exemption. After the Temple was destroyed in 70, Vespasian kept the tax and redirected it: every Jew in the empire, now including women, children and the elderly, paid two denarii a year to the treasury of Jupiter Capitolinus. A tax formerly paid to one's own god was now paid to the god of the people who had destroyed his house. Domitian enforced it aggressively, and Nerva made a point of stopping the abuses, striking coins announcing that malicious accusation under the Jewish tax had been abolished, which is a rare case of an emperor advertising the correction of his predecessor on the coinage.

The arrangement is worth holding onto because it defines the shape of the problem. Rome could tolerate a group that would not sacrifice, provided the group was an ancient nation with its own customs, was identifiable, paid something, and did not recruit.

Why Christians did not fit

Christians failed all four conditions. They were not an ancient nation, and in Roman eyes a novel practice had no claim on respect at all. They were not identifiable, since they came from every people and could not be told apart until they were asked. They refused sacrifice on principle rather than by ancestral custom. And they recruited, across households and classes, which meant the refusal spread.

That combination produced a genuinely novel category, and the state took decades to work out what to charge them with. Tacitus records the first attempt: after the fire of 64, Nero fixed the blame on a group popularly detested, whose founder Christus had been executed under Tiberius by the procurator Pontius Pilatus, and had them killed in ways Tacitus thought excessive enough to produce sympathy. He says they were convicted of hatred of the human race, which is a charge about attitude rather than about any act.

Example. Around AD 112, Pliny writes to Trajan from Bithynia asking how to try Christians, and describes what he has been doing. He asks the accused three times whether they are Christians, warning them of the penalty, and executes those who persist. Those who deny it he tests by requiring them to invoke the gods, offer wine and incense to the emperor's image, and curse Christ. Read the procedure. What is the crime?

The procedure answers the question, and the answer is that the crime is being a Christian. Pliny investigated what Christians actually did and found nothing criminal in it: they met before dawn, sang a hymn to Christ as to a god, bound themselves by oath not to steal, rob, commit adultery or break faith, and met again to eat ordinary food. He tortured two slave women who held office among them and found, in his words, nothing but a debased and extravagant superstition. Having established that the group commits no offences, he goes on executing its members.

That is only coherent if the offence is the refusal itself, and the test he applies makes it explicit, because it tests behaviour rather than belief: sacrifice, and you are released whatever you privately think. The state is not asking for faith and does not care about it. It is asking for a public act of participation and treating persistent refusal as contumacy, which a Roman court could punish almost by definition. Notice how sincerely puzzled Pliny is throughout. He is a competent administrator who has run into a category his law has no name for, and he is asking his emperor for one.

Now you. Trajan replies that Christians must not be sought out, that anonymous accusations have no place and are unworthy of the age, but that anyone properly accused and convicted must be punished unless he sacrifices. Is that tolerance?

How many, and how fast

Nobody counted Christians, and the sources that do give numbers are apologists with an interest in exaggeration. What can be done is a growth model, and it is more informative than it sounds.

Example. Suppose there were about a thousand Christians in the empire around AD 40, and suppose the movement grew at 40 per cent a decade, which is roughly the sustained rate observed in modern new religious movements. Where does that get you?

Forty per cent a decade doubles the number every 2.06 decades, so about every twenty-one years. Starting from a thousand in AD 40 gives around 7,500 by AD 100, 40,000 by 150, 218,000 by 200, 1.2 million by 250, and 6.3 million by 300, which is about ten and a half per cent of sixty million people.

That last figure is close to what historians independently estimate for the eve of Constantine, and the whole point of the exercise is that nothing dramatic was required to reach it. No mass conversions, no miracles of scale, no imperial favour: a modest rate, compounded for two and a half centuries. It also explains why the state's attitude changed when it did. At one in five hundred, around AD 200, Christians are a curiosity; at one in ten, around 300, they are a constituency, and any emperor doing arithmetic can see which way the curve is going.

The model is an assumption dressed as a calculation, and its honest status is this: it does not show that Christianity grew at 40 per cent a decade. It shows that a rate low enough to be unremarkable is sufficient to explain the outcome, which removes the need for any special explanation and puts the burden on anyone who wants one.

Now you. Pliny tells Trajan that the temples in his province had been almost deserted, that the sacred rites are now being resumed, and that fodder for sacrificial victims has found buyers again. How much should that be allowed to prove?

Decius, and the first order to everyone

In 250 the emperor Decius issued an edict requiring the inhabitants of the empire to sacrifice to the gods, pour a libation, taste the offering, and obtain a certificate from a local commission recording that they had done so in its presence.

Around forty-five of those certificates survive on papyrus from Egypt, and they are startlingly bureaucratic. A woman gives her name, her father's name, her age and a distinguishing scar, states that she has always sacrificed to the gods and has now sacrificed and poured libations and tasted the offerings in the commissioners' presence, and asks them to certify below. Two commissioners sign. It is a form.

Example. The edict, as far as anyone can tell from its text and from the certificates, never mentions Christians. What was it for?

It was a universal act of supplication, ordered because the empire was in trouble. The 240s had brought defeat on the Danube, a newly aggressive Persia and the beginning of the run of short reigns the next lesson describes, and the Roman response to disaster was always the same: the peace of the gods has been broken, so restore it properly and comprehensively. Decius did on an imperial scale what a city did after a bad omen.

The consequence for Christians was severe and incidental. For the first time the state was not waiting for accusations but requiring an act from everybody and issuing paperwork, so refusal now had a name, an address and a missing document. The persecution that followed produced not only martyrs but a mass of Christians who sacrificed, or who bribed commissioners for certificates without sacrificing, and the church spent years arguing about whether they could be readmitted. That argument mattered more in the long run than the deaths, because settling it forced the church to develop procedures for authority and forgiveness it did not previously have.

Now you. Decius's edict was universal, and the edict of 212 that made everyone a citizen was universal. What does the pairing tell you about where the empire was heading?

The pressure that produced Decius's edict is the subject of the next lesson. Between 235 and 284 the empire lost more emperors than it had had in the previous two centuries, lost one of them to capture by a Persian king, broke into three pieces, and destroyed its own silver coinage. That it survived at all is the surprising part.

For half a century the Roman empire failed at almost everything a state is for, and the interesting question is not why it nearly collapsed but why it did not.

The previous lessons built a picture of a system in balance: a small army on a fixed frontier, a light tax, government delegated to cities, a coinage nobody doubted, and emperors who mostly died in bed. This lesson watches every one of those hold-downs come loose at once, starting with the murder of Severus Alexander by his own troops in 235 and ending with the accession of Diocletian in 284.

The arithmetic of instability

Example. Between 235 and 284 the empire had roughly twenty-six emperors whom the Senate recognised, besides a much larger number of men proclaimed by armies and killed before they were acknowledged. What does that produce, and what does it do to the man in the job?

Forty-nine years divided by twenty-six is an average reign of one year and ten months. Compare the preceding period: from Augustus in 27 BC to Severus Alexander's death, roughly 260 years produced about twenty-five emperors, an average of ten years each, and that period included the civil war of 69. So the rate of turnover multiplied by more than five.

The effect on decision-making is the point. An emperor whose expected tenure is under two years, and who knows that essentially every recent holder of the office was murdered, cannot plan. He cannot let a rival general win a victory, cannot leave a frontier army unpaid, cannot afford an unpopular tax, and cannot be absent from any theatre for long, because absence is where usurpation starts. His rational strategy is to pay the troops immediately, from whatever source is quickest, and to march to whichever frontier is loudest. That describes the actual behaviour of most third-century emperors, and it explains the debasement later in this lesson better than any theory of economic ignorance. The behaviour is not stupidity; it is what a two-year horizon does to anyone.

