In 1899, as the story is usually told, a scholar in Beijing noticed strange scratches on a "dragon bone" that a pharmacy was selling to be ground into medicine. The marks were writing. The bones were the questions that Shang kings had put to their ancestors about 3,200 years earlier, cracked with heat and then read. Chemists had been swallowing the oldest Chinese archive for generations.
That is the ancient historian's situation in one picture. We do not have "the past". We have what happened to survive, mostly by accident, in the places where it survived, and a toolkit for asking what it can and cannot tell us. A graduate of this course keeps the toolkit: where the evidence comes from, how it is dated, how many people it can really count, and which of the big arguments are still open.
This piece teaches how historians of the earliest cities work and what they currently think. It leans on estimates throughout, and says so. It does not argue current politics; where ancient places are claimed by modern ones, it describes the evidence.
Every ancient historian works with a biased sample, and the bias is physical. Baked and sun-dried clay survives fires and floods, so Mesopotamia left several hundred thousand cuneiform tablets, many still unpublished. Papyrus survives only in dry air, so Egypt's desert margins give us scrolls and Greece's damp does not. Maya books were bark paper folded like screens; Spanish clergy burned them in the 16th century, and only a handful survive (three undisputed ones, the Dresden, Madrid and Paris codices, plus the Grolier Codex, which a 2018 study accepted as authentic). The Indus cities left thousands of stamp seals with a script of roughly 400 signs, usually only about five signs long. No one has deciphered it. A lost literature may have been written on cloth, wood or leaves that rotted.
Add the question of who wrote. Temples and palaces kept the archives, so we know a great deal about grain and tribute and very little about how a farmer felt. Kings wrote about their victories. Later writers wrote about earlier kings with their own purposes: the Greek historians who tell the Persian side of the Persian Wars did not read Persian, and no Persian narrative of those wars survives.
The craft is to match the question to the source. Try it with three invented documents. Pick a question, then pick the document you would trust for it.
All three documents are made up to show the reasoning. The ratings are judgments historians would make, not a formula.
Without JavaScript: the receipt is the strong source for quantities and dates because it was made at the time for bookkeeping, not for show. The inscription is the strong source for what the king wanted believed, not for how many died. The late chronicle is weak unless it quotes something older that we can check.
Notice the rule underneath: a source is good or bad for a particular question. The humble receipt is the best evidence we have for ancient economies precisely because nobody was trying to impress anyone with it. A nice example is a papyrus logbook found in 2013 at Wadi al-Jarf on the Red Sea, kept by an official named Merer, which records boats hauling limestone to Giza in the reign of Khufu, about 2560 BCE. It shows the builders of the Great Pyramid as organised, paid labour crews, and it does it by accident, in a diary.
"The year the king built the wall" is not a date. Ancient people counted years by kings, by named years, by officials who gave their names to a year, or from some founding. Historians turn these into our calendar by chaining them. The Assyrians kept lists of officials, one per year. One list notes an eclipse of the sun in the month of Simanu. Astronomers can calculate that the eclipse was on 15 June 763 BCE, which fixes that entry, and by counting along the list in both directions, a stretch of Near Eastern history gets exact years. That single eclipse is a nail in a very large wall.
Other nails: the first securely dated event in Chinese history by the traditional record is the start of a regency in 841 BCE, which the Han historian Sima Qian's chronology supports; before it, early Chinese dates are reconstructions (the state-sponsored Xia–Shang–Zhou Chronology Project of the 1990s and 2000 proposed dates and drew criticism for how it chose among sources). The Sumerian King List shows another trap: it gives kings reigns of tens of thousands of years (one is credited with 28,800), then shrinks to reasonable lengths, and lists as consecutive some rulers who reigned in different cities at the same time.
Physics helps. Radiocarbon dating uses the fact that living things take in carbon-14, which then decays with a half-life of about 5,730 years. Measure how much is left and you can read off an age. Tree rings give a year-by-year check: wood from different trees is matched ring by ring into chains that now reach back about 12,000 years, and the carbon in each dated ring tells us how the atmosphere's carbon-14 changed. That is why a raw radiocarbon age must be calibrated (the standard curve is IntCal20, published in 2020), and why calibrated results come as a range, not a point. Play with the decay and the error.
