# Final Tier D Batch 1 — focused Discovery / Editorial Freeze

Scope: CONVENT / FIL / MARCHAND / MONTRER. Baseline: `1388f4283b37179b4d5eaddc93257333ce21e1c8` (33/42 migrated). Review: Codex, 2026-09-20, under Jinkai Liu's explicit authorization to perform focused review and conditionally migrate Ready entries in the same task. This is not a claim that Jinkai Liu separately reviewed the newly generated candidate assessments.

All four: **Ready to migrate with Featured Pending**. Standard meanings, scoped source history and evidence boundaries can be published without selecting a phonetic winner. No new schema, pipeline or Batch 2. All cross-language Historical Relation: **Not claimed**.

## Frozen historical units

| Entry / ID | Source history and necessary stages | Modern standard | Boundary |
|---|---|---|---|
| CONVENT / LB-en-convent-convention-023 | convenire → conventus (assembly) → ecclesiastical/medieval conventus (community/dwelling) → English convent via Anglo-French and Middle English covent; 4 stages | 修道院／女修道院 | conventio → convention is a separate branch; not custom derived from a monastery |
| FIL / LB-fr-fil-filiere-fille-019 | Latin filum → medieval fil → modern French fil; 3 stages | 线／丝，依语境 | filière derives from fil; fille derives from filia, not thread |
| MARCHAND / LB-fr-marchand-march-021 | mercatus → reconstructed *mercatare / *mercatantem → French marchand (earlier marchedant etc.); 3 stages | 商人／商贩 | marcher follows a separate walking lineage. Reconstruction is not attestation |
| MONTRER / LB-fr-montrer-monitor-018 | Latin monstrare → mostrer/monstrer → modern montrer; 3 stages | 显示／展示／指出，依语境 | monstrare ← monstrum is related more deeply to monere; monitor is a separate monere formation, not a direct ancestor/descendant of montrer |

The chronology above records source forms, not a forced sequence of every associated meaning. No Chinese mapping is required at a bridge stage. Modern Chinese comparisons do not imply contemporaneous contact.

Sources: [Merriam-Webster convent](https://www.merriam-webster.com/dictionary/convent), [CNRTL couvent](https://www.cnrtl.fr/lexicographie/couvent), [CNRTL convention](https://www.cnrtl.fr/definition/convention); [CNRTL fil](https://www.cnrtl.fr/lexicographie/fil), [CNRTL filière](https://www.cnrtl.fr/etymologie/fili%C3%A8re), [Académie fille](https://www.cnrtl.fr/definition/academie9/fille); [CNRTL marchand](https://www.cnrtl.fr/definition/marchande), [CNRTL marcher](https://www.cnrtl.fr/etymologie/marcher); [CNRTL montrer](https://www.cnrtl.fr/definition/montrer), [CNRTL moniteur](https://www.cnrtl.fr/etymologie/moniteur), [MW monster](https://www.merriam-webster.com/dictionary/monster), [MW monitor](https://www.merriam-webster.com/dictionary/monitor). Specific online entries were checked; some CNRTL etymological passages were retrieved through indexed entry text rather than a complete direct page response. No quotation or invented historical IPA is used.

## Bounded candidate table

One focused pass, **12 serious candidates + 8 controls**. Chinese layer throughout: **Modern Mandarin**, date metadata modern reference, historical first attestation Pending. Readings are Hanyu Pinyin; segment comments are limited modern phonetic comparisons. Historical Latin sound fit remains **not_evaluated** where no pronunciation has been frozen.

