Start of the Silk Road: The Greco-Chinese War and the Romans of Gansu

Just over 100 years after the unification of  ‘China’ under the First Emperor, the bellicose Han Dynasty began to extend tendrils of diplomacy, trade and conflict west. The states and peoples encountered offered the first indirect links to the flourishing Western states of Greece and the Roman Empire poised to replace them – the nascent Silk Road – but also an interesting footnote that may have led to the only clash between the the European classical world and imperial China.

China Rising

In the late 2nd century BCE, Han China was in the ascendant. Under the leadership of the ‘martial’ Emperor, Han Wudi, Han generals had repudiated decades of passive acquiescence to take the fight to their nomadic bête noire, the Xiongnu, whose depredations from the north had cast a deep shadow over the cities and people of the central plains. After a series of startling victories the Xiongnu had been driven from the Ordos loops north of the Yellow River, while for the first time Chinese armies began to open up the Hexi Corridor through present Gansu and Xinjiang into the oasis cities of the Taklamakan desert – what would become the Silk Road.

In 104 BCE Emperor Wudi of Han China began his most audacious gambit yet. Under the leadership of general Li Guangli, Han armies set out to conquer the Central Asian kingdom of Dayuan 大宛, separated from the the capital Chang’an by over 2,500 miles of foreboding deserts, mountains and fiercely independent city states. The war is ascribed by Sima Qian to a desire to obtain the 天马, the famed blood sweating ‘Heavenly Horses’ of the Ferghana Valley, but it’s likely geopolitics paid a bigger role – a show of force projection so far into unknown territory to demonstrate the unstoppable reach of the imperial armies.

Han Expansion.png
Han expansion under Wudi – Dayuan is transcribed as ‘Tayuan’  (WikiCommons)

The war dragged on for a further three years. When the Dayuan capital of Ershi 贰师 finally fell to Han forces in 101, it immediately sent shockwaves through the region:


“And after the Ershi General [Li Guangli] conquered the Dayuan, the Western Regions shook with fear, and sent many envoys [to Chang’an] to pay tribute.” (Hanshu 96a)

In doing so a new hegemon was established in Central Asia, the Han both guaranteeing territorial security and stimulating an increasing exchange of goods, information and people between Chinese East Asia, Indo-Aryan Parthia and its satellite states, and eventually Imperial Rome. A little over a century later, the scale of trade led Roman historian Pliny the Elder to lament:

“minimaque computatione miliens centena milia sestertium annis omnibus India et Seres et paeninsula illa imperio nostro adimunt: tanti nobis deliciae et feminae constant.”

“And at the lowest estimate, each year India, the Chinese and the Arabian peninsula draw one hundred million sesterces from our Empire: that’s how much our luxuries and women cost us.” (Hist.Nat.12.41)

Empires’ Most Distant Outpost

So who were the people of Dayuan, whose capital Ershi was to prove a crucial hinge between East and West?

From detailed Chinese accounts, a location roughly in the Ferghana Valley is largely beyond dispute. Straddling modern Tajikistan and Uzbekistan, far from a historic terra incognita this had for centuries been in the orbit of a decidedly foreign kingdom – the Greco-Bactrians, ruled by the descendants of Alexander the Great’s Macedonians who had conquered the area in the early 320s BCE. Bringing his armies into the Ferghana after vanquishing Persia, Alexander would fight a two year guerrilla war against the Scythians, in the process founding a walled city on the Jaxartes (Syr Darya) river which he settled with Greek and Macedonian mercenaries (Arrian Anabasis 4.4). Now modern Khujand, Takijistan, the new foundation would be dubbed Ἀλεξάνδρεια Ἐσχάτη – ‘Alexandria the Furthest’ – and marks the furthest ever Greek expansion into East Asia.

Two centuries later, extensive numismatic evidence shows warrior kings of neighbouring Bactria maintaining fiercely Greek self expression, while archaeological studies of their cities in present Afghanistan such as Ai Khanum attest to deep and sophisticated Greco-Hellenistic identity in Central Asia up until the mid-2nd century. The area around Bactra (modern Balkh, around 300 miles south of the Ferghana Valley) was still demonstrably Greek ruled (at least) at the time of Han Wudi – coins give us the names and identities of his Greco-Bactrian contemporary Heliokles I and his neighbour the Indo-Greek Menander I.

