Bifolia: which leaves were written together

The manuscript is sewn from bifolia: a sheet of parchment folded in half gives two leaves of the book, one nearer the start of the quire and one nearer its end. A test of 24.10.2026 showed that the text remembers this structure. Pages of one bifolium share their whole vocabulary even when other leaves now lie between them: 1.6-1.7 times more than pages from different quires, the same on two independent transcriptions (Zandbergen ZL and Takahashi IT2a). The beginnings of pages are similar across an opening - the two pages seen together in the open book (1.3 times), and most of all at the centre of a quire, where the opening is a single bifolium (2.9 times). Openings whose pages are in different Currier "languages" (A and B) show none of this: those pages were not written together.

Below, each bifolium is shown as an open sheet - as it lay before the scribe before sewing: the inner side and the outer side. The score shows how much closer the vocabularies of its pages are than those of random pages of the same section from other quires. The bifolium vocabulary is a stable trait (median: closer than 85 % of random pairs). The similarity of opening starts is so far only a descriptive number: a first check against the physical evidence of binding (a drawing running across the fold; leaves lost between pages) did NOT confirm it as a pointer to the original opening.

How to read it

• Bifolium vocabulary - percentile among random pairs of pages of the same section from different quires: 0.5 - like random pages, 0.9 - the pages' vocabularies are closer than in 90 % of random pairs. The measure is the shared word stock of the pages of the two leaves (Jaccard).
• Centre of quire - a bifolium whose two halves lie side by side: its inner side is an opening.
• Openings of the quire - neighbouring pages in the present binding and the score of their beginnings (first words of the page); A-B marks pages in different Currier languages.
• Foldouts are shown with the scan of the whole foldout. Scans - Beinecke Library (Yale), reduced; a click opens the full scan.

What we do not claim: that this is a reading of the text - there are no readings; that a low score proves a move - it is a candidate to check against paint offsets, prickings and codicology; that the present order of leaves is original - the page shows it as the binding gives it (ZL markup). These are descriptive numbers and a hypothesis about how the book was made (medium confidence): the scribe worked on separate open bifolia and wrote the start of a page with an eye on the facing page.