Recorded readings
The NEXUS matrix records textual states. A match at a site does not itself identify a copying direction.
Recorded + reconstructed history / identity
The Caxton second edition carries a base-text relation to Cx1 and an attributed correction source from a second, unresolved exemplar. The graph must preserve both without pretending that shared words reveal ancestry.
01 / Bounded result
The source corpus has 58 witnesses and 4032 NEXUS characters. This interface displays seven witnesses and two source-discussed readings to explain the result, not to substitute a miniature corpus for the full one.
02 / Attributed transmission
Every arrow below is attributed to Robinson's published analysis, not observed directly. Solid slate arrows are tree-compatible within this bounded view; the orange dashed correction source is explicitly not.
03 / Recorded reading slice
The two sites are deliberately selection-biased examples from the published discussion. “Not supplied” means the compact source projection contains no count for that cell; it is not zero.
04 / Published correction profile
Robinson reports 207 differences between Cx2 and Cx1 where the Cx2 reading also appears in more than three witnesses. This supports extensive correction from a second exemplar in the published analysis. It is neither 207 copied events nor a Historical Load value.
05 / Similarity guardrail
06 / Evidence ablation
07 / History equivalence
Pn and Wy can share both displayed readings, the published group, and the bounded transmission role while remaining distinct witness records.
08 / Historical Load
null, not zero09 / Reading discipline
The NEXUS matrix records textual states. A match at a site does not itself identify a copying direction.
Exact equality over the selected slice is deterministic but knowingly non-representative.
Copying and correction-source relations come from the published analysis and remain non-observed.
The better copy is not given an invented shelfmark, extant witness identity, or false precision.