Proposed correction notice

Publication identified by author: “Exploratory research awaiting independent verification”, Ian James Prince, Zenodo, published 10 October 2026, version v1. Record DOI and original deposited files remain to be verified. This document is a proposed supplement, not a claim that the Zenodo record has been amended. Proposed correction notice This supplement clarifies the evidential status of exploratory statistical observations reported in the 10 October 2026 record. The original invitation already acknowledged that stronger controls weaken the observed EVA -dy / -ey → qo- association. We retain that qualification and provide more specific methodological context. No decipherment, linguistic interpretation, scribal rule or manuscript-level glyph-generation mechanism has been established by these analyses. The available research dossier reports a Hand 3 folio-and-predecessor-stem matched contrast of approximately +11.176 percentage points (920 matched observations), with a one-sided within-hand permutation p-value of 0.00267 from 3,000 shuffles. This is an exploratory statistic, not a multiplicity-adjusted confirmatory result. The corresponding Hand 2 estimate is approximately +0.971 percentage points across 737 matched observations; the descriptive difference between hand-labelled subsets is not itself a formal test of a hand-by-ending interaction. Sensitivity testing matters. The reported within-line and exact-successor-length randomisation for Test 431 gives p = 0.33813. Accordingly, the association has not been shown to persist independently of the tested local line/length structure. Conditioning on successor length may itself alter the estimand or induce bias; this control is therefore a sensitivity analysis, not proof that the original association is artificial. The analysis used historical EVA transcriptions. Agreement between ZL3b and IT2a cannot by itself establish a one-to-one reading of original manuscript glyphs, especially where transcription conventions differ or omit/merge signs. Later VIGOR/VCT work motivates independent manuscript-image validation, but its subsequent observations must not be retroactively represented as evidence available on 10 October 2026. The archived programme comprises many sequentially developed exploratory tests. The available dossier does not provide a complete clean-environment replication of every historical test, and original transcription files are not included in the proposed public package pending source-rights checks. Absence from this snapshot should not be interpreted as proof that earlier experiments were never conducted. These qualifications do not withdraw the numerical observations. They narrow their interpretation and identify the independent replication and original-image checks required before a mechanism claim can be advanced. Correction ledger Hand 3 -dy/-ey association — Retain as a descriptive matched EVA-transcription association; disclose selection and multiple-testing limitations. Cross-hand difference — Report within-hand contrasts separately; do not claim a tested hand-by-ending interaction. Local structure — Disclose Test 431 exact-length sensitivity p = 0.33813, with caveat about conditioning. Glyph-level interpretation — Distinguish EVA tokens from VIGOR/VCT one-to-one manuscript glyph readings. Reproducibility — Specify which scripts, source texts, metadata and checksums are available; do not label archive fully independently reproduced. Verification before Zenodo submission Confirm DOI and exact v1 deposited files; compare all numbers against the relevant scripts and result CSVs; add script versions, seeds and source-transcription provenance; link v2 to v1 through Zenodo versioning; include this supplement as a dated change log. Do not alter the original v1 description without a transparent revision record.