Repository path: workshop/regimes/R40-matched-shape-cjk.md · rendered 2026-09-09

R40 — the matched shape, forced, from a source that matches by measure instead of by pattern

v0.1, pilot. R40 is R39 moved to Classical Chinese. Two things are carried across verbatim and one thing is new.

Carried verbatim from R39, and this is the point of the regime: the table of English resources that count as a match (§2 below) is R39's table, word for word, unchanged. A verdict written here and a verdict written in T-kalila-fanza-R39-v1 are about the same English question and may be counted together. Had the table been re-tuned for a Chinese source, nothing from the two renderings could be pooled, and pooling them is the reason this regime exists.

New: what counts as matched shape in the source. Arabic muwāzana matches derivational patterns; Classical Chinese 對偶 matches syllable-count and word-class and, in the strict case, matches nothing lexically at all. The class has to be redefined on the source side or the census would be meaningless. It is defined in §1, before any English exists.

1. The source-side class, declared before the enumeration

A locus is admitted iff two or more members stand in matched position and are matched in form by all three of:

• (i) equal character count per member — the Chinese syllable is the measure;
• (ii) matched construction, member for member — verb + object against verb + object, modifier + head against modifier + head;
• (iii) matched word-class at each corresponding position.

Semantic parallelism alone does not admit a locus. Two members that mean answering things but are built differently are not this class, exactly as a doublet of synonyms is not muwāzana.

Admitted loci are then sorted into two sub-classes, kept apart in every count:

Why both, and why apart. R39's inventory tightened its Arabic class to exclude matched frames (kalila-fanza/R39-v1/inventory-frozen.md §0), because Arabic marks the class by pattern and seventeen frame-loci would otherwise have swamped it. Applying that tightening to Chinese would delete 排比, which is not a marginal case here but one of the two ways the language does the thing. So both are enumerated, and every count reports them separately, so that a reader who wants the strict class can have it and the regime is not caught choosing its own denominator.

Reduplicatives (洸洸, 昂昂, 巍巍, 赫赫) are not a matched-shape locus on their own. Where two reduplicatives stand in matched position in a couplet, the couplet is the locus and the reduplication is part of what (iii) matches.

2. What counts as matched shape in English — R39 §2, imported verbatim

Not counted as MATCHED: two members that merely mean parallel things; a doublet of synonyms with no shared form; alliteration alone; a rhyme alone.

3. The rule

1. Enumerate the source's matched-shape loci under §1 and freeze the inventory before any English is written, with each locus assigned to SHAPE‑a or SHAPE‑b, and with every locus the rule declines recorded together with the ground on which it declines it.
2. Render the sense. No fact, event, referent or evaluation added; nothing dropped.
3. At every locus in the inventory, write one of exactly two verdicts, at the same sitting as the rendering and before any comparator is opened: - MATCHED — an English figure of the same kind stands at the same span: two or more members, in matched position, matched in form under §2. The English is quoted and the resource (A)–(D) is named. - IMPOSSIBLE — with the reason, and with the plain English actually written at that span quoted beside it.

There is no third verdict. A reduction is an IMPOSSIBLE and is written down as one. 4. Put no matched-shape figure at any place outside the inventory. 5. Record, at every MATCHED locus, the plain wording refused — the unmatched English the hand would otherwise have written at that same span, carrying the same content — written at the same sitting as the device. 6. Where the plain wording refused is identical to the device, or already matched, the locus is recorded MATCHED — free and is excluded in advance from any paired use, because it supplies no contrast. The exclusion is declared here, before the census, so that it cannot be made afterwards by a hand who has seen how the pairs behave.

4. What R40 is for, and what it is not

For: to produce, from a second and unrelated source language, a set of matched-shape loci at which the same four English resources were tried, so that a finding about which resource is perceptible is not a finding about Arabic.

Not for: ornament, or quality. Rule 4 stands. R40 produces a rendering and a table of verdicts, not a judgement that the rendering is good, and the lead never judges his own translation (charter §5).

A known cost, declared in advance, as at R39: forcing a match where English does not want one is bought with something — a longer phrase, a more Latinate word, an inversion. Rule 5 makes the price visible at every site and the rendering's log reports the total.