Pre-registration: Does Linear A Fall Outside the Mycenaean Range? A Held-Out Test on Thebes, Mycenae and Tiryns (Phase 22.D).

Locked pre-registration for Phase 22.D. Phase 22.C was contaminated by premature execution (see doi 10.5281/zenodo.22748836) and is void as a confirmatory test. This phase repeats the range-non-overlap criterion on a genuinely unseen hold-out: Linear B sub-corpora from Thebes, Mycenae and Tiryns, whose word-final -o rates have never been computed, inspected or reported in this project. The reference corpus will be harvested from DAMOS only AFTER this record is published, and its SHA-256 will be deposited with the results. Hypotheses, statistics, thresholds, validity gates and the full verdict table — including the pre-declared retraction condition — are fixed here. No results. Execution is performed solely by the author; the AI assistant executes nothing. LOCK STATEMENT: this record's publication date and the accompanying OpenTimestamps proof constitute the lock. SHA-256 of 176_phase22D_prereg_v1.md: 63e66197dbcd768c01d35c72002f7c253a8854c1512a815bd9f9e17cb2592926. Version v2 (2026-09-15): OpenTimestamps proofs upgraded to confirmed Bitcoin block attestations. The pre-registration is anchored in blocks 966966, 966969 and 966971; the hold-out corpus (damos_holdout_thmyti.csv) and the runner are anchored in blocks 967020, 967026 and 967042. The pre-registration therefore precedes the hold-out data by 54 blocks (approximately nine hours), so the blindness of the test is established by the Bitcoin blockchain rather than by declaration. Adds the OpenTimestamps proofs for the hold-out corpus and the runner, and an updated 176_phase22D_prereg_v1.PROOF.txt recording all SHA-256 checksums, DOIs and block heights. No change to the pre-registration text, the runner, or the TEST-VOID verdict.