A conversation about ancient books easily turns into a story about fire. The Library of Alexandria, the Heruli, the Crusaders, the barbarians—and a picture takes shape in your head: there was a vast world of texts, then something destructive arrived and burned it down. In fact, that picture is wrong, and wrong at the most crucial point.
Antiquity did not burn. Antiquity was not copied.
The logic is almost boring, like bookkeeping. A papyrus scroll lasts a couple of centuries if left untouched, and noticeably less if it is read. So a text exists only as long as someone keeps copying it. They copy what there is demand for: the school curriculum, beloved authors, a useful profession. Everything else does not perish in flames—it simply stops being copied and quietly crumbles into dust on a shelf. This is not a catastrophe; it is a filter, and it works slowly and invisibly.
Which leaves two questions that seemed more interesting to me than the number itself. How much survived? And why did exactly this survive?
Let me begin with an honest admission: there is no exact figure, and there cannot be one. There are estimates, and they differ by an order of magnitude.
The most frequently cited is around 10%. Its source is not a modern study but a sweeping line in a book from 1965: although much Greek literature has survived, “less than 10% of what was written” made it into modern times. Thirty years later, another author in the same volume adjusted the wording: “perhaps ten percent of the major classical Greek works” has survived. Note the word “major”—it smuggles a selection criterion into the estimate itself.
The harshest figure is roughly 1% of pagan Greek literature. It is older than it looks, and worth tracing. Oxford palaeographer Nigel Wilson, when asked about it, named Pietro Bembo—the Renaissance humanist—as its source. The number came from his “Oration in Defense of Greek Literature,” addressed to the Venetians, and there it is not a measurement but a rhetorical calculation: “...if we count the poets, orators, philosophers, and, in short, the writers of old whose works were read by previous generations of interested students—if we count them all, we shall find that perhaps we do not possess even one percent of them.” Note the construction: “perhaps” and “one percent” stand side by side, because this is a 16th-century estimate, not data. Half a millennium later, it still travels from text to text as a fact.
And Bembo has a metaphor there that I consider the best explanation for this entire story. Literature, he says, is like a field. So long as the field is ploughed, sown, watered, and given everything it needs in due time, it yields a harvest. But when no one tends it, it does not merely cease to bear fruit: “all the estate is not only barren and disordered, but deteriorates day by day and in due course disappears.” A work does not die from a blow—it decays from neglect. Then Bembo makes the move I would hang as an epigraph over the whole history of antiquity’s loss: if the rulers of Italy, as soon as the barbarians ceased their raids, had neglected what remained to them, then “perhaps, by the gods, we would not now possess even the elements of our literature.” The idea that texts are saved not by heroism but by the routine work of copying them was formulated four centuries before papyri confirmed it.
Where did the neat figures come from? In 1948, German philologist Hans Gerstinger tried to count. Before the discovery of the papyri, around 2000 Greek authors were known by name. Complete works had survived from 136 of them (6.8%), and fragments from another 127 (6.3%). And that is counting authors, not works: count titles instead, and the ratio becomes noticeably worse.
And if you descend from authors to the level of genre, the arithmetic changes completely. Take tragedy—probably the best-preserved genre of antiquity. 33 tragedies have come down to us (including satyr plays): 7 by Aeschylus, 7 by Sophocles, 19 by Euripides. The three great tragedians wrote a little more than 300 plays between them. 33 out of 300 is about 10%. But if we count how many plays actually existed: if three tragedies were staged at every City Dionysia between 485 and 406 BC, that gives roughly 948 plays at that festival alone, and 33 out of 948 is 3.5%. Add repertory productions over the following five hundred years, the Lenaea and other festivals, cities, the “book tragedies” of the 4th century, not one of which survived—and the estimate drops toward 0.4%. Against the original ten percent, that is a difference of twenty-five times, and all of it lies in how you define the denominator.
That, to my mind, is the central fact. The argument over the percentage is not an argument about how much was lost, but about what we agreed to count as a unit.
And this is where things get interesting, because in one particular case the mechanism of selection is known with ridiculous precision.
