Repository path: framework/v0.3/entries/HB-rhymed-prose.md · rendered 2026-09-09

Rhymed prose and fixed rhyme: whether to carry the sound, and what each way of carrying it costs

Standing. Written at the project's close (2026-09-09, S257) as a consolidation of the record without the translation limb the procedure's step 6 requires — no fresh passage was translated under this entry, so it stays status: draft and its §5 Application reads "not applied". Evidence: X1a/X1b (eleven published hands, graded mechanically or by blind seats, as each §2 row states) and X2 (fourteen internal experiments on model seats, no human reader); Tier D is NOT PASSED and EXHAUSTED (config/models.md, RS-20260906-tierD-verdict-v3), so nothing here says one rendering is better than another on a jury's word — reading figures are what model seats registered, and craft observations are the lead's own coding, internal-judgment-only. Consolidated from v0.2 §7.24–§7.35, §7.39, §7.44, §7.49, against RS-20260908-chime-slot's corrected number.

1. The problem

Persian and Arabic prose can rhyme — saj': paired cola whose final words chime, at Sa'di's aphorisms and at 137 of al-Ḥarīrī's 139 cola — and their verse binds one rhyme across a poem (the ghazal's qāfiya, the qaṭ'a's run of three or four bayt-ends). English prose does not rhyme, and English end-rhyme sits on a lexical, not a grammatical, ending. A translator decides whether to let the sound go, answer it in prose, lift the passage into verse, or answer its shape another way — and, in verse, whether to hold a run or a monorhyme at all, knowing what the rhyme slot costs the sense. This entry records what eleven published hands did, what forcing each choice on one hand cost, and where the chime is and is not to be found.

2. What published translators do

The regularities:

1. No published hand of either author carries the prose chime. 0 of 128 and 1 of 187 full rhymes, the one at its hand's accident rate — seven hands, two authors, 1767–1899. Two say so in print: Chappelow, "nor shall I imitate the author … our English tongue will not admit of it"; Preston, "rhyming prose is extremely ungraceful in English, and introduces an air of flippancy" (A-hariri-hands §3).
2. What crosses is the doublet. The paired members survive at 0.974 while the sound goes (RS-20260824b); Lane and Burton keep every colon too (RS-20260814, family L).
3. They do not move the figure into verse instead, though a verse line-end chime is registered at 2.5× a prose one (§3 item 3): 4 of 128, all by the hand who prints most of the book as verse and leaves 28 of his 32 saj' loci in prose (RS-20260822c).
4. Where the same books answer, they answer in verse — on Sa'di. Eastwick and Arnold at ~0.9 where they chime at 0.09 and 0.03 in prose. Not general: two of three Ḥarīrī hands print the poems as verse and rhyme 0 of 9 (RS-20260824b).
5. No published hand holds a Persian run at its depth; the tradition's answer is two pairs at the source's distance, or couplets (RS-20260823).
6. A printed refusal is not a description of the page. Chappelow's sits at or below his own chance level, Preston's above it in 3 of 4 cells and above the non-declarer's in both U-COMMA cells; Preston keeps half his positive rule (the fourteen-word line), not the balance (RS-20260826, §7.39).
7. A monorhymer spends the rhyme slot on whatever will rhyme, in both hands: the source's rhyme sense reaches the rhyme at ~0.04; an ordinary sense closes a free line 1 in 4, a rhyming one 1 in 20 (RS-20260902, corrected at RS-20260908). Neither census's hands are independent witnesses (Eastwick~Gladwin DEPENDENT?, A-gulistan-hands §5.2; Preston~Chenery, A-hariri-hands §5).

3. What this project's own practice found

The lead's renderings, labelled subjects never pooled with a hand; contamination as filed:

Beyond the table:

