The Silk Road

Every other subject you might study begins with a thing and then finds a name for it, and this one begins the other way round, with a name coined in Berlin in 1877 that has been looking for its thing ever since.

A word from a railway survey

Ferdinand von Richthofen was a German geologist who travelled in China between 1868 and 1872, largely at the expense of the Shanghai Chamber of Commerce, mapping coal seams and reporting on where a railway might usefully run. The first volume of his China: Ergebnisse eigener Reisen appeared in 1877, and in it, drawing a map of what he took to be the Han dynasty's western communications, he wrote Seidenstrassen, silk roads, in the plural.

His source for the ancient route was not Chinese. It was Ptolemy, quoting the lost geography of Marinus of Tyre, who in turn reported that a Macedonian merchant named Maes Titianus had sent agents east towards the land of the Seres, the silk people. Richthofen was reconstructing a first century route from a second century Greek summary of a first century trader's report, seventeen hundred years later, in order to shade a nineteenth century map.

The singular form, and the romance, came later. Sven Hedin, who had been Richthofen's student, published Sidenvägen, The Silk Road, in 1936 after his own expeditions in the Tarim Basin, and it is his version of the phrase, one road, singular and continuous, that entered every language. By then the term was already doing work that the evidence could not support, and it has not stopped.

Nobody who used it called it anything

The people who actually moved goods across Asia had no word for the whole. This is not an accident of vocabulary but a fact about how the traffic worked, and it is the single most important correction a beginner has to make.

Chinese officials had a term for the region, Xiyu, the Western Regions, and the Hanshu, the history of the Former Han completed around 111 CE, devotes a chapter to it that reads like a gazetteer: fifty-odd kingdoms, each with its distance from Chang'an in li, its households, its population, its fighting men, and the number of days to the next place. That is an administrator's document about frontier management. It names Kashgar and Khotan and Ferghana. It does not name a road, because from the Han point of view there was no road, only a set of places one could reach and, if one were fortunate, tax.

The merchants' vocabulary was smaller still. What survives of their writing, and there is more of it than you would expect, is about the next stretch: this pass, that ford, the price of a camel in Guazhou, a debt owed in Samarkand. A Sogdian trader in Dunhuang in 313 knew about the road to Luoyang and the road home. There is no reason to think he had ever formed a mental image of a route running to the Mediterranean, and no document suggests anyone did.

The traffic was older than the opening

The conventional date for the road's beginning is 138 BCE, when Emperor Wu of Han sent an officer named Zhang Qian west to find allies against the Xiongnu confederation. Zhang Qian was captured almost at once, spent about ten years in Xiongnu hands, married, escaped, went on anyway, and returned to Chang'an in 126 BCE with two survivors of an original party of a hundred and no alliance.

What he brought back instead was information, recorded in chapter 123 of Sima Qian's Shiji, and one detail in it demolishes the conventional date. In Bactria, in what is now northern Afghanistan, Zhang Qian saw cloth from Shu and bamboo canes from Qiong, both from Sichuan in southwestern China. He asked how they had come there and was told that merchants bought them in Shendu, India. So Chinese goods were already reaching Bactria by a route the Han court knew nothing about, through India, before the Han court had sent anybody anywhere.

The lesson generalises. Han armies and Han garrisons changed the volume, the security and above all the documentation of the traffic, and it is documentation that historians mistake for existence. Jade from Khotan was reaching Chinese elites well before the Han: the tomb of Fu Hao at Anyang, sealed around 1200 BCE, held hundreds of jade objects, and much of that jade is nephrite from the Khotan riverbeds. The routes were not opened in 138 BCE. They were noticed.

Example. A textbook says the Silk Road "opened in 138 BCE with the mission of Zhang Qian". Using only what is in this section, say precisely what is wrong with that sentence and what a defensible version would be.

Three things are wrong and one is defensible. The date is wrong as a start date for traffic, since Zhang Qian himself reported Sichuan goods in Bactria on arrival, and Khotanese jade was in Chinese royal tombs a thousand years earlier. The word "opened" is wrong, because a mission that returned with two survivors and no treaty did not open anything; what it produced was a report. And "the Silk Road" is wrong as the object of the verb, since the term did not exist and the thing it names was not a single entity that could be opened or closed. What survives is real and worth saying: from the 130s BCE the Chinese state began to acquire systematic knowledge of Central Asia, and within a few decades it was garrisoning parts of it, which is why the written record starts then. A defensible sentence is that the Chinese documentation of Central Asian exchange begins with Zhang Qian, not the exchange itself.

Now you. A different textbook says the Silk Road "closed in 1453 when the Ottomans took Constantinople". Without knowing anything yet about the fifteenth century, what should make you suspicious of the sentence's form?

A relay, not a journey

Almost nobody travelled the length of it. The traffic moved as a relay, each merchant working a familiar stretch, selling to the next, who sold to the next.

That structure has an arithmetic consequence that explains one of the most quoted facts about the ancient world. Roman writers complained bitterly about the price of silk. Pliny the Elder called it a drain on the empire, and the emperor Tiberius is reported to have persuaded the Senate in 16 CE to ban men from wearing it, partly on grounds of cost. It is tempting to read those complaints as evidence of a punishing journey, and the journey was punishing, but the compounding of ordinary trading margins does most of the work by itself.

Example. Suppose a bolt of silk changes hands twelve times between the loom and the buyer, and each merchant takes a modest twenty per cent over what they paid. By what factor has the price multiplied?

Each hand multiplies the price by 1.2, so twelve hands multiply it by ${1.2}^{12}$. That is 8.92, so the Roman buyer pays close to nine times the Chinese price without a single merchant taking what anyone would call an unreasonable margin. Notice what the calculation shows and what it does not. It shows that a long price gradient is not evidence of extortion or of extraordinary transport costs, because compounding is enough. It does not show that this is what actually happened, because we have no series of prices for the same commodity along the chain, and the twenty per cent is a guess. The value of the model is that it sets a floor: any explanation of high prices at the Roman end has to beat plain compounding before it earns belief.

Now you. A merchant proposes cutting out intermediaries by carrying the goods eight stages himself, taking twenty-five per cent at each of the two ends only. What does the buyer pay relative to the loom price, and why did almost nobody do this?

Four things the name could mean

Given all that, what is left for the term to denote? There are four candidates, and they are not equally defensible.

It could mean a road, a physical continuous route. That reading is simply false, and no serious writer holds it, though the singular form of the name keeps reviving it in popular accounts.

It could mean a set of routes, the actual tracks through the passes and along the desert rims. That is real and mappable, but it is too narrow to carry the subject, because the interesting claims are about what happened along the tracks rather than the tracks themselves.

It could mean the long-distance luxury trade, silk going west and a few dense valuables coming east. This is the usual working definition, and Valerie Hansen's The Silk Road: A New History (2012) is a sustained attack on it, arguing from the excavated documents that this trade was small, intermittent, and swamped in the record by local exchange. She does not deny it existed. She denies that it was big enough to explain anything.

Or it could mean the whole system of exchange across Central Eurasia, goods and people and ideas and pathogens together, from roughly the second century BCE to roughly the fifteenth century CE. That is the definition James Millward defends in The Silk Road: A Very Short Introduction (2013), and it is the one this course uses, because it is the only one under which the important questions are askable. The cost of adopting it is real and worth stating plainly: it is so broad that it risks meaning no more than "Eurasian history", and the discipline it needs is a refusal to let anything into the account without evidence that it actually moved.

Example. Under this course's definition, is the transmission of Buddhism from Gandhara to China part of the Silk Road? Is the spread of the Black Death? Is the Mongol conquest of Persia?

The first two are in and the third is on the boundary. Buddhism qualifies because the mechanism is exchange along the routes: the monks travelled with the caravans, the texts were carried, translated and copied at the oasis towns, and the donors named in the cave inscriptions are merchants. The plague qualifies for the same reason, with the extra advantage that the pathogen leaves physical evidence in the ground, which most cargo does not. The Mongol conquest is different in kind: it is an army moving, not an exchange, and the definition would become useless if any movement across Asia counted. What belongs in the account is not the conquest but its consequences for exchange, the single administration, the postal relay, the forced transfer of craftsmen, which is exactly how a later lesson treats it.

Now you. A critic says the term is Eurocentric: it names the whole system after the one commodity Europeans wanted, defined by a European in a European language, and centres a Chinese and Central Asian world on its western customers. Is the critic right, and does it matter?

Where this leaves us

The name is late, the thing is real but diffuse, and the traffic long predates the documents that record it. A working definition is now on the table, and it is deliberately broad, which means the only thing keeping the subject honest is evidence.

That raises the question the next lesson has to answer before anything else can be claimed. We know about Sogdian family firms and Turfan grain prices and the Buddhist donors of Dunhuang, and we know almost nothing comparable about the wetter, greener, richer lands to the north and south of the desert routes. That is not because less happened there. It is because paper, wood and cloth rot in damp ground and survive for two thousand years in dry sand, and the entire subject is built on a handful of accidents of preservation that a careful reader has to learn to see.

The previous lesson settled on a deliberately broad definition of the subject and warned that the only thing keeping it honest is evidence, so this lesson looks at what that evidence actually is, and it is stranger and thinner than most accounts admit.

The sample nobody took

Start with what a historian would want. To say how much trade crossed Eurasia, you would want customs records from a representative set of crossing points across several centuries, the way an economic historian of eighteenth century Britain has port books and excise returns. Nothing of the kind exists for any part of this subject, at any date, anywhere.

What exists instead is a scatter of documents that survived because of where they were dropped. There is no sampling frame, no way of estimating what fraction of the original writing has come down, and no reason to think that fraction is even roughly constant across regions. A historian of the Silk Road is in the position of an archaeologist trying to describe a city from the contents of four rubbish pits, three of which are in the same street.

That is not a counsel of despair, and the material that does survive is extraordinarily rich. It is a warning about a specific and very common error: reading the density of evidence as the density of activity. The Tarim Basin has produced more everyday documents from the first millennium than the whole of northern India, and the reason has nothing to do with how much happened in either place.

Why the desert keeps things

Organic material survives when the microbes that eat it cannot work, and they need water. Annual rainfall at Turfan is around 16 mm, at Dunhuang around 40 mm, and evaporation exceeds it by orders of magnitude. Paper, wood, silk, leather, grain, human bodies and their clothing sit in that sand for two thousand years and come out legible.

Compare the alternatives. In the Ganges plain, in Sogdiana's irrigated lowlands, in the Chinese south, in the Mediterranean outside Egypt, the same objects are gone within a few centuries. The palm-leaf manuscripts of Indian Buddhism are known largely from copies made in Nepal and Tibet, or from fragments found in Central Asia, which is to say that even Indian religious texts reach us through the desert. Sogdian, the language of the merchants who ran the trade for six hundred years, is known mainly from documents found outside Sogdiana.

Set out plainly, the asymmetry is severe: the places we can see are the dry, marginal, thinly populated ones, and the places we cannot see are the wet, rich, populous ones. Any statement of the form "most of the trade was in the Tarim" is therefore suspect on its face, because the Tarim is where the evidence is.

Example. An excavation report notes that roughly seven hundred documents survive from Niya, a small oasis in the southern Taklamakan abandoned around 400 CE, and that essentially nothing comparable survives from Pataliputra, the great city on the Ganges. What may we conclude about their relative commercial importance?

Nothing at all, and the temptation to conclude something is exactly the error this lesson exists to prevent. The two sites differ in the one variable that controls document survival, which is water, and they differ in it so extremely that the comparison carries no information about anything else. Pataliputra sits in a monsoon flood plain where wood and birch bark and palm leaf decay within generations, and it has also been continuously occupied, which means later building has destroyed earlier layers. Niya was abandoned and never reoccupied, and it is bone dry. To compare the commercial importance of the two you would need a class of evidence that survives equally in both, such as coins, imported ceramics or building remains, and if you use those the ranking reverses at once.

Now you. A historian argues that Buddhism was more important in Central Asia than in India during the fifth century, citing the enormous quantity of Buddhist manuscripts recovered from Central Asian sites and the near absence of contemporary Indian ones. Diagnose the argument.

Four pits

Four bodies of material carry most of the weight, and it is worth knowing them by name, because nearly every specific claim in the rest of this course descends from one of them.

The Niya documents were found by Aurel Stein from 1901 onwards at a site in the southern Taklamakan, capital of a small kingdom the Chinese called Shanshan. There are roughly seven hundred of them, written in Gandhari Prakrit in the Kharosthi script, an Indian language and an Indian script in a town two thousand kilometres from India, which is itself a finding. Many are wedge-shaped wooden double tablets, tied with cord and sealed with clay so that the text inside could not be read without breaking the seal, an envelope made of wood. Their contents are relentlessly local: the sale of a camel, the ownership of a vineyard, a dispute over a runaway slave, an order from the king about monks behaving badly. There is almost no international trade in them at all.

The Sogdian Ancient Letters, five of them substantially complete, were found by Stein in 1907 in a ruined watchtower on the Han wall west of Dunhuang. They are the oldest substantial Sogdian texts known, and they are business and family correspondence written on paper. The reason they are in a watchtower and not in Samarkand is that they were never delivered. A mailbag, abandoned or confiscated, is the single most informative kind of document a historian can find, because it has not been selected for importance by anybody.

The Turfan documents come from the Astana cemetery and other sites in the Turfan oasis, and reach us through a bizarre route: waste paper was recycled into burial goods, so that shoes, hats and figurines for the dead were made from discarded contracts, tax registers, market price lists and school exercises. Peeling apart a paper shoe from a seventh century grave yields the everyday administration of an oasis town.

The Dunhuang library cave, cave 17 of the Mogao complex, was walled up early in the eleventh century and opened in 1900 by a Daoist caretaker named Wang Yuanlu. It held something like fifty thousand manuscripts and printed items, in Chinese, Tibetan, Sogdian, Khotanese, Uighur, Sanskrit and more. Among them is the Diamond Sutra printed on 11 May 868, the earliest complete printed book bearing its own date. Why the cave was sealed is unresolved: the two live proposals are that it was a store of sacred waste, texts too holy to burn and too damaged to use, and that it was hidden in a hurry against an expected invasion. Rong Xinjiang's argument for the sacred waste reading rests on the state of the contents, which are heavily weighted towards damaged and duplicate material rather than the treasures a monastery would evacuate.

Example. Of the roughly fifty thousand items from cave 17, the overwhelming majority are Buddhist scripture, and the secular material, contracts, loan agreements, household registers, lay society circulars, is usually put at a few per cent. Take five per cent as a working figure and say what follows.

Five per cent of fifty thousand is 2,500 documents. That number is the entire documentary base for the social and economic history of one oasis over roughly six centuries, and it is by a wide margin the largest such collection in Central Asia. Two consequences follow. The first is that any statistical claim built on it, average loan interest, the frequency of a contract type, the price of a bolt of cloth, rests on tens or low hundreds of instances once the material is sorted by century and category, which is why the honest literature quotes ranges rather than means. The second is that Dunhuang inevitably becomes the default picture of the whole region, simply because nowhere else can answer the question at all, and a reader should be alert whenever a general statement about Central Asian society turns out to be a statement about one town.

Now you. Suppose the sacred waste explanation for the sealing of cave 17 is correct. How should that change the way you use the collection as evidence?

Dating something nobody dated

Most of this material carries no date, and the techniques used to supply one are worth seeing once, because they decide how much weight a document can bear.

The Sogdian Ancient Letters are the standard case. Letter II describes a catastrophe in China: the last emperor, the writer says, fled Luoyang because of famine, his palace was set on fire, and the city and Ye are no more. It also reports that Indians and Sogdians in Luoyang had all died of starvation. Chinese sources record the sack of Luoyang by Xiongnu forces in 311 CE and the flight of the court, and W. B. Henning argued in 1948 that the letter must have been written shortly after, dating the group to 313 or 314. János Harmatta later proposed 196 CE, matching the letter to an earlier round of disorder at the end of the Han.

The dispute is settled, for most scholars, on the details rather than the general fit: the sequence of events in Letter II matches 311 closely, and the letters' Sogdian is late enough in form to sit uncomfortably in the second century. What matters here is the shape of the reasoning. The letters are dated by tying an internal narrative to an externally dated event, which means the date is only as good as the match, and the whole chronology of early Sogdian trade rests on one letter's account of a fire.

Evidence that is not writing

The subject has been transformed since about 1990 by evidence that no scribe produced. Coin finds map circulation independently of any text. Isotope analysis of human teeth shows where a person spent childhood, which is how cemeteries of apparent locals have turned out to contain first generation migrants. Textile analysis identifies where a cloth was woven from its weave structure and its fibre: the silks excavated at Palmyra in the 1930s include Han damasks whose patterns and technique are Chinese, and that identification is a laboratory result rather than a reading of Pliny.

Example. A silk fragment from a Palmyra tomb is identified as Chinese from its weave structure rather than from its pattern. Why is that the stronger form of identification?

Because a pattern can be copied and a loom setup cannot be, at least not without building the loom. Motifs travel: a Syrian weaver who has seen an imported Chinese textile can imitate its dragons or its cloud scrolls on local equipment, so a Chinese-looking design proves only that somebody had seen Chinese design. The structure of a warp-faced compound weave, by contrast, records the physical arrangement of the loom that made it, the number of warp sets, how the sheds were controlled, the twist and preparation of the thread. Reproducing it requires reproducing the whole apparatus and the skill to work it, which is the point the tenth lesson develops at length. So structure is close to a fingerprint of a manufacturing tradition, and it is the kind of evidence a laboratory can assess without any historical assumption at all. The general principle is worth keeping: prefer the evidence that is hard to imitate over the evidence that is merely characteristic.

Now you. The Niya documents carry no calendar dates that a modern reader can convert directly. How would you date them?

The most dramatic recent case is ancient DNA. A cemetery in the Chu valley of Kyrgyzstan, excavated in the 1880s, has tombstones dated 1338 and 1339 that record deaths from pestilence; in 2022 a team led by Maria Spyrou published Yersinia pestis genomes recovered from teeth in those graves, placing the strain at the root of the lineage that caused the Black Death. That is a chain of evidence, gravestone date to genome to phylogeny, in which no historical text appears at any point, and a later lesson is built on it.

How to read a claim

The working habit this lesson is trying to install is a question to ask of any specific statement about the Silk Road: what is this resting on, where was it found, and how was it dated?

Try it on three claims. "Sogdian merchants had commercial networks reaching from Samarkand to Chang'an" rests on the Ancient Letters plus Sogdian colonies attested in Chinese administrative sources plus tomb reliefs in China, three independent classes of evidence, and it is about as secure as anything in the field. "The Silk Road carried enormous volumes of trade" rests, when you chase it, mostly on Roman moral complaints about luxury and on the presence of exotic objects in elite tombs, neither of which measures volume. "Paper making passed to the west after the battle of Talas in 751" rests on a single anecdote written down nearly three centuries later, and the next chapters of this course take it apart.