Euripides has 19 surviving plays—more than Sophocles and Aeschylus combined. Though ancient authors credited him with 92–95 dramas. Which means that this great tragedian reached us not as a complete collection but by two utterly different routes, and both are worth examining separately.
The first ten are Alcestis, Andromache, Bacchae, Hecuba, Hippolytus, Medea, Orestes, Phoenician Women, Rhesus, Trojan Women. These are the “selected” plays: those prescribed as compulsory reading in late Greek and Byzantine schools. The reason they survived was not fame but the curriculum. And at once it becomes clear why Sophocles and Aeschylus have exactly seven plays each. There were two Byzantine school traditions: one selected seven plays, the other three. For Sophocles and Aeschylus, the two sets overlapped, producing exactly seven. For Euripides, the sets did not overlap—and produced ten. The number of surviving works by Europe’s greatest tragedians was determined by the lesson timetable.
The other nine are Electra, Helen, Heracles, Heracleidae, Suppliants, Ion, Iphigenia at Aulis, Iphigenia in Tauris, Cyclops. Look at the first letters of their titles: E, E, H, H, H, I, I, I, K—eta, eta, eta, kappa. This is the second quarter, or fifth, of the Greek alphabet, and there is only one explanation: this is a volume of the complete works of Euripides, arranged alphabetically, that survived by chance. At some point, one papyrus scroll surfaced and was joined to the ten school plays—and that was all. How, where, and by whom, nobody knows. Researchers phrase the best explanation this way: an act of random alphabetical kindness.
There is one detail in this story that I consider the finest in the whole plot. Cyclops is not a tragedy but a satyr play, the only surviving work in the entire genre. It entered this volume not by literary logic but because it was short and fit at the end of the scroll, where a full tragedy would no longer fit. The only surviving example of an entire genre lived because it happened to be the right size.
And now the consequence. Without these nine “alphabetical” plays, we would not know that whole types of tragedy existed. Helen, in which Euripides’ Helen was never in Troy at all—she was replaced by a phantom—and which ends happily, with comic violence and the couple’s escape home, is a melodrama and a “rescue” play, not a tragedy in our sense. Our definition of classical tragedy—dark, serious, with a bad ending—would have narrowed until it became false. We would define the genre from a random sample.
The second mechanism works in exactly the opposite way: a text survives not because it was preserved, but because it was destroyed—and the destruction stopped halfway through.
Parchment is processed skin, which means expensive. By the 7th–8th centuries, when writing material was scarce, manuscripts by classical authors began to be erased and written over. The scribe was not interested in the content; he was interested in the material. The ancient text was scraped away, washed off, the leaves folded anew, and prayers written over them. This produced a palimpsest, and the paradox is this: mechanical erasure destroyed the visible text but preserved the physical object, which without that second life would either have crumbled away or gone into the fire.
The scale of the effect is clear from the figures: among the sparse corpus of the oldest surviving Latin manuscripts, most texts by classical authors have come down to us precisely as palimpsests. Cicero was restored in exactly this way: De re publica was recovered from a palimpsest in the Vatican Library.
The most beautiful case is Archimedes. A 10th-century Byzantine codex, containing seven treatises, including the Stomachion and The Method of Mechanical Theorems, preserved nowhere else in the world, plus the only Greek version of On Floating Bodies. After Constantinople was sacked by the Crusaders in 1204, the manuscript ended up at the Mar Saba monastery in Palestine—probably carried there precisely to escape the Crusaders, who often equated Greek writing with heresy and burned such books. In a remote monastery, the complicated mathematical manuscript was not appreciated, and in 1229 it was scraped down and rewritten as a liturgical book.
From that point on, it became a matter of the survival of the object itself, and it turned out to be hardy. In 1906, Danish philologist Johan Ludvig Heiberg recognized washed-away mathematics beneath the lines of the prayer book and published the text in 1915. In 1922, the manuscript vanished during the evacuation of the Greek library in Istanbul, and spent the next seventy-odd years with a Western businessman, who in places painted forged images of the Evangelists over it—to raise its price. Mold in the basement ate the collagen; three pages still available to Heiberg in 1908 simply disappeared from the book. In 1998, the codex sold at auction for $2 million.
