The Sogdians: The Silk Road's Merchant Civilisation
Who the Sogdians were, what their 313 CE letters and Afrasiyab murals reveal, why they faded, and where to see Sogdian sites in Uzbekistan and Panjikent.
"I would rather be a dog's or a pig's wife than yours!" That's Miwnay, a Sogdian woman stranded in Dunhuang around 313 CE, writing to the husband who'd left her there with their daughter and no money (translation: Nicholas Sims-Williams). Her letter never reached him. It got no further than a Chinese watchtower on the edge of the desert, and it's a big part of why we know anything personal about the Sogdians.
The Sogdians were an Iranian-speaking people from the Zeravshan valley, the stretch of river that runs through Samarkand and Bukhara, and between roughly the 4th and 8th centuries CE they ran the busiest middle section of the Silk Road. They left no empire and no single monument you can queue for. What they left is scattered, much of it in Uzbekistan, and this page shows where.
Where Sogdiana was, and what it wasn't
Sogdiana wasn't a country in the modern sense. It was a string of oasis city-states along the Zeravshan: Samarkand (the largest), Bukhara downstream, Panjikent upstream in what's now Tajikistan, Kesh (today's Shakhrisabz) to the south, and dozens of smaller walled towns such as Paykend and Varakhsha. Each had its own ruler; Bukhara's called themselves Bukhar Khudat, "lord of Bukhara".
Written references go back to the 5th century BCE, and Alexander took their chief city, Marakanda, in 329 BCE. But the Silk Road story belongs to the 4th to 8th centuries CE, which the Smithsonian's Sogdians: Influencers on the Silk Roads exhibition treats as their golden age.
Think of them less as a nation than as a family firm with branch offices across Asia.
A mail bag that never arrived: the Ancient Letters
In 1907 the archaeologist Aurel Stein was working along the ruined Han-era wall west of Dunhuang when he found a bundle of paper letters in a watchtower roughly 90 km west of the town, just beyond the Jade Gate. There were five more or less complete letters and fragments of others, written in Sogdian. After decades of argument, the dating argued by Sims-Williams and colleagues, 313 or 314 CE, is the accepted one. The University of Washington's Silk Road Seattle edition reproduces his translations of the key letters.
Why do they matter so much? Because they're ordinary. Letter 2 is a merchant's report to his partner, and its envelope says it was to go to Samarkand. It reports Luoyang burned and the emperor fled, then gets on with debts and stock. It reads like a regional manager's memo written during a war.
Letters 1 and 3 are Miwnay's. She had followed her husband Nanai-dhat from Sogdiana to Dunhuang against her family's advice. He then left, and by the time she wrote she'd been stuck there for three years, turned down by relatives and reduced to depending on a local priest.
One practical note: the letters aren't in Uzbekistan. They're held by the British Library in London. You'll see reproductions and references in Samarkand, not the paper itself.
How the trade network actually worked
Sogdians didn't just pass through China, they lived there. By the early 4th century there were Sogdian communities in the Gansu corridor towns (Dunhuang, Wuwei, Jiuquan) with families, local leaders and regular post back to the home cities. Later they turn up across northern China and the steppe as merchants, envoys and interpreters for Turkic rulers.
What moved along the network? The Smithsonian's summary is a good checklist: silk from China, horses from the Sogdian lands, gemstones from India, musk from Tibet and furs from the northern steppe. At home most Sogdians followed a local Zoroastrianism, hence the ossuaries (clay bone boxes) in Samarkand's museums; abroad they carried Buddhism, Manichaeism and Christianity eastward.
Our take: the Sogdians' real product wasn't silk. It was trust. A system of family firms, letters, credit and people who spoke several languages is what let goods move 4,000 km without any single state controlling the road.
The Ambassadors' Painting at Afrasiyab
If you see one Sogdian thing in Uzbekistan, make it this. In 1965 road builders cutting through the Afrasiyab plateau in Samarkand hit painted plaster, and it turned out to be a reception room in an aristocrat's house with murals on all four walls. The main wall shows foreign delegations arriving at the court of Varkhuman, king of Samarkand. The north wall has a Chinese scene with a boat on water; the east wall is usually read as Indian.
The dating trips people up, so here's what the sources say. A Sogdian inscription on the painting names Varkhuman and quotes Pukarzate, a chancellor of Chaganiyan (the region around modern Denau and Termez), announcing his embassy. That embassy is usually placed around 648–651. The painting itself is generally dated to about 660 CE, roughly the time Varkhuman was recognised by the Tang court. So the scene may record an event a decade before it was painted. "Around 660" is the safe answer.
Entry to the Afrasiyab Museum is 80,000 som for foreign adults on the museum-reserve's 2026 ticket listing, and the Samarkand State Museum-Reserve lists opening hours of 09:00–18:00 (last checked: September 2026). Our longer guide to the Afrasiyab mounds and museum covers what each wall shows and how to walk the site.
The script that outlived the language
Sogdian writing came from Aramaic, the administrative script of the old Persian empire. Here's the odd part: long after people stopped speaking Sogdian, its alphabet kept travelling. A cursive Sogdian hand was adapted by the Uyghurs for their Turkic language; the Uyghur script was then taken up for Mongolian in the 13th century, traditionally credited to a Uyghur scribe in Genghis Khan's service; and in 1599 the Manchu leader Nurhaci ordered the Mongolian script adapted for Manchu. Traditional Mongolian script, still taught in Mongolia and used in Inner Mongolia, is in effect a grandchild of the letters Miwnay wrote in.