The next lesson changes tack and asks a question that needs no documents at all. Before asking what moved, ask what could move: a loaded camel has a range, a payload and an appetite, and those three numbers constrain the trade more tightly than any political history does.

Before asking what moved across Eurasia it is worth asking what could move, because the answer can be derived from the biology of a pack animal and a map, without consulting a single document.

The previous lesson argued that the written evidence is thin and unevenly preserved. The argument that follows has the opposite character: it uses almost no documents, it can be checked by anyone with a calculator, and it turns out to predict the contents of the caravans quite well. Where the two kinds of argument agree, the conclusion is about as firm as this subject gets.

How far, in numbers a Han official would recognise

The Hanshu chapter on the Western Regions lists each kingdom with its distance from the capital in li. Shule, which is Kashgar, is given as 9,350 li from Chang'an.

To turn that into kilometres, take the Han li as 300 paces of 6 chi, so 1,800 chi, and the Han chi as about 23.1 cm, a value fixed by surviving bronze and wooden measuring rules. Then one li is $1800\times 0.231=415.8$ m, and the distance is $9350\times 415.8=3\{,\}888$ km. That is Chang'an to the far western edge of the Tarim Basin, with the Pamirs, Sogdiana, Persia and the Mediterranean all still ahead. To Antioch is roughly another 5,000 km by any usable line, so the whole span is something over 8,000 km, and more once the detours around mountains and deserts are counted.

A caravan on level ground makes about 4 km/h and walks for seven or eight hours, so 30 km is a normal day and 35 a good one. At 30 km a day, Chang'an to Kashgar is $3888/30=130$ days of travel, or well over four months allowing for rest days, weather and the wait for a pass to clear. Nobody in the ancient world doubted this was a long way. What the arithmetic adds is that it was long in a very specific way: not a voyage but a wage bill, 130 days of paying men and feeding animals before a single item is sold.

The animal

The pack animal of the eastern half of the route is the Bactrian camel, two-humped, wool-coated, and able to work at temperatures from about minus 30 to plus 40 degrees Celsius, which matters because the Tarim Basin does both within a year. West of Persia the one-humped dromedary takes over, and hybrids of the two were bred deliberately in Central Asia because the first cross is larger and stronger than either parent.

A Bactrian carries 150 to 200 kg on a long march, and traditional caravan practice loaded towards the lower figure, since an overloaded camel goes lame and a lame camel is a total loss. This course uses 150 kg throughout as a conservative working payload.

The camel did something more radical than carry loads: it drove the wheel out of large parts of Asia. Richard Bulliet's The Camel and the Wheel (1975) collected the evidence that wheeled transport largely disappeared from the Middle East and North Africa between roughly 300 and 600 CE, and argued that this was an economic choice rather than a decline. A cart needs a road, and a road needs building and maintaining by someone; a pack camel needs neither, crosses ground no cart can, requires one handler for several animals rather than a driver and a team per vehicle, and eats browse that would not feed an ox. Where the state that maintained the roads weakened, the camel simply won on cost. It is a useful corrective to any account in which technology only moves forward: cities in the Islamic world were rebuilt with lanes too narrow for a cart because nobody needed one.

The appetite

Here is the constraint that shapes everything else. A working Bactrian eats roughly 6 kg of dry fodder a day when there is nothing to browse, and drinks heavily when water is available.

Set that against the payload. A camel carrying 150 kg of its own fodder carries $150/6=25$ days of food. It eats its own maximum load in twenty-five days. Over the 130-day march to Kashgar it would need $6\times 130=780$ kg, which is $780/150=5.2$ full camel-loads, and those five extra camels each eat 6 kg a day as well, and the camels carrying their fodder eat too. The series does not converge into anything useful. No caravan on this route could ever carry the food for its own journey.

That single fact does more explanatory work than any political history. It means the route cannot exist as a route: it can only exist as a chain of places that sell fodder, water and shelter, spaced a few days apart. It means those places have monopoly power over anyone travelling, since a caravan that cannot buy at the next oasis has no alternative and no reserve. And it means the political history of the Silk Road is mostly a history of who controlled the supply points, which is exactly what the Chinese, Tibetan, Turkic and later Mongol contests over the Tarim were about.

Example. A caravan wants to cross a stretch of eight days with no fodder available at all. Using 6 kg a day and a 150 kg payload, what fraction of each camel's capacity is consumed by its own food, and what does that do to the freight rate over that stretch?

Eight days of fodder is $8\times 6=48$ kg, which is $48/150=32$ per cent of the payload, leaving 102 kg of cargo. The freight rate per kilogram of cargo therefore rises by a factor of $150/102=1.47$, close to a fifty per cent surcharge, for a stretch of only 240 km. Two things follow. Long fodderless stretches are punitively expensive rather than merely difficult, which is why routes bend a long way to stay near supply. And the surcharge rises steeply, not linearly, with the length of the gap: at sixteen days the fodder is 96 kg, the cargo is 54 kg, and the rate has nearly tripled. Beyond about twenty-five days the cargo is zero and the crossing is impossible at any price.

Now you. The same camel is worked on a stretch where grazing supplies half its needs. Redo the eight-day calculation and say what kind of country that describes.

The thirst

Water sets the geometry. A Bactrian camel in cool weather can work five days or more without drinking, and rather less in summer heat. At 30 km a day, five days is about 150 km, and that is the maximum spacing between watering points that a caravan can survive.

Now look at the Taklamakan. It is roughly 1,000 km east to west and 400 km north to south, some 337,000 km² of shifting sand, with no permanent water in the interior. A direct crossing along the long axis is 1,000 km, which is over thirty days, which is six times the camel's dry range. The desert cannot be crossed. It can only be skirted, and the water to skirt it comes from the same place at every point: meltwater streams running off the Kunlun range to the south and the Tian Shan to the north, which sink into the sand a short distance out.

So the route splits, not by choice but by hydrology, into a northern rim through Turfan, Kucha and Aksu and a southern rim through Miran, Niya and Khotan, meeting again at Kashgar before the passes. Draw the streams and you have drawn the map. Every well-known Silk Road town in the Tarim sits where a mountain stream reaches the sand, and the two lessons of the last section apply to all of them: they are the only fodder for hundreds of kilometres, and they know it.

What a cargo has to be worth

Put the pieces together and you can derive, roughly but honestly, which goods could cross.

Measure everything in days of unskilled labour, which travels across centuries and currencies better than any coin. Suppose a camel with its share of a driver, its fodder, its market dues and the caravan's protection costs runs at about two labour-days per camel-day. That is a guess, but it is a guess with a floor, since a driver managing three or four camels has to eat, and the fodder alone is a real cost.

Over 130 days one camel-load costs $130\times 2=260$ labour-days, and it carries 150 kg, so the freight is $260/150=1.73$ labour-days per kilogram delivered.

Now compare commodities. If a labourer earns about 6 kg of grain a day, grain is worth $1/6=0.17$ labour-days per kilogram, so carrying it the length of the route costs about ten times what it is worth. Grain cannot cross, at any price, in any century, under any emperor. Silk is the opposite case: a bolt weighing around 0.35 kg represents perhaps fifteen labour-days of sericulture, reeling and weaving, so silk is worth about $15/0.35=43$ labour-days per kilogram, and the freight is $1.73/43$, about four per cent of the value. Silk can cross the whole of Asia and arrive having absorbed a transport cost smaller than a single merchant's margin.

The threshold is around 2 labour-days per kilogram, and the goods actually attested in the long-distance traffic sit far above it: silk, precious stones, gold and silver work, musk, aromatics, high-grade paper, coral, glass vessels, medicines. Nothing bulky, nothing cheap, nothing perishable.

Example. Pliny gives the Roman price of black pepper as 4 denarii a Roman pound. Taking a Roman pound as 327 g, a denarius as 4 sesterces and a soldier's pay as 900 sesterces for about 300 working days, could pepper have crossed Asia overland?

Four denarii is 16 sesterces per Roman pound, so the price is $16/0.327=49$ sesterces per kilogram. At 3 sesterces a day, that is $49/3=16$ labour-days per kilogram, comfortably above the threshold of about 2. Overland freight of 1.73 labour-days per kilogram would consume $1.73/16.3=11$ per cent of the Roman retail value, which is survivable. So the answer is yes in principle, and pepper is the interesting borderline case: dense enough in value to cross, but at a cost that eats a tenth of the price before any merchant's margin, in a trade where the sea alternative cost a fortieth as much per unit distance. That is exactly why pepper went by sea and silk could afford to go either way, and it is a good illustration of what the threshold does. It does not sort goods into possible and impossible so much as it decides which route a good is forced onto.

Now you. A caravan pays a toll of 5 per cent of the cargo's value at each of twelve oases. How much of the cargo's value survives, and what does the answer imply about political fragmentation?

The model can be checked against an independent source. Diocletian's Price Edict of 301 CE fixes maximum prices for goods and for freight, and the classic result drawn from it, by A. H. M. Jones and Richard Duncan-Jones, is that carrying wheat about 300 Roman miles by land roughly doubled its price. Three hundred Roman miles is $300\times 1.48=444$ km, or 15 days at caravan pace. Run that through the model above: 15 days at two labour-days is 30, divided by 150 kg is 0.20 labour-days per kilogram, against a grain value of 0.17. The transport adds about 118 per cent to the price, which is to say it slightly more than doubles it. A crude model built from a camel's appetite lands on the same answer as a Roman price schedule, which is the sort of agreement that should raise your confidence in both.

Example. Lapis lazuli from Badakhshan, in the mountains of northeastern Afghanistan, reached Egypt from at least the third millennium BCE, well before any of the states in this course existed. Does the freight model explain that?

Yes, and cleanly. Lapis is a stone of no practical use whatever, valued entirely for its colour, and the good grade is scarce, which puts its value per kilogram in the same class as silk or above it. A commodity like that pays a freight of two labour-days per kilogram out of small change, so distance is nearly irrelevant to whether it moves; what limits it is demand at the far end, not carriage. This is why the earliest long-distance trade in every region is in gems, metals and aromatics rather than in anything useful, and it is a warning against reading the antiquity of a trade as evidence of a developed route. Lapis crossing 4,000 km in 2500 BCE tells you only that a chain of short exchanges existed, which is a much weaker claim than the one it is usually recruited for.

Now you. Chinese paper, when it began moving west, was expensive and light. Cotton cloth from India was cheap and light. Iron tools were cheap and heavy. Predict which crossed the whole route, which crossed part of it, and which did not.

The limits of the model

This is a model, and it is worth naming what it leaves out before the next lesson uses it.

It ignores everything but freight. Tolls, protection payments and outright robbery could easily exceed transport costs, and they varied wildly with who held the passes, which is a political variable no camel arithmetic can predict. It assumes camels, when the Pamir crossings used yaks and porters, the steppe used horses and ox wagons, and rivers were used wherever they ran the right way. It assumes a merchant pays cash costs, when much of the actual traffic moved as tribute, gift, salary or plunder, and none of those respect a freight rate. And it is calibrated on assumptions, the two labour-days most of all, that are defensible rather than measured.

What it does reliably is separate the impossible from the possible. Nothing in the historical record contradicts it, and where a document seems to show a bulk good crossing a great distance overland, the explanation has always turned out to be that the state was paying, which is the subject two lessons ahead.

The next lesson takes the strongest consequence of the model, that no caravan can feed itself, and follows it to the institution it implies. If the trade depends absolutely on places that sell fodder and water, then those places are not stops on the route. They are the route, and understanding how an oasis worked, how it got its water and what happened when the water failed, explains more about the rise and fall of the traffic than any account of emperors.

A caravan crossing Asia had to buy food and water every few days from someone, which makes the places that sold them the most important institutions in the subject and the ones most accounts skip over in a sentence.

The previous lesson derived the constraint: a camel eats its own maximum load in twenty-five days, so no caravan can carry the supplies for a journey of months. This lesson follows the constraint to what it implies. If the trade depended absolutely on a chain of supply points, then the history of the trade is largely the history of those points, of where their water came from, who controlled them, and what happened when either failed.

Water that arrives from a mountain

The Tarim Basin is an endorheic basin: rivers run into it and none run out. Rain contributes almost nothing, 16 mm a year at Turfan and about 40 mm at Dunhuang, so every drop that matters is snow and glacier melt from the Kunlun range to the south and the Tian Shan to the north.

That gives the water a distinctive behaviour. It appears in early summer as the snowline retreats, runs down the mountain front in short violent streams, spreads across the alluvial fans where the slope flattens, and sinks. Most Tarim rivers simply end in the sand. A handful, chiefly the Tarim itself and the Khotan River, run far enough to matter, and the Khotan River is famously fed well enough in high summer to be crossed by caravans in the wrong season and impassable at the right one.

An oasis, then, is not a spring in a desert. It is a place on an alluvial fan where a mountain stream is still on the surface, or close enough beneath it to be reached, and where the gradient is gentle enough to spread the water over fields. Draw the ring of alluvial fans around the Taklamakan and you have drawn the position of every significant town in the basin, and the reason the route splits into a northern and southern rim.

Tunnels, because the sun takes the rest

Turfan is the extreme case and the most instructive. The basin floor at Aydingkol Lake is about 154 m below sea level, summer air temperatures pass 45 degrees Celsius, and potential evaporation is around 3 m a year. Water carried in an open ditch across that basin is being poured onto a hotplate.

The response was the karez, called qanat in Persia: a gently sloping tunnel dug from the water-bearing gravels of the alluvial fan down to the fields, with vertical shafts every twenty or thirty metres for spoil removal, ventilation and maintenance. Seen from the air the system is a line of small craters marching down the slope. The water runs underground, in the dark, at a gradient shallower than the land surface, so that it emerges at the surface where it is wanted, entirely by gravity, with no lifting and almost no evaporation. Turfan's network is usually given as something over a thousand karez with a total tunnel length of several thousand kilometres, dug and maintained by hand.

Example. Take a karez delivering a modest 20 litres per second, and compare it with the same water carried 20 km in an open canal 3 m wide across the Turfan basin, where potential evaporation is 3 m a year. How much is saved?

The flow is $0.020\times 3600\times 24\times 365=630\{,\}700$ m³ a year, enough to irrigate on the order of 125 hectares. The open canal has a surface of $20\{,\}000\times 3=60\{,\}000$ m², and at 3 m of evaporation a year it loses $60\{,\}000\times 3=180\{,\}000$ m³, which is 29 per cent of the whole flow, before counting seepage into coarse gravel, which in that ground can take as much again. The tunnel loses effectively none of either. So the karez is not a quaint alternative to a canal; it is roughly the difference between irrigating 125 hectares and irrigating 60 or fewer, which in a basin where cultivable land is defined by water is the difference between a town and a hamlet.

Now you. Karez tunnels need constant maintenance: collapse, silting, and the shafts filling with blown sand. What does that imply about the relationship between an oasis's political stability and its agricultural capacity?

What the oasis actually sold

The obvious product is agricultural: wheat, millet, cotton, melons, and above all the grapes of Turfan, whose raisins are attested in the documents and whose vineyards were taxed. But a caravan town's income is not chiefly farming.

It sold fodder, water and stabling, which the previous lesson showed no traveller could do without. It sold fresh animals, and bought exhausted ones cheap, which is a business with a very good margin. It sold lodging, food and repairs. It sold credit: the Turfan and Dunhuang documents are full of loans of grain, cloth and coin, often at interest rates that look startling until you remember the default risk of lending to a man about to walk into a desert. It sold information, on which passes were open and which lord was demanding what. And it sold, or extorted, protection and passage, which is the largest item of all and the one that appears least often in the accounts.

The market dues are what turned a village into a state. A town of a few thousand people sitting where all traffic must pass, selling necessities to people with no alternative, collects a tax on the entire commerce of a continent, and this is why so many Tarim oases were independent kingdoms with royal titles, palaces and diplomatic correspondence quite out of proportion to their size.

Kingdoms of fourteen thousand people

The Hanshu gives figures for each of these kingdoms, and they repay a close look, both for what they say and for how far they can be trusted. Shanshan, the kingdom whose towns included Niya and Loulan, is given 1,570 households, 14,100 persons and 2,912 men able to bear arms. Khotan is given 3,300 households and 19,300 persons with 2,400 able to bear arms. Kucha, the largest in the basin, is given 6,970 households and 81,317 persons with 21,076 able to bear arms.

Example. Check those three sets of figures against each other for internal consistency, and say what you conclude.

Compute the household size and the fighting share for each. Shanshan gives $14\{,\}100/1\{,\}570=9.0$ persons per household and $2\{,\}912/14\{,\}100=20.7$ per cent able to bear arms. Khotan gives $19\{,\}300/3\{,\}300=5.9$ persons per household and $2\{,\}400/19\{,\}300=12.4$ per cent. Kucha gives $81\{,\}317/6\{,\}970=11.7$ per household and $21\{,\}076/81\{,\}317=25.9$ per cent. Household size varies by a factor of two and the fighting share by more than two, and no plausible demography does that between neighbouring towns with similar economies. So the categories are not being counted the same way in each entry, which is what you would expect of figures assembled from different informants at different dates for different purposes: some may count men of military age, others men actually enrolled, others all adult males. The right conclusion is not that the numbers are worthless but that they are order-of-magnitude information. Shanshan had thousands of people and Kucha tens of thousands, and the precision of 81,317 is spurious.

Now you. A modern account writes that "the population of Kucha was 81,317". What is wrong with the sentence, and how would you write it?

The size matters for the politics. A kingdom that can field 2,912 men cannot resist a Han expeditionary force or a Xiongnu raid, and knows it. The consequence, visible throughout the Chinese sources, is a foreign policy of prompt and reversible submission: send a prince as a hostage to whoever is currently strongest, accept a seal and a title, pay, and switch when the balance moves. Han officials wrote about the faithlessness of the Western Regions as a moral defect. It was the only survival strategy available to a town of fourteen thousand people sitting on the only water for three hundred kilometres.

The building that follows from the pace

Between the towns, the same logic produced a piece of architecture. If a caravan makes about 30 km a day, then shelter is useful at roughly 30 km intervals, and that is very close to the observed spacing of the caravanserais built across the Islamic world and Anatolia: a walled courtyard with a single defensible gate, animals and goods stalled below, men above, water in the yard.

The form is a direct expression of the constraints. One gate, because the cargo inside is worth more than the building. A courtyard, because animals must be unloaded, watered and kept together. And the spacing is set by the camel, not by the terrain or the ruler: the Seljuk sultans of Anatolia in the thirteenth century built a chain of hans at roughly a day's march apart, and the Safavid state repeated the exercise across Iran in the seventeenth. Rulers built them because a caravan that reaches shelter each night is a caravan that arrives, and a caravan that arrives pays dues.