The outcome of this odyssey deserves its own line: the author of the palimpsest left part of the text of The Method untouched, and those pages remain lost to this day—among them a result concerning the volume of the intersection of two cylinders, a problem that would surface again only with Kepler’s Stereometry a millennium and a half later. The treatise explaining how Archimedes actually arrived at his results—through physics, centers of mass, and the law of the lever, none of which appears in his rigorous proofs by the method of exhaustion—exists in a single copy, and part of that copy was eaten by a monk’s knife.
A separate, wholly unexpected channel of preservation is trash.
The classical world left us mainly what it carved in stone: official inscriptions, decrees, epitaphs. It left us no actual books: the Great Library of Alexandria and the twenty-eight public libraries of imperial Rome vanished without a trace. What remains are copies of copies.
But it turned out that books survived where nobody intended to preserve anything. Oxyrhynchus, a provincial town five days by water south of Memphis, the “city of the sharp-nosed fish,” lived a Greek-speaking life on Egypt’s edge for a thousand years. It never rains in that part of Egypt: everything above the water table is preserved practically forever. Which is why papyrus survived in the city’s dumps—what no “real” classical monument in Greece or Italy could preserve. Not a library, but rubbish heaps thirty feet deep.
Grenfell and Hunt, two Oxford men, excavated here for six seasons between 1896 and 1907. Estimates of what they found vary, and vary in precisely the spirit of all this material. The project’s official website once gave 400,000 items. Peter Parsons, who sorted through the finds himself, wrote in a 2007 book of “700 boxes of papyri, which may be estimated at 500,000 pieces and scraps,” recovered for £4000. And Dirk Obbink, in 2016, spoke of “more than a million fragments”—from postage-stamp-sized scraps to a long uninterrupted scroll. By 2019, meanwhile, the series The Oxyrhynchus Papyri had published 5476 entries. Five and a half thousand out of a million still waiting to be catalogued and deciphered.
And this is what those scraps contain. Oxyrhynchus returned authors known in classical times and wholly submerged in the Middle Ages: the songs of Sappho, the “sitcom” of Menander, the learned elegies of Callimachus. Menander is a story of his own: for a long time, his comedies were known only through fragmentary quotations, but in 1952 a papyrus containing the almost complete text of Dyskolos, the only New Comedy to survive intact, was recovered from Egypt—and published in 1958. Molière staged The Misanthrope knowing only the plot theme of this play: it had gone under long before his time.
But what shines brightest in the Oxyrhynchus scraps is not literature, but life. Because dull things end up in the dump too. We know where the fisherman Tonis, the embroiderer Athenchis, the dyer Anicetus, and the greengrocer Philammon lived. We know that on November 2, 182, an eight-year-old slave named Epaphroditus leaned out of a bedroom window to watch the castanet players in the street, fell, and was killed. We know that Sabina struck Syra with a key and put her to bed for four days—ancient keys, as the author notes, were good heavy objects. The classical tradition reached us through school lessons and imperial libraries; these people reached us through household papers, tax receipts, and shopping lists, stolen from time only because nobody thought to throw them away twice.
Let us assemble the picture.
Ancient literature did not perish in a single blow, but through a combination of practical causes, and that matters: the loss of most of the Greek and Latin corpus is explained chiefly by prosaic things—the cost of writing materials, the change in format from scroll to codex, altered literary tastes. Religious motives existed too, especially where classical and “heretical” texts were concerned, but as background rather than the sole cause. There was also the crisis of the 3rd century, plus specific destructions of Christian writings before Constantine and retaliatory ones afterward.
But the main engine was the change of format. When the codex finally displaced the scroll in the 4th century, what was decided was this: they copied not what was better, but what there was demand for. Everything else did not physically survive—the scroll had to be recopied, which meant someone had to think it worth doing.
So the right metaphor is not fire. It is a lottery with known rules, in which tickets were handed out along three entirely different lines: ten school plays, nine random letters of the alphabet, and mountains of trash at the edge of the Egyptian desert. Aeschylus, Sophocles, and Euripides reached us not as the summit of Greek drama, but as what was read in Byzantine schools, what fit at the end of a scroll, and what lay in a dump. We lost almost everything, and we lost it for boring reasons: because nobody paid for it.