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.
Answer
The difference is in how the two routes were paid for, and it is a clue worth following. Southern India traded by sea with the Roman world directly, in a two-party relationship where Roman buyers had a persistent deficit: they wanted pepper, gems and textiles, and India wanted rather little that Rome made, so the balance was settled in coin, and Pliny complains about exactly this drain. Coin that arrives to settle a deficit and is not spent on return cargo stays where it lands, often melted or hoarded as bullion, which is what the south Indian hoards are. The overland route, by contrast, was a relay: no Roman merchant paid a Chinese seller, so Roman coin had no reason to travel the full distance, and each leg was settled in whatever the local medium was, which in the Tarim meant Chinese coin, bolts of silk and grain. The absence of Roman coins in Central Asia is therefore evidence about the structure of the trade rather than about its volume, and it is the sort of negative evidence that is easy to misread as showing that nothing crossed.
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.
Answer
The strongest objection is selection: both are Chinese-administered oases at the eastern end of the system, both preserved documents because they were dry, and both were provincial towns rather than commercial capitals. Samarkand, Merv, Balkh, Kashgar and Chang'an, where the large-scale business would have been transacted, have left no comparable documentary hoard, so the argument compares the recorded commerce of villages against an imagined commerce of cities and finds the villages wanting. There is a second and subtler version: a long-distance trade run by literate merchants keeping their own records in their own language, and settled between principals rather than through local markets, would generate very few documents in a Chinese prefectural archive even if it were enormous. Neither objection defeats her, because her claim is comparative and modest, that the visible traffic is mostly local and that the long-distance component is smaller than the traditional picture assumes, and no evidence has been produced on the other side. What the objections do establish is that the finding is a lower bound on long-distance trade rather than a measurement of it, and that a single archive from a major entrepot could shift the picture substantially.
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 bolts. At 0.35 kg a bolt that is 140,000 kg, or 140 tonnes, which is 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?
Answer
You would want a series of the same measurement at the same place over time. The realistic candidates are customs or market tax receipts from a fixed point, which would give value passing and sometimes quantity; coin finds by stratum, which track circulation independently of any text; distributions of cheap, identifiable, indestructible imports such as glazed ceramics, which is exactly how maritime trade volumes are estimated; and shipwrecks, where the whole cargo is preserved at once. The overland route supplies almost none of these. There are no runs of customs records anywhere in Central Asia. The bulk goods that would leave ceramic evidence could not cross overland at all, by the freight argument, so the trade is nearly invisible in the one class of material archaeologists can count. And a caravan that fails does not sink intact: it is unloaded and dispersed. This is why the honest literature argues about the scale of the overland trade with orders of magnitude rather than percentages, and why the maritime evidence, when a later lesson reaches it, is so much better.
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.