Comparison of the results — three similarity maps
Compare manuscripts' textual proximity (shared apparatus readings with regular spelling tendencies neutralised) against their orthographical proximity (substitution-rate profiles). The first two networks are laid out independently from their respective evidence; the third superposes the two sets of links on the textual layout, making divergent affiliations visible.
Normalisation settings (all built-in equivalences active by default)
Two proximity networks and their comparison
Identical manuscript sets, node labels and nearest-neighbour rules are used in all three maps. Each has independent zoom, pan and Fit controls. Positions in the two individual maps are calculated separately, from textual or orthographical relationships.
1 Proximity according to shared readings
Textual proximity after normalising regular spelling conventions. Links join the closest neighbours under the currently selected coverage-adjusted textual ranking.
2 Proximity according to orthographical tendencies
Orthographical proximity from recurrent spelling rates. Links join the closest orthographical neighbours.
3 Superposition: comparison of both networks
The third network uses the positions of Map 1 as a common reference. Blue and orange lines represent independent neighbour selections and purple lines are shared by both analyses. Zoom with mouse wheel, pan by dragging, or use the + / − / Fit controls. Links show relationships, not a stemma.
Most significant discrepancies
The discrepancy score is the difference between the pair's textual and orthographical neighbour ranks, averaged from both manuscripts' perspectives. When sorting positive discrepancies, the ranking also gives priority to substantial Bayesian affinity and well-attested overlap, rather than treating a near-zero score as a strong textual relationship. Positive values mean textually closer than their spelling habits suggest. Only pairs with sufficient evidence in both analyses are ranked.
Specific comparison: MSS 2000 and 2065
These witnesses have a documented exemplar–copy relationship, but no connection is imposed on either model. Their textual and orthographical ranks are calculated independently.