Now you. Nearly all of these men were raised by frontier armies. From the army's point of view, was proclaiming its own commander irrational?

Two enemies, and a new kind of one

The external situation had genuinely changed, and one change was qualitative. In 224 the Parthian dynasty that had faced Rome for nearly three centuries was overthrown by Ardashir, founder of the Sasanian house, and the new Persian state was more centralised, more ideological and far more aggressive. Parthia had been a loose confederation that Rome could usually buy or divide; Sasanian Persia campaigned with intent, took cities, and deported their populations to found new ones.

On the northern frontier the change was in scale and organisation. The small groups Tacitus had described were coalescing into larger confederations with names Rome had not used before, the Goths on the lower Danube, the Alamanni and the Franks on the Rhine, better armed and capable of holding together long enough to break through and stay. In 251 the emperor Decius was killed with his army at Abritus by Goths under Cniva, the first Roman emperor to die in battle against a foreign enemy.

Behind both sat something the sources describe and cannot quantify. A pandemic beginning in the late 240s, called after the bishop who wrote about it, ran through the empire for over a decade, following the plague of the 160s within living memory of its effects. Mortality figures for either are guesses, and the honest statement is that two major epidemics in a century almost certainly reduced the population and the tax base, and that nobody can say by how much.

260

If one year has to carry the crisis, it is 260. The emperor Valerian, campaigning against Shapur I near Edessa, was taken alive. No Roman emperor had ever been captured.

Example. For this event we have Roman accounts and a Persian one: Shapur's own trilingual inscription carved on the wall of a building at Naqsh-e Rustam, listing his victories, claiming to have taken Valerian with his own hands along with a great army, naming the Roman commanders captured, and listing the cities he took. How should the two be used?

By reading each for what its author had no reason to lie about, and then testing the overlap. The Roman sources are compressed, embarrassed and vague about the circumstances, and later Christian writers turn Valerian's fate into a moral about a persecutor's end, which is a strong reason to distrust their details. Shapur's inscription is triumphal propaganda, so its numbers and its claim of personal capture are exactly what a sceptic should discount.

But the two agree on the facts each side would have preferred otherwise: the Romans concede the capture, which is the last thing they would invent, and Shapur's list of cities can be checked against archaeology and in several cases is confirmed by destruction layers and by the appearance of Roman-style deportee settlements deep inside Persia. Where a hostile source and an interested source agree against their own interests, the agreement is strong. What remains uncertain is everything else: whether Valerian was captured in battle or seized during negotiations, how large the army was, and how he died. That is the normal shape of a good result in this subject, a secure core surrounded by an unresolved margin, and it is worth more than a confident narrative.

Now you. Within months, the Rhine and Gallic provinces recognised their own emperor, Postumus, and the east came under the control of Palmyra. Neither declared independence from Rome. What were they doing instead?

The coinage eats itself

The empire's silver coin had been diluted before. What happened now was different in kind.

Caracalla had introduced a new coin, the antoninianus, tariffed as two denarii but containing about one and a half denarii of silver, which is a devaluation dressed as a new denomination. Its silver content then fell steadily: about half silver under Caracalla, around 40 per cent by the 240s, and by the late 260s a copper coin with a silver wash on the surface, containing perhaps two per cent silver, which wore off in use.

Example. Set the late antoninianus against the Augustan denarius and work out what the state gained.

The Augustan denarius held about 3.8 grams of fine silver, so a Roman pound of 327 grams yielded about 86 of them. A late antoninianus of roughly 3.5 grams at two per cent holds about 0.07 grams of silver, so the same pound yields something like 4,700 coins. The state could produce about fifty-five times as many coins from the same bullion as it could two and a half centuries earlier.

That is the whole attraction and the whole disaster in one figure. The gain is immediate: this year's donative is paid. The cost arrives when the coins accumulate, and prices in Egypt, which had held around seven or eight drachmas for a measure of wheat for two centuries, are running many times higher by the end of the century. The lag is what makes the policy so dangerous, because each emperor experiences the benefit within his two-year horizon and the cost falls on somebody else's reign.

Now you. Why could an emperor who understood the problem not simply stop?

What it looked like on the ground

The literary sources for this half-century are unusually bad, which makes the material evidence carry more weight than usual.

Coin hoards, meaning caches buried and never retrieved, spike sharply in the 250s to 270s across the western provinces. A hoard is deposited when its owner is frightened and stays in the ground when the owner does not come back, so the distribution maps insecurity, with the useful caveat that debased coin was also hoarded because nobody wanted it and good coin because everyone did.

Cities built walls. The most eloquent case is Rome itself, which had been undefended for four centuries because the idea of an enemy reaching it was absurd, and which Aurelian enclosed from 271 with a circuit of about 19 kilometres and some 380 towers, a tower every 50 metres. Across Gaul, towns built small, thick circuits that enclosed a fraction of their built-up area, using tombstones and broken monuments as material. That last detail is worth noticing: a town in a hurry, dismantling its own cemetery for stone, tells you more about the mood of the 270s than any surviving history of the period.

Why it did not end

It should have ended, and the reasons it did not are institutional rather than heroic.

The men who saved it came from one place. Claudius, Aurelian, Probus, Carus and Diocletian were all professional soldiers from the Danubian provinces, risen through the ranks in an officer corps that had recently been opened to them, because Gallienus had excluded senators from legionary commands around 260. That decision, made to reduce the pool of plausible usurpers, had the effect of replacing an aristocratic officer class with a career one, and the career one turned out to be extremely good at the job.

The parts also stayed compatible. Because the breakaway states had reproduced Roman institutions rather than replacing them, Aurelian could recover Palmyra in 272 and Gaul in 274 and simply resume administering them, which is why he could be called restorer of the world on his coins without much exaggeration. He was murdered the following year by his own officers over a misunderstanding, which is the period in miniature.

And the empire shed what it could not afford. Dacia, the province whose gold had paid for Trajan's forum, was evacuated in the 270s and its population moved south of the Danube, shortening the frontier. The Antonine settlement was not restored anywhere.

What the crisis proved is that the light-touch empire of the first two centuries could not survive two simultaneous frontiers, and that the coinage could not fund an army large enough for them. Both conclusions point the same way: the state had to become bigger, more expensive and far more intrusive. The next lesson is about the man who did that deliberately, comprehensively, and with a plan for the succession that failed within eighteen months of his retirement.

The empire that came out of the third century was rebuilt from the frame outwards by a soldier from Dalmatia who held power for twenty-one years and then retired, and almost every feature of the later Roman state is his.

The previous lesson left three conclusions from the crisis. One frontier could be held by the old army and two could not. A coinage could not be debased indefinitely without ceasing to be money. And an emperor with a two-year horizon cannot govern. Diocletian, proclaimed in 284, addressed all three, and the solutions cost more than the problems had.

Four rulers

The first move answered the political problem directly. In 285 Diocletian made a fellow officer, Maximian, his colleague, and in 293 he added two junior emperors, Constantius and Galerius, each with a territory, a court, an army and a capital: Trier, Milan, Sirmium, Nicomedia. Rome was the capital of none of them.

The logic was the one the previous lesson identified. If a frontier army's incentive to proclaim its own commander comes from the emperor's absence, then have an emperor present on every frontier. If a capable general is a threat, make him a colleague. Diocletian and Maximian took the divine surnames Jovius and Herculius, presenting themselves not as two men who had agreed to share but as a single sacred authority in two persons, with the Caesars beneath them.

