How oral poets remember thousands of lines

Traditions across the world contain long narrative poems performed from memory — epics running to many thousands of lines, recited over hours or days.

The obvious question is how. Rote memorisation of that quantity of material, transmitted accurately across generations without writing, strains belief.

The fieldwork that answered it

In the 1930s, the American scholars Milman Parry and Albert Lord went to the Balkans to study living oral epic performance among singers who were frequently not literate.

What they found was that the singers were not reciting memorised texts. They were composing during performance, using a system of ready-made components.

Recording the same singer performing the same story on different occasions produced versions that differed substantially in wording and length while telling the same tale. Asking a singer whether he had changed anything produced the answer that he had sung it exactly the same — because from his point of view, he had. The story was the same.

The building blocks

Formulas. Fixed phrases that fit the metre and express a recurring idea. A stock phrase for dawn breaking, for a hero speaking, for a horse being saddled. The singer holds a large inventory of these, and each is metrically shaped to fit a particular position in the line.

This is why the epithets in Homeric poetry are so repetitive — the swift-footed one, the wine-dark sea. They are not stylistic decoration; they are metrically useful components that let a performer complete a line while thinking about the next one.

Themes. Larger recurring units — an arming scene, a feast, a council, a duel, a journey. Each can be expanded or compressed depending on the occasion, the audience and how long the singer wants to go on.

The story pattern. The overall structure, which is what the singer actually remembers as fixed.

Composition in performance means assembling these under the constraint of the metre, in real time, at speaking speed. It is closer to fluent speech in a highly constrained register than to recitation.

Why it works

The metre does a great deal of the work. A rigid metrical pattern constrains what can come next, and formulas exist precisely because they fit. The constraint that looks like a difficulty is actually the support.

The instrument helps too. Many traditions accompany performance on a stringed instrument, whose rhythm keeps the line moving and gives the singer a moment to think.

Audience knowledge matters. Listeners know the story, so a singer who compresses or expands is not confusing anyone. Performance is about the telling rather than the plot.

What this did to classical scholarship

The findings were applied to the Homeric poems, and they reframed a long-running argument about their composition. The formulaic density of the Greek epics looks precisely like the product of an oral tradition, which changed the terms of the debate about how those texts came to exist in the form they did.

The approach has since been applied to a wide range of traditions — Old English verse, medieval European material, Central Asian and South Asian epics, African oral narrative — with varying degrees of fit. Not every oral tradition works the same way, and the model has been refined.

The traditions that do memorise exactly

Worth noting as a contrast, because they exist and they work differently.

Some traditions transmit texts with extreme fidelity, using deliberate techniques to ensure accuracy: recitation in multiple orders, systematic checking, and training regimes lasting years.

Vedic recitation is the standard example, where methods of reciting the text in permuted sequences were developed specifically to detect and prevent error, and the transmission of the material across a very long period appears to have been remarkably precise.

So oral tradition is not one thing. It ranges from fluid recomposition to near-perfect fixity, depending on what the tradition is for. A story performed for entertainment tolerates variation; a text believed to require exact wording does not, and the transmission techniques differ accordingly.

What is lost with literacy

Writing down an oral epic produces a text, and a text is a snapshot of one performance.

Everything that made the tradition a living practice — the adjustment to audience, the expansion and contraction, the singer's own contribution — is fixed and removed. Once a written version exists and acquires authority, singers may begin to defer to it, which ends the tradition as composition and turns it into recitation.

Collectors have repeatedly observed this happening in traditions they were documenting, which is an uncomfortable position to be in.

Where the practice survives

Living oral epic traditions persist in several regions, including parts of Central Asia, South Asia, West Africa and the Balkans, though the number of performers is generally declining and the audiences have other options.

In several places there has been deliberate support — training programmes, recognition schemes, festivals — which sustains performers and changes the context in which they perform.

The technique itself, though, turns out to be more widely distributed than the epic traditions. Anyone who improvises in a fixed form — freestyle rap, extempore verse contests, liturgical improvisation — is doing something structurally similar: composing at speed within constraints, using a stock of ready components. The tradition is old and the capacity is not going anywhere.