Example. Caravanserais on the great routes sit roughly 30 km apart, and rulers built them at their own expense and often charged nothing for a night's stay. Why would a state pay for something it does not charge for?

Because the return arrives as tax rather than as rent. A caravan that reaches a defensible walled compound each night is a caravan that arrives with its cargo, and a cargo that arrives pays market dues, customs and the multiplier of local trade, all of which the ruler collects. Charging for the shelter itself would collect a trivial sum and would divert traffic to a rival's road, whereas free shelter attracts traffic away from competitors, which is the real prize when routes have alternatives. There is also a reason that has nothing to do with revenue: a chain of garrisoned buildings a day's march apart is a military communication system and a way of holding a road, and it doubles as a display of the ruler's reach in country where his authority is otherwise abstract. This is the same logic behind free ports and state-funded roads generally, and it is worth noticing that the private version, an inn charging for beds, coexisted with it everywhere and served the traffic the state was not courting.

Now you. Why did so many oasis rulers, whose kingdoms could not have resisted anyone, nonetheless mint their own coins?

When the water moved

Oases fail. Niya was abandoned around 400 CE, with houses left standing, documents left in place and orchards left to die, which is the pattern of a slow withdrawal rather than a sack. Loulan, on the edge of the Lop Nur, was abandoned in the fourth century as well. Sven Hedin, who found Loulan in 1900, proposed that the Lop Nur was a wandering lake that shifted position over centuries as sediment blocked one channel and opened another, and that the town died when its water went elsewhere.

The honest position is that the cause is not settled and probably was not single. The rivers of the southern Tarim demonstrably shift, and the desert is demonstrably advancing at some sites. But the fourth century is also when Chinese garrisons and the subsidies that came with them were withdrawn from the region during the collapse of the Western Jin, which is precisely the moment when a marginal irrigated settlement loses both the labour and the money that keep its channels open. Climate, hydrology and politics all point the same way at the same time, and no evidence yet separates them cleanly. Anyone who tells you an oasis died of climate change is choosing among live hypotheses.

Further west the same logic operated on a grander scale. Merv, on the Murghab river in what is now Turkmenistan, was one of the largest cities on earth in the twelfth century, fed by a canal network drawing off the river and administered by a dedicated official with, according to the geographers, thousands of workers under him. Its walls enclose several hundred hectares, and it is the clearest case in the subject of an enormous city that exists only because a river was engineered. A later lesson returns to what happened to it in 1221.

The model of the last two lessons is now complete: a physical constraint, a chain of supply points, a set of small polities living off passage dues. It predicts which goods could cross and which could not. The next lesson tests that prediction against what was actually found in the ground and written in the market lists, and the test half succeeds and half fails, in a way that redirects the whole subject.

The last two lessons built a model of what could cross Asia overland and never checked it against anything found in the ground, which is what this lesson does.

The model says that a good had to be worth more than about two labour-days per kilogram to survive the journey, so the long-distance cargo should be dense in value, light, durable and non-perishable, and everything else should move only locally. The test has two parts, and they give different answers. On composition the model passes comfortably. On scale it fails, and the failure is the most important finding in the modern study of the subject.

Silk, going west, in the ground

Start with the commodity the whole thing is named after. The best physical evidence is not Chinese and not literary: it is a set of textiles excavated from the tower tombs of Palmyra in Syria in the 1930s and published by Rodolphe Pfister between 1934 and 1940. Palmyra is a caravan city in the Syrian desert, its wealth built on the run between the Euphrates and the Mediterranean, and its tombs are dry.

Among the fragments are silks whose weave structure is Han Chinese, patterned damasks of a type made in China and not, at that date, anywhere west of it, one of them carrying Chinese characters. That identification is technical rather than interpretive: the loom setup needed to produce a warp-faced compound weave of that kind is diagnostic, in the way a signature is. So Chinese silk was in the Syrian desert in the first three centuries CE, and it is there as cloth in a grave rather than as a claim in a text.

The same textiles carry a second finding. Pliny, in the Natural History, says that the cloth of the Seres was taken apart and rewoven in the Roman world, unravelled and combined with linen thread to make something lighter and more transparent than the original. Some Palmyra fragments show exactly that: Chinese thread in a non-Chinese weave. The cargo was not simply consumed at the far end. It was raw material for a Syrian industry, which is a considerably more interesting fact about the trade than the traditional image of a Roman matron in a Chinese robe.

Glass, going east

The traffic in the other direction is best seen in glass, and for the same reason. Chinese glassmaking existed but was different in composition and inferior in clarity for large vessels, so Roman and later Sassanian glass is identifiable in Chinese and Korean contexts by chemistry as well as by form.

The tomb of Feng Sufu, a nobleman of the Northern Yan who died in 415 CE, excavated in Liaoning in northeastern China, contained glass vessels of Roman or Near Eastern manufacture. The royal tombs at Gyeongju in Korea, from the fifth and sixth centuries, have produced Roman glass cups that travelled the full width of the continent and then across a sea. And the Shosoin repository at Nara in Japan, sealed in the eighth century, still holds a Sassanian cut-glass bowl.

Around those two headline commodities sits a longer list that the model also predicts: jade from the Khotan riverbeds, which had been moving east since before 1200 BCE and is found in quantity in the tomb of Fu Hao at Anyang; Mediterranean coral, prized in both India and China; aromatics, frankincense and myrrh from Arabia, camphor from Southeast Asia, musk from the Tibetan plateau; gemstones; medicinal drugs; fine metalwork; and, appallingly, people.

Example. A Roman glass bowl is excavated from a fifth century Chinese tomb. A guidebook says it proves the existence of a flourishing trade route between Rome and China. What does the object actually establish?

It establishes that the object moved, which is much less than it sounds. A single artefact is consistent with at least five histories: it was traded through many hands over decades, it was a diplomatic gift, it was loot, it was carried by a migrant or an envoy, or it was an heirloom already two centuries old when buried. Nothing about the bowl distinguishes these, and the deposition date is an upper bound on manufacture rather than a date of transfer. To get from objects to a trade route you need a pattern rather than an instance: many examples, a distribution that thins with distance in a regular way, evidence of repeat supply rather than one arrival, and ideally something at the other end going the other way. Roman glass in East Asia is in fact a reasonable case by those standards, since there are dozens of finds across several centuries, but the guidebook sentence would be wrong for any single object, and single objects are what such sentences usually rest on.

Now you. Roman coins, mostly gold and silver of Augustus and Tiberius, have been found in more than a hundred hoards in southern India, and hardly any in the Tarim Basin. Both regions traded with the Mediterranean world. Explain the difference.

People as cargo

One category on that list needs saying plainly rather than folding into a phrase about goods, because it was among the most valuable things moved along these routes and it is routinely omitted from accounts of them.

The Turfan documents include sale contracts for human beings, drawn in the same form as contracts for camels and land, with warranties of title and provisions for guarantors. Sogdian merchants dealt in slaves as part of ordinary business, and the practice is attested in the Ancient Letters and in Chinese records of the trade. The direction of the traffic was mostly from the steppe and the frontier zones into the settled empires at either end.

The largest and best documented version ran west. From the ninth century the Islamic world imported Turkic slaves from Central Asia in enormous numbers, and the Samanid state at Bukhara made the trade a major source of revenue, taxing every slave sold south. What made this different from slavery elsewhere is what the slaves were bought for: Abbasid and later rulers built their armies out of them, the ghulam or mamluk system, on the reasoning that a soldier with no local kin was loyal only to his master. The reasoning failed in the most spectacular way possible. Turkic military slaves and their descendants ended up commanding the armies, then appointing the caliphs, then ruling in their own right, from the Ghaznavids to the Mamluk sultanate of Egypt, which governed from 1250 to 1517. A commodity moved along these routes and became the ruling class at the far end.

A second stream ran from eastern and central Europe through the Volga and the Black Sea into the same markets, and its scale is one reason the Radhanite and Rus trade networks existed at all. The freight model of the third lesson applies without modification, and the reason is bleak: a human being is high in value, walks without needing to be carried, and can be sold anywhere.

The list that shows what a market really sold

Now the second half of the test, which is where the picture changes. The Turfan documents include a fragmentary schedule of official market prices from Xizhou, the Tang prefecture at Turfan, dated 743. It lists something like 350 commodities, most of them in three grades, with a price for each.

Its very existence is informative: a government thought it worth fixing and recording prices for hundreds of goods in a frontier market, which implies both a functioning market and an administration that intended to tax it. But the contents are what matter. The list is dominated by ordinary things. Grain, flour, several grades of local cloth, felt, hides, iron goods, dyestuffs, livestock, and yes, silk in several grades, along with a modest number of imports. It is the price list of a provincial market town in an agricultural oasis, not the tariff of an international emporium.

The other Turfan and Dunhuang documents point the same way. Loan contracts are for a few bolts of cloth or a few bushels of grain. Sale contracts cover camels, donkeys, land, and human beings. Travel permits, the Tang guosuo, record a man, his servant, and four animals going to the next prefecture. There is very little in this enormous documentary base that looks like intercontinental commerce, and a great deal that looks like a farmer borrowing seed.

Example. The 743 list gives most commodities in three grades, upper, middle and lower, each with its own price. What does that detail tell you?

More than it looks. Grading implies that the goods are heterogeneous and that quality disputes were common enough to need an official answer, which is what you would expect of hand-made cloth, livestock and grain of variable condition. It implies a market with enough repeat trade to make standard categories worth defining, since a one-off exchange would simply be haggled. It implies literate officials present at transactions, because a grade is useless unless somebody assigns it. And it tells you what the price list was for: not to inform buyers, who could see the goods, but to give the state a defensible basis for assessing taxes in kind, valuing government purchases and settling disputes. Read that way the document is a fiscal instrument rather than a commercial one, which is a good general suspicion to hold about any price list surviving from a pre-modern administration.

Now you. Hansen's argument rests largely on Turfan and Dunhuang. Name the strongest objection to generalising from them, and say whether it defeats her.

Valerie Hansen's The Silk Road: A New History (2012) made this the centre of the argument. Her conclusion, stated bluntly, is that the Silk Road was not a great artery of commerce but a network of short local trade routes along which small quantities of goods moved intermittently, and that the largest single mover of goods was not private trade at all.

How small is small

It is worth making the scale concrete, because "smaller than we thought" is the sort of phrase that slides past without registering.

Example. In 773 the Uighur khaganate sent 10,000 horses to the Tang court at the standing rate of 40 bolts of silk each. Convert that into caravans, taking a bolt at 0.35 kg and a camel-load at 150 kg.

The payment is $10\{,\}000\times 40=400\{,\}000$ bolts. At 0.35 kg a bolt that is 140,000 kg, or 140 tonnes, which is $140\{,\}000/150=933$ camel-loads. A large caravan runs to perhaps a hundred camels, so the entire silk payment for a year's horse purchase, a transaction big enough that Tang officials wrote about it as a national burden, is about nine or ten caravans. Spread over a year, that is one caravan every five or six weeks. The single largest silk transfer in the region is, physically, a trickle. The reason it mattered so much to the Tang treasury is that silk was expensive, not that it was bulky, and the two are constantly confused.

Now you. Suppose you wanted to measure the volume of the overland trade rather than argue about it. What evidence would you look for, and why is so little of it available?

Where the test leaves the model

The composition test passed. Everything found moving long distances overland is high in value per kilogram, and there is no counterexample: no case anywhere of grain, timber, ordinary pottery or cheap cloth crossing Asia by land as commerce. The physical model predicted the cargo list correctly from a camel's appetite.

The scale test failed, and in the direction the model could not have anticipated. The freight arithmetic sets a ceiling on what can move; it says nothing about whether anyone bothers. What the documents show is that the ceiling was rarely approached, that most recorded exchange was between an oasis and its neighbours, and that intercontinental commerce was episodic.

Two objections should be lodged before this becomes an orthodoxy in your head. The first is the survivorship problem from the second lesson: the documents that survive come from a handful of oasis towns, and the great trading centres, Samarkand, Merv, Balkh, Chang'an, have left nothing comparable, so we may be measuring the commerce of provincial villages and calling it the commerce of a continent. The second is that Hansen's argument, correctly, concerns quantity of goods, and quantity of goods is not the same as significance. Something moved along these routes that changed the history of half the world, and it did not weigh anything: Buddhism, Manichaeism, Christianity, Islam, paper, sugar, algebra and the plague were all carried by a traffic too small to show up in a freight statistic.

That leaves a specific puzzle, and the next lesson answers it. If private long-distance commerce was thin, and yet enormous quantities of silk are attested in Central Asia, in tombs, in payments, in temple inventories and in tax registers, then somebody other than merchants was moving it. The largest supplier of silk to Central Asia was the Chinese state, paying its own soldiers and buying horses it could not breed, and once that is understood the whole economy of the region looks different.

The previous lesson ended with a puzzle: private long-distance trade looks thin in the documents, and yet Central Asia is full of Chinese silk, so somebody other than merchants was moving it.

The answer changes how the whole subject looks. Silk was not only a luxury fabric. In China it was money, in the strict sense of a standardised unit that things were priced in and debts were settled with, and the largest single holder of it was the state, which took it in tax and had to spend it somewhere. Where it spent much of it was the western frontier.

What money has to be

A commodity works as money if it is durable, portable, divisible, hard to counterfeit, widely wanted, and above all standardised, so that a quantity of it can be named in a contract without inspection. Cattle fail on divisibility, grain on durability, and most cloth on standardisation, since a length of homespun is worth whatever the buyer thinks of it.

Chinese silk was engineered past that last obstacle by the tax system. Tang law defined the bolt, pi, as 40 chi long by 1.8 chi wide. At a Tang chi of about 30 cm that is 12 m by 0.54 m, an area of 6.5 m², and at the weight of a plain tabby weave a bolt comes out at roughly a third of a kilogram. Every adult male taxpayer owed cloth to the state annually under the zu yong diao system, grain and labour service and cloth being the three components, and cloth delivered in payment of tax had to meet the statutory dimensions or be refused. A standard enforced on millions of households every year produces something a merchant in Kucha can accept sight unseen from a stranger.

Add the other property that made it necessary. China ran a chronic shortage of copper coin. Minting never kept pace with the economy, coin was heavy relative to its value, and it circulated unevenly, so large payments in cash were physically awkward and often impossible on the frontier. A currency that is also a valuable export commodity, and that weighs a third of a kilogram per unit rather than several kilograms, solves a real problem.

Example. Test the bolt against the properties above, and identify the one on which it is weakest.

It is durable, since silk stored dry lasts for centuries, and the Palmyra and Astana finds prove it. It is portable at an extreme: a camel-load of 150 kg is about 430 bolts, so a fortune moves on one animal. It is divisible in a rough way, since cloth can be cut, though a cut bolt is worth less than its share of a whole one, which is a genuine defect. It is hard to fake in the sense that weave quality is visible to anyone in the trade. It is universally wanted from the Pacific to the Mediterranean. Its weakness is supply control: unlike a coinage, its quantity is set by how much silk the empire's households happen to weave and how much the state chooses to take in tax, so it cannot be managed, and its value against grain and coin drifted considerably across the Tang. It is commodity money with all the strengths and the one central weakness of the type.

Now you. The Turfan and Dunhuang documents price small purchases in copper coin, medium ones in grain, and large ones in bolts of silk. Why would a single market use three currencies at once, and what does that tell you about the economy?

What else was circulating

Silk was not alone, and the mixture in circulation is itself evidence about how the region worked.

Sasanian silver drachms, struck in Persia, circulated widely in Central Asia and turn up in hoards and in Chinese tombs, valued as bullion of known fineness rather than as the coin of any authority the holder recognised. Byzantine gold solidi reached China too, and here the finds are unusually informative: a substantial proportion of the pieces recovered from Chinese burials are imitations or thin one-sided copies, made to be buried rather than spent, which tells you the object had become a prestige token whose function was symbolic by the time it arrived. Chinese copper coin circulated in the Tarim wherever Chinese administration reached and thinned out sharply beyond it.

The Tang eventually produced an instrument that points at where all this was heading. Around the beginning of the ninth century merchants and provincial offices began using feiqian, flying cash: a merchant deposited coin in the capital, received a certificate, travelled, and drew the equivalent at the other end, so that the money made the journey and the metal did not. It is a draft rather than a banknote, since it represents a specific deposit, and it exists for exactly the reason this lesson has been describing, that moving value physically across a continent is expensive and dangerous. The Song turned the same principle into printed paper currency, and it was Yuan paper money that astonished Marco Polo enough for his description of it to become, six centuries later, one of the arguments that he had really been there.

What forty bolts bought

Two documents from Turfan put the currency to work, and they sit uncomfortably beside one another.

A contract of 731 records the sale of a Sogdian slave woman in the Turfan market. The price is 40 bolts of silk. The document is a routine commercial instrument: it names the parties, warrants that the seller has good title, provides for a guarantor, and follows the standard form of a sale of livestock or land, which is how the law classified what was being sold.

And 40 bolts of silk was also, in the same century, the standing rate for one horse from the Uighurs.

The equivalence is worth sitting with, because it is not a rhetorical flourish but a price. It also demonstrates the point of the section: bolts function as a genuine unit of account across categories of goods that have nothing else in common, which is what money does and what a mere luxury commodity does not.

At the labour valuation used earlier in this course, roughly fifteen labour-days to produce a bolt, 40 bolts is 600 labour-days, about two years of a working life. That is the order of magnitude of a horse, a person and a small farm all at once, and it tells you both how expensive horses were and how cheaply a human being could be bought at the edge of an empire.

The fiscal engine

Now the argument that reorganises the subject. The Tang held the Tarim Basin through the Anxi Protectorate, headquartered at Kucha, with garrisons at the Four Garrisons of Kucha, Khotan, Kashgar and Karashahr, and a standing force usually put at about 24,000 men. Those men had to be paid, fed and clothed at a distance of nearly four thousand kilometres from the capital, in a region that produced very little the army needed and no coin at all.

They were paid substantially in cloth. The Turfan documents show the machinery: government offices buying grain, hiring animals, settling accounts, issuing travel permits, all in a mixture of coin, grain and bolts, and the bolts arrive from the east because the state's tax system delivers them there.

Example. Suppose each of 24,000 soldiers received the equivalent of a dozen bolts a year in cloth, as pay and issue combined. How much silk is that, physically, and how does it compare with what private trade could plausibly have carried?