Whether this counts as dividing the empire is worth settling now, because it recurs to the end of the course. It did not. There was one empire, one body of law, one citizenship and one coinage, and legislation issued by any emperor ran in all the territories. What was divided was the work. That distinction holds through 395 and beyond, and much of the confusion about the fall of the west comes from people at the time and afterwards treating a division of labour as a division of states.

More provinces, more officials

The administrative reform was equally direct. Provinces were subdivided from about fifty to close to a hundred, grouped into twelve dioceses under officials called vicars, and above them regional prefectures. A governor now ran a much smaller territory, and, crucially, ran no troops: civil and military commands were separated, so that a man with soldiers had no tax revenue and a man with tax revenue had no soldiers.

The cost was a bureaucracy of an order the earlier empire would not have recognised. An earlier lesson counted around 350 senior officials for sixty million people; the fourth-century establishment ran into tens of thousands of salaried posts, and the change is visible in the sources as a new kind of complaint, about fees, delays, and men who must be paid to move a document. Contemporaries thought it enormous. Measured against the Chinese comparison from that earlier lesson, it was still small.

An army twice the size, and how to feed it

Example. John the Lydian, using official material, gives Diocletian's army as 435,266 soldiers with 45,562 in the fleets. Set that against the quarter of a million of the early empire and against a tax base that two plagues and fifty years of disorder had probably shrunk.

The total is 480,828, or roughly twice the force of the first century. An earlier lesson found the old army absorbing over half of an imperial revenue near 800 million sesterces, when the economy was at its largest and the coinage was sound. Now the army is doubled, the population is smaller, and the silver coin has become a copper token. Lactantius, a hostile witness, says the number of men receiving was becoming greater than the number paying, which is rhetoric, but the direction is right and no arrangement built on cash wages could survive the arithmetic.

So the state stopped relying on cash. The annona militaris, formerly an emergency requisition, became the ordinary way of supplying the army: grain, meat, oil, wine, uniforms, animals and labour, assessed on the provinces and delivered in kind. Soldiers were fed and clothed from what was collected rather than paid to buy it. Once you see that the coinage had failed as a store of value, this is not a step backwards into barbarism but the only workable response: a state whose money is unreliable moves to goods, and it has to build the machinery to assess and move them, which is where the new bureaucracy comes from.

Now you. Why not simply raise cash taxes and pay the army properly?

Counting everybody

Requisition in kind needs a much more detailed register than a cash levy, because the state must know what each district can actually produce. The answer was a new assessment built on two units: the iugum, a notional unit of land adjusted for quality and crop, so that a given area of vineyard, olive grove and arable counted differently, and the caput, a unit of labour. Liability was set per unit, and the assessment was revised on a fixed cycle, which became the fifteen-year indiction and remained a way of dating documents for centuries.

Example. In 297 the prefect of Egypt, Aristius Optatus, published an edict announcing the new census, and it survives on papyrus. It says that the emperors, having learned that the public levies were being made capriciously, so that some were let off lightly and others overburdened, have decided to fix a rate for each aroura of land according to its quality and for each head of the rural population, and it instructs everyone to pay promptly and warns officials against exaction. Read it as evidence.

Three things are visible at once. First, the reform is presented as a remedy for unfairness, and the complaint it names is real: the old system delegated assessment to local men who protected their own. Second, the units are land measured by quality and people counted by head, exactly the iugum and caput pairing, so a document from an Egyptian village confirms the structure that late Roman law codes describe from the top down. Third, and most revealing, the prefect finds it necessary to tell taxpayers to pay before they are compelled and to warn collectors off extortion, which tells you what everyone expected to happen.

The caution is that Egypt is the province whose documents survive and the province with the longest tradition of detailed registration, so it is both the best evidence and the least typical. What this edict establishes securely is that the reform was real, dated and implemented on the ground, which is more than a law code can prove on its own. What it cannot establish is that Gaul or Africa was assessed the same way or as thoroughly.

Now you. The new system taxes a piece of land together with the people who work it. What follows for those people?

The price edict, and why it failed

In 301 Diocletian issued maximum prices for over a thousand goods and services throughout the empire, with death as the penalty for exceeding them. The preamble blames profiteers, and it is worth noticing what it does not blame.

The text survives in dozens of fragmentary copies, all from the Greek-speaking provinces, and it is the best price list from the ancient world. A farm labourer may be paid 25 denarii a day plus food; a modius of wheat is capped at 100 denarii; a pound of purple-dyed silk, the most expensive item listed, at 150,000.

Example. Explain the failure, using the wheat price.

At the capped prices, four days of a labourer's wage buys a modius of wheat, and a pound of purple silk costs six thousand days of it. The ratios are informative and the failure is not about ratios. A maximum price binds only where it sits below what a seller would otherwise accept, and if it does, the seller's options are to sell at a loss, sell elsewhere, sell secretly, or not sell. Wheat does not appear on the market at 100 denarii because a farmer told to accept less than the grain is worth stops bringing it in. Lactantius says exactly this: much blood was shed over trifles, goods disappeared from sale, and the law was eventually allowed to lapse.

The deeper point is in the preamble. Diocletian attributes rising prices to greed, and there is no reason to think he was being cynical, because nobody in antiquity had a theory connecting the quantity of coin to the level of prices. He had spent his reign issuing increasingly nominal money and was seeing the consequence, and he identified the visible party, the merchant raising his price, rather than the invisible one, the mint. Blaming the intermediary for a monetary phenomenon is a mistake with a very long history, and it is not fair to call it a Roman one.

Now you. Diocletian also launched the most systematic persecution of Christians the empire ever attempted, from 303. Given everything else in this lesson, why then?

The one part designed to work

Diocletian's most original act was to retire. In May 305 he abdicated, compelled Maximian to do the same, and the two Caesars became Augusti with two new Caesars beneath them. The point was to convert the succession from a lethal accident into a procedure: merit, seniority, promotion and a living predecessor to supervise the handover. Nothing like it had been attempted since Augustus, and it addressed the deepest problem this course has, which is that the position was never an office.

It lasted a little over a year. Constantius died at York on 25 July 306, fourteen and a half months after the abdication, and his troops immediately proclaimed his son Constantine, who had no place in the scheme. In October, Maximian's son Maxentius was proclaimed at Rome, seventeen months and three weeks after the abdication. By 310 there were four or five men claiming imperial rank and the wars between them ran for eighteen years.

The reason for the failure is not obscure, and it is the same one the whole course keeps returning to. The tetrarchy allocated the succession by merit as judged by the senior emperor, and the armies allocated it by inheritance and presence. When Constantius died in front of his army with an adult son beside him, the army applied its own rule, and there was no institution, no law and no ceremony capable of saying that the army was wrong. Diocletian, living in his fortified retirement palace on the Dalmatian coast, was asked to come back and refused, reportedly telling Maximian that if he could see the cabbages he had planted with his own hands he would not talk to him about power.

What he had built survived the collapse of the way he had arranged to pass it on. The provinces, the dioceses, the tax in kind, the separated commands and the enlarged army were all still standing a century later. The next lesson follows the man who won the wars, kept every one of those structures, and added the one thing Diocletian would have found unimaginable: an emperor who backed the church he had tried to destroy.

Twenty-two years after the empire set out to destroy the Christian church, a Roman emperor summoned its bishops to a conference at his own expense and paid for their travel on the imperial post.

The previous lesson ended with Diocletian's succession scheme collapsing within eighteen months, and with the army at York proclaiming a dead emperor's son on 25 July 306. This lesson follows that son. Constantine's reign is where the reconstructed state of Diocletian acquires the two features that define the rest of Roman history: a capital in the east and a church inside the government.