1. The chime is not where a translator is told to look. Not in the ordinary word (RS-20260821c); not among its synonyms — 2 full rhymes in 2,173 cross-member pairs, +0.042 on the STEM class (RS-20260822); not among the words already in the draft — gain 0.000 at 50 seat-loci (RS-20260823c). What produced the 12 R45 chimes was a joint act: a different word and a different member. Declining the available chime (8 refused at the same 29 loci) was as much of the craft as taking it.
2. Relax to a near-chime and you are finding English, not the figure: near-chime reach is ~0.7 at rhymed loci and at the cola Sa'di left plain (0.708 / 0.708; 0.420 / 0.556 on a second span) (RS-20260822, RS-20260823c).
3. What a slant is worth is decided by position (25 constructed loci, 3 seats): full rhyme +0.578 above a matched floor; slant at a verse line-end +0.542, at a prose colon-end +0.222; 35 of 37 slant detections called "half"; one seat in three matched spelling (sword / word: pooled 0.722 vs 0.778, that seat 0.833 vs 0.333) — say registered, not heard (RS-20260822b). A pararhyme spelled unlike (full / coal) registered 0 in 28 in prose; spelled alike (full / gull) +0.286 reading, +0.071 listening (RS-20260824, NEEDS-REDESIGN).
4. The price the tradition named is measurable as a registration, not a verdict. R48 against its de-chimed twin, 15 segments: 3 blind seats registered the rhymed text as lighter in manner (levity +1.47 of 10, 12/13 segments) and more manner-forward (+1.29, 13/15); their GRACE score (−0.60) is a quality rating, X3, not carried. The estimand is a rhyme-forward policy, not the sound alone. The largest registration sat on a chime R43 found by accident (dine / opine): the price is charged on any chime that lands (RS-20260825b; reception untested).
5. Refusing the chime is a different act from not reaching for it. R43 fell into 29 near-echoes and 5 identical colon-ends unasked; R50, which checked every pair, has 5 and 0, and 0·0·0 in 139 pairs on a fresh maqāma — while Preston, who printed the refusal, has 6 and 7 (§7.34, RS-20260826).
6. Run depth: four is reachable and five ran out (R44, one hand, one rhyme family); the price is repetition; the English ‑ed preterite is a grammatical rhyme family and held a four-deep run clean when the sense wanted past verbs (R47). A chime at the Persian's distance is registered at +0.214 against +0.321 adjacent — but the gap lives in the control cells and two seats of three supply it: take registered at both placements (RS-20260824c).
7. Monorhyme: availability is close to constant (one rhyme sense in five gathers into one English rhyme, median ~0.22; Leaf did not choose his poems for it, P = 0.855); what varies is whether the hand goes looking (RS-20260830b). The R57 and R59 limbs are description only.

4. The options

• Drop the sound, keep the doublet. The tradition's practice on both authors; costs the figure, keeps the members (0.974), pays nothing in sense. Not a licence: the sound reaches English at zero.
• Chime in prose. Reachable under instruction at 0.625 (Sa'di) and 0.477 (al-Ḥarīrī, 3 full rhymes in 65); costs fidelity in 41 of 65 substitutions; seats registered lighter, more manner-forward English at every landed chime; the search is joint reselection; refusals are part of the craft.
• Half-chime. A third of a rhyme at a prose colon-end, most of one at a verse line-end; spelled unlike, nothing in prose (§3 item 3).
• Lift into verse. Registered at 2.5× the prose value, taken at 4 of 128 saj' loci; on Sa'di the same hands rhyme their verse at ~0.9, on al-Ḥarīrī two of three do not.
• Preston's lineated, balanced, unrhymed cola. The record's one positive prescription; the length half (≤ 14 words) is on his page, the balance half is not but is reachable by rule (R50); pushes diction toward commoner words; unused by any hand this project has read since 1850.
• A verse run: hold it (reachable to depth four; price repetition; registered at ~⅔ of the adjacent placement, §3 item 6; the couplet form was not measured), or the tradition's two pairs / couplets.
• A monorhyme: the rhyme slot is spent (1 in 20 against 1 in 4 free; coverage near 0.04); load the free part of the line (§5 item 10). Whether the price is worth paying turned on telling the reader what the source does (HB-telling-the-reader).