Labs report conventional ages using the older Libby half-life of 5,568 years and then calibrate; this toy uses the physical half-life only to show the shape. The point: the same small measurement error is a few decades of uncertainty for a young sample and hundreds or thousands of years for an old one.
Without JavaScript: half the carbon-14 left means about 5,730 years; a quarter means about 11,460; one eighth means about 17,190. Because the curve flattens, a fixed measurement error costs far more years in old samples.
The Egyptian case shows the method working. Egyptologists dated the Old Kingdom from king lists and a few astronomical anchors; a 2010 radiocarbon study of short-lived plant remains from known reigns (Bronk Ramsey and colleagues, Science) largely agreed with the historical dates, with differences of a few decades. Two independent methods meeting is what confidence looks like in this field.
Cities and states appeared in several places that did not know each other. That is the first thing to see, and it is why this course is not about one cradle. Uruk in southern Mesopotamia was a real city by about 3300 BCE, with the earliest known writing tablets from its temple archives. Egypt was unified under one crown around 3100 BCE. The Indus cities (Harappa, Mohenjo-daro, Dholavira) were at their mature phase from about 2600 to 1900 BCE, with planned streets, drains and standard weights, and no obvious palaces. On the coast of Peru, Caral and its neighbours (the Norte Chico) built monumental platforms from roughly 3000 BCE with no pottery. In China, Erlitou (c.1900–1500 BCE) and the Shang (c.1600–1046 BCE, with the writing on the oracle bones) followed millennia of Neolithic towns like Liangzhu. In Mesoamerica the Olmec built San Lorenzo and La Venta from about 1200 BCE. In West Africa, the Nok culture made iron and astonishing terracotta figures between about 900 BCE and 300 CE. Aboriginal Australians and the voyagers of the Lapita culture in the Pacific had different, enormously successful ways to organise people. Not every society built a city, and not building one was not failing.
Move the slider and read what was going on, in parallel, in eight regions. The bars show when each entry is dated to; where bars overlap, two things were happening at once.
Dates are conventional, rounded and sometimes disputed (the Mesopotamian ones use the "middle chronology"). Early dates in China and the Americas are approximate.
| Around | West Asia | Egypt, Nubia, Africa | South and East Asia | Mediterranean and Americas |
|---|---|---|---|---|
| 3300 BCE | Uruk city, early writing | Egypt unified c.3100 | Liangzhu towns in China | Caral rises, c.3000 |
| 2500 BCE | Early city-states | Giza pyramids | Indus cities mature | Minoan Crete |
| 1200 BCE | Collapse of old palace world | Ramesses III; Nok begins | Shang at Anyang | Olmec San Lorenzo |
| 300 BCE | Hellenistic kingdoms | Meroë in Nubia | Maurya; Warring States | Rome expands; Preclassic Maya |
| 200 CE | Parthians, then Sasanians | Aksum rising | Han ends; Kushans | Roman Empire; Teotihuacan |
Ancient numbers are of two kinds. Some are claims in a text: Herodotus gave Xerxes' army as over five million fighting men, which no one believes. Others are estimates built from archaeology and a few censuses. The honest ones come with a range. The world's population around 1 CE is commonly put at roughly 200 to 300 million, and published estimates run from about 170 to 400 million. The Roman Empire at its height is put at about 50 to 70 million people. A Han census of 2 CE survives and gives roughly 58 to 60 million (the figure depends on how the Book of Han is read), one of the very few real counts from the ancient world, and it counted households for tax, so it undercounts the people the state could not reach.
For a city with no census, the recipe is area × density: measure the walled area from survey, then multiply by a plausible number of people per hectare taken from places where we do know. The trouble is that the plausible range is wide. Slide it.
Illustrative numbers. Real estimates for Uruk, Mohenjo-daro or Rome rest on the same recipe, with their own debated areas and densities, and they stay uncertain by a factor of two or more.
Without JavaScript: 100 hectares at 150 people per hectare is 15,000 people; the same area at 50 or 400 per hectare is 5,000 or 40,000. The density guess matters as much as the survey.
Ancient people also counted with other bases. Mesopotamian scribes wrote numbers in base 60, which is why we still have 60 minutes and 360 degrees. A school tablet from about 1800–1600 BCE in Yale's collection (YBC 7289) shows a square with its diagonal marked 1;24,51,10 in base 60. The reader below turns that into decimal.