From York to the Tiber

Constantine spent six years consolidating the north-western provinces while the tetrarchy tore itself apart, and in 312 he invaded Italy against Maxentius, who held Rome and Africa. The armies met just north of the city on 28 October. Maxentius, who had a defensible city and chose to fight in front of it, was driven back onto a bridge of boats beside the broken Milvian Bridge and drowned in the Tiber. His head was carried through Rome the next day.

The battle mattered because it gave Constantine Rome, Italy and Africa, and it is remembered because of what he is said to have seen before it.

Three accounts of one moment

Example. We have three near-contemporary treatments of the Milvian Bridge. Lactantius, writing about three years afterwards, says Constantine was warned in a dream to mark the heavenly sign of God on his soldiers' shields, and describes the mark as a letter X with its top bent over. Eusebius, writing after Constantine's death in 337 and saying the emperor swore to it personally, describes a cross of light in the sky above the sun, with words telling him to conquer by this, seen by the whole army, followed by a dream in which Christ appeared. The Arch of Constantine, dedicated in Rome in 315, says the victory was won by the prompting of the divinity and by greatness of mind, and carries no Christian symbol anywhere on it. Set them side by side.

The pattern is the clearest example of source growth in this course. The earliest account is the most modest: a private dream and a mark on shields. The latest is the most spectacular: a public daytime miracle witnessed by an army, told at second hand a quarter of a century later by a bishop writing a panegyric of a dead emperor he admired. Between them, and closest of all to the event, stands a monument voted by the Senate three years afterwards which describes the divine help in terms so carefully unspecific that a pagan senator and a Christian courtier could each read it as their own.

The right conclusion is not that Eusebius lied. It is that a story of this kind acquires detail, witnesses and precision as it is retold, and that the historian's rule is to prefer the earliest version and to treat every later improvement as improvement. Notice also what all three agree on, which is that Constantine attributed the victory to a god and said so publicly and immediately. That much is secure, and it is the part that had consequences.

Now you. Does it matter whether Constantine was sincere?

What was actually decreed, and when

Two corrections belong here, because the popular account of Constantine legalising Christianity is wrong in both date and instrument.

Christianity had already been legalised. In April 311, dying and in the middle of the persecution he had pressed hardest for, Galerius issued an edict permitting Christians to exist again and to rebuild their meeting places, on condition that they pray for the emperor's safety. That is the legal turning point, and it was made by a persecutor, on the grounds that the policy had not worked.

What happened in 313 was not an edict and was not issued at Milan. Constantine and Licinius met at Milan and agreed a policy; what survives is a letter Licinius sent to provincial governors afterwards, preserved in Latin by Lactantius and in Greek by Eusebius. It goes further than Galerius by granting freedom of worship to everyone and ordering the restitution of confiscated church property at state expense, which is the operative and expensive part. Calling it the Edict of Milan is a convenience that has misled generations about what kind of act it was.

The coin that lasted seven hundred years

At Trier around 310, Constantine began striking a new gold piece at 72 to the Roman pound.

Example. Work out its weight, compare it to Augustus's aureus, and explain why gold could be stabilised when silver could not.

A Roman pound of 327 grams divided by 72 gives 4.54 grams. Augustus had struck 42 to the pound, or 7.79 grams, so the new coin is 42 per cent lighter. That is not a debasement in the third-century sense, because the metal is pure and the weight is stated by the standard: everyone knows a solidus is one seventy-second of a pound, and the coin can be checked on a balance. What had destroyed the silver coinage was not lightness but the pretence that a copper-washed token was still a denarius.

Gold could be held to a standard for two reasons the previous lessons supply. The state controlled the mines and had just confiscated the accumulated treasures of temples across the empire, so it had bullion. And the gold coinage did not have to bear the mass payments: an army fed and clothed by requisition in kind does not need to be paid in millions of pieces of good money. Gold could therefore be reserved for taxes the state wanted in gold, for donatives, for salaries of officials and for large transactions, where a reliable coin is worth most. The solidus held its weight and fineness essentially unchanged for over seven hundred years, which makes it the most durable currency standard in recorded history.

Now you. Who gained from a stable gold coin, given that the everyday currency remained a copper token that continued to lose value?

A new Rome

In 324 Constantine defeated Licinius and became sole ruler, and in 330 he dedicated a new capital on the site of the Greek town of Byzantium, on the strait between Europe and Asia.

The choice was strategic before it was anything else, and the previous lessons explain it. The frontiers that mattered were the lower Danube and the Persian border, and the site sits between them, on the crossing every land route uses, with a deep sheltered harbour and water on three sides. It could be supplied by sea from Egypt, and it was defensible in a way Rome was not.

It was also a Rome. It had a senate, a corn dole, hills counted to reach seven, and the emperor's residence, and Constantine stripped statues from the cities of the east to furnish it. What it did not have was the old aristocracy, the old priesthoods or the old memories, which was much of the attraction. Rome itself remained ceremonially supreme and politically peripheral, a place emperors visited, and by the end of the century some did not visit at all.

The church as a public institution

The most consequential decisions were administrative, and they show best in what they broke.

Example. In 313 Constantine exempted Christian clergy from the compulsory public duties owed by men of property, principally membership of city councils. Within a few years he was legislating to stop wealthy men entering the clergy. Read the sequence.

The exemption was meant as a favour to a small body of men, and it collided with the situation the previous lesson described: council membership had become a hereditary burden that the propertied were desperate to escape, and the state was already legislating to stop them escaping by other routes. Ordination now offered a legal exit, so people took it, and Constantine had to rule that decurions and those eligible to be decurions could not be ordained.

Two things follow. The first is that a privilege granted to a religious body immediately became a fiscal problem, which is what happens when a church acquires legal standing inside a state that assigns duties by status. The second is more general and worth carrying: legislation is evidence of behaviour, and repeated legislation is evidence of failure. The fourth-century codes return to this question again and again, which means men kept doing it. If you want to know what people were actually doing in the later empire, read what the emperors kept forbidding.

Now you. In 325 Constantine summoned bishops from across the empire to Nicaea, paid their travelling costs, attended in person and pressed them towards agreement. Why would an emperor involve himself in a dispute about the relationship between God the Father and God the Son?

What he did not change

It is worth ending with the continuities, because the temptation is to treat 312 as the hinge of everything.

The state Constantine ran was Diocletian's, essentially unaltered: the provinces and dioceses, the prefectures, the separation of civil from military command, the tax in kind, the enlarged army, the court ceremonial and the remote sacred emperor. He kept the persecutor's machinery and pointed it in a different direction, which is one reason the change was possible at all.

Nor did he make the empire Christian in any straightforward sense. Traditional cults continued and were funded; the Senate at Rome remained largely pagan for two more generations; and the growth curve of the previous lesson, running at around a tenth of the population in 300, does not reach a majority by his death in 337 on any reasonable assumption. What he did was make Christianity the religion of the state's rewards, which changed the incentives of everybody with a career.

He also left the question of doctrine unsettled. Nicaea produced a formula, and it did not produce agreement, and the empire spent the next fifty years arguing about a single word with emperors on both sides. That argument, and what it did to the relationship between bishops and the state, is the subject of the next lesson.

An empire that had spent three centuries not caring what anyone believed spent the fourth century legislating about a single Greek word.

The previous lesson left Constantine with a church inside his government and a doctrinal quarrel he had convened a council to settle. This lesson follows what the settlement did not settle, and traces the consequence that outlasts everything else in this course: an institution with its own hierarchy, its own law, its own revenues and its own claim to authority, operating inside the Roman state and eventually outliving it in the west.