The spoken language has one living descendant. Yaghnobi, spoken in the Yaghnob valley of Tajikistan's Zeravshan mountains and by resettled communities further south, is generally treated as a surviving form of a Sogdian dialect. Speaker estimates run to around 12,000, though counts vary and many of those people also use Tajik daily.
An Lushan: the Sogdian who nearly broke the Tang
The most famous Sogdian in history probably never saw Sogdiana. An Lushan was a Tang general of Sogdian and Turkic parentage (his surname "An" was one of the Chinese names given to people from Bukhara) who commanded frontier armies in the northeast. In 755 he rebelled, took Luoyang, declared himself emperor of a new Yan dynasty in 756, and was murdered in 757. The war dragged on until 763 and gutted the Tang state.
For Sogdians in China, the aftermath was bad. Foreign-looking culture fell out of favour, there were killings of "Western barbarians" in some cities, and many families of Sogdian origin blurred or rewrote their ancestry.
So why did the Sogdians disappear?
Short answer: they didn't, exactly. They stopped being called Sogdians.
The Arab conquest came first. Qutayba ibn Muslim took Samarkand in 712, and ten years later the ruler of Panjikent, Dewashtich, who'd tried to hold out, was captured and executed in 722. His archive, hidden in a castle on Mount Mugh in the mountains east of Panjikent, turned up in 1933 and is one of the largest groups of Sogdian documents ever found. Panjikent limped on until the 770s, then emptied.
Then came the slower change. Under the Samanids, ruling from Bukhara in the 9th and 10th centuries, New Persian (the ancestor of modern Tajik) became the language of court, literature and religion; Rudaki, often called the first great Persian poet, was born near Panjikent. Sogdian faded from everyday speech, and Turkic dynasties later redrew the language map again. Today's Tajik and Uzbek populations of the Zeravshan valley are, in large part, the Sogdians' descendants under new names, which is a theme our piece on Samarkand's Tajik-speaking side picks up.
Where to see Sogdian traces today
Be realistic: Sogdian sites are mostly mud-brick weathered into mounds, and the Hermitage in St Petersburg holds large parts of the Panjikent and Varakhsha murals. What's left still rewards anyone who's read the story first. In 2023 UNESCO inscribed the Silk Roads: Zarafshan-Karakum Corridor, a cross-border listing covering Uzbekistan, Tajikistan and Turkmenistan that includes ancient Panjikent, Paykend, Varakhsha and Kafir Kala among its components.
Kafir Kala
This fortress guarded the southern edge of the Samarkand oasis, and it burned during the Arab conquest, some time between 712 and 738. The fire is why it's famous. In 2017 excavators in the citadel's small throne room found carved wood that had charred rather than rotted: a door panel with 46 human figures in four tiers worshipping the goddess Nana, seated on a lion throne. After restoration it was shown at the Louvre from November 2022. The panel isn't displayed at the fortress, which lies about 12 km south of central Samarkand, so what you'll see there is the citadel mound and the view across the oasis. Take a taxi and bring water; there's no shade.
Varakhsha and Paykend, near Bukhara
Both sit at the desert edge of the Bukhara oasis and both died when water stopped reaching them. Varakhsha was the palace town of the Bukhar Khudat. Its Red Hall had a frieze 9.02 m long showing a rider on an elephant fighting beasts and monsters, according to Tashkent's Center for Islamic Civilization, and the main fragment is now in the Hermitage. Paykend is better exposed and has its own museum; our Paykend day-trip guide covers the taxi logistics. If you only have time for one, pick Paykend.
Panjikent, across the border
This is the most complete Sogdian town anywhere. Excavations started in 1946 and ran for decades under Aleksandr Belenitsky and then Boris Marshak, who directed the work from 1978 until he died on site in 2006. You walk real streets past house walls and temple foundations, for a small fee in somoni. Most of the original paintings are in the Hermitage and in Dushanbe's National Museum of Antiquities, but the Rudaki museum in town has finds and copies. It's a day trip from Samarkand if you can enter Tajikistan visa-free; see our border crossings guide.
Frequently Asked Questions
Who were the Sogdians in simple terms?
They were an Iranian-speaking people from the oasis cities of the Zeravshan valley, chiefly Samarkand, Bukhara and Panjikent. Between the 4th and 8th centuries CE they were the Silk Road's leading merchants, running family trading firms with communities across northern China and the steppe. Their language served as a common tongue of trade across much of Central Asia.
Are Uzbeks and Tajiks descended from the Sogdians?
In large part, yes, though not neatly. The people of the Zeravshan valley kept living there after the Arab conquest; they switched first to Persian (the ancestor of Tajik) and, in many areas, later to Turkic speech. So today's Tajiks and Uzbeks around Samarkand and Bukhara both carry Sogdian ancestry. Only the Yaghnobi still speak a language descended from Sogdian.
Where can I see the Sogdian Ancient Letters?
Not in Uzbekistan. The letters Aurel Stein found in 1907 near Dunhuang are held by the British Library in London, which has also made them available as digitised images. In Samarkand you'll only meet reproductions; for original Sogdian objects, go to the Afrasiyab Museum.
Is the Afrasiyab Museum worth it if I only care about Sogdians?
Yes. It's the single best Sogdian stop in Uzbekistan because the Ambassadors' Painting is displayed on the site where it was found, and the museum also holds ossuaries, coins and pottery from the old city. Allow about 45 minutes inside, then walk up onto the mounds for the view south over Samarkand.
If you're putting together a history-led route, pair this page with our era-by-era Silk Road history of Uzbekistan. Knowing the Sogdians first makes Bukhara's Samanid brick and Samarkand's Timurid tile read as later chapters of one valley's story.