# T5 receipt test: VERDICT (29 August 2026, 23:41 ACST)

Order of record, from git: freeze committed 23:28:55 (ba6904a); scoring
launched 23:31 on that commit and committed 23:37:32 (0a9d37a); verdict
committed 23:41:17 (aa6975d). See DEVIATIONS.md entry 1 on the freeze header's
clerical timestamp.

Written under the readings frozen in FROZEN.md section 6 before any scoring.
Result files: results/ITERATION-1.md, results/t5_results.json,
results/background_rate.md. Deviations: none (DEVIATIONS.md is empty).

## Result

Study status: INFERENTIAL (background 1.17 per cent per cluster; expected
false hits 0.070 over six clusters, 0.035 over the three testable rows).

| # | Cluster | Pin | Status | Nearest miss (first failing criterion, margin) |
|---|---|---|---|---|
| 1 | Cohanim J1 CMH (L858 path) | J-Z1884 | FAIL | J-Z1884 itself: (d) stem 60 SNPs vs 36 expected, ratio 1.67 (3.34 against the half-expectation bar). Pre-registered expected failure; every other node on the 28-node path fails harder, mostly on window. |
| 2 | Cohanim J2a (J-M410*) | J-M410 | UNTESTABLE | pin is 19,000 ybp; the informative segment is unknown |
| 3 | Ashkenazi Levite R1a | R-Y2619 | FAIL | R-F1345 (3900 [3300-4500], 3 SNP-defined children, in window): (c) no ancestor CI reaches 7120; nearest is 380 years short. Under the frozen 500-year relaxation the anchor becomes R-M198 and (d) still fails: stem 76 SNPs vs 58 expected. R-Z2124 and R-Z2122 miss by the same 380 years. |
| 4 | Cohanim J-M12 | J-M102 | UNTESTABLE | pin is 15,600 ybp |
| 5 | Cohanim J-M318 | J-M318 | FAIL | J-Z423 (3500 [2900-4200], anchor J-Z387 6600 [5700-7500], gap 3100): (d) stem 64 vs 38 expected, ratio 1.66 |
| 6 | Samaritan priestly Cohen | E-M78 | UNTESTABLE | pin is 13,100 ybp |

Testable rows: 3 of 3 FAIL, 0 HIT. Supplementary non-counting rows (the
Z18271 branch for the CMH; the preprint-only J-FGC4975 path for the J2a
cluster): no hit either. O8 consistency check passed both ways. Rate 80.61
reproduced; all six pin identities asserted before scoring.

Secondary arm: 38 T4 PRIMARY matches resolve to 23 distinct anchors; 0
survive the compound two-daughter test. Only two sisters anywhere have 20 or
more SNP-defined leaves (J-Y20492 under J-Z8096, 34 leaves; H-Z5885 under
H-SK1225, 24 leaves) and both are one-SNP children with zero gap to their
anchor, so no stem-to-gap ratio exists for them. The three sisters with an
in-range ratio (C-F11898 1.21, G-Z31455 1.03, G-Y36036 1.02) have 6, 3 and 3
leaves. No geography check was delegated because there was nothing to check.

## Reading that applies: ALL-FAIL (verbatim from the freeze)

"All-fail: adverse for the conjunction (Architecture B and attested-cluster
continuity), with the full disjunction stated: the surviving patriline may
not be among the attested clusters or may be unsampled; the signature may be
invisible at dump resolution, as T4 showed it can be; or Architecture B is
wrong, and weight shifts back toward A, with the fork to be revisited."

And from the standing form: "If T5 fails across all attested clusters, this
datum legitimately weighs back toward Architecture A and the fork is
revisited with data."

Applied to what was actually measured:

1. The conjunction (Architecture B AND Noahic continuity through an attested
   Jewish patriline cluster) found no support in the three clusters the dump
   can test. The expected number of chance hits was 0.035, so a hit would
   have been informative; its absence is a real, if modest, negative.
2. The disjunction's branches, weighed against this run's specifics:
   - "Not among the attested clusters or unsampled": live. Three of six
     admitted clusters could not be tested at all because their only
     peer-reviewed pin is a 13,000 to 19,000-year-old haplogroup. The
     peer-reviewed literature simply does not yet locate the J2a Cohen, J-M12
     Cohen or Samaritan Cohen lineages at SNP resolution. The one SNP-level
     source that would help (Lipson et al., bioRxiv Dec 2025) is a preprint
     and was excluded; its proposed J2a pin scored as a supplementary row and
     also failed.
   - "Invisible at dump resolution": live, and T4 already showed the frozen
     stem statistic does not discriminate on this dump (38 short versus 40
     long tails). The Levite row is the clearest case: R1a-Z93's whole
     expansion is young, and nothing on that path is anywhere near a
     7,120-year anchor, so the row fails structurally rather than narrowly.
   - "Architecture B is wrong": the reading says weight shifts back toward A
     and the fork is revisited with data. That is now on the table, and it is
     Mike's call, not this run's. The strongest form of A recorded on 29
     August (Adam = L858, Noah = Y2919) remains on the record as the option
     to reach for if the fork reopens; it carries its own known cost (a
     lineage that sheds side-branches all along its stem).
3. The Cohen tension named in the standing form is sharpened, not resolved:
   the J1 CMH path fails exactly where it was pre-registered to fail (L858,
   ratio 1.67), and nothing above L858 rescues it. Held as founder-not-name,
   the row is a pre-registered non-event; held at full strength (founder =
   Aaron), it is a direct miss of the entry signature on the best-attested
   priestly line.

## Reading that did not apply: HIT

"Hit: a receipt candidate, probability language only, no name attached, with
a CI-width check and a report of which burst-child the cluster descends
through." Recorded as not applicable; no node on any counted or supplementary
path passed.

## Two findings for Mike that are not part of the verdict

- The vault's standing form says "the J1 Cohen cluster sits inside P58 inside
  L858". In the dump the nesting is the reverse: L858 sits inside P58
  (P58 9200 > Z1865 7300 > YSC0000234 5700 > L858 4400). The spec's
  pre-registered path was honoured regardless; the sentence in the vault
  should be corrected.
- The Levite R1a cluster is, on the papers' own account, an Ashkenazi founder
  event enriched in Levites rather than a Levite-exclusive line (Behar 2017).
  It was admitted because the spec named it; its row's evidential weight for
  "attested-cluster continuity" is accordingly lower than the Cohanim rows'.

## What was not done

No website change (none authorised). No literature check for a hit (none
occurred). No geography delegation (no surviving anchor). The frozen cluster
list stays closed; a rerun with the Lipson et al. pins becomes possible only
once that paper is peer-reviewed, and would need a new freeze.