That is $24\{,\}000\times 12=288\{,\}000$ bolts, or $288\{,\}000\times 0.35=100\{,\}800$ kg, about 101 tonnes, which is $100\{,\}800/150=672$ camel-loads a year. Set that beside the private trade visible in the same documents, which consists of individuals moving with three or four animals and loans measured in single bolts. The state was putting several hundred camel-loads of currency-grade silk into the Tarim Basin every year as a routine administrative operation, and doing it whether or not any merchant sold anything. Whatever the private trade was, this was almost certainly larger, and unlike the private trade it is attested by the documents rather than inferred from luxury goods in graves. The assumption of twelve bolts is a guess and the conclusion is not sensitive to it: at six bolts a man the flow is still 336 camel-loads.

Now you. If the fiscal engine account is right, what should have happened to the oasis economies when the Tang lost control of the corridor after the An Lushan rebellion of 755, and is that a prediction you could test?

Horses, and why China had to buy them

The other half of the fiscal engine is what the silk was for, and it is the most durable structural fact in the whole of Chinese frontier history.

China could not breed cavalry horses in the numbers it needed. The reason is agricultural: horses of military quality need extensive pasture, and Chinese land under the plough is worth far more in grain than in grass, so a horse raised in China is competing with wheat and losing. Add the disease environment of the wet south and the loss of the northern pastures whenever the frontier moved, and the arithmetic never improved. The one place with limitless pasture and expert horsemen was the steppe, occupied by the very people the cavalry was needed against.

So the pattern repeats for two thousand years: China buys horses from its enemies and pays in silk. The Han bought peace and horses with the heqin arrangement, and the Hanshu's record of the annual gifts to the Xiongnu, tabulated by Yu Ying-shih in Trade and Expansion in Han China (1967), shows the payment rising from around 8,000 rolls of silk in 51 BCE to around 30,000 in 1 BCE, a factor of nearly four in fifty years. Emperor Wu tried the alternative and sent armies to Ferghana in 104 and 102 BCE for the famous "heavenly horses", at a cost the Shiji makes clear was ruinous and a return of a few thousand animals, most of which died on the road. Buying proved cheaper than conquering, which is a conclusion the Tang reached independently.

Under the Tang the buyer's position got worse. The Uighurs had helped suppress the An Lushan rebellion and were owed for it, and in 773 they delivered 10,000 horses and demanded payment at 40 bolts a head, 400,000 bolts. The 754 census counted about 9.07 million registered households; if each owed something on the order of half a bolt a year, the empire's whole annual silk revenue was a few million bolts, so a single year's horse purchase absorbed something like a tenth of it. Tang officials wrote about it as extortion, and in a sense it was, but the court paid because the alternative was to face Turkic cavalry without cavalry of its own.

Example. The Han payments to the Xiongnu rose from around 8,000 rolls of silk in 51 BCE to around 30,000 in 1 BCE. What is the annual rate of increase, and what does the shape of the series suggest?

The ratio is $30000/8000=3.75$ over fifty years, so the annual growth rate is ${3.75}^{1/50}=1.0268$, about 2.7 per cent a year. That is a steady compounding escalation rather than a step change, which is the signature of a recurring negotiation in which one party can raise its demand a little at each renewal and the other pays rather than fight. Two readings are available and both are probably right in part. From the Han side, this is the cost of a policy that worked: fifty years of relative peace on the northern frontier bought at a price that grew slowly and predictably, which is cheap compared with the campaigns of Emperor Wu. From the Xiongnu side, it is a tribute stream extracted by a confederation whose leaders needed silk to distribute to their own subordinates, which means the demand grew with their internal political needs rather than with anything the Han did. The number to hold on to is the rate: a 2.7 per cent annual escalation is invisible in any one year and quadruples the bill in a lifetime.

Now you. Chinese officials repeatedly described these payments as tribute received from barbarians, while the recipients understood them as tribute paid by China. Which was it, and does the question have an answer?

What this does to the picture

Three consequences follow, and they should be held together.

The trade of Central Asia had a state at the middle of it. The silk that reached Sogdian merchants, Uighur khagans and eventually Persian and Byzantine markets was largely silk that had entered circulation as Chinese tax revenue paid out to soldiers and allies. Private commerce was real and was substantially downstream of a fiscal flow.

The relationship was not one-sided in the way the phrase "Chinese silk exports" suggests. China was the party with the structural weakness, buying a strategic necessity it could not produce from suppliers who could and did raise the price.

And the currency function explains the survival pattern of the evidence. Silk turns up in Central Asian graves, temple inventories and contracts not because Central Asians were unusually fond of Chinese fabric but because that is what money looked like there.

The limits are worth stating. This picture is built on two document hoards, Turfan and Dunhuang, both in the zone the Chinese state administered, so it may over-represent the state precisely because the state generated paper. The strongest form of the argument, that private long-distance trade barely mattered, goes further than the evidence can carry, and Hansen's critics are right that a trade conducted by illiterate carriers, settled in cash, and taxed by nobody would be invisible in exactly this way.

Still, somebody had to carry the goods, whoever was paying. The next lesson is about the people who did it: a single Iranian-speaking people from the valleys around Samarkand who ran the relay for six hundred years, and whose own letters, left in a mailbag near Dunhuang, show how a trading diaspora actually worked.

Somebody had to carry the goods, whoever was paying for them, and for roughly six hundred years the people who did it across the middle of Asia were the Sogdians.

The previous lesson argued that much of the silk moving into Central Asia entered circulation as Chinese state spending. That says who supplied the currency and says nothing about who did the work. This lesson is about the merchants themselves, and it is the one place in the subject where we can watch a commercial system from the inside, because a bag of their letters was abandoned in a watchtower and never delivered.

A people without a state

Sogdiana was the land between the Oxus and the Jaxartes, the Amu Darya and Syr Darya, centred on the Zarafshan valley in what is now Uzbekistan and Tajikistan. Its cities were Samarkand, Bukhara, Panjikent and Kesh, and it was never once, in its entire history, a unified independent state.

It was a province of the Achaemenid Persians, then taken by Alexander, then held in turn or in part by Seleucids, Kushans, Hephthalites, the Turkic khaganates, and finally the Arabs. Between and beneath those overlords, the Sogdian cities ran themselves as an aristocratic patchwork of walled towns and fortified estates, each with its own ruler, its own coinage, and no common authority.

That absence is not a footnote. A people with a small, fertile, densely irrigated homeland, a position exactly halfway between China, India, Persia and the steppe, and no empire to defend or feed, is a people with every incentive to become intermediaries and no capacity to become anything else. What Sogdiana had to sell was position and organisation.

How we know where they went

The single most useful piece of evidence about the diaspora is a naming convention. Sogdians settling in China took Chinese surnames drawn from their city of origin, in a fixed and well-attested scheme: Kang for Samarkand, An for Bukhara, Shi for Kesh or Tashkent, Mi for Maimargh, He for Kushaniya, Cao for Ishtikhan. Chinese sources call them the Nine Surnames of Zhaowu.

The consequence for a historian is remarkable. Chinese administrative records, tomb inscriptions and census fragments are full of ordinary Chinese names, and a subset of them announce that their bearer's family came from a specific Central Asian city. It becomes possible to map a diaspora through the bureaucracy of the country it settled in, which is not possible for any other trading people of the period.

Example. A Tang register from a garrison town lists a militia unit whose members include several men surnamed Kang, two surnamed An and one surnamed Shi. What can you conclude, and what can you not?

You can conclude that a Sogdian community was settled there rather than passing through, and that it had been there long enough to be enrolled in a Chinese institution and taxed like everyone else. Names in a register are evidence of residence and of legal integration, and militia service in particular means these men were treated as subjects rather than as foreign visitors. What you cannot conclude is that any of them was a merchant. The surname records an ancestral city, not an occupation, and by the eighth century many families of Sogdian origin in China were soldiers, farmers, officials, grooms, musicians or entertainers, several generations from anyone who had traded anything. An Lushan, whose rebellion nearly destroyed the Tang in 755, carried the surname of Bukhara and was a frontier general. The convention tells you about origin and settlement, and inferring a caravan from a surname is exactly the overreach it invites.

Now you. The Chinese bureaucracy created an official post, the sabao, from the Sogdian term for a caravan leader, and made it a ranked position responsible for the resident foreign merchant community. What does the existence of that office tell you?

The mailbag

The five Sogdian Ancient Letters, found in a watchtower west of Dunhuang and dated to 313 or 314, are the earliest substantial evidence of how this trade was organised, and their value comes from the fact that they never arrived.

Letter II is from a merchant named Nanai-vandak, writing east to west to his employer or patron in Samarkand. It is a business report in the middle of a catastrophe. He describes the destruction in China, the emperor's flight from Luoyang because of famine, the burning of the palace, and the deaths by starvation of Indians and Sogdians in the city. Then, without pause, he turns to accounts: what is owed, who holds what, what he has sent, what should be done with a sum of money, and a request for instructions.

That mixture is the point. This is a firm with a principal at home and agents distributed across thousands of kilometres, sending consignments on credit, keeping accounts across years, and expecting written instructions to arrive from a city three months away. It is not adventurers with a camel. It is a business organisation, and one recognisably like the family firms of the medieval Mediterranean.

Letter III is not a business letter, and it is the reason the collection is famous outside the field. Miwnay, a Sogdian woman stranded in Dunhuang with her daughter Shayn, writes to her husband Nanai-dhat, who has left them without support. She has appealed to the local Sogdian community for help and been refused. She would rather be a dog's or a pig's wife, she writes, than his. A separate letter from the daughter to her mother's relatives asks them to intervene.

Both letters were in the same undelivered bag, and both stayed there for sixteen centuries. The system that moved silk across a continent also stranded a woman in a desert town on the far side of it, and the same community network that enforced commercial obligations declined to enforce a domestic one.

The problem a diaspora solves

Why should a trade of this kind be run by one ethnic group at all, rather than by whoever has capital? The answer is a problem in enforcement, and it can be made precise.

A merchant in Samarkand who wants goods sold in Chang'an must put them in the hands of an agent 3,500 km away, in a jurisdiction whose courts he cannot use, whose language he may not speak, and where his agent could simply sell the cargo and disappear. No contract helps, because no authority spans both ends.

Example. An agent holds a cargo worth 100 and earns a commission of 10 a year for continuing to work for the firm. Treat theft as taking the 100 now and losing the commission stream forever. At what discount rate does honesty stop paying, and what does the answer imply?

The value of the honest stream is $10/r$ for a discount rate $r$, so honesty beats theft when $10/r>100$, that is when $r<0.10$. At a 5 per cent discount rate the stream is worth 200 and the agent stays honest; at 10 per cent it is worth exactly 100 and he is indifferent; at 20 per cent it is worth 50 and stealing is clearly better. Since a merchant in a violent, plague-ridden, politically unstable world discounts the future heavily, and 10 or 20 per cent is if anything optimistic, the pure business relationship does not deter theft. Repeat dealing alone is not enough, which is the whole difficulty of long-distance trade before the modern state, and it means the mechanism that actually held the Sogdian network together must have added something to the agent's payoff calculation.

Now you. What could a diaspora community add that raises the cost of theft enough to close that gap?

The language that outlived the traders

The colonies ran in a chain: Sogdiana itself, then Semirechye and the Chu valley, then the Tarim oases, Turfan, Dunhuang, and on into China proper, with substantial communities at Chang'an and Luoyang. Étienne de la Vaissière's Sogdian Traders (2002, English translation 2005) assembled the chain from Chinese, Sogdian, Arabic and archaeological sources, and it is the standard reconstruction.

Along that chain Sogdian became the working language of the road, and this had a consequence no merchant intended. Sogdian was written in an alphabet descended from Aramaic. The Uighurs, whose khaganate dominated the eastern steppe from the eighth century and whose contact with Sogdian merchants and Manichaean missionaries was constant, adopted that alphabet to write Turkic. In 1204 Chinggis Khan had the Uighur script adopted to write Mongol, and in the sixteenth century the Manchus took the Mongol script to write Manchu, in which form it was the writing of the Qing empire into the twentieth century. The chancery script of the largest land empire in history is a direct descendant of the alphabet used by Samarkandi merchants to keep their accounts.

Example. The Ancient Letters were never delivered. What does an undelivered mailbag give a historian that a carefully preserved archive does not?

Freedom from selection. An archive exists because somebody decided its contents were worth keeping, and that somebody was usually an institution keeping records that served it: title deeds, tax registers, official correspondence. Everything else was thrown away, which is why so much of the surviving documentation of the ancient world is administrative and so little of it is ordinary. A mailbag is the opposite: it contains whatever happened to be travelling on one day, chosen by nobody, mixing a merchant's account of a national catastrophe with a woman's letter about her husband abandoning her. That is as close to a random sample of what people actually wrote as this subject ever gets. The catch, and it is a real one, is that a single day's sample from a single route is tiny, so the letters are unusually unbiased and unusually few, and any generalisation from five documents needs the caution the second lesson recommended.

Now you. Almost all the surviving Sogdian documents were found outside Sogdiana. Why, and what does that do to the picture we have of the Sogdians?

The end

The Arab conquest of Sogdiana, campaigned by Qutayba ibn Muslim from 705 and completed over the following decades, ended Sogdian political independence. One of the sharpest documentary windows on it is the archive of Divashtich, the last ruler of Panjikent, found on Mount Mugh in Tajikistan in the 1930s: letters, legal documents and accounts from a man losing a war, ending with his capture and execution around 722.

The commercial network outlasted the homeland by decades rather than centuries. The An Lushan rebellion of 755 made Sogdian identity dangerous in China, and many families concealed their origins. In Central Asia, conversion to Islam and Turkicisation absorbed the Sogdians into the populations that succeeded them, and their commercial role passed to Muslim merchants who used Persian and Arabic. The language itself did not quite die: Yaghnobi, spoken today by a few thousand people in a high valley in Tajikistan, is its direct descendant, the last living remnant of the tongue that once ran a continent's trade.

The Sogdians left one more thing behind, and it was not commercial. Their caravans carried missionaries and monks alongside the cargo, their colonies supported temples, and Sogdians appear in the record as translators of religious texts into Chinese in numbers far out of proportion to their population. The next lesson takes up the largest of those transmissions, Buddhism moving from India to China along the same roads, and asks the question the freight model cannot answer: why should a religion travel with a trade at all?

Chinese silk went west and Buddhism came east along the same roads, and the second of those is much harder to explain than the first, because most religions did not travel at all.

The previous lesson described the merchant network. This one asks what rode on it. The temptation is to treat religious transmission as automatic, an inevitable side effect of contact, but the comparison cases forbid that. Confucianism, the dominant intellectual system of the empire at one end of the route, went essentially nowhere. Daoism barely left China. Brahmanical Hinduism, whose homeland was the source of Buddhism itself, did not move overland into China in any comparable way. Something specific about Buddhism made it portable, and the specifics are institutional rather than spiritual.

What we are trying to explain

Fix the phenomenon first. The first securely dated Buddhist translation activity in China is that of An Shigao, a monk of Parthian origin, at Luoyang from 148 CE. Within two centuries there were monasteries across northern China, and by 401, when Kumarajiva was installed at Chang'an with state backing and a translation bureau, the Chinese state was funding the project directly. That is a foreign religion, in a foreign language, from a country most Chinese would never see, becoming a major institution of Chinese life over roughly two hundred and fifty years.

The route is not in doubt. The transmission ran through Gandhara and Bactria, across the Pamirs, along the Tarim rims through Kucha and Khotan, through Dunhuang, and down the Hexi corridor. Its physical remains are the cave complexes strung along exactly that line: Bamiyan, Kizil near Kucha, the Mogao caves at Dunhuang, Yungang, Longmen. It is the same line as the caravans, the same oases, the same passes.

Monasteries as infrastructure

The first mechanism is unromantic and probably decisive. A Buddhist monastery on a trade route provided precisely what a caravan needed.

It offered lodging and shelter, in a walled compound, at a fixed point, to travellers who were strangers. It offered storage for goods. It offered water and, at some sites, medical care. And Indian and Central Asian monasteries lent money, taking deposits as permanent endowments, the akshaya nivi, whose principal was preserved and whose interest funded lamps, repairs and offerings. A monastery is a durable institution holding capital in a world of unstable dynasties, which makes it a bank whether or not it wishes to be one.

Put that beside the previous lesson's enforcement problem. A merchant with goods to store, money to leave somewhere, and a need for shelter every few days in territory where he has no relatives is a merchant with a use for an institution that exists in every oasis, speaks a shared religious language, and cannot easily abscond because it owns immovable buildings. The monastery network solved logistics and trust simultaneously, and it was already spread along the route.

Donation as a doctrine merchants could use

The second mechanism is doctrinal, and it is what allowed the first one to be paid for. Buddhism offers merit, punya, accumulated by giving, and it accepts giving from anyone. A merchant with money and no priestly status can convert wealth into spiritual standing directly, by endowing a cave, a stupa, a lamp, a text or a meal for monks, and the transaction requires no ritual qualification, no birth, no purity, and no permission.

This is not a modern reconstruction; it is written on the walls. The cave complexes carry donor inscriptions and donor portraits by the thousand, naming the people who paid and what they paid for. At Mogao near Dunhuang there are 492 decorated caves cut over roughly a thousand years, from the traditional first excavation in 366 CE onwards, an average of one new decorated cave every two years, sustained across dynasties and changes of ruler. That is a capital programme, and it was funded by donors who wanted their names on it.

Compare what the competition offered a merchant. Confucian practice centres on family ritual, state ceremony and office holding, none of which a foreign trader can acquire and none of which travels. Brahmanical religion at that date tied ritual standing to birth and attached real purity costs to long travel, and it had no mechanism for converting an outsider. Buddhism was, in the specific sense that matters for transmission, the only universalist and monastic option on the route, and universalism plus monasticism is exactly the combination that turns trade contact into institutional spread.

Example. A merchant endows a cave at Dunhuang with his name and portrait in it. List what he receives, in this world and the next, and say which items the competing traditions could have matched.

He receives merit, which is the stated purpose and should not be treated cynically, since he almost certainly believed it. He receives standing in a community that extends the length of the route, which is worth money: a known donor is a known man in every monastery from Kucha to Chang'an, and that is credit and hospitality. He receives the goodwill of a local institution whose abbot deals with the local ruler, which is protection. He receives a permanent, public, dated record of his family's wealth in a place everyone important will see. And his descendants inherit all of it. Of these, the competing traditions could match almost none for a foreign merchant: Confucian standing runs through examination and office, closed to him; local Chinese cults are local; and no available system let a foreigner buy religious standing with cash and have it recognised two thousand kilometres away. The donation mechanism is not incidental to the transmission. It is the funding model.

Now you. The earliest translators of Buddhist texts into Chinese are An Shigao, a Parthian, Lokaksema, a Kushan, Kang Senghui, from a Sogdian family, and Dharmaraksha, of Yuezhi descent born at Dunhuang. Not one of them is Indian. What does that tell you about how the transmission worked?

