+++
title = "They Left the Other Ninety"
description = "The most visible site can carry the best measurements and the worst representativeness. Civilizational collapse studies and site records both require the writer to say what was sampled, what survived and what remains unopened."
date = 2026-08-19
draft = false

[taxonomies]
topics = ["ancient-civilizations", "archaeological-sites"]
kinds = ["survivorship", "attestation"]

[extra]
tier = "public"
schema_type = "Article"
canonical = ""
hero = "articles/they-left-the-other-ninety/hero.webp"
hero_alt = "Partly catalogued rack of rolled archaeological survey plans"
diagram = "diagrams/they-left-the-other-ninety.svg"
podcast = ""
mechanism = "Archaeological sites record which ground was excavated and how it was dated, while collapse studies distinguish proxies, surveys, archives, and skeletal evidence; in both systems a precisely measured fragment does not automatically represent the whole."
discriminating_test = "Map observed coverage against the target universe, then compare the conclusion after weighting or excluding the best-covered segment; dependence on that segment supports an excavated-part diagnosis."
false_friend = "A genuinely concentrated system can also have a few dominant units, but independent coverage of peripheral units establishes that their omission cannot change the conclusion."
symptoms = [
  "A few well-covered issuers stand in for an entire supply chain",
  "A market-wide thesis rests on firms with the longest history",
  "A clean panel result comes from the most disclosed regions only",
  "A site percentage is reported without saying whether conservation, funding, or deliberate restraint produced it",
  "A regional proxy is presented as evidence for a whole civilization",
]
sources_verified = false

[[extra.faq]]
q = "How can a site be well measured but still unrepresentative?"
a = "A measurement can be exact about the ground it covers and silent about the rest. Göbekli Tepe was roughly 10% excavated in 2021 because the research strategy preserves unexcavated ground for future techniques. Pompeii is roughly two-thirds excavated, with major work effectively halted since 1960 for conservation. Those percentages describe different decisions and cannot establish that either exposed portion is a miniature of the whole site."

[[extra.faq]]
q = "Does a precise drought record prove why an entire civilization declined?"
a = "No. Evans et al. reconstructed annual precipitation declines of 41–54% from gypsum isotopes in Yucatán lakes in Science in 2018, with peak years reaching reductions of up to 70%. That is strong evidence for those lake basins. It becomes a civilizational explanation only after a separate regional-representativeness argument connects that record to the settlements, institutions, and populations being described."

[[extra.faq]]
q = "What should happen when a collapse argument rests on a count of destroyed sites?"
a = "Each destruction event needs its own date and causal support before the count is treated as a system-wide result. Of 153 events once attributed to the Late Bronze Age Collapse around 1200 BCE, a re-examination found 61% were misdated or had not occurred as claimed. The broader eastern Mediterranean disruption can remain real while many concrete-looking proof points fail individually."

[[extra.faq]]
q = "Why is a surviving archive not a complete record of an economy?"
a = "The Linear B tablets give unusually precise terminal evidence because fires baked unfired clay hard enough to preserve it. About 4,158 tablets survive from Knossos and 1,026 from Pylos, whose final entries date to around 1200 BCE. That precision depends on writing, a preservable medium, excavation, and decipherment; the archive records an administration that survived those conditions, not every activity around it."

[[extra.faq]]
q = "Can an unobserved financial segment be investigated after the first analysis is complete?"
a = "Sometimes. A financial gap can be reopened through retained filings, contracts, or vendor archives when those records still exist, so a coverage failure may be reversible. The central tomb mound of Qin Shi Huang has never been opened because painted surfaces and organic artifacts degraded when earlier material was exposed to air. There, obtaining the missing observation can destroy the evidence that made it valuable."

[[extra.sources]]
title = "\"4.2-kiloyear event,\" Wikipedia"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/4.2-kiloyear_event"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Ancient Egypt,\" Wikipedia"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/Ancient_Egypt"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Bosnian pyramid claims,\" Wikipedia"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/Bosnian_pyramid_claims"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Bronze Age collapse,\" Wikipedia"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/Bronze_Age_collapse"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Dorak affair,\" Wikipedia"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/Dorak_affair"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Enamel hypoplasia,\" Wikipedia"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/Enamel_hypoplasia"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Fall of the Western Roman Empire,\" Wikipedia"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/Fall_of_the_Western_Roman_Empire"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Göbekli Tepe,\" Wikipedia"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/GÃ%B6bekli_Tepe"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Knossos,\" Wikipedia"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/Knossos"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Linear B,\" Wikipedia"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/Linear_B"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Machu Picchu,\" Wikipedia"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/Machu_Picchu"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Maya collapse,\" Wikipedia"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/Maya_collapse"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Meghalayan,\" Wikipedia"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/Meghalayan"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Piltdown Man,\" Wikipedia"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/Piltdown_Man"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Pompeii,\" Wikipedia"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/Pompeii"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Stonehenge,\" Wikipedia"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/Stonehenge"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Terracotta Army,\" Wikipedia"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/Terracotta_Army"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Troy,\" Wikipedia"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/Troy"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Çatalhöyük,\" Wikipedia"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/Ã%87atalhÃ%B6yük"
kind = "tertiary"
fetched = "2026-08-13"
+++

In 2021 roughly 10% of Göbekli Tepe had been excavated.

