while-overlap / while-contrast — did ‘while’ mean at the same time, or ‘whereas’?

Deterministic screens SCREEN PASS

These are code-based surface checks, not a measured robustness result or proof that readers understand the construct.

• one-edit corruption min distance 1 while-overlap → while overlap (d=1 · visible) while-contrast → while contrast (d=1 · visible) while-overlap → while (d=8 · camouflaged) while-contrast → while (d=9 · camouflaged) while-overlap(E; C) → while-overlap(non-interval-ref; C) (d=16 · visible)
• slot cross-product min distance within slot 6
• transform screen no collision in the fixed transform list (finite-list floor, not proof of transform safety)
• background collision floor COMPUTED — no collision in the fixed 229-word list No fixed-list background collision found. Reported, never gates: some constructs choose a collision deliberately, but voters should see it chosen. FLOOR, not a verdict: the word list proves membership and cannot prove non-membership, so hits here are real and a clean result is not evidence of safety (ordinary words absent from a fixed 229-word list — `unless`, `given`, `except` — read clean and are not).

Server-computed from the construct's own declared surface; the attacks are derived from the slot, never chosen by the proposer. Reproduce any of it: python3 measure.py (the reference harness).

Predicted measurement its falsifier

PRIMARY CLAIM CARRIER: preregister at least 144 fresh consequence scenarios, balanced 72 temporal and 72 contrastive, across operations, monitoring, contracts, scientific summaries, scheduling, safety instructions, product comparisons, and ordinary coordination. Each world fixes the relevant event interval, both clause truth values, and whether timing or comparison is the load-bearing relation. Randomize readers across three arms: the registered form, deliberately ambiguous bare `while`, and complete careful English using `during a nonempty part of` or `whereas` with the same facts. Ask held-out questions that do not repeat marker words: whether a scheduler must overlap actions, whether either clause can occur at a different time, whether both clauses are asserted, and whether one clause is merely a time anchor. The declared `comprehension_accuracy_delta` is registered form minus the balanced bare-`while` arm, not registered form minus careful English. Prediction: at least +25 percentage points overall, at least +20 points in each relation stratum, and at least 90% absolute exact relation-plus-entailment accuracy for each marker. Complete careful English is a ceiling and information-equivalence control: report it separately, and flag a deficit greater than 5 points as a usability warning rather than relabelling it as success on the bare-English claim. Report every form × domain × question-type cell. REFUTED if either marker fails 85% absolute accuracy, improves by less than 10 points over bare `while`, induces temporal-overlap answers on more than 10% of contrast cases, induces contrast answers on more than 10% of temporal cases, or routinely imports causation, full-duration coverage, preference, or exception semantics. A ceiling-bound or chance-bound arm is unresolved, not a win. PREREQUISITE: on a separate frozen set of at least 48 complete semantic pairs, measure `token_delta` for complete marked sentences against their complete careful-English mappings under current cl100k_base, o200k_base, and p50k_base. The least-favourable tokenizer mean may be positive but must be at most +4 tokens. Cost against bare `while` is diagnostic only because bare `while` omits the load-bearing distinction. ROBUSTNESS: test hyphen-to-space, case folding, dropped suffixes, swapped clause order, missing or non-interval event references, negated clauses, nested reported speech, both relations holding in the same world, and speech-to-text loss. Hyphen loss may fall back to direction-preserving ordinary wording; dropping `overlap` or `contrast` must reopen ambiguity rather than silently selecting a reading. Verify gold answers against frozen interval and clause records, not annotator intuition. Adoption remains separate evidence: zero non-author use in a current post-ratification scan counts against flagship status.

Measurement

Comprehension accuracy: no settled result

Technical aggregate assessment: unmeasured. Results concern the recorded comparisons and populations. Token cost, comprehension and declared-plan completion are separate questions.