It Dated the Afternoon

In 1935 Milman Parry recorded Avdo Mededovic, an illiterate singer in Yugoslavia, performing an epic of some twelve thousand lines.

The recording is a date. It is not the date of the story.

Parry had spent three summers recording guslari around Dubrovnik with Albert Lord, and the work left thousands of texts and several thousand aluminium discs, now held at Harvard. It is one of the best-documented bodies of oral performance anywhere.

None of it turns 1935 into a beginning.

It makes 1935 the date at which one performance became a recoverable witness.

That distinction goes missing fast when the thing under inspection is a market maxim rather than a song. Somebody says the maxim is old. The earliest thing anyone turns up is a research note from eleven years ago. The note gets read as an origin, or its absence gets read as a refutation, and both conclusions ask the record for something it does not hold.

A recording fixes a performance, not a tradition #

Lord went back to Yugoslavia at the start of the 1950s, and in The Singer of Tales he set out what the material showed: oral epic is composed anew at each performance, out of metrically fitted formulas and type-scenes.

A singer can experience every performance as the same song while no two performances are word-for-word alike.

So a field-collection date has a narrow meaning and should be given one. It identifies when a collector heard one instance, from one performer, under one set of circumstances. It does not date the formulas, the type-scenes, the previous tellings, or the moment somebody first called the material traditional.

The record can be vast and still be late. Ireland’s National Folklore Collection holds millions of manuscript pages, and its Schools’ Collection was gathered at the end of the 1930s by more than fifty thousand schoolchildren across some five thousand schools. Those dates describe a national collection programme carried out over three years.

They are not the birth dates of anything the children heard at home.

An unrecovered predecessor is a gap in attestation. It may be a good reason to reject a tidy origin story. It is not a licence to collapse transmission into nonexistence.

A tradition is documented as variants #

Francis James Child published The English and Scottish Popular Ballads between 1882 and 1898, numbering 305 ballad narratives. A Child number does not assert that its texts are identical. It groups several recorded versions Child judged to be one underlying ballad despite differences in wording, verse order and plot.

Barbara Allen, Child 84, carries more minor textual variation than almost anything else in the catalogue. Charles Seeger later assembled a collection of its versions and variants from thirty American recordings made over two decades, and the plural in the title is doing real work.

The object being documented moves.

Tehrani’s study of Little Red Riding Hood made that countable, coding dozens of plot variables across dozens of documented variants from thirty-odd populations. A branching descent model explains much of the pattern; borrowing and convergent change explain the rest.

The result does not pick an authentic wording and discard the others. It measures how much shared history the variants can actually carry.

The operating rule follows directly. A claim under audit needs its versions kept verbatim, dated, and linked to each other. A phrase added in a slide deck, a qualifier dropped in a note, a source name attached after the fact — none of that is editorial noise. It is the material a transmission chain gets built from.

An old source proves only what it says #

The opposite error starts with a source that genuinely is old. Its age lends the modern account an air of depth, and the wording gets quietly strengthened on the way through.

Zecharia Sitchin’s The 12th Planet appeared in 1976 and proposed that visitors from a planet on a long elliptical orbit engineered humanity and founded Sumerian civilisation. The book’s date and its contents are separate facts from the validity of the reading it offered. Michael Heiser identified specific mistranslations, including Sitchin’s rendering of the Sumerian sign DIĜIR, against meanings attested in bilingual dictionaries compiled by ancient scribes.

Being an old book does not make it a source for a stronger claim, however many later accounts cite it.

Erich von Däniken supplies the simpler version of the test. Chariots of the Gods? presented the Ica stones as ancient evidence of advanced technology, and the stones were traced to a living modern carver producing them for the tourist trade. The book remains a perfectly good witness to what its author asserted in 1968. It is not evidence that the stones are ancient, and nothing can make a modern embellishment older than the book reporting it.

Collection and compilation leave different dates #

Elias Lönnrot made eleven field trips from 1828, mainly in Karelia. One singer, Arhippa Perttunen, supplied some four thousand verses across two days on a single visit. The Old Kalevala appeared in 1835; the New Kalevala, nearly twice the length, in 1849.

Those are unusually good dates, and they date four different things: fieldwork, one performance, an editor’s arrangement, and two published editions.

The continuous national epic was never a discovered manuscript. It is a construction out of collected material, arranged and expanded by Lönnrot, and the record says so plainly.

Here the easy audit has to stop being easy. Oral tradition persists without a written archive. A market narrative can leave extensive private electronic traces that are inaccessible rather than absent, and no search procedure converts either case into a recoverable early witness.

The restraint required is modest. Mark the first witness as the first witness. Mark a claimed origin as claimed. Mark each leap between them as supported, inferred, or unaccounted for.

Then leave the chain broken where it is broken.

Avdo Mededovic sang into a machine in 1935, and that machine is the reason anyone can date anything about him at all.

It dated the afternoon.

It did not date the song, and no better equipment would have. A file that says where it stops is worth more than one that closes.