Almost everyone has seen, even if they do not know it, a scene from the Bayeux Tapestry: a standing warrior who brings his hand to his face, with an arrow embedded in his eye, beneath the embroidered word HAROLD. It is one of the most famous images of the Middle Ages, the one that fixes in the collective memory how the last Anglo-Saxon king died at the Battle of Hastings in 1066: from an arrow to the eye. The image is so sharp that it appears to be a document. And perhaps it is not. Perhaps that arrow was not embroidered by anyone in the eleventh century, but by a hand in the nineteenth century.
The Bayeux Tapestry has not arrived intact since it was embroidered, nearly a thousand years ago. Like any object of such antiquity, it has lost threads, suffered stains and tears, and has been repaired several times. There is a documented restoration in 1842, and before that, records of what it was like are preserved: the engravings commissioned by Montfaucon in 1730 and, above all, the meticulous drawings made by Charles Stothard between 1816 and 1819. Those records are the only window we have into the previous state of the piece. And here the interesting part begins.
Historian Martin Foys has maintained that the arrow in Harold’s eye was not present, or at least not in that form, prior to modern interventions. According to his reading of ancient records, what occupied that figure was a different element, possibly a spear or a shaft above the head, rather than an arrow entering the face. The restoration purportedly converted that ambiguous element into the unequivocal arrow we recognise today. It is a hypothesis, and I state this clearly: there is no unanimous consensus among specialists, and the history of how Harold died remains uncertain, as contemporary chronicles do not even agree on whether an arrow was the cause of his death. Nevertheless, the hypothesis is serious, and its implications are profound.
For if Foys is correct, even in part, then the tapestry’s most iconic image—the one that illustrates textbooks and documentaries, the one that everyone knows represents the King’s death—would not be an eleventh-century testimony. It would be a nineteenth-century interpretation embroidered over a lacuna, a restorer’s conjecture rendered in thread, and subsequently repeated for a century and a half as if it were the original. It was not merely a gap that was filled: the most famous moment in the entire work may have been fabricated.
This is more unsettling than a clumsy restoration. When a reintegration is noticeable, when it is evident that a portion is newer or of a different hue, the observer recognises the presence of a later hand and judges accordingly. Yet when the reintegration is proficient, when it merges with the remainder and coincides with what we expected to see, it ceases to be perceived as an addition; it becomes canonical. The arrow in the eye aligned so precisely with the established narrative of Harold that no one had cause to doubt it. The addition did not survive despite being invisible: it survived because it was invisible, and because it confirmed the narrative.
I recently wrote in this notebook about restoration using artificial intelligence, concerning the danger of a machine completing what is missing and our labelling that as restoration. Harold’s arrow demonstrates that the problem is not one invented by AI: it is as old as restoration itself. A nineteenth-century restorer, with needle and thread and with all the good faith in the world, could do exactly what we fear from the machine, fabricating what did not exist and presenting it with the same texture as the authentic. The difference is that, at that time, there was no digital record to leave evidence, so the conjecture dissolved into the work and became indistinguishable from it. It took nearly two centuries and a stubborn historian to separate them again, and even then, not entirely.
All of this does not make me distrust the tapestry, which remains a marvel, nor does it serve to accuse restorers who worked with the logic of their time. Rather, it shows us something about how what we take for inherited truth is formed. Many of the images we believe we receive intact from the past passed through intermediate hands that retouched them, completed them, and adjusted them to what was expected of them. A trace does not always remain. Harold’s arrow has been looking at us for a thousand years, or so we believed; it may be that it has actually been looking at us for two hundred, and that what fixes that gaze is not the memory of a battle, but the conjecture of someone who wished to conclude it.
On the open conversation
This text begins with a verifiable academic hypothesis, that of Martin Foys regarding the arrow in Harold’s eye in the Bayeux Tapestry, to consider how a subsequent reintegration can become indistinguishable from the original and be canonised as if it had always been there. I mark the thesis as what it is, a solid hypothesis without unanimous consensus, and I do not arbitrate the historical question of how Harold died, which the sources of the period themselves leave open. It links to what I wrote about restoration with AI: the problem is older than the machine. If anyone wishes to intervene from the fields of medieval history, textile conservation, or museology, this notebook remains open.
Sources
Martin K. Foys, 'Pulling the Arrow Out: The Legend of Harold’s Death and the Bayeux Tapestry', in The Bayeux Tapestry: New Interpretations (Boydell Press, 2009).
Records prior to the 1842 restoration: engravings by Bernard de Montfaucon / Antoine Benoît (1730) and drawings by Charles Stothard for the Society of Antiquaries (1816-1819), both with their own interpretative reconstructions.
Musée de la Tapisserie de Bayeux: 2020 conservation study and restoration project planned following the 2027 reopening.
Regarding the historical uncertainty of Harold’s death at Hastings (1066): discrepancy in the chronicles close to the event.