Silver tetradrachm of the Bactrian king Eucratides I (r. 171-145 BCE) (WikiCommons)

The etymology of Dayuan (literally ‘The Great Yuan’) gives us our first clue that the people the Han encountered were related to their Greek neighbours south of the Pamirs. Yuan (perhaps /ʔʉɐn/ or /ĭwɐn/ in Middle Chinese) sounds like the contemporaneous Sanskrit ‘Yavana’, the common term in Indian texts to refer to the Into-Greek and Greco-Bactrian kingdoms of Central Asia (itself ultimately deriving from the Greek ethnonym of Ἴωνες /iōnes/ ‘Ionians’). Indeed, the descriptions given of Dayuan by the envoy Zhang Qian, who Wudi had sent to explore the region, could describe a Greek-like people (although equally fits Indo-Iranian Sogdians or Bactrians):


“Around the Yuan [region] people make wine from grapes, with the rich storing up over 10,000 stones [a liquid measure] which can last for ten years without spoiling. The people are very fond of wine... Although the states from Dayuan west to Anxi [Parthia] have different languages, their customs are similar and languages mutually intelligible. Their people all have deep eyes, profuse beards; their merchants are excellent and will fight for tuppence…” (Shiji 123)

Some historians of Han China (e.g. Loewe, Dubs) discount the idea that the Dayuan encountered by Zhang Qian and Wudi’s forces still retained any Greek character, and for good reason. Zhang Qian’s mission to Central Asia and Wudi’s invasion was in part to seek allies created by a seismic political shift over the preceding decades: the ingress of a new nomadic people, the Indo-European Yuezhi 月氏 (ancestors of the Kushans and called the Tokharioi in Greek sources on the end of the Greco-Bactrians). Displaced by the Xiongnu from their home in Gansu in the 170s, masses of Yuezhi and other displaced peoples had poured into Dayuan and would eventually snuff out the last Greco-Bactrian king, Heliocles I in the 130s.

The loss of Greek-led political sovereignty over the wider Ferghana and Bactrian regions is therefore fairly clear cut. But culturally and perhaps ethnically, there are indications that Dayuan continued to hold onto its Greco-Macedonian roots.

For starters, Zhang Qian reports that Dayuan was comprised of walled cities, ruled by local kings (unlike the nomad people pouring into the region, but also broken up from the grand Hellenistic Kingdoms they succeeded). Most interestingly he also notes its customs, governance and cities were the same as those of Daxia 大夏 to the south – the Chinese term for the Greco-Bactrian kingdom which coins show had remained Greek-ruled until less than a decade before his survey.

This brings us to point two, the issue of time. In 208 BCE, Seleukid King Antiokhos III journeyed north east from his capital in Syria (his famous ‘anabasis’) to reimpose rule on the breakaway Greco-Bactrian kingdom of Euthydemus. The kingdom he encountered in Afghanistan, as related by Polybius (11.34), is a fully functional Hellenistic ‘great power’ led by Greek speakers leading Greek-style armies. He was met with a prescient prediction from King Euthydemus:

“πλήθη γὰρ οὐκ ὀλίγα παρεῖναι τῶν Νομάδων, δι᾽ ὧν κινδυνεύειν μὲν ἀμφοτέρους, ἐκβαρβαρωθήσεσθαι δὲ τὴν χώραν ὁμολογουμένως, ἐὰν ἐκείνους προσδέχωνται.”

“For the nomad horde is considerable, threatening both of us, and should they be let in its commonly agreed our lands will be utterly barbarised.” (Plb.11.34.5)

Within a century this menace would arrive.  What is interesting for our purposes is that according to Polybius, almost within Zhang Qian and Han Wudi’s parents’ lifetimes Central Asia hosted a distinctly Hellenistic kingdom still tied to its Mediterranean metropoleis, contrasted in traditional Greek racial terms with the Yuezhi and Saka who would topple it. To put another way: chronologically, Zhang Qian could (just about) have met an elderly local who as a child witnessed Antiokhos’ entry to Bactra. And their borders lay within a week’s journey of the Ferghana Valley city attacked by the Han, with which it shared close historic political links: Apollodorus of Artemita (contemporaneous to Wudi) is even recorded as saying that the Greco-Bactrian kingdom extended as far as China (Strabo.11.11.1) , which allowing for embellishment may imply continued claim over their northern neighbours well into the 2nd century. So we are not talking about vastly different epochs between Han entry into Central Asia, and previous Greek hegemony.

Our third reference is Quintus Curtius, the (likely 1st century CE) Roman historian of Alexander the Great. There are numerous historiographic issues with Curtius’ works, but he clearly had access to extensive sources and lived alongside well-informed authors on the east such as Strabo, Pliny and Apollodorus. When describing Alexandria the Furthest in the Ferghana, Curtius notes:

“….quorum posteri nunc quoque non apud eos tam longa aetate propter memoriam Alexandri exoleverunt.”