One word

The dispute Constantine walked into had been started by an Alexandrian priest, Arius, who argued that if the Son was begotten by the Father, then there was a time when he did not exist, so he is the first and highest of created things rather than God in the same sense as the Father. This was not a quibble. It concerned whether Christians worshipped one God or something closer to two, and whether the figure at the centre of their cult was divine or exalted.

The council at Nicaea in 325 answered with a creed containing the word homoousios, meaning of the same substance as the Father. It was a decisive word and an unfortunate one, because it appeared in no scripture, had been used by heretics, and meant different things to Greek and Latin speakers. In the following decades a rival formula circulated, homoiousios, of like substance, differing by a single letter, and the contrast became the standard illustration of theological hair-splitting.

Two honest qualifications. The one-letter contrast belongs to the 350s rather than to Nicaea, so quoting it as the substance of the original dispute compresses thirty years. And the number of bishops present, traditionally 318, is a figure borrowed from the book of Genesis; contemporary estimates run nearer 250 to 300, and fewer than ten came from the Latin west, which matters for how the western church later regarded a decision taken almost entirely in Greek.

Fifty years of imperial theology

Nicaea did not end the argument; it made the emperor a party to it. Constantine himself recalled Arius from exile within a few years, and was baptised on his deathbed by a bishop from the other side. His son Constantius II, sole ruler from 353, backed formulas that avoided homoousios and used councils and exile to enforce them, and by 359 a pair of imperial councils had produced a compromise creed that the Nicene party regarded as a defeat. Jerome later wrote that the whole world groaned and was astonished to find itself Arian.

Example. Athanasius was bishop of Alexandria for 45 years from 328 and spent about 17 of them in exile, expelled five separate times by four different emperors. What does that career show?

That doctrine had become a function of who held power, and that it could not be settled by holding power. Thirty-eight per cent of an episcopate spent in exile is a measure of how completely theological allegiance now determined civil status, since these were not doctrinal punishments imposed by a church but banishments ordered by emperors. Each expulsion was an imperial policy and each return was another.

The second half is the important half. Five expulsions and five returns mean the machinery could remove a bishop and could not replace him, because Athanasius's authority rested on his standing in Alexandria and in the wider church rather than on his tenure of a post. An emperor could take the building; he could not take the following. That is the structural novelty of the fourth century, and it is why the relationship between emperors and bishops develops the way it does. The state was dealing for the first time with an organisation whose leaders it had not appointed and could not simply dismiss.

Now you. How much should Jerome's line about the world waking up Arian be allowed to prove?

Julian, the control experiment

In 361 the throne passed to Constantine's nephew Julian, who had been raised Christian, had turned privately to the old gods and philosophy, and now announced it. He is the closest thing this course has to a controlled test of how deep the change had gone, and his reign as sole emperor lasted about twenty months.

Example. Julian restored funding to temples, recalled exiled bishops of every faction in order to let Christians quarrel, and forbade Christians to teach classical literature on the grounds that a man who thinks Homer's gods false should not be paid to expound Homer. He died in Persia in June 363. What did the twenty months establish?

Chiefly that the old religion could not be restored by an emperor's decision, which is the mirror image of the finding about Christianity from the persecutions. Julian's letters record his frustration at what he found: temples without worshippers, priests who had to be instructed in how to behave, and, revealingly, an attempt to organise pagan charity because he could see that the Christians' care of the poor was winning converts and that traditional cult had no equivalent. He was trying to build an institution, and the point about traditional Roman religion, established several lessons ago, is that it had never been one.

The teaching ban is the sharpest of his measures and the most instructive. It shows that Julian understood the real mechanism: not belief but access, since classical education was the gateway to every career in the empire, and excluding Christians from teaching it would in time exclude them from the elite. Even his Christian opponents thought it his cleverest and cruellest act. It was revoked immediately after his death, which is the other finding: a policy that depends entirely on one man lasts as long as he does. On both sides, the fourth century keeps demonstrating that religious change ran deeper than any emperor's power over it.

Now you. In 384 the pagan senator Symmachus petitioned for the restoration of the Altar of Victory to the Senate house, arguing that Rome's greatness had been built under these rites, that an old religion deserves respect, and that so great a mystery cannot be reached by one road alone. Ambrose of Milan answered him and won. Why did that argument fail?

The bishop as a civic power

While the theology was being fought over, the office of bishop was quietly turning into the most important local institution in the empire.

An earlier lesson showed government delegated to city councils and the notables who staffed them, and the lesson after Diocletian showed those men fleeing an office that had become a hereditary liability. Something had to fill the space, and the bishop did. He controlled substantial revenues, in property given by emperors and by the rich. He ran the distribution of food and money to the poor, which no pagan cult had ever done and which made him a patron on the old Roman model with a far broader clientele. He could hear civil cases between consenting parties, with the state enforcing his judgement, so he was a judge. He interceded with governors for prisoners and taxpayers, ransomed captives, and spoke for his city.

By 400 the bishop of a provincial town was the man a community sent to negotiate, the man who could raise money in an emergency, and the man with a permanent building, a staff and a written record. He was, in the functional sense, what the decurion had been, with the difference that his position was for life and his authority came from outside the imperial hierarchy.

Ambrose and Theodosius

Example. In 390 a riot at Thessalonica killed a Roman commander, and Theodosius ordered a reprisal in which a large crowd was massacred in the circus, thousands according to the sources. Ambrose, bishop of Milan, withdrew from the emperor's presence, wrote to him privately that he could not offer the sacrifice while the emperor came to the church with the blood of an innocent people on his hands, and Theodosius performed public penance before being readmitted to communion. Why is this new?

Because no authority inside the Roman world had ever been able to impose a penalty on an emperor. The Senate could flatter or condemn after death; armies could kill; but no institution could require a reigning emperor to submit to a discipline and have him comply. Ambrose had no troops, no office and no legal power over Theodosius. What he had was control of access to a rite the emperor believed he needed, and the standing to make the exclusion public.

Read carefully, though. Ambrose's letter is respectful, offers a precedent for royal repentance, and gives the emperor a route back, and Theodosius's submission was politically manageable and arguably even useful, since it presented him as pious rather than as a butcher. Both men were skilful and neither was testing a theory of church and state. But the precedent was set and remembered for a thousand years, and the reason it could be set is structural rather than personal: once the emperor's own legitimacy is bound up with a church, the men who run that church have leverage over him that no subject previously had. The empire had acquired a second authority inside itself, and had done so as a side effect of adopting a religion for reasons of state.

Now you. Theodosius died in January 395, leaving Arcadius, aged about eighteen, in the east and Honorius, aged ten, in the west. Historians often call this the division of the empire. Should they?

What a Christian empire had become

By the end of the century the legal position had reversed completely. Theodosius's law of 380 ordered all peoples to hold the faith professed by the bishops of Rome and Alexandria and named everyone else heretics; the council of 381 confirmed the Nicene position; and laws of 391 and 392 prohibited sacrifice, closed temples and forbade even private household cult. Sixty-eight years after Constantine's victory, the religion that had been a capital offence was compulsory and the traditional rites were criminal.

The enforcement was patchier than the laws suggest, and the reversal was not as total as a reading of the code implies. Temples went on functioning in places, rural cult persisted for centuries, the Senate at Rome retained pagan members, and the state had neither the officials nor the appetite to police belief in villages. What the laws did achieve was to remove the traditional cults from public life and from public money, which for a religion that had always been civic was fatal in a way that persecution of individuals would not have been. The old gods did not lose an argument; they lost their funding and their place in the calendar.

