New evidence for the existence of “lost” ancient medieval manuscripts, long forgotten from memory, is now being uncovered by researchers using agent-based computer simulations.
The findings, reported by Jean-Baptiste Camps and colleagues, relied on a novel approach to uncovering evidence for these long-lost texts: a complexity science approach that allowed them to estimate the volumes of lost ancient literature that once existed, by inferring references to such texts from existing chivalric narratives beginning in the 12th century.
The team’s findings also bring to attention a new concern for historians, since according to their simulations, as much as 60% of all ancient texts from this period, and more than 95% of manuscripts once believed to have existed, may have been lost to time. The findings were recently reported in PNAS Nexus.
Evidence of “Lost” Medieval Manuscripts?
The names we commonly associate with epic tales passed down to us from ancient literature are not the only heroes of yore that existed—they are merely the ones whose stories survived.
For every Gilgamesh, Thor, and King Arthur, it is possible that several more heroic characters from undiscovered ancient epics also exist. However, based on Camps and his team’s findings, they are probably unrecoverable stories, lost to time apart from scant references that occasionally appear in the few surviving manuscripts that have made their way down to us throughout time.
Some of the problems that have contributed to this concerning loss of ancient knowledge are technological. In the days before the invention of the printing press, texts had to be copied manually—a process that led to both innovations, but also errors and inconsistencies as these early copies were made.
From existing copies of surviving ancient texts, scholars can often infer a significant amount of information based on subtle (or in some cases, not-so-subtle) changes that appear from one version to the next. Philologists, researchers who study the structure, historical development, and relationships between languages, refer to the evolutionary “trees” encompassing such texts and their evolution through time as stemmata.
What’s ‘Stemmata’ With You?
Since modern stemmata are based only on existing copies, they lack the full evolutionary history of texts and fail to account for versions that have been forever lost, which may have contained significant differences or additional information.
Jean-Baptiste Camps and his colleagues’ new model has revealed a key discovery: that the early years of a text’s creation, and the number of copies that were made during that time, are of crucial importance to their survival.
As one might guess, the fewer copies that are made, the more endangered the text becomes. However, another implication based on the team’s modeling is that if no copies of the original work are made, all the surviving derivatives generally form what scholars recognize as a subsidiary branch of that manuscript’s family tree.
And therein lies the issue: one that has sobering implications for the preservation of knowledge from ancient times.
Roland’s Unknown Song
As an example, the researchers point to the Song of Roland, an 11th-century account of the Frankish military leader Roland’s deeds at the Battle of Roncevaux Pass. The events in question took place in AD 778, during the reign of Charlemagne, and the account remains the oldest single surviving major work of French literature.
However, the version of the tale we know today comes to us from several different manuscript copies that were made over time. While this speaks to the tale’s popularity in the Medieval world of the 12th to the 16th centuries, it also underscores the fact that its oldest, original versions are probably unknowable to us today.
But how is such knowledge lost to time? Arguably, one of history’s most famous examples involves the famous burning of the library at Alexandria—a story which is itself somewhat contested by modern scholars, in that rather than one catastrophic loss of information resulting from the destruction of a single building, a range of factors over an extended period are more likely to have resulted in the loss of this wealth of knowledge from the ancient world.
This is merely one example; a range of other instances involving random occurrences, accidents, and cataclysmic events point to additional reasons for the fragility of knowledge. One that Camps and his colleagues point to is obvious: the Black Death, which, in addition to raging through Europe and decimating human populations, also likely contributed significantly to the destruction of ancient manuscripts.
Knowledge must be protected, in other words, and the team’s recent findings only underscore how fragile it is—and also how easily it can be lost.
The team’s study, “On the transmission of texts: Written cultures as complex systems,” appeared in PNAS Nexus on July 7, 2026.
Micah Hanks is the Editor-in-Chief and Co-Founder of The Debrief. A longtime reporter on science, defense, and technology with a focus on space and astronomy, he can be reached at micah@thedebrief.org. Follow him on X @MicahHanks, and at micahhanks.com.