“…whose descendants even now, because of their memory of Alexander, after such a long period of time have not disappeared.” (Plb.11.34.5)

Greek-style soldier on the Sampul tapestry, with centaur above – Xinjiang Museum, photo WikiCommons


The Latin here is not immediately obvious – “posteri…exoleverent” could be just referring to local recensions of Alexander myths in Central Asia. However, contemporary Latin authors mainly use the verb much more concretely to describe the loss of cultural rites – e.g. clothing fashions (Tac.Ann.14.21), traditional religious practices (Tac.Ann.11.15), old civic customs (Suet.Gal.4). A very plausible reading therefore is therefore that around the reign of Claudius (43 CE) Curtius believed the Ferghana valley still held Greeks who in some way retained their Alexandrian cultural identity; and even if his sources were out of date, this may make the Greek identity of the Dayuan people encountered by the Han a century prior to Curtius writing plausible .

The final point is a tantalisingly hard to date tapestry (left) excavated in the 1980s in Lop county, southern Xinjiang (near present Khotan). The contents are overtly Greek – a diadem wearing, Caucasian (note the blue eyes) soldier clutching a spear with tip much like a Macedonian sarissa, and with a centaur motif on the band above.

For me, this shows that at some point there were Macedonians trading produce with or influencing artisans in Central Asia – overwhelmingly likely neighbouring Greco-Bactrians. And, given trade routes (below) and the intervening Pamir range, these Macedonians were far more likely to be the Ferghana valley based Alexandrians described by Curtius than those from further south.

The easiest trade route from Khotan runs north to the Ferghana Valley cities of Andijon and Kokand, before heading on to Buchara and eventually Baktra. Khujand, the site of Alexandria the Furthest, lies on the road between Samarkand and Kokand (image WikiCommons)

The devil is in the detail: due to looters destroying the archaeological context, date estimates for the tapestry range between 3rd century BCE to 4th century CE. Earlier proves nothing for our argument – its beyond dispute that Greeks were in the Ferghana around 300 BCE, and this is a long time from Wudi’s invasion. For historical reasons I am more inclined to a date in the centre of this range shortly before Han expansion – when trade routes to the Tarim basis oasis cities were being opened up, but before the nomad incursions into Central Asia. If so, this is perhaps the strongest evidence for a Greek identification with the Dayuan – and the tantalising possibility of serried Macedonian phalanges meeting Li Guangli’s Han troops in 104.

So where was Ershi? As I have alluded to above, in principle I think the most likely identification is Khujand – i.e. Alexandria the Furthest. Archaeological and historical evidence tells us this was the biggest Greek colony in the Ferghana, and if we agree on a Greek identity for ‘Great Ionian’ Dayuan it would be the most logical location for the capital. The 3rd century Weilue 魏略 by Yu Huan 魚豢, an account of later Silk Road routes, almost certainly mentions a powerful Alexandria the Furthest in the Dayuan region (see the excellent linguistic argument for its identification under §24 here). However, when describing Dayuan the Hanshu chronicles actually refers to a 貴山 Guishan as the capital (HS.96 “大宛國,王治貴山城”), elsewhere in the section only referencing 貳師 Ershi obliquely as Li Guangli‘s title (the ‘Ershi General’) in his war against Dayuan. Conversely, the Shiji of Sima Qian clearly describes Ershi as a city, origin of the heavenly horses and the target of Li Guangli’s campaign (e.g. Shiji.123: 宛有善馬在貳師城“). Finally, another part of the Hanshu (in the annals of Wudi’s reign) describes Guangli executing a king of Dayuan, presumably after conquering a capital at Ershi (HS.6 “貳師將軍廣利斬大宛王首”) and the choice of a new ruler from local nobles. 

I don’t know what to make of this – there have been numerous suggestions for the locations of Guishan and Ershi (tending to assign Khujand, i.e. Alexandria the Furthest, to at least one); perhaps Ershi was strategic military fortress-cum-bolthole, or we are dealing with two competing toponyms for the same place (cf. the well known problem with historic Troy as Ilion/Wilusa or Troy/Troia). From my inexpert Classical Chinese, I would also note that Ershi 貳師 literally means ‘second [military] division’ and I do wonder if there has been textual corruption or poor readings of the source material originally referring to a military detachment or commandery.