The empire that emerges is recognisably Diocletian's state with a church fused into it: the same provinces, the same tax, the same army, plus a parallel hierarchy of bishops with revenues, courts and moral authority over the emperor himself. That is the institution that will still be standing in the west when the imperial government is not. Before that happens, the peoples on the other side of the Danube have to arrive, and the next lesson asks who they actually were and how many of them there were, because the answer is much smaller than the outcome suggests.

The peoples who took the western empire apart were a fraction of one per cent of its population, and any explanation of what happened has to start by facing how few of them there were.

The previous lesson left the empire administratively split in 395 with two incapable emperors, a church fused into the state, and the same frontiers Diocletian had reorganised. This lesson looks across those frontiers. It asks who the Goths and Vandals actually were, how many of them there were, and why a word like invasion fits the events badly enough to mislead.

Who was over there

The first difficulty is that we have no account written by any of them. Everything comes from Roman authors, and the best of them, Ammianus Marcellinus, is a Greek-speaking soldier from Antioch writing in Latin about people he mostly met in war.

The names in those sources are unstable. Tacitus in the first century lists dozens of small German peoples; by the third century the sources name a few large ones, the Goths, the Alamanni, the Franks; by the fifth, Visigoths, Ostrogoths, Vandals and Burgundians. The old view read this as migration, whole nations moving with their languages and customs intact from somewhere else. The evidence does not support it. Archaeology on the lower Danube and in Ukraine shows a mixed material culture with strong Roman elements and no sign of a population arriving en masse, and the written names correspond poorly to anything a spade can find.

The better model is that these were political groups rather than nations: coalitions formed around a successful war leader and a core of families carrying a tradition, which could absorb outsiders quickly and dissolve quickly. Some historians go further and argue the groups were essentially manufactured by contact with Rome, since a confederation is what you need in order to negotiate with, raid or serve an empire. The argument is unresolved and the honest position is that both processes were probably at work. What is not in doubt is that a Gothic army of 400 contained men whose grandfathers had not been Goths.

What the frontier did to them

Three centuries of living beside the empire changed them profoundly, and the changes ran the way the empire's own institutions pushed.

They traded. Roman goods, coins and prestige objects appear far beyond the frontier, and the Roman customs posts of an earlier lesson existed because there was traffic to tax. They were paid. Subsidies to friendly chiefs were standard Roman practice and a cheaper alternative to campaigning, and a chief who could distribute Roman gold to his followers had an advantage over one who could not, which is itself a force pushing towards larger and more stable groupings. And they served: whole units were recruited from beyond the frontier, and by the fourth century the Roman army in the west was full of them, with men of Frankish, Vandal and Gothic descent commanding it.

Example. Around the 340s a bishop named Wulfila, of Cappadocian descent from a family taken captive in a Gothic raid, was consecrated to work among the Goths, devised an alphabet for their language and translated most of the Bible into it. He worked while the imperial court held to non-Nicene formulas, and the Gothic church that resulted was Homoian, which later Roman writers called Arian. Trace the consequence.

The immediate consequence is that the Goths became Christians on the empire's terms, using an alphabet built from Greek and Latin letters, with a Bible whose vocabulary had to borrow Greek theological words. That is Romanisation, arriving through a missionary rather than through conquest. Wulfila is reported to have left the books of Kings out of his translation on the grounds that the Goths were warlike enough already, which if true tells you he was making choices about what kind of people he was forming.

The delayed consequence is the important one. The creed the Goths adopted was the imperial orthodoxy of the 340s and 350s, and the previous lesson showed that orthodoxy reversing under Theodosius. So the Goths were left holding a version of Christianity the empire had come to define as heresy, purely because of when they converted. For the next two centuries, Gothic and Vandal rulers in the western provinces were separated from their Roman subjects by a doctrinal line neither had chosen, which impeded intermarriage, gave Roman bishops a reason to look elsewhere for a protector, and turned religious difference into the marker of who was and was not a Goth. A quarrel over one Greek word in the 320s shaped the politics of Spain and Africa a century and a half later.

Now you. Ammianus says an attempt was made to count the Goths crossing the Danube in 376 and was given up, and he quotes Virgil on numbering the sands of the Libyan desert. How should the numbers in these sources be handled?

376, and the crossing

In 375 or 376 a Gothic group under pressure from the Huns arrived on the north bank of the lower Danube and asked to be admitted to the empire. The emperor Valens agreed. This was neither unprecedented nor foolish: settling defeated or displaced groups inside the frontier was old Roman practice, it produced taxpayers and recruits, and Valens was preparing a Persian campaign and wanted the manpower.

The administration of it was catastrophic. Ammianus, no friend of the Goths, describes Roman officers exploiting the refugees systematically: food withheld and then sold at extortionate rates, so that families traded their children into slavery for dog meat. Disarmament, which was the condition of admission, was not enforced because the officers were busy profiteering. Within two years the settlement had become a war.

Adrianople

Example. On 9 August 378, Valens attacked the Gothic force near Adrianople without waiting for the western army marching to join him. He was killed with, in Ammianus's account, two thirds of his army and thirty-five senior officers, and Ammianus compares the defeat to Cannae. Was it decisive?

Take the arithmetic first. A field army of the size usually reconstructed for this campaign, perhaps 20,000 men, losing two thirds, means something over 13,000 dead. Against a total imperial establishment of the order of 400,000 that is a bad afternoon, not a mortal wound, and the empire had absorbed worse: the comparison Ammianus reaches for, Cannae in 216 BC, was followed by Roman victory, as the earlier course on the Republic shows in detail. Rome went on for another 98 years in the west after Adrianople, and in the east for far longer.

What was lost that mattered was not the men but the settlement. Valens was dead, the eastern field army was wrecked at the moment it was needed against Persia, and the Goths were inside the empire, undefeated and unresettled, in provinces that could not feed them. Theodosius, appointed to deal with the problem, spent four years trying and failing to destroy them militarily, which is the real evidence of what the defeat cost. Adrianople was decisive in the sense that it removed the option of solving the problem by force, and that forced the treaty described next. Battles are rarely decisive because of casualties; they are decisive when they eliminate a choice.

Now you. In 382 Theodosius made a treaty admitting the Goths to Thrace as a body, under their own leaders, not distributed among landowners, and obliged to serve as allied contingents under their own commanders rather than being enrolled in Roman units. What was new about it?

How many of them there were

The best figure we have for any of these groups comes from the Vandals. When Geiseric took his people from Spain to Africa in 429 he had them counted, and the number reported is 80,000, including women, children and slaves.

Example. Test that against the empire.

Against a population of perhaps 50 million by this date, 80,000 is 0.16 per cent, one person in six hundred. If one in five was a fighting man, which is generous for a group that includes families and slaves, the army is around 16,000. The late Roman establishment on paper ran to several hundred thousand men, so this force is perhaps four or five per cent of it.

With those 16,000, Geiseric took the richest provinces of the western empire, held them for a century, built a fleet, sacked Rome in 455 and destroyed the largest expedition the two empires ever mounted against him. The disproportion is the central fact of the fifth century and it is what makes the usual explanation useless: an empire of 50 million was not overwhelmed by 80,000 people. It has been argued that Geiseric inflated the figure to impress, which if true makes the point sharper still. Something other than numbers is doing the work, and the next lesson is about what.

Now you. Given all this, what is wrong with the word invasion?

What this leaves to explain

Two facts now sit uncomfortably together. The peoples involved were few, familiar, partly Romanised and mostly admitted. And within seventy years of Adrianople the western provinces were governed by their kings.

Nothing in this lesson explains that, and it is important to see that nothing does. There was no barbarian population surge, no new military technology, no religious motive, and no evidence of any intention to destroy the empire; Gothic and Vandal kings issued Roman-style law, used Roman officials and dated by Roman consuls. The east faced the same peoples on the same river, and in the case of the Huns faced worse, and survived.

