Formulaic oral composition
A Socratic walk-through of formulaic oral composition — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #How did singers hold epics of thousands of lines steady across generations with nothing written down?
The Iliad runs to some fifteen thousand lines of strict metre, and the tradition that produced it had no writing to lean on. The natural assumption is that singers memorised it, word for word, and passed the memorised text down. That assumption deserves a poke before we build on it. Have you ever heard anyone recite fifteen thousand lines of anything? And if the text were fixed by memory, why would the same singer, asked twice, produce versions of visibly different length?
Reasoning it through
REASONING #Milman Parry began not with a theory of memory but with an oddity in the words themselves. Homer's epithets do not behave the way a poet's word-choices normally do. Achilles is swift-footed when he is sitting still and sulking; ships are swift while drawn up on a beach; the sea is wine-dark at times when its colour is beside the point. If these were chosen for meaning, they are frequently wrong. So what are they chosen for?
Sort them by their metrical shape and the pattern appears. For a given hero, in a given grammatical case, filling a given stretch of the hexameter line, there is usually one standard phrase — and, strikingly, rarely a second phrase of the same metrical shape doing the same job. Parry called this extension and thrift: the system covers each common slot, and covers it once. That is not how a poet chooses adjectives. It is how a craftsman keeps a set of parts.
Ask what problem such a kit solves. Composing in a strict metre is hard; composing in a strict metre at speaking tempo, in front of an audience, with no chance to stop and revise is a different order of difficulty. The singer has perhaps two seconds to produce a metrically perfect line that also continues the story. He cannot search for words. What he needs is a stock of ready-made phrases already shaped to fit the gaps, so that when the line requires a certain rhythm ending at a certain point, a phrase of exactly that rhythm is available.
So the epithets are not decoration and not padding. They are the components that make real-time composition possible at all — which explains why they sit where they sit and why they are indifferent to context.
Albert Lord then supplied what Parry's argument still needed: living singers to watch. Working with Parry in the 1930s and continuing afterwards, he recorded South Slavic guslari performing long heroic songs, and found they were not reciting but composing, in performance, every time — assembling lines from formulas, scenes from typical sequences such as arming or feasting, and whole songs from familiar story-patterns. Two performances of "the same song" by the same singer could differ substantially in length and wording, and the singer would insist, sincerely, that he had sung it exactly as before.
That last detail is the key to the original question, because it dissolves it. What stayed steady across generations was not a text. It was a system: a metre, a stock of phrases fitted to it, a repertoire of scenes, and a set of story shapes. Hand a trained singer that system and he can produce a song of any length on demand, one that his audience recognises. Nothing needed to be held word for word, because nothing was.
The analogy
THE ANALOGY #Think of a jazz musician playing standards. He is not recalling a recording note by note, nor inventing from nothing. He holds a tune's harmonic shape and a large vocabulary of licks already fitted to particular chord changes, and he assembles a performance in real time. Two renditions differ in length and detail while being, to him and to the audience, the same piece.
A jazz musician knows he is improvising and prizes novelty, whereas the epic singer understood himself to be transmitting inherited truth without alteration — so the tradition's own self-description is exactly the thing the evidence contradicts.
Clarifying the model
THE MODEL #Three refinements matter.
Formulaic does not mean mechanical. The system operates at several levels at once — phrase, scene, story-pattern — and skilled singers were judged on how they handled all three. A kit of parts constrains a builder without building for him.
Nor does formulaic mean unchanging. A tradition that composes afresh each time is precisely one that drifts: material enters, scenes swell, names shift. Stability of system is compatible with considerable movement in any given song.
And the honest disputes. Parry's thrift — one phrase per slot — has been shown to be less tidy than he claimed, with metrically equivalent alternatives sitting side by side more often than the strict version allows. The related idea that high formula density proves a text was orally composed has largely broken down, since lettered poets working in a traditional idiom produce formulaic verse too, and no agreed density threshold separates the cases. How the Homeric poems themselves reached written form — dictated, or composed by a lettered poet steeped in the oral idiom — remains genuinely unsettled.
Which gives the falsification test. If formulas exist to solve real-time metrical composition, epithet choice should track the metrical slot rather than the narrative moment, and the phrase-stock should be organised by rhythmic shape. It would be refuted by finding epithets selected for contextual aptness, or by finding that live singers hold their songs stable word for word between performances — and Lord's recordings are precisely where that second refutation could have happened and did not.
A picture of it
THE PICTURE #How to readThis is a packet diagram, normally used for the byte layout of a network message, borrowed here because a metrical line is likewise a fixed-width frame divided into slots. Read left to right as the line is sung: the singer's own material occupies the opening stretch, the caesura is the natural break inside the line, and the closing stretch is repeatedly filled by a ready-made noun-epithet phrase of exactly the right rhythm. The widths stand for metrical extent, not syllable counts, and the point is the shape of the constraint — the end of the line is where the traditional stock does the work, freeing the singer to steer the story in the part that comes before it.
What became clearer
WHAT CLEARED #The premise of the question was wrong in an instructive way. Nothing was held steady across generations except a technique. Singers did not store epics; they stored a metre, a phrase-stock fitted to it, typical scenes and story shapes, and rebuilt the song at every performance. The odd epithets that look like careless decoration are the visible joints of that machinery — and the machinery exists because composition happened at the speed of speech.
Where to go next
ONWARD #- How typical scenes and story-patterns work above the level of the formula.
- Why the oral-formulaic test for detecting orality in written texts lost its authority.
Key terms
TERMS #| Term | What it means |
|---|---|
| Formula | in Parry's sense, a group of words regularly used under the same metrical conditions to express a given essential idea. |
| Thrift (economy) | the tendency of a formulaic system to hold roughly one phrase for each metrical slot and function. |
| Composition in performance | building the song afresh during the singing, rather than reciting a memorised text. |
| Guslari | South Slavic epic singers, accompanied on the gusle, whom Parry and Lord recorded in the 1930s. |
Every term the collection defines is gathered in the glossary.