The translation outruns every Greek manuscript

Claim (verbatim)

Proclus' Commentary on the Parmenides - late Neoplatonism's masterwork on the One - survives in Greek in a tradition with a single wound: every Greek manuscript breaks off at the same point in book VII, before the conclusion of the first hypothesis. The ending exists anyway, because William of Moerbeke's Latin translation, made in the 1280s from a Greek exemplar older and fuller than the archetype behind our entire Greek tradition, keeps going past the break to the end of the first hypothesis - the stretch where the One is beyond name, knowledge and being. That Latin-only tail was first printed by Klibansky and Labowsky in Plato Latinus III (1953); Carlos Steel's editions of the commentary (Leuven, 1982-85, and the subsequent Oxford critical edition) use Moerbeke as a primary witness throughout precisely because his exemplar predates the surviving Greek line, and Moerbeke's notorious word-for-word literalism makes the lost Greek retrovertible almost clause by clause. Nicholas of Cusa read and annotated this Latin; the negative theology of the Latin West thus fed on pages of Proclus that no surviving Greek manuscript contains. The mechanism is the archetype funnel: a tradition that narrows to one damaged Byzantine copy shares that copy's missing final quires forever after, while a translation made outside the funnel samples an older textual state - so a thirteenth-century Latin friar can outrank the entire Greek manuscript tradition as a witness to a Greek philosopher's conclusion.

Prediction clause (verbatim)

Prediction: in the witness registers of the modern editions, every catalogued Greek manuscript of the commentary will terminate at the same locus in book VII with none carrying any part of the final section, while the Latin tradition carries a continuous further stretch amounting to at least 5% of book VII's length in the editions' pagination (primary clause: the shared Greek terminus plus the >=5% Latin-only tail; the verdict follows it); secondarily, the apparatus of the last Greek-extant stretch will cite Moerbeke's readings against the united Greek consensus at multiple loci; Greek excerpts and scholia count as continuation only if they carry post-truncation text, and the test voids unless the shared terminus can be confirmed for at least 90% of catalogued Greek witnesses.

Kill-dataset (verbatim)

Kill: Steel's Leuven edition (1982-85) and the Oxford critical edition of Proclus' In Parmenidem with Klibansky-Labowsky's Plato Latinus III - a shared-terminus census over the Greek witnesses plus a length measurement of the Latin-only tail.

Novelty / leakage triage

already answered in the literature

The shared Greek terminus and the Latin-only ending are the published discovery of the very editions the kill names: Klibansky-Labowsky printed the pars ultima from Moerbeke in 1953 as text no Greek manuscript carries, and Steel's editions establish that the entire Greek tradition descends from one truncated archetype while Moerbeke's exemplar predates it, citing his readings against the united Greek consensus throughout. The tail is printed in extenso, so the >=5% share of book VII is page-arithmetic on the editions with the threshold comfortably cleared; the quantification adds a formalization, not a finding.

• R. Klibansky & C. Labowsky (eds.), Parmenides usque ad finem primae hypothesis nec non Procli Commentarium in Parmenidem, pars ultima adhuc inedita interprete Guillelmo de Moerbeka (Plato Latinus III, London, 1953)
• C. Steel (ed.), Proclus: Commentaire sur le Parmenide de Platon, traduction de Guillaume de Moerbeke, 2 vols (Leuven, 1982-1985)
• C. Steel et al. (eds.), Procli In Platonis Parmenidem commentaria, 3 vols (Oxford, 2007-2009)

Predictions

No prediction registered yet.