Nonetheless, I think we can still draw some tentative conclusions. Firstly, the Kingdom of the Heavenly Horses had until very recently been Greek ruled, and likely still maintained a recognisably Greek character (and perhaps even Greek style army) when the Han went to war with it. Secondly, the capital of Han Wudi’s furthest western conquest was by historic accident also the site of Alexander the Great’s furthest eastern colony – and may well have been the same city taken in 101. 

Crassus’s Lost Legions

For the final part, we jump forward 70 years to a battle on the Han frontier, and to the 1940-50s and the theory of American Sinologist Homer H. Dubs.

Dubs noted (see bibliography) an unusual passage from historian Ban Gu in the Hanshu, recalling the Battle of Zhizhi in 36 BCE between the Han and Xiongnu somewhere on the Talas River in southern Kazakhstan. As the Xiongnu forces leave their city;


“…and over one hundred horse riders arrayed below the city, and over a hundred infantry came out of the gate in fish scale formation, and practised their manoeuvres” (HS.70)

Dubs’ leap (allegedly spurred on by great William Tarn) was to first assume the unusual formation above could refer to the Roman testudo ‘tortoise’ formation of interlocking shields. But whence Romans in Kazakhstan? Dubs then pointed to the Battle of Carrhae in 53 BCE, 17 years previously, where over 10,000 Romans soldiers were taken prisoner after the Parthians obliterated an army led into Syria by the Roman triumvir Marcus Licinius Crassus. Pliny refers in passing (Hist.Nat.6.18) to King Orodes bringing these prisoners to Margiana (Merv, present Turkmenistan), while Horace laments on their marriage to Parthian barbarae (Odes.3.5). Dubs contended that these Romans fled their captivity, enlisting as mercenaries in Xiongnu service and leading to their appearance at the Battle of Zhizhi.

The second leap was what happened to these soldiers. The Hanshu records that after the battle, defeated Xiongnu troops were distributed to allied kingdoms in the region. Dubs noted that a 5 CE Han register lists a town called Liqian 驪靬 in Gansu; Liqian is a non-Chinese loanword used elsewhere to refer to Rome, from which Dubs hypothesised this was where the lost legionaries were settled. This part of the story has had a long life, cropping up regularly in newspapers and journals, spurred on by a prevalence of supposedly Caucasian features around the site of Liqian in modern Yongchang County, Gansu (which has recently erected a theme park-style ‘European town’ and even funded a terrible romcom inspired by the myth). 

While his paper is a cracking read, needless to say there are numerous problems with Dubs’ imaginative theory both from Roman history and modern evidence (genetics show the people of Liqian, while maybe connected to the ancient Caucasian people of western China such as the Yuezhi, are certainly not Romans). However, in light of the above discussion on Dayuan, there are a few intriguing further considerations to note:

  • The Hanshu puts the battle of Zhizhi as near Kangju – the same territory as Yu Huan’s Weilue (above) sites Alexandria the Furthest. Modern Taraz, a proposed site for the battle, is only 200 miles from Khujand and Dayuan.
  • Dubs etymologises Liqian, which is used elsewhere in Han documents for Rome, as a transcription of Alexandria (with the Egyptian metropolis used as metonym for the whole Empire). This must be right; but there is equally no reason it cannot also be right for Alexandria the Furthest.
  • Li Guangli may have defeated Dayuan 70 years previously, but this was not a Mongol-style obliteration – the territory was left intact, with a new local king chosen and a tribute relationship with Chang’an.

In his 1957 article Dubs roundly dismissed a Greek identity for the fishscale formation at Zhizhi and the ancestors of Liqian, asserting that interlocked Macedonian aspis shields would not achieve this image. Perhaps he is right; I would also contend we should be flexible in visualising a three-character Chinese hapax legemon. Either way, my sense is we should first look to the embers of Greek civilisation in Central Asia, and their fascinating possible interactions with a newly risen China, when mapping out the hints and suggestions of east meets west in the first century BCE.


Most of the above comes from my own (perhaps hopelessly naive) reading of primary source material, but I also found the following useful when putting together:

Twitchett, D and Loewe, M (eds.) (1986) The Cambridge History of China I: The Chi’n and Han Empires, 221 BC-AD 220 

Sampson, G. (2008) Defeat of Rome in the East: Crassus, the Parthians, and the Disastrous Battle of Carrhae, 53 BC

Dubs, Homer H. (October 1957). “A Roman City in Ancient China”. Greece and Rome Vol.4 no.2  

Hill, John E. (2004) The Peoples of the West from the Weilue 魏略 by Yu Huan 魚豢: A Third Century Chinese Account Composed between 239 and 265 CE Quoted in zhuan 30 of the Sanguozhi Published in 429 CE (Draft English translation) at 




Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s