The route in stone

The physical residue of all this is a line of excavated cliffs, and it is worth treating as evidence in its own right rather than as illustration.

At Bamiyan in central Afghanistan, two standing Buddhas of 55 m and 38 m were cut into a cliff face in the sixth and seventh centuries, at a pass on the route between Bactria and India, and destroyed by the Taliban in March 2001. At Kizil near Kucha, several hundred caves carry painted cycles in a style that fuses Indian, Iranian and local elements. At Mogao near Dunhuang the caves run from the fourth century to the fourteenth. Further east, at Yungang and Longmen inside China proper, the same tradition appears at imperial scale under the Northern Wei and the Tang.

Two things follow from the distribution. The first is that it maps the route: the complexes sit at passes, at oasis towns and at the entrances to the Chinese heartland, in the order a traveller would meet them, which is about as direct a demonstration of the transmission path as could be asked for. The second is economic. Cutting and painting a cave is a large, slow, skilled capital project, and a chain of them across two thousand kilometres represents wealth being converted into permanent religious infrastructure by people who had wealth to convert, which in an oasis town means merchants and officials. The art historian's evidence and the economic historian's evidence are the same evidence, and the donor inscriptions are where they meet.

Going the other way for the source

The relay produced its own correction. From the fourth century Chinese monks began making the journey in reverse to get texts and authority at first hand, and their travel accounts are among the best sources for the whole region.

Faxian left in 399 at about sixty years old, walked across the Taklamakan and over the Hindu Kush, spent years in India and Sri Lanka, and returned by sea in 414, writing a record of Buddhist kingdoms that is still the best single description of northern India in his day. Xuanzang left in 629 against an imperial ban on travel, spent sixteen years away, studied at the great monastic university of Nalanda, and returned in 645 with 657 Sanskrit texts. Yijing went by sea in the 670s and left an account of monastic practice in India and Sumatra.

Xuanzang's return shows what the Chinese state had come to think of all this. He left as a criminal and came back to an official reception, imperial patronage and a translation bureau.

Example. Xuanzang returned in 645 and died in 664, and in that time his bureau produced translations amounting to 1,335 fascicles. What follows from the rate?

The rate is $1335/19=70$ fascicles a year, more than one a week sustained for nineteen years. No individual, however gifted, translates philosophical Sanskrit at that rate, so the number is evidence about organisation rather than about Xuanzang. The Chinese translation bureaus were production lines with defined roles: someone recited the Sanskrit, someone rendered it orally, someone wrote it down, someone checked the meaning against the source, someone polished the Chinese style, and someone verified the finished text. Kumarajiva's operation at Chang'an from 401 worked similarly, with a large staff. So the correct picture of Buddhist transmission is not a lone monk with a bag of scrolls but a state-funded institution converting foreign texts into Chinese at industrial rates, and this in turn is why the transmission accelerated so sharply once emperors decided to fund it.

Now you. The first translations date from 148 CE. Is that a good estimate of when Buddhism reached China?

Example. The Mogao caves were cut over roughly a thousand years, from 366 to the fourteenth century, at an average of one decorated cave every two years. What kind of institution does that rate imply?

It implies continuity of a very unusual kind. One cave every two years, sustained for a millennium, is not a burst of piety but a permanent works programme, which requires that the following all persist across dynasties, conquests, changes of language and changes of ruling religion: a body of monks to commission and consecrate, a supply of donors with money, a resident workforce of stonecutters, plasterers and painters with the skills passed down, quarrying and pigment supply, and a legal or customary arrangement securing the site. When Dunhuang passed to Tibetan control in the 780s and later to local Chinese and then Uighur rulers, the cutting continued. That is the signature of an institution that had become independent of any particular state, funded by a local society that regarded endowment as normal. It is also, incidentally, why the caves are such good evidence: a continuous series lets you watch styles, donors and languages change against a fixed background.

Now you. Buddhist monasteries in Central Asia and China accumulated land, mills, oil presses and dependent labourers, and lent at interest. Is that a corruption of the religion or a condition of its survival?

What arrived was not what was sent

The final piece is honesty about the product. "Buddhism reached China" compresses six centuries of difficult, contested and sometimes badly executed translation.

The early translators had no Chinese vocabulary for Buddhist concepts and borrowed Daoist terms, a practice later called geyi, matching concepts. Dharma was rendered with dao. Nirvana was glossed with wuwei, the Daoist non-action. The result was intelligible and wrong, and it produced a first century of Chinese Buddhism that read as a variety of Daoism. Kumarajiva's bureau after 401 is the turning point, both because his Chinese was better and because he worked with a team that argued about meaning, and his versions of the Lotus and Diamond Sutras are the ones chanted in East Asia today.

What emerged was not Indian Buddhism relocated. Chinese Buddhism developed schools with no Indian counterpart, above all Pure Land, with its salvation by faith in Amitabha, and Chan, which became Zen in Japan. Both are legitimate developments and neither could have been predicted from the texts that arrived. Transmission along the Silk Road was never copying. It was translation in the full sense, and what came out the other end was a new thing built with imported parts.

The mechanism explained here, portable institutions, universal membership and a doctrine that lets outsiders buy in, was not unique to Buddhism, and three other faiths used the same roads with the same tools. Two of them reached China, acquired imperial patronage, and vanished; one arrived last and stayed. Comparing all four is the fastest way to see which parts of the mechanism actually mattered, and that is the next lesson.

Four other religions travelled the routes described in this course, and their different fates are the closest thing the subject has to a controlled experiment on why anything spreads.

The previous lesson gave an account of Buddhism's transmission built on institutions: monasteries as infrastructure, donation as a way for outsiders to buy standing, and eventually state funding. If that account is right it should predict the other cases, so this lesson runs it against them. It succeeds in three and fails informatively in the fourth, where the missing ingredient turns out to be law.

Zoroastrianism: carried everywhere, adopted nowhere

Zoroastrianism was the religion of the Persian empires and of the Sogdians themselves, which means it travelled with the most successful merchant network on the road for six hundred years and gained essentially no converts along it.

The evidence that it travelled is now very good, and much of it is recent. Two tombs excavated in Xi'an, that of An Jia in 2000 and that of a Sogdian called Wirkak, whose Chinese name was Shi Jun, in 2003, both of men who died in 579, carry carved and painted funerary furniture with unmistakably Zoroastrian imagery: priests with mouth coverings tending fire altars, and the soul's crossing of the bridge of judgment. These are Sogdian community leaders in the Chinese capital, buried according to their ancestral religion, in a hybrid of Chinese and Iranian funerary form.

And there the religion stopped. It supported the diaspora and converted no Chinese. The reasons fit the previous lesson's framework exactly and in reverse. Priesthood was hereditary, so an outsider could not enter the institution. Ritual purity requirements were demanding and territorially specific. Fire temples were tied to place and community rather than run as a network of hostels open to strangers. And there was no doctrine of merit purchasable by a donor from outside the community. Zoroastrianism was the religion of a people, and religions of a people move where the people move and no further.

Manichaeism: designed to travel, and it did

Mani, born in 216 CE in Mesopotamia and killed in prison under the Sasanian king Bahram I around 274, did something no earlier founder had done. He wrote the canon himself rather than leaving it to disciples, and he designed the religion for translation, deliberately synthesising elements of Zoroastrian, Christian and Buddhist material and instructing that his books be rendered into other languages.

It worked. Manichaean texts survive in Middle Persian, Parthian, Sogdian, Uighur, Chinese, Coptic and Latin, and the German expeditions to Turfan between 1902 and 1914 recovered Manichaean manuscripts and painted book pages of extraordinary quality from Kocho. Merchants carried it: Sogdians were the principal transmitters east, and the religion appears in China from the late seventh century.

Its great moment came in 762, when the Uighur khagan Bögü, campaigning in China to help suppress the An Lushan rebellion, encountered Manichaean clergy at Luoyang and adopted the religion. Manichaeism thereby became the state religion of a major steppe power, 488 years after its founder was executed as a heretic, and the Uighurs then leaned on the Tang, whom they had rescued, to tolerate Manichaean temples in Chinese cities.

Then the patron fell. The Kirghiz destroyed the Uighur khaganate in 840, and in 843 and 845 the Tang court, under Emperor Wuzong, moved against foreign religions and against Buddhism above all. The traditional figures from the 845 edict are 4,600 monasteries suppressed and 260,500 monks and nuns returned to lay life. Buddhism was too large and too deeply rooted to be destroyed and recovered. Manichaeism, whose Chinese existence rested on Uighur protection, did not, surviving only as a hunted southern Chinese sect for a few more centuries.

Example. The 845 figures are 4,600 monasteries and 260,500 monks and nuns. Check them for plausibility and say what they establish.

Divide: $260\{,\}500/4\{,\}600=57$ religious per monastery, which is entirely reasonable for a mixture of large urban establishments and small rural ones, so the two figures are at least consistent with each other. Against a Tang population of the order of 50 million, 260,500 is about 0.5 per cent of everyone, a very large ecclesiastical establishment but not an impossible one. What the numbers establish is a lower bound on how far Buddhism had penetrated Chinese society by the ninth century: not a foreign cult with a following but an institution large enough that suppressing it was a fiscal policy, and indeed the confiscation of monastic land and metal was among the court's motives. What they do not establish is accuracy. They come from the state that carried out the suppression and had an interest in the size of its own achievement, and no independent count exists.

Now you. Manichaeism collapsed in the east within a few decades of losing Uighur protection, while Buddhism survived a much more determined persecution in the same years. What does the comparison show, and what does it not?

The Church of the East: an empire of bishops with few subjects

The Christianity that travelled these roads was not the Christianity of Rome or Constantinople. It was the Church of the East, centred on Mesopotamia, using Syriac as its liturgical language, and separated from the imperial churches after the christological disputes of the fifth century, for which it is often labelled Nestorian.

Being outside the Roman empire turned out to be an advantage. The church was free to expand eastwards into Sasanian and later Islamic territory, and it did so along the trade routes with a proper hierarchy: metropolitan sees at Merv, Herat and Samarkand, bishops as far as Kashgar and eventually China. Its missionary Alopen reached Chang'an in 635 and obtained an edict of toleration in 638.

The most famous evidence is a stone. The Xi'an stele, erected in 781, records the arrival of the religion and its imperial reception, in Chinese with a long list of clergy names in Syriac down the sides. It was buried at some point and rediscovered in the 1620s, 842 years after it was cut, whereupon it caused a sensation among Jesuits in China who had no idea Christians had preceded them by a millennium.

Example. What does the stele establish, and what does a careless reader take from it that it does not support?

It establishes a great deal: that an organised church with a clergy, a hierarchy and Syriac liturgy existed in the Tang capital in 781, that it had imperial permission and grand enough patrons to commission a monument of that quality, that it dated its own arrival to 635, and, from the Syriac name list, that its personnel were substantially foreign. What it does not establish is size. A stele is an advertisement, commissioned by the community it commemorates, and it is designed to make a small well-connected group look like a movement. There is no independent evidence of large numbers of Chinese converts, the church is largely invisible in Chinese sources that are not about itself, and it disappeared from China proper after the ninth century suppressions with very little trace. The correct reading is a well-organised, well-connected minority religion of foreign merchants and clergy, which is exactly what Zoroastrianism also was, and the stele's grandeur is evidence of ambition rather than of scale.

Now you. Christian gravestones in Syriac script have been excavated in the Chu valley of Kyrgyzstan, dated by their own inscriptions to the 1330s. What does that tell you about the church's real geography?

Islam: the one that stayed

Arab armies took Sogdiana between about 705 and 715, and after the Tang defeat at Talas in 751 no Chinese army returned west of the Pamirs. But conquest is not conversion, and the interesting fact is how long the second took. Central Asia's population became substantially Muslim over the following three centuries, not the following three decades.

The turning point among the Turkic peoples came with the Qarakhanids, whose ruler Satuq Bughra Khan converted in the mid tenth century, and Ibn al-Athir reports the conversion of two hundred thousand tents of Turks around 960. Popular conversion followed through Sufi orders, above all the tradition associated with Ahmad Yasavi, who died at Turkistan around 1166 and whose shrine Timur later rebuilt on a colossal scale.

Now apply the framework. Islam had everything the previous lesson identified: universal membership open to anyone who says the words, a route for lay donation through waqf endowments, an institutional network of mosques and lodges along the roads, and after the conquests, state patronage on a scale nobody could match.

It also had something none of the others did, and it is the piece that makes the difference. It had portable law. The Islamic legal system supplied standard forms for commercial partnership, the mudaraba under which one party supplies capital and the other labour and the profit is split on agreed terms, instruments for transferring value at a distance, rules of evidence, and, above all, judges whose jurisdiction was recognised anywhere in the Muslim world. Recall the problem of the seventh lesson: an agent 3,500 km away cannot be sued, so trade must rely on kinship and community sanction, which cannot recruit outsiders. Islamic commercial law dissolves that constraint. A merchant in Bukhara and an agent in Cairo were under the same law, could contract in a form both understood, and could take a dispute to a court either would recognise.

That is why the Muslim merchant network eventually replaced the Sogdian one rather than competing with it: it offered the same trust guarantee without requiring shared ancestry, so it could absorb anyone, including the Sogdians themselves, who converted and disappeared into it.

Example. Return to the seventh lesson's calculation, where an agent holding a cargo worth 100 and earning 10 a year steals whenever his discount rate exceeds 10 per cent. How does Islamic commercial law change the payoff?

It adds a term the Sogdian solution could not: legal recovery from a stranger. Under a mudaraba the relationship is a recognised contract with defined shares, witnesses and a written instrument, and a judge in either city will hear a claim on it, so an agent who absconds faces not only the loss of future commission but a real probability of being pursued, of having assets seized, and of a judgment that follows him anywhere in the Muslim world. Suppose that probability is only one half, and the whole cargo is recovered when it applies: the expected gain from theft falls from 100 to 50, and honesty now wins at any discount rate below 20 per cent instead of 10. The decisive advantage is not the size of the improvement but its portability: the diaspora solution achieves a similar deterrent only among kin, while the legal solution works between two men who have never met and share nothing but a religion. That is why the Muslim commercial network could recruit from any population it converted, and it is a large part of why it displaced the Sogdian one rather than merely competing with it.

Now you. Islam arrived in Central Asia with armies in the eighth century and the population became substantially Muslim only across the tenth to twelfth. What does that gap tell you about conquest and conversion?

Judaism, and the exception on the steppe

Two others belong in the picture. Jewish merchants, the Radhanites described around 870 by the Abbasid postmaster Ibn Khordadbeh as speaking Arabic, Persian, Greek, Frankish, Andalusi and Slavic and travelling from the Frankish lands to China, are the clearest documented case of a trading diaspora spanning the whole system, and Jewish communities in China are attested at Kaifeng by a stele of 1489. Judaism spread as a people rather than as a mission, like Zoroastrianism, and its network was commercial rather than evangelical.

Which makes the Khazars a genuine anomaly. The Khazar khaganate, a Turkic power controlling the lower Volga and the north Caucasus and sitting astride the routes between the Islamic world, Byzantium and the northern forests, adopted Judaism as the religion of its ruling elite at some point in the eighth or ninth century. The sources are difficult, chiefly the Hebrew correspondence attributed to King Joseph and the Andalusi official Hasdai ibn Shaprut in the tenth century, plus Arabic geographers, and the depth of the conversion beyond the court is unclear. But the direction of choice is intelligible on this lesson's terms and is often read that way: a state wedged between a Christian empire and a Muslim caliphate, both of which treated conversion as a claim to overlordship, took the one prestigious literate monotheism that came with no imperial patron attached.

The Khazar case cuts against the framework in one way and supports it in another. It shows that a religion with no mission apparatus can nonetheless be adopted by a state, so patronage does not require a missionary. It also shows the limit of patronage without a mechanism for spreading among ordinary people: when the khaganate was destroyed by the Rus in the 960s, Khazar Judaism went with it, exactly as Uighur Manichaeism went with the khaganate in 840.

What the comparison says

Set the five side by side and the pattern is clear enough to state. A religion spreads along a trade route when it accepts outsiders without qualification of birth, gives them a way to convert money into standing, maintains an institution physically present along the route, and finds a state to fund it. Zoroastrianism and Judaism failed the first two by design and travelled without spreading. Buddhism, Manichaeism and Christianity had all four and spread. Islam had all four plus a legal system that solved the merchants' own central problem, and it is the only one of the five that still dominates the region.

The honest limit is that patronage is contingent. Bögü might not have met a Manichaean priest at Luoyang; Wuzong might not have needed monastic silver in 845. The framework explains which religions were capable of spreading, and a good deal of which ones actually did comes down to which ruler was persuaded in which decade.

Faiths moved as texts, and a text needs something to be written on. The next lesson takes up the hardest category of all to move, technique, using paper as its main case, and finds that an object, a recipe and a skill travel at three entirely different speeds.

The hardest thing to move along these routes was not a heavy cargo but a way of doing something, and measuring how slowly technique travelled is the most surprising arithmetic in the subject.

The previous lesson followed religions along the roads and noted that they moved as texts, which requires something to write on. This lesson takes paper as its main case, and uses it to separate three things that get lumped together as "the transmission of technology" and that behave completely differently.

Three things, three speeds

An object moves by being carried. A bowl, a bolt of silk or a sheet of paper needs a merchant, a pack animal and a buyer, and nothing else. It requires no shared language and no cooperation from the maker.

A recipe moves as information. It needs someone who can write it down, someone at the far end who can read it, a shared enough vocabulary that the words mean the same thing, and a person who wants it. That is a much higher bar, but it is still a bar that a document can clear.

A skill moves only inside a person. Making paper by hand involves judgments that nobody can write down: how long to ret the fibre, when the pulp feels right, how to shake the mould so the fibres interlock evenly, how to judge the sizing. This is what Michael Polanyi called tacit knowledge, the part of competence that the competent cannot articulate, and it moves when a trained body moves and stays long enough to train other bodies.

The three speeds differ by orders of magnitude, and the rest of the lesson measures them.

Paper as an object

Paper existed in China well before the routes in this course were busy. Fragments of a paper map from Fangmatan are dated to the early second century BCE, and the Hou Hanshu credits Cai Lun in 105 CE with improvements that made it cheap: bark, hemp, rags and fishing nets pulped into a usable sheet.

As an object it moved fast. The Sogdian Ancient Letters of 313 or 314, found in a watchtower west of Dunhuang and written by merchants, are on paper: a Chinese material in the hands of Iranian-speaking traders two centuries after Cai Lun, being used casually for private correspondence rather than treasured as an exotic. Nobody in Samarkand needed to know how paper was made in order to write on it.