Without JavaScript: 1 + 24/60 + 51/3600 + 10/216000 = 1.41421296, and the true square root of 2 is 1.41421356. The scribe was right to six decimal places, about six ten-millionths too low.
Why did people ever gather in cities, pay taxes and obey kings? Nobody knows the full answer, and the field has several serious theories. Karl Wittfogel (1957) argued that big irrigation works demanded bosses; the evidence for Mesopotamia and Egypt is mixed, because many early canals were local. Robert Carneiro (1970) argued that where farmland was hemmed in by mountains or desert, losers could not leave, so they submitted. James C. Scott (Against the Grain, 2017) stressed that early states were built on grain, which is visible, countable and taxable, unlike roots or herds. David Graeber and David Wengrow (The Dawn of Everything, 2021) argued that early cities often had no obvious kings and that people experimented with many ways of organising; critics replied that the evidence is thinner than the book's confidence and that some of its examples are contested.
Evidence keeps moving. At Göbekli Tepe in Anatolia, monumental stone enclosures were built around 9500 BCE by people who, as far as we know, were not yet farmers, which upsets the old order of "farming, then monuments". Writing was invented independently more than once: in Mesopotamia by about 3300 BCE and Egypt soon after, in China by about 1250 BCE (oracle bones), in Mesoamerica by roughly 500 BCE. That several unconnected peoples reached the same tools is the strongest argument that cities are not a freak accident.
The ancient world is a story of endings as much as beginnings, and historians argue about them in public. Around 1200 to 1150 BCE the palace world of the eastern Mediterranean broke: the Hittite capital was abandoned, Ugarit was destroyed, Mycenaean palaces fell, and Egypt, which survived, recorded fighting off the "Sea Peoples" in Ramesses III's temple at Medinet Habu. Eric Cline's 1177 B.C. (2014) argues for a systems collapse: earthquakes, drought, migration, rebellion and broken trade each hitting an interdependent network. Other scholars weight individual causes differently, and the climate evidence is itself argued.
Why did the western Roman Empire end?disputed Edward Gibbon (1776–89) blamed moral and military decay and Christianity. Peter Brown (1971) and the "late antiquity" school saw a long transformation, not a fall. Bryan Ward-Perkins (2005) argued the end was violent and that living standards really collapsed, citing the archaeology of pottery and building. Peter Heather (2005) stressed external pressure, especially the Huns pushing others into the empire. Kyle Harper (2017) added plague and climate: the Antonine Plague (from 165), the plague of Cyprian (c.249–262) and the Justinianic plague (from 541).
What is not in dispute: in 476 Odoacer removed the last western emperor, Romulus Augustulus, and the central government in the West ended; the eastern empire, which we call Byzantine, continued to 1453. And the Roman story is only one of several endings of this period. The Han state fell in 220 CE and China split into rival states for centuries; the Gupta kingdom, the Kushite kingdom of Meroë and the empire of the Parthians each ended on its own timetable. A graduate sees "the fall of Rome" as one case among many, not as the template.
The ancient world is reconstructed from a biased sample of surviving things, dated by chains with a few firm anchors, and counted with ranges. Cities and states arose independently in several places, and the big arguments, from why states formed to why Rome fell, are open in useful ways.
Try this: next time you see a headline claiming a "5,000-year-old" discovery, ask which kind of dating it used and whether it gives a range.
Sources. Assyrian eponym lists and the eclipse of 15 June 763 BCE; the Sumerian King List (Electronic Text Corpus of Sumerian Literature, Oxford); radiocarbon half-life 5,730 years and IntCal20 (Reimer et al., Radiocarbon 2020); Bronk Ramsey et al., "Radiocarbon-based chronology for dynastic Egypt," Science 2010; Wadi al-Jarf papyri (Tallet 2017); the 2018 Grolier Codex study (Carter et al., Maya Archaeology); Van De Mieroop (2015); Cline (2014); Gibbon, Brown, Ward-Perkins, Heather and Harper as cited in the text; Scott (2017) and Graeber and Wengrow (2021); Göbekli Tepe (Schmidt; Dietrich et al.); census figures from the Book of Han as discussed in Scheidel and Hopkins; YBC 7289 (Yale Babylonian Collection). Population ranges are from standard compilations (McEvedy and Jones 1978; HYDE). Dates for the Shang and early Zhou are conventional.