A whole poet, passed through a single lost book — so that book’s gaps and slips became everyone’s inheritance
R847’s longer ending won by being in the exemplars: present in the book in front of each copyist, it needed no persuasion. Its closing turn asked for the opposite geometry — not a variant spreading through many copies, but a whole thing surviving through a single ancestor, so that one copy’s accidents became every later copy’s inheritance. The Roman poet Catullus is that shape, almost perfectly.
One glimpse, then silence. Catullus wrote in the first century BC and was read for a time after his death, but his poems were never a school text, and the trail thins fast. A bishop of Verona, Ratherius, is reported to have been reading him around 965 — a single glimpse, in the poet’s own home town — and then, for centuries, effectively nothing. Whatever copies once existed had, by the early fourteenth century, narrowed to the edge of vanishing.
The one book. What resurfaced at Verona, early in the 1300s, was a single manuscript, known to scholars as V, the Codex Veronensis. It held the collection essentially whole — in modern counting, a little over a hundred poems, more than two thousand lines — but not intact: even in summary accounts it is described as complete save for roughly a dozen gaps of a line or more. V itself then disappeared, before the century was out. It has never been found again. Everything since is reconstruction from what was copied out of it in the brief window while it was available.
Three copies carry the world. Apart from one much older exception — a ninth-century manuscript holding poem 62 alone — every surviving manuscript of Catullus descends, directly or through further lost intermediates, from V. Three fourteenth-century copies are of the first rank: G, now in Paris, dated 29 October 1375; O, now at Oxford; and R, now in the Vatican, written around 1390 and not recognised for what it was until 1896, when it was found in the Vatican collections and eventually collated in full in 1970. Stemma diagrams draw lost copies between V and these three, and their exact relations are still not completely settled. That only tightens the point: the bottleneck may not even be one step. It may be a short chain of single steps, each one a place where the whole tradition passed through one pair of hands at a time.
What a bottleneck does. Mark’s longer ending, R847’s case, could in principle be checked against copies that did not have it — the early codices ending at verse 8 are a surviving control group. A bottleneck leaves no control group. A gap in V is a gap in every descendant: there is no sibling line to fill it from, so editors mark the hole or guess across it by conjecture. An error in V is harder still, because it does not look like an error from inside the tradition — all three first-rank witnesses can agree on a reading, in perfect good faith, and the agreement proves only that V had it, not that Catullus wrote it. Agreement downstream of a bottleneck is one voice, counted three times. The modern text of Catullus is therefore a portrait of two people at once: the poet, and the unknown state of the one book at Verona — its damage, its omissions, its scribe’s habits — with no clean way, at the places where they diverge, to tell which portrait you are looking at.
The toy. Schematic only — a stemma walk, not a census of lost antiquity, whose true numbers are unknown. Follow a single reading through the tradition. The point is the asymmetry: a reading present in V is transmitted to effectively 100 percent of the later tradition, and a reading present only in a lost sibling of V is transmitted to 0 percent of it — regardless of which one the poet wrote. Survival through one ancestor does not average the population. It selects it, silently, in full.
Before the bottleneck: an unknown number of ancient copies, differing among themselves in ways we can no longer recover.
What is actually kept. Liberty’s copies multiplied a figure; Mark’s ending multiplied a variant. Catullus was not multiplied at the decisive moment at all — he was narrowed, to one book, in one town, for part of one century, and the multiplication only resumed after the narrowing had already decided what could be multiplied. That his poems feel so immediate is part of the strangeness: the voice comes through loud and intact in most places, and in a handful of places what comes through is a hole shaped exactly like the damage in a book nobody has seen since the 1300s. A bottleneck does not merely risk a thing. It edits the thing’s future population, permanently, and leaves no record of the edit — except the population itself.