How much of human history has been lost?

What survives is not a sample of what existed. Papyrus lasts in the parts of Egypt where it never rains, and the famous sites of Greece and Italy could not preserve it. UNESCO set up its programme for documentary heritage in 1992 “to guard against the collective amnesia”, and directs its preservation work particularly at places affected by conflict or natural disaster.

That is a distortion in the record before any historian touches it. Which materials last, and where, is a question about chemistry, weather and war rather than about importance.

Some of what was written is known only by name. The Suda, a tenth-century Byzantine encyclopedia, credits Sophocles with 123 plays. Seven survive, and of the rest only fragments remain.

The Oxyrhynchus papyri come from the rubbish dumps of a town in Egypt, and the project holding them calls the collection the largest of its kind in the world. They have given back authors famous in antiquity and lost in the Middle Ages, Sappho and Menander among them, alongside private letters, shopping lists and tax returns. They survived because a town threw them away in a place where it never rains.

Hundreds of scrolls carbonised by the eruption of Vesuvius in 79 CE have stayed sealed and unreadable since they were dug up in the eighteenth century. High-resolution X-ray scanning combined with machine learning now reads them without unrolling them: in 2026 the surviving core of one scroll, PHerc. 1667, became the first to be read end to end. This is genuinely new: material that was written off as destroyed is producing readable Greek.

Estimating what has been lost requires estimating what existed, which is only possible where independent records mention it. For societies without that scaffolding, the size of the original corpus is unknown, and so the fraction lost is unknown.

Linear A, the Indus Valley script, Rongorongo and others remain undeciphered. For the Indus symbols the argument is not only about reading them but about whether they encode speech at all: Farmer, Sproat and Witzel hold that they were not evolving in linguistic directions after at least six hundred years of use and worked like the non-linguistic sign systems of the Near East, and Rao and colleagues find the conditional entropy of the sequences closer to natural languages than to the non-linguistic systems it was tested against. Where a script does carry language and cannot be read, we know a society wrote without knowing what it said.

A society that built in perishable materials and was never mentioned by literate neighbours can leave almost nothing behind. There is no method for counting such societies. The evidence that one existed arrives when it is found, and finds still arrive: fieldwork and lidar in the Upano Valley of Amazonian Ecuador have revealed a dense system of urban centres occupied from around 500 BCE.

Knowledge transmitted orally (navigation, medicine, ecology, law, history) lasts only as long as each generation passes it on, and where that holds it can last a very long time. Aboriginal Australian stories from twenty-one places along the coast describe the sea rising over land, and the study cited here argues that in most instances they plausibly refer to events more than about seven thousand years ago, when the sea reached its present level. That raises the question of how much else was once carried this way and is now gone.

If most human experience was never recorded in a durable medium, no future technology recovers it. What can be reconstructed is limited to what left physical traces, bones, tools, isotopes, genomes, sediment.

Ancient DNA, lidar survey under forest canopy and imaging of damaged documents have each reached evidence that was inaccessible rather than absent. A genome sequenced from a finger bone in a Siberian cave identified the Denisovans, a group distinct from Neanderthals and modern humans. A lidar survey of 2,144 square kilometres of northern Guatemala revealed interconnected Maya settlement under the forest. Near-infrared imaging recovers writing on the Dead Sea Scrolls that has faded to invisibility. The boundary between lost and unread has moved repeatedly, and there is no reason to assume it has stopped moving.