The standard version of the Roland Roncesvalles myth goes like this: a medieval poem turned a small ambush into a holy war, swapping the real attackers for an invented army of Muslims, and historians eventually set the record straight. That version skips a step. Before any poem existed, Charlemagne’s own court chroniclers had already tampered with what happened on August 15, 778, in the Pyrenees pass known in Spanish as Roncesvalles and in Basque as Orreaga. They did it not by inventing a new enemy, but by refusing to mention the defeat at all.
This matters because the popular framing of “the epic lied, the history books corrected it” hides a more uncomfortable and more interesting fact: the very first written record of the battle is not neutral. It was produced by people with a direct stake in how Charlemagne’s reign looked on paper, and it says nothing happened. The correction came only decades later, from a revised edition of that same court record. The wholesale invention of a Muslim army came later still, roughly three centuries after the ambush itself. Understanding Roncesvalles honestly means tracking three separate versions of the story, not two.
This account draws on the two earliest surviving written sources for the battle, the Royal Frankish Annals and Einhard’s Vita Karoli Magni, cross-checked against Encyclopaedia Britannica’s entry on the battle and the modern scholarly editions of both texts (Scholz and Rogers’s standard English translation of the Annals, and Rosamond McKitterick’s research on how and when the Annals were revised). Nothing below is drawn from the Chanson de Roland itself as a factual source; the poem appears only where it is explicitly identified as legend.
What Actually Happened on August 15, 778
Charlemagne entered the Iberian Peninsula in 778 after local Muslim rulers in the northeast, including the governor of Barcelona, offered him their submission in exchange for military help against a rival emir in Córdoba. The campaign did not go as planned. Charlemagne’s army besieged Zaragoza for more than a month and never took the city; the local governor who was supposed to hand it over changed his mind and refused. Charlemagne accepted a payment of gold and a handful of prisoners instead, then turned north to withdraw across the Pyrenees.
Before he left, he moved against the Basques of the region, whom he suspected of sympathizing with the Muslim rulers he had failed to conquer. He ordered the walls of the Basque city of Pamplona torn down, and several other towns nearby were razed as well, to prevent them from being used against the Franks again.
The Basques did not forget it. As Charlemagne’s army crossed back over the mountains through the pass at Roncesvalles, a Basque force that knew the terrain far better than the retreating Franks struck the rear of the column in the evening of August 15. The rearguard, separated from the main army by the terrain and the ambush itself, was surrounded and killed to the last man. The attackers stripped the abandoned baggage train, which likely still held gold from the Zaragoza settlement, and vanished into the mountains before morning, leaving the Franks nothing to pursue.
Among the dead were three named Frankish nobles: Eggihard, steward of the royal household; Anselm, count palatine; and a man recorded only as Hruodlandus, prefect of the Breton March. That third name is Roland. Almost everything else claimed about him as a historical person, his age, his appearance, his relationship to Charlemagne, is absent from the record. The man later remembered as the greatest of Charlemagne’s knights survives in contemporary history as a single job title attached to a casualty list.
The Cover-Up That Came Before the Legend
The first written account of Charlemagne’s reign to address 778, the Royal Frankish Annals, does not mention Roncesvalles at all. The Annals were compiled year by year by clerics with close access to the royal court, and modern scholarship on the text (Scholz and Rogers’s 1970 critical edition, and Rosamond McKitterick’s subsequent analysis of the Annals’ composition) treats their early sections as functioning partly as royal propaganda: defeats were routinely left out or reframed as victories. The Annals apply the same treatment to a real Frankish defeat at the Battle of Süntel against the Saxons in 782, which the original text also describes in terms that obscure the loss.
The Roncesvalles disaster was restored to the record only in a revised edition of the Annals, compiled sometime after Charlemagne’s death in 814 and closely associated with, though not conclusively proven to have been written by, his biographer Einhard. That revised text states plainly that the Vascones, the Basques, set an ambush in the heights of the pass, that the Franks were better armed and no less brave but were undone by the terrain and the Basques’ different style of fighting, and that “the majority of the paladins that the King had placed in command of his forces” were killed.
Even in this corrected version, written to acknowledge a defeat the original annalists had chosen to suppress, the attackers are still, correctly, Basques. There is no Muslim army anywhere in the ninth-century record. Einhard’s own biography of Charlemagne, the Vita Karoli Magni, completed by 833 at the latest and drawing on the revised Annals, gives the fullest ninth-century account and names the Basques as well. The first myth of Roncesvalles, in other words, was not who attacked. It was that anything happened at all, and it was corrected by the same institution that created it, decades later, once the men who needed the cover-up were dead.