So the question the next lesson has to answer is not why the barbarians came but why the west could not absorb them when it had absorbed everything else for four hundred years, and why the east could. The answer turns out to be about revenue.

Two halves of one empire, with the same institutions, the same law, the same enemies and the same religion, went in opposite directions within a single lifetime, and the difference is mostly a matter of which provinces each of them kept.

The previous lesson established that the peoples involved were few, familiar and mostly admitted rather than invading, and that nothing about them explains the outcome. This lesson supplies the mechanism. It is fiscal, it is visible in the laws the western government issued about itself, and it runs as a loop rather than a sequence.

The loop

State the empire's structure as three dependencies, all of them established in earlier lessons. The army exists to hold territory. The army is paid, fed and clothed out of taxes. The taxes come from territory.

That is a loop, and while it turns in the right direction it is extremely stable: a province held yields tax, which pays soldiers, who hold the province. Turn it the other way and it is just as stable. A province lost stops yielding tax, so fewer soldiers can be paid, so the next province is harder to hold. Nothing external is needed to keep it running once it starts, and the groups on the losing side of it do not have to be numerous or formidable. They only have to be there.

The western empire entered that loop in the first decades of the fifth century. Britain was let go around 410, with the emperor reportedly telling its cities to see to their own defence. Much of Spain and large parts of Gaul passed out of effective control in the following decades. And in 439 the Vandals took Carthage.

What Africa was worth

Africa was the richest tax territory the western government still controlled, exporting grain and oil on the scale the hill of broken amphorae in an earlier lesson recorded, and it had been untouched by war for centuries. Its loss is documented not by a historian's lament but by the western government's own legislation, which is far better evidence.

Example. In 445 a law of Valentinian III reduced the tax owed by Numidia and Mauretania Sitifensis to one eighth of the previous assessment, on account of the devastation. The year before, another law had stated that the treasury could not supply the army with food or clothing, and that what could be raised from the landholders was barely enough for the soldiers to subsist on. Read the two together.

The remission is the more informative document. Cutting an assessment to one eighth means writing off seven eighths of the revenue of provinces still nominally in Roman hands, which no government does for effect: it does it because the money is not there and pretending otherwise produces neither revenue nor compliance. So a treasury already short is now formally recognising the loss of the greater part of what remained in the region, in addition to everything the Vandals held outright.

The second law shows the consequence arriving on schedule. This is not a chronicler complaining about decline; it is the imperial government, in its own legislation, telling landowners that the army cannot be fed. Both halves of the loop are visible in the same archive within twelve months, and neither has anything to do with a battle.

Now you. The western empire still had Italy, and Italy contained some of the largest private fortunes in the ancient world. Why could the government not simply tax them harder?

Generals who made emperors

As money became the binding constraint, the men who commanded what army remained became the government. From the 450s the western throne was in the gift of a single soldier, Ricimer, of Suevic and Gothic descent, who held the senior military command for sixteen years, raised at least four emperors and disposed of three of them, and never took the title himself because he could not: his ancestry barred it and he did not need it.

An earlier lesson identified the empire's founding weakness as the fact that the position of emperor was never an office and had no rule of succession. In the fifth-century west that weakness reached its logical end. The emperor was whoever the man with the troops named, the man with the troops was whoever the largest surviving field army followed, and the field armies were increasingly composed of allied contingents under their own leaders, on the model of the treaty of 382. The chain from tax to army to political control had not broken; it had simply been taken over at the point where the payments were made.

The last attempt, costed

In 468 the two halves combined for one enormous effort to retake Africa, funded overwhelmingly by the eastern government and commanded by Basiliscus. It reached Cape Bon, was destroyed by Vandal fireships, and achieved nothing.

Example. The sources give the cost as between 64,000 and 130,000 Roman pounds of gold. Convert it and set it beside two other figures from this course.

At 72 solidi to the Roman pound, 100,000 pounds is 7.2 million solidi, and at 327 grams to the pound it is 32.7 tonnes of gold. The lower figure gives about 21 tonnes and the higher about 43, so the honest statement is that the expedition cost somewhere between twenty and forty tonnes of gold.

Two comparisons make it legible. Trajan's Dacian windfall, on the emended figure this course used earlier, was about 164 tonnes, so this single failed expedition consumed something like a fifth of the greatest treasure Rome ever captured. And the annual tribute the eastern empire had been paying Attila a generation earlier, 2,100 pounds of gold, would have to be paid for close to fifty years to reach it.

The importance is what the failure implies. This was not a doomed venture: it was well funded, jointly mounted, and aimed at the one province whose recovery would have restarted the loop in the right direction, since Africa's revenue could have paid a western army again. When it failed there was no second attempt, because there was no second sum of that size. The west did not fall because of a battle; it ran out of the ability to try.

Now you. The eastern empire had paid Attila 2,100 pounds of gold a year and had been terrified of him. Set that against the senatorial income above. What does the comparison say about the east?

Why the east held

Three advantages, none of them a matter of virtue.

Geography. The east faced one land frontier that mattered, the lower Danube, plus Persia, which was a conventional state that could be negotiated with and was often distracted. Its capital sat behind the Theodosian walls and a strait, and could be supplied by sea; Constantinople was not taken by anyone until 1204. The west faced the Rhine, the upper Danube, the Alpine passes and, once the Vandals had a fleet, the whole Mediterranean coast.

Revenue. Egypt alone was a tax province of the first rank and remained in eastern hands until the seventh century. The east lost no significant territory during the period in which the west lost nearly everything, so its loop kept turning the right way, which is the entire argument of this lesson stated once more.

Structure. The eastern aristocracy was largely a service aristocracy, made by office in Constantinople and dependent on the emperor, rather than a class of vast independent landholders like the western senate. It was easier to tax and less able to obstruct, and no eastern general ever acquired the position Ricimer held in the west.

476

On 4 September 476 a commander of the Italian field army named Odoacer deposed the boy Romulus, mockingly called Augustulus, who had been installed by his father the previous year. Odoacer did not appoint a successor. He sent the western imperial regalia to the emperor Zeno in Constantinople, with the message that the west no longer needed an emperor of its own, and asked for the rank of patrician and authority to govern Italy.

Example. Explain why this date is both a reasonable end point and a poor description of an event.

Poor as an event, because almost nothing happened. There had been no effective western emperor for years; the legitimate one, Julius Nepos, was alive in Dalmatia and outlived the deposition by four years, and Zeno's reply pointed this out. Odoacer governed Italy with the Roman senate, the Roman civil administration, Roman law and the consulship still functioning, and dated documents by consuls in the usual way. No contemporary describes 476 as the end of the Roman empire, because from Constantinople's point of view the empire plainly still existed and had merely stopped having a second emperor, which had happened before.

Reasonable as an end point, because of the one thing Odoacer did not do: he did not put up a replacement. Every previous general in his position had needed an emperor as a source of legitimacy, and this one calculated that he did not, and was right. That marks the moment the western imperial office ceased to be worth maintaining even for the men who had been controlling it. Everything the office had been for, an army to command and revenues to distribute, was gone; and Odoacer had recognised that the useful thing to be was the man with the soldiers, with the title supplied from the east.

Now you. Historians increasingly avoid the word fall. Given this lesson, is that reasonable?

What is left to settle

The mechanism is now on the table. The west lost the provinces that paid for its army; an army that cannot be paid attaches itself to whoever can pay it; and the groups that could pay it became the governments of Spain, Africa, Gaul and Italy. The east kept its provinces and therefore kept its army and therefore kept its state.