Paper as a technique

Now the technique, and the story everybody tells. In July 751 an Abbasid army defeated a Tang force under Gao Xianzhi at Talas, in what is now Kyrgyzstan, when the Karluk contingent changed sides. The standard account says Chinese prisoners taken there included papermakers, who were brought to Samarkand and taught the craft, from which it spread across the Islamic world and eventually to Europe.

Treat that claim the way the second lesson recommended: what does it rest on? It rests essentially on al-Tha'alibi, who died in 1038 and was therefore writing about 280 years after the event, and no contemporary source connects the battle to papermaking at all. Meanwhile paper as an object had been in Sogdiana for four centuries, and paper manufacture in Central Asia may well predate the battle. The Talas story is a good story with one late source, and the honest summary is that Samarkand became a major paper centre in the later eighth century, that the technique came from China at some point, and that the battle is a convenient marker rather than a demonstrated cause.

What is solid is the westward chain of first attestations. Samarkand in the later eighth century. Baghdad, where a paper manufactory is recorded in 794 or 795 and where the Abbasid chancery switched from papyrus and parchment to paper within a generation. Egypt by around 900. Muslim Spain, with Xativa named as a paper town by al-Idrisi in the 1150s. Italy, with Fabriano producing paper by around 1276. France, with a mill at Troyes by 1348.

Example. Samarkand to Fabriano is about 4,440 km in a straight line. Take 751 and 1276 as the endpoints and work out how fast the technique travelled. Then compare that with how fast a caravan moved.

The elapsed time is $1276-751=525$ years, so the rate is $4440/525=8.5$ km per year. A caravan covering 3,888 km in 130 days of travel is moving at $3888/130\times 365=10\{,\}900$ km per year if it kept going. The technique travelled about a thousandth as fast as the goods on the same roads. That ratio, roughly three orders of magnitude, is the single most useful number in this lesson, and it should be held against every casual statement that the Silk Road transmitted technology. It transmitted objects quickly, information slowly, and skill barely at all.

Now you. Brahmagupta wrote the decimal place-value system with zero into the Brahmasphutasiddhanta in 628 at Ujjain. It reached Baghdad in Arabic translation in the 770s, al-Khwarizmi wrote his manual of calculation with Indian numerals around 825, and Fibonacci's Liber Abaci introduced it to Latin Europe in 1202. Ujjain to Pisa is about 6,310 km. Compute the rate and compare.

Why the slowness is not stupidity

The reason a technique cannot simply be copied is that it is never the technique alone. Chinese paper was made from bast fibres, mulberry and hemp, sized in a Chinese way and formed on a Chinese mould. The Islamic world had different raw materials and switched to linen rags, which behave differently in the pulp and need different treatment. Europe's papermakers used rags too but sized with animal gelatine to take iron gall ink, and mechanised the pulping with water-powered stamping mills, and added the wire mould that leaves a watermark.

Each of those transfers was a re-invention with the original as a hint. The receiving culture had to find local materials that would work, adapt the tools, and rebuild the surrounding trades that supply and use the product. That is why a technique cannot travel in a letter, and why the century-scale delays are not evidence that anyone was slow-witted.

Silk going the other way

The mirror case is sericulture, where China had the monopoly and lost it, and the sources are stories.

Xuanzang, writing in 644 about his travels, records a Khotanese legend: the king of Khotan asked for a Chinese princess in marriage and had her told that there was no silk in his kingdom, so that she smuggled silkworm eggs and mulberry seeds out in her headdress, which the border guards dared not search. A painted wooden panel found by Aurel Stein at Dandan-Uiliq, near Khotan, shows a seated princess with an elaborate headdress and an attendant pointing at it.

Procopius, writing in the sixth century, tells a different one: that around 552 monks came to the emperor Justinian from a land called Serinda and offered to bring silkworm eggs, and did, concealed in hollow canes, whereupon Byzantine silk production began.

Both are good stories and neither is good evidence. One is a legend recorded as such; the other is a court historian's account of an event he did not witness. What archaeology adds is that silk weaving, as opposed to silkworm rearing, was practised outside China much earlier, that wild silk species were used in India and the Mediterranean independently, and that the leak was probably gradual rather than a single dramatic theft.

The deeper point is agronomic and it undercuts the whole secrecy framing. Sericulture needs mulberry trees, which take years to establish, a climate the worms tolerate, and an enormous supply of attentive seasonal labour at exactly the moment when the leaves are right. The barrier to producing silk outside China was never chiefly that nobody knew the trick. It was that the whole agricultural system had to be built, which is the general shape of technique transfer stated once more.

Example. In 647 the Tang court sent envoys to Magadha in India to learn the boiling of cane sugar, and the Tang histories record it as a deliberate mission. What kind of transfer is this, and why does it work faster than paper's did?

It is a state buying a skill by moving the people who have it, which is the fastest mechanism available before modern documentation. It works faster than diffusion for three reasons. Somebody has decided in advance that they want it, so no persuasion is needed. The state can pay for travel, for practitioners, and for the years of failure while the process is being made to work on local cane. And an emperor can order the surrounding conditions into existence, land, labour and equipment, rather than waiting for a market to justify them. It is worth noticing that this is not an exotic ancient practice: it is what every industrialising state has done since, and it is the mechanism behind most of the fast transfers in the historical record. Diffusion is slow; recruitment is quick.

Now you. Chinese blue and white porcelain of the Yuan period was painted with cobalt imported from Persia and made largely for Middle Eastern buyers, which is why the great collections are at Topkapi in Istanbul and Ardabil in Iran. Classify what moved in this case.

Example. Classify each of these by the taxonomy at the top of the lesson and predict how far each travels: a finished porcelain bowl, a written formula for gunpowder, and the ability to throw a two-foot porcelain vase on a wheel.

The bowl is an object and travels as far as anyone will pay to carry it, which for Chinese ceramics meant every coast of the Indian Ocean and eventually every court in Europe. The formula is a recipe, and it travelled: descriptions of incendiary and explosive mixtures appear in Chinese texts, then in Arabic works, then in Latin ones within a couple of centuries, because saltpetre, sulphur and charcoal in stated proportions is exactly the kind of knowledge that survives being written down and read by a stranger. The throwing skill is tacit and travels only with a potter, and even then only if he stays long enough to train apprentices and finds clay that behaves like the clay he learned on. The prediction, which holds, is that porcelain objects are found across Eurasia, gunpowder recipes appear in three language traditions within a few hundred years, and porcelain manufacture stays at Jingdezhen until Europeans reverse-engineer it independently in the eighteenth century after decades of state-funded experiment. The three speeds are not a metaphor; they are three different transport problems.

Now you. Europeans reproduced hard-paste porcelain at Meissen in 1708 after years of experiment funded by a ruler who had the researchers effectively imprisoned. Which of this lesson's mechanisms does that illustrate?

What never crossed at all

The negative cases matter as much as the positive ones. Printing from carved blocks was in routine use in China from the eighth century, and the Diamond Sutra printed at Dunhuang in 868 is a fully commercial product of a mature industry. Europe had contact with China throughout the Mongol period, and European printing nevertheless begins in the 1450s with an independent invention using movable metal type, an alphabet and a press derived from wine and oil presses. Whatever crossed, it was not the technique.

The same is true of many other things: the wheelbarrow, the segmental arch bridge, the sternpost rudder, the crossbow all have complicated and disputed diffusion histories, and the pattern is that where a device is simple enough to be understood from seeing one, it moves; where it requires an industry, it does not.

The finding to take away is a limit rather than a marvel. These routes were a superb channel for objects, a decent channel for texts and religions, and a poor channel for skill, and popular accounts that credit the Silk Road with transmitting technology across Eurasia are describing a process that ran at about ten kilometres a year.

There was one exception to all of this, and it needed no channel at all. The next lesson is about the period when the entire route lay under a single authority for the only time in its history, and about what that authority made possible, including one transfer that travelled faster than any merchant and killed on arrival.

Everything in this course so far has assumed a route crossing dozens of independent polities, each taking its cut and enforcing its own law, and for about a century that assumption was false.

The Mongol conquests put the entire span from Korea to the Black Sea under a single family. That makes the thirteenth and fourteenth centuries the one period where the effect of political fragmentation on Eurasian exchange can be estimated, by the crude but genuine method of taking it away. The results are instructive in both directions, because the same century that produced the safest roads in the subject's history also produced the worst destruction in it, and then killed a third of the population of the western end.

What was unified

Temujin was proclaimed Chinggis Khan in 1206. Within seventy years his successors held China, Central Asia, Persia, Mesopotamia and the Russian steppe, some 24 million km² at the greatest extent, roughly five times the area of the Roman empire at its height and the largest contiguous land empire there has ever been.

It did not stay one empire. After the succession struggle of 1260 it functioned as four: the Yuan in China and Mongolia, the Ilkhanate in Persia, the Chagatai khanate in Central Asia and the Golden Horde on the Russian steppe. They fought each other, sometimes bitterly. But they shared a ruling family, a language of administration, a legal tradition and a set of institutions, and for most of the period a merchant crossing between them was crossing between relatives rather than enemies.

The yam

The institution that mattered most for exchange was the postal relay, the yam. Stations were placed roughly a day's ride apart along the main routes, each obliged to keep horses, fodder, food and lodging ready, supplied by the local population as a tax in kind. Yuan records indicate something like 1,400 stations in China alone. Authorised travellers carried a paiza, an inscribed tablet of wood, silver or gold according to rank, which entitled the bearer to draw on the stations.

The gain is not a modest improvement, and it can be quantified.

Example. A single rider with one horse sustains about 40 km a day over a long journey, because the horse must rest and graze. A relay courier changing horses at every station manages 300 km a day. Karakorum to Tabriz is about 4,630 km. Compare the two.

The lone rider takes $4630/40=116$ days, close to four months. The relay courier takes $4630/300=15$ days. The ratio is 7.5, and the reason is worth stating precisely: the limit on a long ride is the animal's endurance, not the rider's, so an institution that supplies fresh animals converts a biological constraint into a logistical one. Nothing about the technology changed, no faster horse was bred, and no road was built. What changed is that the state guaranteed remounts, which is an administrative achievement rather than a technical one. The consequence for the empire is that orders could reach a frontier and a reply come back within a season, and the consequence for merchants, who could not use the yam but travelled roads that its stations made safe and supplied, was almost as large.

Now you. The yam was paid for by requisitioning horses, fodder and labour from the populations near the stations rather than out of a central treasury. What would you expect that to do over time, and how would you check?

What the merchants said

Two kinds of evidence show that the traffic responded. The first is institutional: Mongol elites invested in trade directly through the ortogh, partnerships in which a prince or a khatun supplied capital to an association of Muslim or Uighur merchants who traded it and returned an agreed share. That converts the ruling class from predators on trade into stakeholders in it, which is a rare and valuable alignment.

The second is a handbook. Francesco Balducci Pegolotti, an agent of the Bardi company of Florence, compiled La pratica della mercatura around 1340, a merchant's reference of weights, currencies, tariffs and routes. On the road from Tana on the Sea of Azov to Cathay, he reports that it is perfectly safe by day or by night, according to the merchants who have used it. He lists the stages, from ox wagon to camel wagon to pack ass, with the days each takes, and the total runs to something like eight or nine months of travel. He advises taking a dragoman, servants who know the language, and provisions, and he notes what will and will not sell.

The last clause of his safety claim is the important one: according to the merchants who have used it. Pegolotti never went. He was compiling from informants for a commercial house, in a manual whose purpose was to encourage and equip trade, and by the time the book was in circulation the political conditions it described were already dissolving. It is excellent evidence that Florentine merchants believed the road across Asia was open in the 1330s, and no evidence at all about how many took it.

Example. How much weight can Pegolotti's statement that the road is perfectly safe by day or by night bear?

Less than it is usually made to carry, and more than nothing. Weigh it as you would any testimony: the author never travelled it, he names no informant, he is writing a promotional and practical manual for a banking house with money to place, and the manuscript circulated after the conditions it describes had begun to fail. Against that, he had no reason to lie to his own colleagues, who would lose money if he were wrong, and the detail elsewhere in the book, currencies, weights, tariffs, what sells where, is accurate enough to be used as a source for Mediterranean commerce. So the sentence is good evidence that Florentine merchants in the 1330s believed the Asian route was open and usable, which is itself a fact about the Mongol peace worth having, and it is no evidence about how many merchants went, what they carried, or whether the road really was safe. The distinction between evidence of belief and evidence of fact is one to keep, because a great deal of what survives from this period is the first kind and gets quoted as the second.

Now you. European accounts of the Mongols shift from horror in the 1240s to something close to enthusiasm by the 1290s. What changed, and what should you infer from the shift?

The travellers

The Mongol century produced the only substantial body of eyewitness travel writing across the whole route. John of Plano Carpini was sent by the pope in 1245 and William of Rubruck by Louis IX in 1253, and both wrote sober, observant reports on Mongol society. Ibn Battuta, who left Tangier in 1325 and travelled for nearly thirty years, covered the Islamic world and beyond. Rabban Bar Sauma, a Uighur Christian monk born near modern Beijing, went the other way: sent west by the Ilkhan in 1287 as an envoy, he met the Byzantine emperor, the king of France, and Edward I of England at Bordeaux, where he gave the English king communion.

And Marco Polo, in Asia from 1271 to 1295, dictated in a Genoese prison the book that shaped the European imagination of the east for two centuries. Whether he went at all has been argued: Frances Wood's Did Marco Polo Go to China? (1995) pointed to his silence on tea, chopsticks, foot binding and the Great Wall, and to his absence from Chinese records. The strongest response, Hans Ulrich Vogel's Marco Polo Was in China (2013), works from the opposite end, showing that his descriptions of paper money denominations, salt production and the salt revenues are accurate in detail, sometimes in ways no other western source records and that match Chinese fiscal documents. The current balance of opinion is that he was there, that he served the Mongol administration in some capacity, and that his book is padded with material from other people's accounts.

Notice the asymmetry, because it is more revealing than any individual traveller. Many Europeans went east and wrote about it; very few Asians went west and did. That is not a comment on curiosity. Europe was the poor, peripheral, fragmented end of the system with everything to gain from contact, and the flow of travellers ran up the gradient of wealth, as it usually does.

Specialists, moved at swordpoint

The previous lesson concluded that skill moves only inside a person, and slowly. The Mongols found the exception: move the person by force.

Conquered cities were systematically stripped of craftsmen, who were spared the general massacre and deported to work for the khan. Persian and Central Asian artisans were resettled in Mongolia and China; Chinese and Khitan specialists went west. In 1267 an astronomer named Jamal al-Din presented Kublai with Persian instruments and a calendar, and an Islamic astronomical bureau operated alongside the Chinese one at the Yuan court. Rashid al-Din, the Ilkhanid vizier, assembled at Tabriz a world history that included a section on China and a compilation of Chinese medicine in Persian, working with Chinese informants at his court. Most vividly, when the Mongols were stuck before the fortress city of Xiangyang, they sent for engineers from the Ilkhanate who built counterweight trebuchets of Islamic design, and the fall of the city in 1273 opened the Yangzi and finished the Song.

That is a technique crossing 5,000 km in a couple of years, against the ten kilometres a year of the previous lesson. The mechanism is the one identified there, deliberate recruitment rather than diffusion, applied by a state that did not need to persuade anybody.

The cost

None of this should soften what the conquest was. Merv surrendered in 1221 and was destroyed; Nishapur likewise. The chronicles give figures: Juvayni, writing a generation later under Mongol employment, says 1.3 million died at Merv, and Ibn al-Athir gives 700,000.

Example. The walled area of Sultan Kala, the Seljuk city at Merv, is about 400 hectares. Test Juvayni's figure against it.

At 1.3 million people in 400 hectares the density would be $1\{,\}300\{,\}000/400=3\{,\}250$ per hectare, which is 32,500 per km². Pre-modern cities of mud brick and courtyard houses run to perhaps 100 to 200 people per hectare, and even the densest quarters of the most crowded pre-industrial cities do not approach a tenth of Juvayni's implied figure. Taking 200 per hectare as a generous ceiling gives about 80,000 inhabitants, so the chronicle number is too large by a factor of roughly sixteen. What the arithmetic does not show is that nothing happened. Merv was a great city and afterwards it was not: the excavations show contraction, the irrigation system that made the oasis possible failed, and the site of the medieval city was never rebuilt to the same scale. The correct statement is that a catastrophe occurred whose magnitude the sources cannot measure, and that the chroniclers' numbers are literary devices for enormity rather than counts.

Now you. Why would a historian in Mongol service like Juvayni inflate the number of people his employers had killed?

Why it stopped

The unity dissolved fast. The Ilkhanate broke up after 1335, the Chagatai khanate split, the Yuan were driven out of China in 1368 and the Ming that replaced them regarded the steppe as an enemy rather than a partner. The successor states fought over the same routes their predecessors had guaranteed.

And in the middle of that, between 1346 and 1353, the plague arrived at the western end of the system and killed something between a third and a half of everyone in Europe, the Middle East and North Africa. The two facts are related, though not in the crude way often implied, and the relationship has been transformed in the last few years by evidence that has nothing to do with chronicles at all.

That is the next lesson: the best documented cargo the routes ever carried, and the one nobody chose to send.

Everything else in this course moved because somebody wanted it to, and the exception killed more people than all the trade ever enriched.

The previous lesson described the century when the routes were safest and most integrated, and ended with the epidemic that arrived at the western end of that system between 1346 and 1353. This lesson is about how we know what it was, where it came from, and how much of the popular story, that Silk Road trade brought the Black Death to Europe, is actually established. The answer is: more than it was ten years ago, and still not all of it.

What the disease is

Yersinia pestis is a bacterium whose natural home is not humans at all. It lives in populations of wild burrowing rodents, marmots, gerbils, ground squirrels, in which it circulates via fleas, and in which it persists indefinitely in what epidemiologists call a reservoir. Central Asia has one of the world's major reservoirs, in the marmot populations of the Tian Shan and neighbouring ranges.

Humans are an accident. When a rodent population crashes, its fleas seek new hosts, and a flea that bites a human transmits the bacterium into the lymphatic system, producing the swollen nodes, the buboes, that give bubonic plague its name. Untreated, it kills something over half of those infected. If the infection reaches the lungs it becomes pneumonic plague, which spreads directly from person to person in droplets and is almost invariably fatal without antibiotics. If it reaches the blood it becomes septicaemic and kills faster still.

The transmission mechanics matter for the history. Bubonic plague travels the way rats and fleas travel, which means in cargo: grain, cloth bales, and the holds of ships. Pneumonic plague travels the way people travel. So a disease of Central Asian marmots has two quite different routes into a European city, and both of them run along the channels this course has been describing.