That figure gets read as a promise: another 90% of the same thing, waiting underground.

The record is more cautious. The exposed enclosures are a fraction of a mound some three hundred metres across, and the small excavated share reflects a deliberate strategy of keeping ground intact for techniques that do not exist yet. The percentage measures work done against a research plan. It is not a sample certificate for the rest of the mound.

Pompeii supplies the opposite-looking number. About two-thirds of the walled town has been excavated, and major work effectively stopped after 1960 in favour of conserving what exposure had already made fragile.

Two figures, two regimes of restraint, and neither one says the observed ground has the distribution of the unobserved ground.

## An exact fragment can carry the wrong scope

Collapse histories acquire the same problem whenever a particularly good record becomes the civilisation.

The southern Maya lowlands did decline. Monumental inscriptions at Tikal, Calakmul, Copán and Palenque fell from roughly forty a year around 750 CE to none by 900. But Chichén Itzá, Uxmal and later Mayapán prospered in the northern Yucatán through the Postclassic, and the last independent Maya kingdom did not fall until 1697.

The received version calls that a Maya collapse.

The evidence supports a southern-lowland sequence.

The wider label has been repeated for a century by people who were not making anything up. Each inherited a finding about the most legible cities and carried the geography outward with it. More than eighty explanations have been proposed for the southern decline alone, which is what a record looks like when it is being asked a question it cannot answer.

The Late Bronze Age case shows why visible ruins make dangerous representatives. Of 153 destruction events attributed at one time or another to the collapse window around 1200 BCE, a re-examination found that some 61% were misdated or had never occurred as originally described.

The multi-state disruption of the eastern Mediterranean does not vanish with those entries. The point is narrower and harder: a burned layer has a date only once it has earned one, and a collection of burned layers becomes a regional claim only after each site has been checked separately.

## A proxy records its basin first

The 4.2-kiloyear event has a formal anchor. In 2018 the International Commission on Stratigraphy ratified the Meghalayan Age, using a stalagmite from Mawmluh Cave as its boundary reference point for a drought beginning around 4,200 years ago.

That is a named geological boundary.

It is not a demonstration that every drought near it was one event.

The standing objection is that supporters have combined evidence of droughts and wet periods sometimes centuries away from the proposed event. A reference marker in one cave makes a chronology more inspectable. It cannot turn Sumer's agricultural collapse, Egypt's droughts and the Indus monsoon shift into one regional process.

Some records are impressively fine-grained and still local. Isotope work on gypsum crystals from Yucatán lakes has reconstructed annual precipitation declines through the Maya Terminal Classic drought, building on a lake-sediment record established in the 1990s.

The later precision changes the claim about rainfall in those basins. It does not remove the step from basin to civilisation, and nothing in the proxy can take that step for you.

Settlement evidence carries a different boundary again. Obsidian hydration dating puts the Copán Valley population near 28,000 in the fifth century, 15,000 by 900, and under a thousand by 1200. That curve can show gradual demographic change where a destruction layer cannot.

It depends entirely on a systematically surveyed valley.

A region without comparable walking and mapping has no comparable curve. It does not therefore lack a comparable decline.

## The best terminal date may be a preservation accident

Administrative archives look cleanest of all. Thousands of Linear B tablets survive from Knossos and about a thousand from Pylos, and they survive because the fires that destroyed the palaces baked their unfired clay.

Pylos's final entries on wool, grain and military matters date to just before the destruction that stopped the archive.

That is a powerful terminal observation, and it is conditional on five things at once: writing existed, records were made on clay, fire preserved rather than consumed them, the archive was excavated, and the script was deciphered.

The Indus script and the pre-literate Olmec heartland do not meet that chain. Silence in their records is built differently from silence at Pylos, and reading the two the same way is the error.

Dates at a site have the same property. Radiocarbon work narrowed Machu Picchu's occupation to roughly 1420 to 1530. Pompeii is dated by an entirely different route — Pliny the Younger's eyewitness letters, checked against sealed wine jars and seasonal clothing, which moved the consensus eruption date from August into late October.

A date is not a property of a site detached from its method. It is the output of a particular surviving record, and the method belongs in the field beside it.

## Recoverable, or gone for good

Unexcavated ground can be left closed precisely because opening it is irreversible. Nobody gets a second attempt, so the discipline has learned to leave the best material for the people with the better instruments.

A financial coverage gap is usually not like that. It can often be reopened from retained filings, contracts or vendor archives, at a cost, by somebody who decides it matters.

So the diagnostic is not that a segment is missing. It is whether the missingness is recoverable, and whether the result still stands once the best-covered segment has been separated out from the target universe.

Ninety per cent of Göbekli Tepe is still in the ground, and not because nobody has got round to it.

It was left there for instruments that have not been invented yet, by people who had watched what happened to sites that came up too fast.

The ten per cent is a number. The decision behind the other ninety is what says what the ten per cent is allowed to mean.