Claims against their evidence (stage E1, section 1)

Written at E1b-04, 27 October 2026. Contract section 1 names two kinds of claim:

• a pattern claim rests on counts across rooms, with their denominators;
• a words claim rests on what members wrote, through exemplars that resolve and the return's verdicts.

For every proposition (propositions.jsonl) and every branch (branches-D5.jsonl), this table gives:

• the kind declared;
• the evidence named;
• the rival;
• whether the validation the claim received is of its kind.

A claim validated with the other kind's evidence would be a defect. No claim is.

The sources:

• audit-E1/claims.py (E1a) lays out the fields.
• audit-E1/claims_returns.py (E1b-04) counts the return's verdicts on each arrow. It reads the last ruling in adjudications.jsonl of every return line on the arrow, from both the first return and the re-check, and counts the accepted lines by op.
• The arrows' support quotes were resolved against the store at this leg: 57 quotes, all of which resolve.

The model is the D3 freeze, 46384686…. The held proposals for P05 and A11 are not part of it.

The propositions

An arrow's order count reads "holds of both". "Both" is the number of rooms carrying both of its elements. "Holds" is the number in which the first concept of the one comes before the first concept of the other. The null is the 95th percentile of 500 shuffles within each room (seed 20261014). The return's accepted lines are summed over the proposition's arrows.

The bracket each sits in (prior.md section 4):

• The order counts run over the whole record. The pre-event week, placement (hours 0 to 73) and aftermath (73 to 97) are counted together.
• Only A20 (N05) is also read inside a bracket, before hour 73.
• Every counted arrow but A12 and A21 carries the onset contrast's early and late counts (early_late).

Not done, against prior.md section 3:

• No arrow's order test was run at held size, although section 3 says the population pattern is "supported by a pattern that holds in counts over the analysed set and at held size". The one finding tested at held size is F-009 (the play frame does not delay the canvas work).
• The order was read against section 4's brackets for A20 alone.
• Both gaps are recorded in defects.jsonl (line 114).

The branches

Every branch rests on its shared key in the ledger for membership, and on its account's quotes for what it is. The 35 accounts cite 38 to 89 concepts each, and those resolved at E1b-01. The kind of each warrant is therefore both, as prior.md section 3 asks of forms and subforms. "Form" means the branch passed D5's form test; "grouping" means its members share the key but no condition was stated that explains them jointly. Local validation is established for none.

The verdicts are the first return's. branches-D5.jsonl names no proposition that a branch holds or fails.

prior.md section 3's declaration, against what each level rests on

• The population pattern.
  • Declared: counts with their denominators; the order on the clock against section 4's brackets; the onset contrast; the organizing end under section 5; support at held size; the words secondary.
  • Rests on: counts with their denominators and an order beating a within-room null, for 18 of the 20 arrows the propositions rest on. A11 is not beaten, and A12 has no count; A21 is in no proposition and has no null test. The onset contrast's counts are recorded for 19 of the 20, all but A12, and the organizing end is A1's field.
  • Not done: the held-size test, and the brackets for every arrow but A20.
  • The words: the 57 support quotes, secondary as declared.
• The forms and subforms.
  • Declared: both kinds, the shape primary for membership and the words primary for what the form is.
  • Rests on: just that. The keys of D4 give membership, and the accounts' cited concepts give the words.
  • Status: two of 35 branches pass the form test; 33 are groupings, reported as such.
  • The six room forms of D4a-05 to D4a-08: they rest on the same two kinds, by changes and codes, with their accounts (room-form-*-account.md).
• The particular histories.
  • Declared: the words primary: a dated sequence of the room's own turns that the model's elements reach.
  • Rests on: the room models of D2 and D3 and the coders' case accounts.
  • The element-reach half of section 3's local measure was computed at E1a (audit-E1/element_reach.py). Of 5,380 second-order themes that hold a coordinating turn, in 696 rooms, 4,759 are reached by a standing element (88.46 per cent), or 5,239 with the candidates (97.38 per cent). 297 rooms are reached in full, or 581 with the candidates.
  • Not computed: the other half, a validated adaptation or a theme listed as unresolved with its reason. So local validation is not established.
• The method question's verdict is the work of section 3 of the audit.

The switch records of literature-C3.md, against the findings they superseded