Three pandemics

Plague has broken out of its reservoirs three times on a scale that reorganised societies. The first, the Justinianic plague, is dated from 541 CE, when it appears at Pelusium in Egypt, and recurred in waves for two centuries. The second is the Black Death of 1346 to 1353 and the long series of returns that followed it into the eighteenth century. The third began in Yunnan in the nineteenth century, reached Hong Kong in 1894, where Alexandre Yersin identified the bacterium that now carries his name, and went on to kill something like twelve million people, mostly in India.

The third pandemic is the one that made the other two intelligible, because it was observed by people who could culture bacteria, and it established the rodent and flea cycle that nobody had previously suspected.

Why the old evidence was not enough

Until very recently the second pandemic was known entirely from written accounts, and they are less solid than their vividness suggests.

Gabriele de' Mussi provides the famous origin scene: the Mongol army besieging the Genoese trading post of Kaffa in the Crimea in 1346, stricken with plague, catapulting its own dead over the walls, and the infection passing to the Genoese, who carried it home by ship. It is a superb story, it is the source of every account of the plague's arrival in Europe, and de' Mussi was in Piacenza at the time, writing from report.

More seriously, the identification of the disease was contested for decades. Beginning in the 1980s, several historians and zoologists, Graham Twigg and later Samuel Cohn among them, argued that the Black Death could not have been bubonic plague, on grounds that look strong from the documents: it spread far faster than modern plague does, it spread in winter and in cold northern countries where the rat flea is inactive, and no chronicler mentions the mass rat deaths that precede modern outbreaks. Their candidates ranged from anthrax to an unknown viral haemorrhagic fever.

That debate could not be settled with texts, and it was not settled by argument. It was settled by digging up the dead.

The genomes

In 2011 a team led by Kirsten Bos recovered and sequenced Yersinia pestis DNA from skeletons in the East Smithfield emergency cemetery in London, dug in 1348 and 1349 specifically for plague victims. The pathogen was plague, and the case was closed. Subsequent work confirmed Y. pestis in a sixth century cemetery at Aschheim in Bavaria, doing the same job for the Justinianic pandemic.

Then, in June 2022, a team led by Maria Spyrou published in Nature a result that fixed the origin. Two cemeteries in the Chu valley of northern Kyrgyzstan, at Kara-Djigach and Burana, excavated in the 1880s, belong to a Syriac Christian community of the kind the ninth lesson described. Their tombstones are dated and inscribed, and among them is an unusual cluster from 1338 and 1339, several of which give the cause of death as pestilence. Teeth from burials in that cluster yielded Y. pestis genomes.

The genomes did something the tombstones could not. Placed on the family tree of the bacterium, the Kara-Djigach strain sits immediately ancestral to the polytomy, the sudden fourfold branching that biologists call the big bang, from which the lineages responsible for the Black Death and for most modern plague descend. And its closest living relatives circulate today in marmots in the Tian Shan, a few hundred kilometres away.

So the second pandemic's point of departure is now located to a region and dated to within a year or two, by three independent lines, a dated inscription, a genome and a modern animal reservoir, none of which is a chronicle.

Example. The Chu valley is about 3,170 km from Kaffa, and the plague appears there in 1338 to 1339 and at Kaffa in 1346. Kaffa to Messina is about 1,810 km, covered between the siege and October 1347. Compare the two speeds and interpret.

The overland leg is $3170/7=450$ km a year. The sea leg is about 1,810 km in roughly a year, four times faster, and the pattern continues: from Messina in October 1347 the disease reaches London, about 1,920 km away, within about a year, having gone by ship along the Mediterranean and Atlantic coasts before spreading inland. So the plague behaved exactly like the freight of the third lesson. It crawled overland, where it depended on caravans, local rodent populations and short hops between oases, and it accelerated the moment it reached a maritime network, where a ship carries rats, fleas and infected crew a thousand kilometres in a fortnight with no intervening stops. This is a useful corrective to the phrase "the Silk Road brought the Black Death". The overland routes brought it out of its reservoir and moved it slowly westward over years; the sea lanes of the Black Sea and the Mediterranean did the killing work, in months.

Now you. The 1338 tombstones were excavated in the 1880s and the inscriptions were published then. Why did it take until 2022 to learn what they meant?

Example. The sceptics' best points were that the Black Death spread too fast for rat-borne plague, spread through northern European winters when rat fleas are inactive, and produced no reports of mass rat deaths. The genome proves it was plague. How are those observations now explained?

By accepting the observations and revising the transmission model rather than the identification, which is the right order once a direct test exists. Three mechanisms do the work. Pneumonic plague passes directly between people in droplets, needs no rodent at all, and spreads as fast as people move, which covers the winter outbreaks and much of the speed. Human ectoparasites, the human flea and the body louse, can transmit the bacterium between people, and modelling published by Katharine Dean and colleagues in 2018 found that human parasite transmission fitted the observed spread of medieval European outbreaks better than a rat-based model did. And the absence of reported rat deaths is weak evidence in any case, since medieval chroniclers were writing about human catastrophe and had no reason to record rodent mortality. What the episode shows is not that the sceptics were foolish but that their argument was the strongest available until a method arrived that could test the premise directly, and the correct response to that method was to keep their observations and drop their conclusion.

Now you. Plague returned to Europe repeatedly until the eighteenth century and then stopped, without antibiotics, vaccines or any understanding of the cause. What kinds of explanation are on the table, and how would you choose between them?

How much died, and the case that is not proven

Mortality in the second pandemic was catastrophic by any measure and is imprecise by all of them. The usual range for Europe is between a third and a half of the population; Ole Benedictow has argued for around sixty per cent. English manorial records, which are the best series available anywhere because they record tenant deaths for fiscal reasons, indicate something like forty to forty five per cent in the first wave. For the Islamic world and North Africa the documentation is thinner and the estimates correspondingly looser. Recovery to pre-plague population levels took roughly two centuries.

Now the honest complication, because the first pandemic is a live dispute and reasoning about it well is more instructive than the settled case.

Yersinia pestis in sixth century Bavaria is proven. What is not proven is that the Justinianic plague depopulated the Mediterranean world. In 2019 Lee Mordechai, Merle Eisenberg and colleagues published an argument in PNAS that it did not, and their method is the interesting part: rather than counting chroniclers' corpses, they examined proxies that would have to move if a third of the population had died. Pollen sequences showing whether cereal cultivation contracted. Coin production and mint output. The volume of papyri and of inscriptions. Legislation, which responds to labour shortage. Across these, they found continuity rather than collapse, and argued that the pandemic's demographic impact has been assumed from Procopius rather than demonstrated.

The argument is contested and the last word has not been said. What it models is the right procedure: a pathogen's presence and a pathogen's consequences are two separate claims, requiring two separate bodies of evidence, and confirming the first tells you nothing about the second. It is exactly the discipline this course has tried to apply to trade, where the presence of a Roman glass bowl in a Chinese tomb was likewise not evidence of volume.

Example. Someone argues that because the Black Death began in Central Asia and arrived in Europe by trade routes, the Silk Road caused the Black Death. Assess the causal claim.

Break it into steps and check each. That plague emerged from a Central Asian reservoir is now well evidenced. That it reached the Crimea and then Europe along established trade and military routes is highly likely and consistent with everything known, though the eight years between the Chu valley and Kaffa are documented by nothing at all. That the routes were necessary is much weaker: reservoirs also exist elsewhere, human populations have contacted marmots for millennia, and outbreaks had presumably occurred and burned out locally many times without a pandemic. What trade supplied was not the pathogen but the connectivity, the chain of dense, connected, rat-infested settlements down which a local outbreak could propagate instead of dying out. So the defensible statement is that the routes were the mechanism of spread rather than the cause of emergence, and that the maritime network did most of the spreading. As so often in this subject, the strong version of the claim is popular precisely because it is dramatic, and the defensible version is more interesting.

Now you. What would count as evidence that plague also travelled these routes before 1338, leaving no pandemic behind?

What it changed

The consequences at the European end are outside this course's territory but not outside its argument. A third or more of the labour force died while the land, the tools and the livestock remained, which raised the value of a working pair of hands and lowered the value of an acre. Real wages in England roughly doubled over the following century, serfdom in western Europe decayed because a landlord who insisted on services found his tenants had somewhere better to go, and the relative price of labour and land shifted in a way that some economic historians trace forward for centuries.

The point to carry away is about the subject rather than about Europe. The plague is the best evidenced thing that ever moved along these routes: better dated than the silk, better localised than the Buddhism, and better quantified than the trade. It is also the one nobody was trying to move, which is a fair summary of how the largest consequences of exchange usually arrive.

The last three lessons have all pointed at the same overlooked place. Freight went four times faster by sea, the plague went four times faster by sea, and the porcelain of the tenth lesson went entirely by sea. The next lesson goes there properly, and finds that the maritime route was carrying more than the overland one for most of the period this course covers.

For most of the period this course covers, more goods moved between China and the west by sea than by land, and the reason is a freight rate.

The last few lessons kept pointing here. The plague accelerated fourfold when it reached shipping. The porcelain of the technique lesson went east to west entirely by water. The freight model of the third lesson excluded from the overland route precisely the goods that dominate the maritime archaeology. This lesson gives the sea its own treatment, and the effect is to shrink the caravan route to its real proportions without diminishing what was distinctive about it.

Wind that reverses

The Indian Ocean has a property no other ocean has: its winds reverse twice a year. From roughly June to September the southwest monsoon blows steadily from Africa and Arabia towards India; from November to March the northeast monsoon blows back. A sailing vessel can therefore make an open-sea crossing in one season and return in the other, without coasting.

The Periplus of the Erythraean Sea, a merchant's handbook written in Greek around the middle of the first century CE, describes the route in operational detail: which ports to use, what to carry to each, what to buy, when to sail and how long each leg takes. It attributes the open-water crossing to a navigator named Hippalus, and whether or not one man deserves the credit, the effect is clear in the traffic. Strabo reports that in his day 120 ships a year sailed from Myos Hormos on the Egyptian Red Sea coast for India, where under the Ptolemies only a handful had gone.

Note what the monsoon does to the shape of the business. The wind is a timetable, not a convenience: miss the season and you wait months, so cargoes, credit and warehousing all had to be organised around a fixed annual cycle. That is a different commercial rhythm from the caravan trade, and it favours large operators with capital to leave idle.

What one cargo was worth

The Roman trade with India is the best quantified exchange anywhere in this subject, thanks to two very different documents.

Pliny the Elder supplies the rhetoric. In the Natural History he complains that India drains fifty million sesterces a year from the empire, and elsewhere that India, the Seres and Arabia together take a hundred million, all of it to pay for luxuries that in his view do the Roman character no good. He is a moralist writing a diatribe, and the figures have often been dismissed as invention.

Then there is a contract. A papyrus in Vienna, known as the Muziris papyrus, records the terms of a loan for a cargo shipped from Muziris on the Malabar coast to Egypt on a vessel called the Hermapollon, in the second century CE. It values the cargo at 1,154 talents and 2,852 drachmas, and it specifies the Roman quarter-tax, the tetarte, on arrival.

Example. Take the drachma as equal to one sesterce in this Egyptian context, and legionary pay in the second century as 1,200 sesterces a year. Value the cargo and the tax on it, then test Pliny's hundred million against it.

The cargo is $1154\times 6000+2852=6\{,\}926\{,\}852$ drachmas, about 6.93 million sesterces. That is $6\{,\}926\{,\}852/1200=5\{,\}770$ soldier-years of pay in one ship's hold. The quarter-tax alone is 1.73 million sesterces, or 1,440 soldier-years, from a single arrival, which tells you why the Roman state was interested in the Red Sea. Now Pliny: his hundred million is $100\{,\}000\{,\}000/6\{,\}926\{,\}852=14.4$ cargoes of this size. Against Strabo's 120 ships a year, that is one ship in eight carrying a Hermapollon-sized cargo, which is entirely plausible and possibly conservative. Pliny's figure, long treated as a rhetorical exaggeration, turns out to be the right order of magnitude, and against a total imperial revenue usually estimated at something like 800 million sesterces it represents about an eighth, which is large, striking and not absurd.

Now you. What is the strongest objection to using one papyrus this way, and how much does it damage the conclusion?

The freight rate that explains everything

The mechanism behind all of this is one ratio. Working from Diocletian's Price Edict of 301 CE, A. H. M. Jones and Richard Duncan-Jones estimated the relative cost of moving goods by sea, by inland waterway and by road at roughly 1 to 5 to 40.

Take that seriously for a moment. Forty to one means that carrying a cargo 1,000 km by sea costs about what carrying it $1000/40=25$ km by road costs. Distance almost stops mattering on water and dominates everything on land. Every consequence in this course follows from that single ratio: which goods can move, how far, in what quantity, and why the same commodity is a luxury at one end of a continent and a commonplace along a coast.

Sixty thousand bowls

The eastern half of the maritime route was invisible until divers found it. In 1998 fishermen located a wreck off Belitung Island in Indonesia: a sewn-plank vessel of Arabian or Indian construction, sunk in the 820s or 830s, carrying a cargo of Chinese goods. Among them were roughly 60,000 ceramic pieces, mostly bowls from the Changsha kilns in Hunan, packed in storage jars, along with fine gold and silver work. One bowl carries an inscription with a date corresponding to 826.

Example. Take a Changsha bowl at 0.4 kg and use the overland freight figure of 1.73 labour-days per kilogram derived earlier. What would it have cost to send this cargo by caravan?

One bowl weighs 0.4 kg, so its overland freight is $0.4\times 1.73=0.69$ labour-days, most of a labourer's day's wage to move one cheap bowl. The whole cargo is $60\{,\}000\times 0.4=24$ tonnes, which is 160 camel-loads, and the freight bill is $24\{,\}000\times 1.73=41\{,\}500$ labour-days, about 138 working years. A Changsha bowl was a mass-produced object worth a small fraction of a day's wage. Its overland transport would have cost several times its value, which is not a difficulty to be overcome by clever merchants; it is a prohibition. So this cargo could never have existed on the land route, in any century, and every generalisation about Silk Road trade built from caravan evidence is a generalisation about the small, dense, expensive minority of what actually moved between these regions.

Now you. The Belitung ship was built in the Persian Gulf or India and sank in Indonesian waters loaded with Chinese goods. What does the combination tell you about who was running this trade?

Example. A merchant sailing from Egypt to India must leave on the southwest monsoon between roughly June and September and can only return on the northeast monsoon between November and March. Work out the round trip and say what it does to the business.

Leaving in July and arriving perhaps two months later puts him in India in September, and the earliest useful return wind is November, so a fast round trip is about a year and a slow one is two. Compare the caravan, which takes about 130 days each way and can in principle set off again whenever it is loaded. The consequences are commercial rather than nautical. Capital is tied up for a year at a time, so the trade needs financiers rather than traders with their own purse, which is exactly what the Muziris papyrus documents: a loan against a cargo, with terms. The waiting months must be spent somewhere, which produces permanent expatriate merchant communities at the far end rather than passing visitors, and those communities appear in the record at Indian ports, at Guangzhou and later everywhere in the Indian Ocean. And the annual rhythm makes information old: a merchant chooses his cargo on prices a year out of date, which raises risk and rewards anyone with a resident agent. The monsoon, in other words, produced the institutions of the trade as directly as the camel produced the caravanserai.

Now you. The Roman trade with India ran a persistent deficit settled in gold and silver coin, and Pliny complains about it. Was he right to?

The scale later on

The maritime trade grew steadily and the archaeology grows with it. The Nanhai One wreck, a Southern Song merchantman raised whole off the Guangdong coast in 2007, has yielded well over a hundred thousand items. Song and Yuan China ran maritime trade superintendencies at Guangzhou, Quanzhou and other ports specifically to tax this traffic, and Quanzhou, the Zayton of European writers, was described by both Marco Polo and Ibn Battuta as one of the greatest ports on earth.

The most spectacular episode was also the least commercial. Between 1405 and 1433 the Ming sent seven fleets into the Indian Ocean under the admiral Zheng He, reaching Hormuz, Aden and the East African coast, with the first expedition reported at 317 ships and about 27,800 men. The Ming Shi gives the largest vessels as 44 by 18 zhang, which at a Ming zhang of about 3.11 m is 137 m long and 56 m in the beam.

Those dimensions deserve the treatment this course has given other chronicle numbers. A wooden hull of 137 m is far longer than anything Europe built before iron: HMS Victory, launched in 1765, is 69 m. And the proportions are wrong, since a length to beam ratio of $44/18=2.4$ describes a floating rectangle rather than a ship. Naval architects have generally concluded that the figures are exaggerated or refer to some other measure, while accepting that the treasure ships were genuinely very large by contemporary standards. The voyages themselves are not in doubt; a stele the fleet erected records them.

What happened next is the more important fact. The voyages were state enterprises for prestige, diplomacy and tribute rather than trade, they were enormously expensive, they were opposed by officials who thought the money belonged on the northern frontier, and after 1433 they simply stopped. Private Chinese maritime trade continued and often thrived, legally or otherwise, but the state withdrew from the sea at the moment Portuguese ships were beginning to work down the African coast.

Why the sea is under-taught

If the maritime route carried more, why is the overland one famous? Three reasons, and none of them is evidential.

The name is one. Richthofen coined a term for a land route, and the phrase "maritime Silk Road", now used freely and adopted into Chinese policy language, is a back-formation from it.

The sources are another. Overland travel produced travellers' tales, and shipping produced bills of lading and then sank. Until scuba diving made wrecks accessible from the 1960s, and marine archaeology matured after that, the physical record of the sea route was inaccessible while the desert record had been lying in museum drawers since 1907.

And the third is romance, which is not nothing: camels, deserts, passes and lost cities are simply better material than a monsoon timetable.

The corrective is not to swap one route for the other. It is to hold both, and to say what each did. The sea carried the volume: ceramics, spices, timber, cotton, bulk goods, and the great majority of the value. The land carried what could pay its freight, and it also carried the things that had no weight, monks, scripts, faiths, techniques and, once, a bacterium.

That leaves the ending, and the ending most people have been taught is wrong. The next lesson takes apart the story that the Ottomans closed the road in 1453, and replaces it with a slower and better evidenced account in which the overland trade outlived its own supposed death by four hundred years.

The most widely believed fact about the Silk Road is that it was closed in 1453 when the Ottomans took Constantinople, forcing Europeans to seek a sea route to the Indies, and it is wrong in at least three separate ways.

The first lesson of this course flagged the sentence as suspicious on its form, because a diffuse relay system running through dozens of polities cannot be switched off by one conquest at one end. This lesson supplies the evidence, replaces the story with a better one, and arrives at an ending date four centuries later than the traditional one.

The story, and where it comes from

The standard chain runs: Constantinople falls in 1453; the Ottomans block or ruinously tax the eastern trade; spice prices in Europe rise; Portugal and Spain are driven to find a sea route; Vasco da Gama reaches Calicut in 1498; Columbus sails west in 1492; the Age of Discovery begins and the overland trade dies.