| Entry | Chinese / reading | Target stage and scoped sense | Semantic | Phonetic | Decision / counterevidence |
|---|---|---|---|---|---|
| CONVENT | 会 huì | conventus ASSEMBLY; 集会/会合 | Direct | Not evaluated (Latin) | Shortlist; not kuài accounting and not modern nunnery |
| CONVENT | 聚 jù | ASSEMBLY; 聚集 | Direct | Not evaluated | Shortlist; action/result distinction |
| CONVENT | 合 hé | ASSEMBLY; 聚合 | Partial | Not evaluated | Reserve; not gě measure; object combination need not be human assembly |
| CONVENT | 困 kùn | Sound control; 困住/困难 | Mismatch | Not evaluated; only modern /k/ + nasal interest | Reject as mapping; confinement does not establish assembly |
| CONVENT | 修道院 xiūdàoyuàn | Semantic control: modern convent | Direct modern; partial/narrow for generic assembly | Weak modern reference; historical not evaluated | Standard translation, not an ancient general-assembly equivalent |
| FIL | 线 xiàn | Modern fil, flexible strand/thread/wire | Direct | Weak | Shortlist; contexts decide line/wire meanings |
| FIL | 丝 sī | Fine filament / silk | Partial | Weak | Shortlist; material narrower than generic fil |
| FIL | 缕 lǚ | Thread / filament | Direct | Weak | Reserve; noun versus classifier must be separated |
| FIL | 肥 féi | Sound control: fat | Mismatch | Partial onset only | Reject; /f/ match does not yield THREAD |
| FIL | 细丝 xì sī | Compositional semantic control | Partial | Weak | Fine-strand subtype, not every figurative fil |
| MARCHAND | 商 shāng | Modern merchant noun | Direct | Partial, local/noninitial only | Shortlist; /ʃɑ̃/ ≠ /ʂɑŋ/, /maʁ/ unmatched; exclude dynasty/quotient/discussion |
| MARCHAND | 贩 fàn | Reselling trader | Partial | Weak | Shortlist; narrower small/peddling trader sense |
| MARCHAND | 贸 mào | mercatus commerce / trade | Partial | Not evaluated (Latin) | Historical-stage shortlist; activity, not a merchant person |
| MARCHAND | 马 mǎ | Sound control: animal | Mismatch | Partial onset only | Reject; horse-riding merchant is a scene, not etymology |
| MARCHAND | 商人 shāngrén | Modern semantic control | Direct | Weak | Standard lexical equivalent, not automatically Featured |
| MONTRER | 示 shì | Modern SHOW / MAKE KNOWN | Direct | Weak | Shortlist; no import of ritual graph history or qí reading |
| MONTRER | 显 xiǎn | Make/show visible | Partial | Weak | Shortlist; being apparent and showing something are not identical syntax |
| MONTRER | 指 zhǐ | POINT OUT | Partial | Weak | Reserve; showing need not involve a pointing gesture |
| MONTRER | 梦 mèng | Sound control: dream | Mismatch | Partial onset/nasal impression | Reject; seeing in a dream is not making visible |
| MONTRER | 展示 zhǎnshì | Modern semantic control | Direct in display sense | Weak | Does not cover every montrer construction |

Mainland-first: the concrete [汉典](https://zdic.net/) **basic/detailed** entries for 会、聚、合、困、线、丝、缕、肥、商、贩、贸、马、示、显、指、梦 were inspected. Links and claim scopes are in each record. These are mediated online entries, **not** verified printed 《现代汉语词典》《汉语大词典》《汉语大字典》 editions/pages. The syndicated 國語辭典 column is excluded. Compound controls 修道院、细丝、商人、展示 are explicitly editorial standard/compositional controls; a dedicated printed-entry verification is Pending. No historical date, dialect or reconstruction is inferred from a modern spelling/reading.

## Freeze / stop decision

- Featured: **Pending for all four**, independently of Reviewed entry and Published availability.
- Reader display: Standard + Featured Pending + two meaningful stage comparisons. Other candidates and controls remain in Research/Evidence.
- Old observations are preserved verbatim in `source.raw_note` and the reversible `legacy_migration` archive. The misleading combined semantic chains are replaced with explicit branching boundaries.
- CONVENT: reject “custom from monastery” as historical derivation; retain shared convenire family.
- FIL: retain fil → filière; reject fille as a thread-family continuation.
- MARCHAND: downgrade walking/travelling association to Research/cognitive scene; no marcher ancestry.
- MONTRER: retain deeper Latin family relation, reject direct montrer → monitor derivation.
- No further candidate search merely to create Featured mappings. Large print dictionaries, earliest Chinese attestations and historical IPA remain Pending.

Consonant features were recorded only after semantic retrieval: modern k/h generation interest in CONVENT, f control for FIL, m/f interest for MARCHAND, m control for MONTRER. The existing group inventory is not redefined. A group feature on a modern form does not score an untranscribed Latin stage. This bounded sample does **not** establish improved precision or recall; four deliberately selected sound controls are semantic mismatches, not an estimated false-positive rate for an entire language. No group-derived candidate earns Featured status.

## Implementation contract

Use existing `historical_stages + mappings` model: 13 stages, 5 stage-mapping records, 20 candidate/control records. `data/review/final-lexical-freeze.v1.json` is the fixed evidence/editorial contract, independent of schema validity. Builds cannot regenerate it. Negative tests cover false family ancestry, reconstructed → attested, Pending → direct, forced Featured, sound-control promotion and unauthorized cross-language origin. The validator is a scoped editorial safeguard, not a general-purpose etymological truth engine.

No Batch 2 work. The remaining five entries (FORM / SIGN / PRESS / ABOVE / LIGHT) stay unchanged.