The Oldest Manuscript of The Song of Roland Is Actually an English Performance Copy
The Oxford Digby 23 manuscript is written in Anglo-Norman, and scholars date it to between 1140 and 1170 by tracking dialect drift — the most complete version of this French epic was performed for the English.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
The most complete version is an English manuscript
The Song of Roland is a French epic that took shape around the 12th century, belonging to the genre Chanson de Geste (songs of deeds), and it spread with 12th-century French cultural hegemony to regions including Britain. But its most complete and oldest surviving version is not from France itself — it is an English manuscript, catalogued as Oxford Digby 23. This manuscript was originally bound together with a completely unrelated book by someone later, which is why it is called Digby 23 B, and A is that "other book." Nobody knows who bound them together; possibly some monk felt both were his favourite books.
Every stanza is a camera position
The poem is organised in laisses (stanzas) of varying length, and each laisse is like a camera focus: when the scene or the mood shifts, you change laisse. The whole poem is in decasyllabic meter, cut into four plus six in performance: the first four syllables rise, then a dramatic pause, then the last six syllables fall, overall close to iambic turning into trochaic. Recited over and over this becomes hypnotic, and the poem runs to four thousand lines — hours of recitation.
At peak emotion, the same action is performed three times
This kind of epic has a convention that is hard for any reteller to handle: when the action is especially intense or the emotion especially strong, it uses consecutive laisses to tell the same action from three angles, a kind of poetic slow motion. For example, Roland's refusal to blow the oliphant is told three times, with almost no gap between them. This is also why it reads like an action film or a comic: scenes cut hard, with no softening transition — for instance, "Roland dies, his soul ascends to heaven" is followed directly by "the Emperor arrives at Roncevaux."
The performer's name is left at the end of the manuscript
Digby 23 is very likely a real minstrel's working copy. Line 4002 of the poem, its final line, says roughly: "here ends the song of deeds; Turoldus will not tell more." This line appears before Charlemagne's lament, so it reads more like the performer saying: this is too sad, I cannot go on. The name Turoldus does not appear in the poem, and he is not a known scribe, so scholars lean towards him being the owner of this manuscript, the performer of this performance.
Its pocket-book size gives away that it was for work
The manuscript is only about six and a half inches by four and a half inches, a pocket book you could carry on your person, not a wealthy man's finely bound volume. It is itself heavily worn and not refined, and it contains some small annotations of unclear meaning. Together these features point to a professional performer on the move: he might have travelled alone with an instrument, or been the lead singer or narrator in a troupe, with someone in front of him doing the physical performance. Such people were jongleurs — the origin of the later word juggler.
The dialect pins this performer to England
The manuscript is not in the French of central France, nor continental French, but in Anglo-Norman, which suggests this performer was probably Anglo-Norman, performing in Britain or at least in Normandy. The dialect features are very specific: continental French O and OU are replaced by U, so Roland becomes Rulant, l'or becomes l'ur; Anglo-Norman French was also losing final soft e. These changes even begin to break the originally strict decasyllabic meter, leaving some lines with only nine syllables or swelling to eleven.
Accent drift alone can date a manuscript
Scholars' grasp of this dialect is fine enough that they can date the manuscript to between 1140 and 1170 based on the degree of accent drift. The method is similar to scholars reconstructing Shakespearean pronunciation, tracking regular changes in syllables and sounds. Poetry is especially suited to this kind of reconstruction, because meter and rhyme expose which sounds were dropped and which were added. For many jongleurs, French was a second language anyway, spoken with a dialect accent, and they spelled by sound.
A French epic, an English artefact
Although this is a French poem about an old French king, this manuscript is an English thing, performed for Britons at the time. There is no equally complete other manuscript to compare line by line, but there are abundant records of The Song of Roland being sung before and after this manuscript, and scholars can use them to judge where this performer departed from what the genre should look like. The episode goes on to cover the poem's symbolic system, how its audience was constructed, and what the music sounded like.
In their own words · checked verbatim
our definitive version of the Song of Roland appears to be a British copy. It's the most complete version and the oldest version.
Each leise is like a camera focus. You see a scene or the scene has emotional tone. As soon as the tone shifts or the scene shifts, you have a new leise.
when something gets really emotional or the action gets really intense, they will basically do the slow-mo matrix version of poetry, which is you see the same shot from three different angles.
So the bard is like, I've gone as far as I want to go. My voice is wrecked. I'm getting a drink. Y'all on your own.
It's like six and a half inch by four and a half inch. So it's like a pocket copy.
It seems to be a Anglo-Norman dialect of French. So we think he was probably Anglo-Norman and performing in Britain or at least Normandy.
they're pretty sure they can date the thing just based on this accent and how far it's drifted, which places this as between 1140 and 1170.
even though this is a French poem about an old French king, this is a British piece and it would have been performed to Britons.
Figures
| Total lines of The Song of Roland | 4000 lines | 4:08 |
| Number of the manuscript's final line | line 4002 | 9:29 |
| Size of the Digby 23 manuscript | about 6.5 inches × 4.5 inches | 10:31 |
| Possible age of the oral tradition | by the 12th century it may already have been 150 years old | 5:18 |
| Dating of the manuscript | 1140–1170 | 14:34 |
Glossary
- Chanson de Geste
- A 12th-century French epic genre; literally "song of deeds."
- laisse
- The stanza unit of The Song of Roland, of varying length, one laisse to a scene or a mood.
- decasyllabic meter
- Ten syllables per line, cut into four plus six in performance, with a pause in the middle.
- oliphant
- The horn Roland refuses to blow in the poem, meant to call Charlemagne for help.
- jongleur
- A medieval professional travelling performer; the word juggler comes from it.
- Anglo-Norman
- The French dialect used in Britain after the Norman Conquest, already markedly different from continental French.
How to listen
Listeners interested in medieval manuscripts, oral epic and dialect reconstruction; anyone who wants to see how a single manuscript gets reconstructed into a specific performance setting.
If you only want the political subtext of The Song of Roland, you can skip it — that part is in the next episode.