It is a wonderfully tidy causal chain, it appears in textbooks in every European language, and it has the shape of a good exam answer. Test it clause by clause.

Three ways it fails

Constantinople is not on the road. Look at the map the third and fourth lessons built. The overland route to China runs through Persia, Merv, Bukhara and Samarkand, then the Pamirs and the Tarim. Its western outlets are the Black Sea ports, the Levantine coast at Aleppo and Tripoli, and Alexandria via the Red Sea. Constantinople is a great commercial city, and it is a terminus for Black Sea and Mediterranean shipping rather than a link in any chain to Asia. Taking it changes who taxes one branch of the western end.

The Ottomans wanted the trade. This is the decisive objection. A state that has conquered a commercial crossroads does not destroy the commerce it has just acquired; it taxes it, and it protects the merchants who generate the revenue. The Ottomans took Syria and Egypt in 1516 and 1517, which gave them the Levant and Red Sea routes as well, and they ran them as a revenue system, maintaining caravan roads, building caravanserais and negotiating trade capitulations with Venice and later with France and England. Ottoman customs registers are one of the better sources for sixteenth century trade volumes precisely because the trade was going on.

The numbers refuse to cooperate. If the route had been closed, the spice trade through the Levant should have collapsed and stayed collapsed after 1498. Frederic Lane's work on the Venetian records showed the opposite: after a genuine disruption in the first decades of the sixteenth century, when the Portuguese arrival was new and shocking, the Levant spice trade recovered, and by the 1560s Venice was again importing pepper through Alexandria in quantities comparable to those before da Gama. The Portuguese never had the ships to seal the Red Sea, and Asian and Muslim merchants continued to move spices north through Hormuz, Basra and Suez in competition with the Cape route for another century.

Example. Suppose you wanted to test the closure story properly. What evidence would settle it, and which way does the available evidence point?

The claim is about quantities, so it needs a quantity series: the volume or value of eastern goods arriving in Europe, at the same places, before and after 1453 and again before and after 1498. Two series exist in usable form. Venetian records of pepper and spice imports through Alexandria and Beirut, which Lane and later scholars have worked through, show disruption in the early sixteenth century and recovery by the middle of it. Ottoman customs registers for the Syrian and Egyptian ports show a functioning taxed trade throughout. Both point the same way: no closure in 1453, a real but temporary dislocation around 1500 caused by the Portuguese rather than the Ottomans, and continued competition between the two routes for most of the sixteenth century. The evidence that would support the traditional story, a sustained collapse in eastern goods reaching Europe after 1453, has been looked for and is not there.

Now you. If the Levant route was still working, why did Portugal invest so heavily in finding a way round Africa?

What actually shifted

Discard the closure story and a slower, less dramatic account is left, in which several things moved at once.

The Mongol integration ended. The Ilkhanate dissolved after 1335, the Chagatai khanate split, the Yuan were expelled from China in 1368, and the successor states fought each other over the same corridors their predecessors had guaranteed. Timur, who died in 1405, rebuilt an empire and made Samarkand spectacular, but he did it by conquest and by deporting the craftsmen of other cities, and his state fragmented after him. The single most valuable thing the Mongols supplied, a merchant's ability to cross a continent under one law, was gone and did not return.

China turned away. The Ming inherited hostility to the steppe, restricted foreign trade to a tributary framework, banned private maritime trade for long periods, and after 1433 gave up long-range naval expeditions. A frontier held defensively rather than administered is a frontier over which trade is smuggled rather than taxed.

The Ottoman and Safavid states, both formidable and both prosperous, fought each other repeatedly from 1514 onwards, and the transit route through Iran ran through the war zone. Shah Abbas tried to reroute Iranian silk exports around Ottoman territory by sea through Hormuz, which is a good indication of how much the overland corridor had come to depend on the state of relations between two empires.

Undercut rather than blocked

And the sea kept getting cheaper, which is the argument of the previous lesson repeated at a later date.

Larger hulls, better rigs and better navigation lowered the cost per tonne. The Cape route removed the transhipments that had made the old way expensive: goods travelling from India to Antwerp through the Red Sea had to be unloaded, carried overland, reloaded, taxed and handled at every break, whereas a Portuguese carrack loaded at Goa was unloaded at Lisbon. And after 1600 the Dutch and English East India Companies brought something new to the problem, which was permanent pooled capital. A joint stock company can keep ships, warehouses, factories and agents standing between voyages, absorb the loss of a fleet without dissolving, and finance a trade whose capital turns over once a year, which is precisely the difficulty the monsoon imposes.

Against that, the overland route offered a chain of separate merchants, a dozen tax authorities, and a freight rate forty times higher per unit distance. It was not blocked. It was undercut, and the process took two centuries rather than a year, which is why no dramatic date marks it.

The caravans that did not notice

Here is the fact that should finish the traditional story. Long after the road was supposedly closed, it was carrying tea.

Russia and Qing China settled their frontier and their commerce by the treaties of Nerchinsk in 1689 and Kiakhta in 1727, and the second established a designated trading point at Kiakhta on the Mongolian border. Through it ran one of the great caravan routes of history, carrying Chinese tea north and west across Mongolia and Siberia to Moscow, a journey of roughly 11,000 km taking about sixteen months, with Russian furs, cloth and later manufactures going the other way. By the first half of the nineteenth century Kiakhta was handling a large share of Russia's entire eastern trade, and Russian tea drinking, a national habit to this day, was supplied by camel.

Example. The Kiakhta tea caravans ran roughly 11,000 km in about sixteen months. Compare with the third lesson's caravan arithmetic and say what the comparison shows.

Sixteen months is about 486 days, so the average is $11\{,\}000/486=23$ km a day including every stop, wait and winter, against the 30 km a day of pure travel used earlier. Nothing had changed. A nineteenth century tea caravan moved at the same speed, by the same means, with the same constraints, as a Tang dynasty silk caravan a thousand years before, and it was commercially viable because tea, like silk, is light, dense in value, durable and wanted at the far end. The comparison shows that the overland trade never became technically obsolete: the freight model that permitted it in the eighth century still permitted it in the nineteenth. What ended it was not that it stopped working but that something else began working far better.

Now you. Tea reached Russia overland by camel while reaching Britain by sea in clippers. What does that pairing tell you about how routes are chosen?

The real ending

If the caravans were still crossing Asia in the 1850s, what finished them? Three things in one generation, and they are all industrial.

The Suez Canal opened in 1869 and cut the sea distance from London to Bombay by around forty per cent, from something over 10,600 nautical miles round the Cape to about 6,300. Steamships, which no longer had to wait for a monsoon or beat against a wind, turned an annual cycle into a schedule. Between them, the canal and the steam engine did to maritime freight what the previous four centuries had only begun.

Railways did the same on land, and killed the caravan directly. The Trans-Caspian railway reached Samarkand in 1888, and the Trans-Siberian, begun in 1891 and through to the Pacific in the following decades, destroyed the Kiakhta tea trade by carrying the same cargo at a fraction of the cost and time.

Example. The Suez Canal cut the London to Bombay voyage from about 10,670 nautical miles round the Cape to about 6,330. If freight cost scaled with distance, what would that do, and why is the real effect larger?

The distance falls to $6332/10667=0.59$ of its previous value, a saving of about 41 per cent, so a purely distance-proportional freight cost would fall by roughly two fifths. The real effect was considerably larger, for reasons that compound. The canal route suits steamships, which can hold a course through the Red Sea where sailing vessels struggle, so it accelerated the switch from sail to steam on the eastern run and steam removed the monsoon timetable entirely. A shorter voyage means each ship completes more voyages a year, so the effective capacity of the fleet rises without a single new hull. Faster passage means less capital tied up in cargo afloat and less spoilage. And insurance and financing costs fall with time at sea. The general point matters more than the number: transport improvements do not simply reduce a freight bill, they raise the frequency and reliability of the service, and it is frequency and reliability that reorganise trade.

Now you. Given all this, what is the best short answer to "when did the Silk Road end?"

And the political geography closed. Russia took Tashkent in 1865, Samarkand in 1868, Khiva in 1873 and Merv in 1884; the Qing reconquered Xinjiang and made it a province in 1884. Central Asia stopped being a set of independent polities through which merchants passed and became the interior of two empires, then, in the twentieth century, the two sides of a sealed and militarised Soviet-Chinese border.

That is the ending: not 1453 but roughly 1869 to 1904, not a conquest but a set of engines, and not a closure but an obsolescence. It is less satisfying than the traditional story and it has the advantage of being supported by evidence.

One question remains, and it is the one the first lesson opened with. If the thing itself faded slowly and unremarkably in the age of steam, why does the name exist, and why is it more famous now than at any point in the past? The last lesson follows the term from Richthofen's map to a trillion-dollar infrastructure programme, and closes the account with a balance sheet of what actually moved and what it changed.

The strangest fact about the Silk Road is that the phrase is more widely used now than at any point in the two thousand years it supposedly describes.

The first lesson established that the name is modern and asked what it could honestly denote. The intervening lessons answered that question with evidence. This one follows the term itself from a German geologist's map to a state policy with its own logo, because a concept that has been recruited this often is one whose uses a reader needs to be able to see. It closes with the balance sheet.

From a railway survey to a bestseller

Ferdinand von Richthofen wrote Seidenstrassen in 1877 in the first volume of his study of China, in a passage reconstructing the Han dynasty's western communications. He was a geologist assessing coal and routes, and the phrase was descriptive rather than evocative, in the plural, and confined to a specific ancient network.

It took roughly sixty years and a different kind of writer to make it famous. Sven Hedin, Richthofen's student, had spent decades exploring the Tarim, finding Loulan and mapping the Lop Nur, and in 1936 he published Sidenvägen, the singular Silk Road. It was written for the public, it was translated widely, and it is essentially the source of the modern image: one road, continuous, ancient, romantic. The academic literature followed the popular usage rather than the other way round.

What was taken, and the argument about it

Between 1900 and 1930 the Tarim Basin was systematically emptied of portable antiquities by expeditions from six countries, and it is impossible to write the history of the subject without confronting that, because nearly every source this course has cited came out of the region in those years.

Aurel Stein, working for the British Indian government, made four expeditions and at Dunhuang in 1907 persuaded the caretaker Wang Yuanlu to part with several thousand complete manuscripts, several thousand fragments and hundreds of paintings, for a sum variously reported and in any case trivial. Paul Pelliot arrived in 1908, could read Chinese, spent three weeks in the cave choosing, and took several thousand of the most valuable items. German expeditions under Albert Grünwedel and Albert von Le Coq worked Turfan and Kucha between 1902 and 1914 and removed wall paintings physically, cutting them out of the rock at sites including Bezeklik and shipping them to Berlin. Russian, Japanese and Swedish expeditions took their share. The Dunhuang library is now divided between London, Paris, Beijing, St Petersburg and Tokyo.

The defence offered at the time, and sometimes still, is preservation: that the material was neglected, at risk from local reuse and from a collapsing state, and that removal to European museums saved it.

Example. Von Le Coq's Bezeklik murals were cut from the cave walls and installed in Berlin. Many were destroyed by Allied bombing between 1943 and 1945. What does that do to the preservation argument?

It refutes the strong version of it empirically, which is rare and valuable in an argument that is usually conducted on speculation. The claim was that removal reduced risk, and in this case removal concentrated an irreplaceable collection in a single building in the capital of a country that was bombed flat, while the caves they came from still stand. That is not hindsight bias: the case shows that "safer in a European museum" was an assumption about the future rather than a fact, and assumptions about the future are exactly what was being wagered with other people's heritage. A weaker version survives and should be granted: some material genuinely was at risk, some was recovered rather than looted, and the scholarly editions, catalogues and conservation that followed made this history knowable at all. The honest position holds both, and notices that the people whose ancestors made these objects were not asked, which is the part no version of the preservation argument addresses.

Now you. The International Dunhuang Project, begun in 1994, photographs the dispersed collections and publishes them in a single free online archive. What problem does that solve, and what does it not?

The name acquires a policy

After the Second World War the term became institutional. UNESCO ran an Integral Study of the Silk Roads from 1988, with international expeditions along the desert, maritime and steppe routes, explicitly framed as a project about dialogue between cultures. In 2014 a stretch of the network, the Chang'an to Tian Shan corridor, was inscribed on the World Heritage list as a transnational serial property of 33 sites shared between China, Kazakhstan and Kyrgyzstan.

Example. The 2014 World Heritage inscription covers 33 separate sites in three countries under one name. What does that form of listing concede about the object, and what problem does it create?

It concedes the argument of the first lesson. A World Heritage property is normally a place: a building, a city, a landscape with a boundary. The Silk Roads could not be listed that way because there is no there, so the nomination had to be serial, covering scattered components, and transnational, covering more than one state, and even then it takes in only one corridor of one branch. The form of the listing is an admission that the subject is a network and a process rather than a site. The problem it creates is where to stop: if the Chang'an to Tian Shan corridor qualifies, so in principle do routes through Iran, India, Syria, Mongolia and the whole maritime system, and each addition requires another set of governments to agree on which of their ruins count. Further corridors have indeed been proposed since. That is not a criticism of the listing, which protects real sites and funds real conservation, but it is a good illustration of what happens when an administrative system designed for places is applied to a concept that was invented to describe movement.

Now you. Central Asian states have adopted Silk Road framing enthusiastically in their tourism, museums and national narratives. Why would they, and what does it cost them?

Then, in 2013, the phrase became the name of the largest infrastructure programme in the world. In September of that year, in Astana, Xi Jinping announced a Silk Road Economic Belt; in October, in Jakarta, a twenty-first century Maritime Silk Road. Together they became the Belt and Road Initiative: ports, railways, pipelines, roads and power stations financed and often built by Chinese firms across Asia, Africa, Europe and Latin America, with well over a hundred countries signing cooperation agreements and cumulative commitments generally estimated in the high hundreds of billions of dollars.

Example. Ask the analytical question rather than the political one: what work is the historical name doing in that context?

It supplies a frame, and the frame carries claims that the policy itself would have to argue for. The Silk Road connotes antiquity, so a new programme becomes a restoration of something old rather than an innovation. It connotes mutuality, since trade is exchange between willing parties, which positions the participants as partners rather than as borrowers or hosts. It connotes peace, because the popular image is caravans and monks rather than tribute, conquest and slavery, all of which this course has shown were also part of it. And it connotes centrality for the region the routes ran through, which is attractive to Central Asian states that spent the twentieth century as somebody's periphery. Note that this is not a claim about whether the programme is good or bad, which is outside the competence of a history course. It is an observation that the name was chosen from many possible names, that its connotations are favourable and specific, and that a reader who knows the history can see which parts of it the frame includes and which it leaves out. It is also worth knowing that the name is contested rather than owned: the United States announced a New Silk Road strategy for Afghanistan in 2011, and Turkey, Japan, India, Iran and the European Union have all used Silk Road language for their own connectivity projects. The phrase is a shared asset because it is nobody's and everybody wants it.

Now you. A museum label reads: "The Silk Road, a great highway of peaceful exchange linking East and West for two thousand years, brought silk to Rome and Buddhism to China." Rewrite it in one or two sentences you could defend.

What the name hides

Before the account is closed, it is worth listing what the phrase smuggles in, because a reader who has come this far can now see each item and few readers can.

It centres silk, one commodity among many, and one whose largest single mover was a government paying its own soldiers rather than a merchant selling to a foreign buyer. It centres the land, which carried the smaller share of the volume and by far the smaller share of the value. Its singular form implies a route, when the reality is a shifting web whose branches opened and closed with the politics of whoever held the passes. Its definite article implies one continuous thing, when the traffic was episodic and depended on who was paying for it in any given century.

And it implies exchange between willing parties, which is the most consequential of the distortions, because a great deal of what moved along these routes was not trade. Tribute moved, and protection payments, and the salaries of garrisons. Deported craftsmen moved, and hostage princes, and enslaved people sold under warranty in the market at Turfan. Armies moved. A bacterium moved. The word "exchange" covers all of it and softens most of it.

None of this makes the term unusable, and this course has used it throughout. It makes it a term to be used with the qualifications attached rather than a description that can be relied on to be accurate on its own.

The balance sheet

This is what the evidence supports, stated as plainly as it can be.

Goods. The long-distance overland traffic was small in volume, high in value per kilogram, and intermittent. It could not have been otherwise: a camel eats its own load in twenty-five days. Most exchange recorded in the documents is local, most of the value that moved between the ends of Eurasia went by sea, and in Central Asia the largest single injection of wealth was Chinese state spending on garrisons and horses rather than private commerce.

Faiths. This is the largest genuine transfer the routes accomplished, and it reshaped the religious map of half the world: Buddhism into China, Korea and Japan, where it remains; Islam into Central Asia, where it remains; Manichaeism and the Church of the East across the same ground, where they do not. The mechanism was institutional rather than mystical: universal membership, donation, a physical network of houses along the road, and a patron state.

Technique. Real, consequential and slow. Paper took about five centuries to travel from Samarkand to Italy, Indian numerals about six to reach Pisa, both at roughly ten kilometres a year, a thousandth of caravan speed, because a skill moves only inside a trained person. The exception is deliberate transfer, whether by an emperor sending for sugar boilers in 647 or a khan deporting engineers, and it is faster by orders of magnitude.

Disease. The best evidenced thing that ever moved along these routes, and the most consequential. A cemetery in Kyrgyzstan dated 1338 and 1339, and genomes published in 2022, place the origin of the second plague pandemic in Central Asia, and it went on to kill something between a third and a half of the population of Europe and the Middle East. Nobody chose to send it.

What the routes did not do. They did not create a Eurasian economy: the volumes were far too small, and no region's prices, wages or living standards can be shown to have moved with another's. They did not transmit technology in any general way; most techniques never crossed at all. And they were not a highway, a golden age, or a single thing with a beginning and an end.

What you can do now

You can take a claim about the Silk Road and ask what it rests on, where the evidence was found and how it was dated, which is the habit the second lesson tried to install and the one that does most of the work. You can estimate from first principles which goods could cross a continent overland and which could not, and check the answer against an excavated inventory. You can explain why a religion spreads along a trade route while its neighbour does not, and why a technique takes five hundred years to move the same distance a bolt of silk covers in four months. You can say what is known about the origin of the Black Death and how it came to be known. And you can read a sentence like the one on that museum label and take it apart.

That is more than most people who have heard of the Silk Road can do, and it is worth finishing with the reason the subject repays it. The routes did not make the ancient world into one place. What they did was demonstrate, over fifteen centuries and with a few hundred camels a year, that no part of Eurasia was ever quite sealed off from the rest, and the most important things that crossed were the ones nobody could carry.