Three Centuries Later, the Basques Became 400,000 Saracens
The transformation most people associate with Roncesvalles, the swap from Basque ambush to Muslim invasion, does not appear anywhere until the Chanson de Roland, composed in Old French around the year 1100. The oldest surviving manuscript of the poem dates to the following century. That gap matters: the poem was written roughly three hundred years after the event it describes, by which point no one alive had any direct memory of it, and Western Europe’s relationship with the Muslim world had changed dramatically with the launch of the First Crusade in 1096.
In the poem, the small Basque raiding party of the historical record becomes an army of four hundred thousand Saracens. Roland acquires a magic sword, Durendal, and a war-horn called an oliphant, both entirely absent from any earlier source. The poem’s most famous scene, in which Roland refuses to sound the horn and summon Charlemagne’s main force until it is too late to save his men, has no basis in Einhard or the Annals; it is a literary invention built to dramatize pride and honor, not a remembered fact.
The recasting of the villain from Basque to Muslim did real cultural work for an audience being asked to think of holy war against Islam as a defining Christian duty. It also required ignoring an inconvenient detail from the actual campaign: in 778, Charlemagne had marched into Iberia at the invitation of Muslim rulers who wanted his help against a rival, and the Basques who ambushed his army were still largely outside Christianity themselves at the time. The poem’s framing of a clash between Christendom and Islam bears little resemblance to the tangled, opportunistic alliances of the real campaign that preceded the ambush. None of that diminishes the Chanson de Roland’s importance; it remains the oldest surviving major work of French literature and shaped centuries of European ideas about knighthood and loyalty. It simply is not a historical account of what happened at Roncesvalles, and it was never meant to be one.
A Second Ambush at the Same Pass Built a Kingdom
The Basques did not stop at one victory. In 824, a larger Carolingian force attempting the same crossing was ambushed at the same pass by a Basque army, this time allied with the Banu Qasi, a powerful Muslim frontier family based along the Ebro. Unlike the 778 rearguard action, which allowed Charlemagne’s main army to escape while the rearguard was destroyed, the 824 battle trapped and routed the entire Frankish force. Its commander, Count Aeblus, was captured along with a fellow Frankish vassal, Aznar Sánchez. This second battle receives far less attention than the first, largely because no poet ever wrote an epic about it, but the historical record treats it as the more consequential defeat: a larger Frankish army lost more men, and lost them for good, with no rearguard sacrifice left to romanticize.
That victory helped consolidate the independent rule of Íñigo Arista, whose alliance with the Basques and the Banu Qasi against renewed Carolingian expansion laid the groundwork for the Kingdom of Pamplona, the political entity that would eventually grow into the Kingdom of Navarra. Two ambushes in the same mountain pass, forty-six years apart, form the real throughline of Roncesvalles: not a single legendary battle, but a sustained pattern of Basque resistance to Frankish control of the Pyrenees, resistance that eventually built a kingdom rather than simply defending a mountain crossing.
The pass and the Camino de Santiago route that later ran through it are their own subject; this article is concerned only with what can and cannot be verified about the events of 778 and how the record of them changed over time.
Frequently Asked Questions
Did Muslims attack Charlemagne’s army at Roncesvalles?
No. Every ninth-century source, including Einhard’s Vita Karoli Magni and the revised Royal Frankish Annals, identifies the attackers as Basques (Vascones), acting in retaliation for Charlemagne’s destruction of Pamplona’s city walls. The idea of a Muslim army comes only from the Chanson de Roland, a poem composed roughly three centuries after the battle.
Who really killed Roland at Roncesvalles?
A Basque force ambushed Charlemagne’s rearguard as it crossed the pass on August 15, 778, killing every man in it, including Roland, described in Einhard’s account only as “prefect of the Breton March.” No historical source identifies an individual killer; the ambush overwhelmed the entire rearguard rather than singling anyone out.
Is the Song of Roland historically accurate?
No, and it was never intended to be a historical record. It was composed around 1100, roughly three hundred years after the battle it depicts, and changes the attackers from Basques to a fictional army of 400,000 Muslims while inventing details such as Roland’s sword Durendal and his war-horn. It is a foundational work of medieval literature, not a chronicle.
Why does the Royal Frankish Annals not mention Roncesvalles?
The original Royal Frankish Annals, written by clerics close to Charlemagne’s court, functioned in part as propaganda and routinely omitted Frankish military defeats, including Roncesvalles and a separate 782 loss to the Saxons at Süntel. A revised edition compiled after Charlemagne’s death in 814 restored the Roncesvalles defeat to the record, correctly naming the Basques as the attackers.
Every article on the Encierro blog is authored or reviewed by active bull runners with direct experience in Pamplona.