Two Taliesins, one attested and one imagined

Claim (verbatim)

The Book of Taliesin (Aberystwyth, NLW, Peniarth MS 2) holds two very different poets under one name. Ifor Williams, in Canu Taliesin (1960), isolated about a dozen poems as the plausibly genuine sixth-century work of a historical bard - praise-poems to Urien Rheged, his son Owain, and Gwallog, kings of the Old North. The rest of the manuscript, edited by Marged Haycock in Legendary Poems from the Book of Taliesin (2007) and Prophecies from the Book of Taliesin (2013), belongs to a legendary, encyclopaedic, prophetic Taliesin - a persona who boasts of transmigration and hidden knowledge. The two strata differ not only in matter but in their anchorage to the world outside the poem. The historical praise-poems name patrons who recur in the genealogies of the Old North and in the Historia Brittonum's account of Urien; the legendary poems refer inward, to a mythos with no independent record. That difference is a measurable asymmetry of external corroboration, and it bears on loss: the historical stratum is the visible corner of a court-poetry that was largely not preserved, while the legendary stratum is a later accretion that grew where the record was thinnest. Restated: among the roughly twelve historical Taliesin poems a large majority name a patron or dynasty independently attested in the early genealogies or annals, whereas an equal-sized sample of the legendary poems yields near-zero such external anchors. One Taliesin is pinned to history; the other floats.

Prediction clause (verbatim)

Prediction: scoring each poem for whether it names a person or dynasty independently attested outside the Book of Taliesin (in the Harleian genealogies, the Historia Brittonum, or the Annales Cambriae), at least two-thirds of Ifor Williams's roughly twelve historical praise-poems will carry such an external anchor, while a matched sample of the legendary and prophetic poems (Haycock 2007, 2013) will carry one in fewer than 15 percent of cases (primary clause: the corroboration gap - at least 0.67 for the historical stratum against below 0.15 for the legendary; the verdict follows it). Anchoring requires a named person or dynasty, not a place-name or a common epithet. The test voids if the two strata cannot each supply at least ten scorable poems.

Kill-dataset (verbatim)

Kill: the two Taliesin strata as delimited by Ifor Williams's Canu Taliesin (1960) and Marged Haycock's editions (Legendary Poems from the Book of Taliesin, 2007; Prophecies from the Book of Taliesin, 2013), each poem's named figures checked against the Harleian genealogies (BL Harley 3859), the Historia Brittonum, and the Annales Cambriae - compare external-anchor rates between the strata.

Novelty / leakage triage

anticipated in the literature — this exact test has never been run

The two-strata division is exactly Ifor Williams's and Haycock's published work, and the historical poems' external anchorage - Urien and Owain in the Historia Brittonum's North-British chapter and the Harleian genealogies - was Williams's own selection criterion, so the historical side of the gap is close to definitionally satisfied in print. The legendary side is not: the prophetic stratum names independently attested dynasts (the Armes Prydein tradition in the same manuscript invokes Cadwaladr and Cynan, both anchored in the genealogies and annals), so whether a matched legendary-and-prophetic sample stays under 15 percent anchored depends on sample construction and has never been scored. The comparison as an explicit two-rate test is un-run, and the second rate is where it could die.

• Ifor Williams, Canu Taliesin (Cardiff, 1960); English version by J.E. Caerwyn Williams, The Poems of Taliesin (Dublin: DIAS, 1968)
• M. Haycock, Legendary Poems from the Book of Taliesin (Aberystwyth, 2007)
• M. Haycock, Prophecies from the Book of Taliesin (Aberystwyth, 2013)

Predictions

No prediction registered yet.