That explains the political outcome and leaves the larger question open. An empire is not only a government, and the previous lessons measured a Roman world of long-distance trade, large cities, mass-produced goods and a shared material culture. Whether that world went down with the tax system, and how anyone can tell, is the subject of the final lesson.

The question of what fell when Rome fell has been answered three different ways in the last three centuries, and the disagreement is mostly about which thing is being asked after.

The previous lesson supplied a mechanism: the western empire lost the provinces that paid for its army, and an army that cannot be paid becomes somebody else's. That explains a political outcome. It does not say whether anything larger ended, and this lesson tests that against evidence that is independent of politics, using the same method as the earlier lesson on the imperial economy: several proxies, each stated with its defect.

Three answers

Edward Gibbon, publishing across twelve years from 1776, wrote a narrative of decline running to 1453 and summarised its western half as the triumph of barbarism and religion. He had grounds for both terms and a personal stake in the second: the empire's adoption of Christianity diverted resources and attention, produced quarrels that consumed emperors, and, in his reading, drained the civic virtue that had made Rome formidable.

Henri Pirenne, writing in the 1930s, moved the ending by two centuries. The Germanic kings, he pointed out, preserved Roman institutions, Roman law and the Mediterranean trade that had always defined the Roman world; what broke that world was the Arab conquests of the seventh century, which turned the Mediterranean from a Roman lake into a frontier and pushed the centre of western Europe northwards. His formulation was that without Muhammad, Charlemagne would be inconceivable.

The third answer, dominant in the later twentieth century, is that the question is wrong. On this view the period is a transformation rather than a fall: barbarians were accommodated rather than invaders, elites adapted, the church carried Roman culture forward, and late antiquity was a creative age in its own right rather than the wreckage of a better one.

Weighing them

Example. Assess Gibbon's two words as causes.

Barbarism does not survive the previous lesson. The peoples involved were few, largely Romanised, mostly admitted by treaty and often already in Roman service, and the same peoples faced the eastern empire without destroying it.

Religion is more interesting and still fails as a primary cause, for a reason that is decisive and simple: the eastern empire was more Christian, earlier, more thoroughly, and with far more disruptive theological quarrels, and it survived for another thousand years. Any factor shared by both halves cannot explain a difference between them. The church did absorb resources and did compete for the loyalty of able men, and Gibbon is right that both happened, but the comparison shows it cannot have been what tipped the balance.

What deserves credit is Gibbon's method rather than his conclusions. He looked for structural causes running across centuries instead of blaming individuals, and he framed the problem as needing an explanation at all, where his predecessors had treated Rome's fall as the natural death of any great thing. His answers are wrong in ways his own approach eventually exposed.

Now you. How would you test Pirenne's thesis?

The test in pottery

Pottery is the historian's luckiest material. It is cheap, it breaks constantly, the pieces are indestructible, and the shapes and fabrics change often enough to be dated closely and traced to their kilns. That combination makes it the best surviving indicator of how complex an economy was.

Example. In Roman Britain, wheel-thrown pottery was made in industrial quantities, distributed across the province and used by ordinary farming households; so were mass-produced roof tiles. By the middle of the fifth century, wheel-thrown pottery has effectively ceased to be made in Britain, tile production stops, and coins go out of circulation. What does that show, and what is the honest limit?

It shows the disappearance of specialisation and distribution, not merely of wealth. Mass-produced pottery requires potters who do nothing else, kilns run continuously, a means of moving the output tens or hundreds of miles, and customers with something to trade for it. When it vanishes, every one of those conditions has failed together, and the replacement is handmade, locally fired ware of a quality below what pre-Roman Iron Age Britain had produced. That is the striking part: the material culture of the fifth century falls not to the level before the Romans arrived but below it.

Two honest limits. First, Britain is the extreme case, the province that had been most dependent on military spending and lost it most abruptly, and Italy, Africa and the eastern Mediterranean show far gentler curves; a claim about Britain is not a claim about the empire. Second, pottery measures complexity, not happiness. It tells you the economy simplified. It does not tell you that any individual was worse off, which the next question takes up.

Now you. Some skeletal evidence suggests that people in post-Roman northern Europe were, if anything, slightly taller than under Roman rule. Does complexity equal welfare?

The other proxies agree

The pottery does not stand alone, and the corroboration comes from the same independent directions used earlier in this course.

The Greenland ice, which recorded lead at four times natural background around the beginning of our era, falls back to near-background levels in the post-Roman centuries and does not return to Roman levels for well over a thousand years. Coinage tells the same story in a different register: the disappearance of small-denomination bronze is more informative than any change in gold, because low-value coin is what an economy uses when ordinary people buy ordinary things, and its absence means everyday exchange has stopped being monetary. Cattle bones from Britain shrink after the Roman period, reverting towards pre-Roman size, which points to the loss of managed breeding kept up by a market for meat and hides.

And the city of Rome itself, which had held something near a million people when the empire was at its largest, is usually reckoned at around 100,000 by 500 and perhaps 30,000 in the centuries after the Gothic wars of Justinian's reconquest: a fall of more than thirty to one. No western European city reached a million again until London around 1800.

The empire that did not end

Meanwhile the empire continued. The eastern half, governed from Constantinople, ran without a break in legitimacy from Constantine to 29 May 1453, when the walls were breached by Mehmed II, which is 977 years after the deposition in Italy and 1,480 years after Augustus's settlement.

Its inhabitants called themselves Romans throughout, in Greek, and it is worth registering how odd our habit of calling them Byzantines is: the word was coined by a scholar in the sixteenth century and no citizen of that state ever used it. In the sixth century Justinian retook Africa and Italy, and had the whole of Roman law codified into the compilation that carries it to the present day. The consulship, an office created when the kings were expelled a thousand years earlier, was last held in the west in 541.

Example. List what survived the western collapse, and then say what all the survivals have in common.

The law survived, reissued by Germanic kings for their Roman subjects and codified in the east, and revived in the Italian universities from the eleventh century as the foundation of continental civil law. The church survived with its provinces mapped onto the old imperial ones, its language Latin, and its bishops still doing the civic work the previous lessons described. The eastern state survived. The name survived: the Ottoman sultan took the title Caesar of Rome, the western kingdom crowned in 800 called itself the Roman empire until 1806, and Kaiser and Tsar are both the word Caesar.

What they have in common is that none of them required a tax system. A legal code is a text and can be copied. A church is funded by donation, land and its own authority. A title is a claim. Every one of these could be maintained by a society far poorer and simpler than the empire that produced it, which is precisely why they lasted.

Now you. So state, in one sentence, what actually ended in the west, and then say what the sentence is doing.

Ending

The empire ran, in the west, for five centuries after Augustus, and it ran on an arrangement that the first lessons of this course laid out and the last ones watched come apart. A monarchy with no name and no rule of succession, so that the army decided. An army that was the largest expense any pre-industrial state had taken on. A tax of about five per cent, light enough to be tolerable and too light to leave any margin. Government delegated to a couple of thousand cities whose elites wanted the job. A citizenship extended until everyone had it. A frontier that stopped moving, so nothing came in from outside. Every one of those is visible in the fifth century, doing the opposite of what it had done in the first.

What is worth keeping from this is not the date. It is the habit of asking, about any large and apparently permanent structure, what it is actually made of and what is paying for it. The Roman empire looked to its inhabitants like a fact about the world, in the way such things always do, and it turned out to be a machine with a specific input. When the input stopped, the parts that needed it stopped too, and the parts that did not are still here: the law in the courts, the church in the towns, the calendar, the alphabet, the roads under the roads, and the word Caesar in three languages.