5. Guidance

For a translator

1. Enumerate the rhymed loci from the source and freeze them before opening any English; grade any chime you claim bearer-to-bearer, never clause-end to clause-end (men / women at Sa'di's S11 is not carriage; note (bra)). — evidenced (FA→EN, AR→EN), method, internal-judgment-only.
2. Letting a saj' go is the tradition's practice, not a lapse: 0 of 128 and 1 of 187 in seven hands. Decide it as a policy, not by default. — evidenced (FA→EN, AR→EN).
3. Keep the doublet whatever you do with the sound; do not simplify the paired members (0.974 kept by hands who dropped every chime). — evidenced (AR→EN); FA→EN untested.
4. Do not look for the chime in the ordinary word, its synonyms, or the words already in your draft — all three measured at or near zero. If you chime, expect to reselect word and member together, and keep an AVAILABLE-REFUSED column: the refusals are half the craft (R45: 12 taken, 8 refused, 29 STEM loci). — evidenced (FA→EN, AR→EN).
5. Price a prose chime before taking it: 33 chimes against 2 cost fidelity at 41 of 65 substitutions — evidenced (AR→EN, one hand); seats registered the text as lighter and more manner-forward (+1.47 / +1.29 of 10, one maqāma), charged on any chime that lands — description; whether a reader minds is reception, untested.
6. If you write a half-chime, price it by position (§4), prefer the spelled-alike one, and call it seen, not heard. A near-chime never tells you the author's figure is there: it is there at the plain cola too. — evidenced (FA→EN, constructed loci).
7. If you refuse the chime, check your page: an ordinary careful rendering fell into 29 near-echoes and 5 identical colon-ends; a printed refusal did not stop Preston's page chiming above its own chance. — evidenced (AR→EN).
8. Where the rhyme was, a length discipline is the recorded remedy: one colon to a line, none over about fourteen words; balance may be added by rule (R50); no page shows it noticed. — evidenced (AR→EN) as description; reception untested.
9. For a run, hold it to depth four and use a grammatical family (‑ed) where the sense allows; expect five to force repetition; held at the Persian's distance, a chime registers at ~⅔ of its adjacent-line-end value; the couplet form itself was not measured. — evidenced (FA→EN, one hand, seven loci).
10. For a monorhyme, count the chime slot as spent: an ordinary sense reaches it about 1 in 20 against 1 in 4 at a free line-end (factor ~5.5, two hands, 182 located positions of 964); the displaced sense goes nowhere in particular; put the load where the line is still free, and do not print a self-report that your rhyme family holds the senses. — evidenced (FA→EN, Leaf 1898, Payne 1901).

For a pipeline

1. Enumerate and freeze from the source: saj' loci classed STEM / AFFIX / CONTROL, verse runs with depth, qāfiya and radif positions. — untested as a pipeline step (done by hand).
2. Set the policy parameter per passage: drop+doublet | prose-chime (R43/R45) | rhyme-first (R48) | balanced-period (R50) | run-held (R44) | monorhyme (R57/R59). Default: drop+doublet with R50's length cap. — untested.
3. Render, logging at every locus TAKEN / AVAILABLE-REFUSED / UNREACHABLE with the word bought, before any check exists. — untested.
4. Check: grade bearer-to-bearer with tools/rhyme_pairs.py (CMUdict General American; declare heteronyms (brk); limits in §6); clause-end chime ratio against the text's own bearer-permutation null (a refusing text ≈ 1.0 or below); for R50, no colon over 14 words; for a monorhyme, count senses displaced from the chime slot. — untested.
5. Addition check: a source-reading seat with planted three-word additions, to catch content bought for a rhyme (8/8 and 7/8 caught by hand). — untested.

Human entry points. Step 2 is a declared-purpose decision (is the passage's sound what the reader is owed) and routes to a person; without one the default is the tradition's. Step 3's refusals are craft judgements a person may take over; unattended, the run takes every chime the grader passes and the refusal count is zero. Whether to tell the reader the source rhymes is family K's entry point.

Application. Not applied: written at close-out without a translation limb.

6. Not evidenced, and open

• Two pairs, four authors, eleven hands. FA→EN (Sa'di, Hafez) and AR→EN (the Nights, al-Ḥarīrī) only. Sa'di's saj' loci are aphorisms and his controls narration; al-Ḥarīrī's unrhymed prose is six words long — no causal claim about verse-setting is possible (note (bqr)).
• No human reader anywhere. Every reading figure is model seats; the flippancy result is a reception claim on a non-reception instrument; human readers stay named-not-built.
• The grader is a dictionary: CMUdict's General American, UNKNOWN counted as unanswered (every hand rate is a floor); rime riche excluded; been / demesne not rhymes though the tradition's are. The detector on Leaf and Payne misses about a quarter of rhyming lines.
• One hand under each regime, contaminated high on every Sa'di span but R47's (suspected); the four/five boundary is one hand's reach in one rhyme family; the depth-2 runs were never coded.
• RS-20260824 stands under NEEDS-REDESIGN: eye vs ear is an association, not a decomposition; nothing was measured at a verse line-end.
• Open: (i) does the rhyme-slot cost differ with and without an English radif — unresolved at RS-20260908 (+0.157 vs +0.241, interval crossing zero, 8 renderings); (ii) what "moved into verse" means for a hand who lineates without rhyming; any such rate is unsafe until settled (RS-20260824b §6); (iii) whether a reader hears run placement or depth — what was measured is preference under side-by-side comparison (RS-20260824c); (iv) Preston's light-matter exemption — untestable on his author (his most comic maqāma read light 2 of 7); (v) whether bearer and position measures should differ for verse end-rhyme (Eastwick's V09: 1 vs 2; RS-20260823 §8); (vi) whether a word deep in a clause can be moved to the rhyming position (the realisability check was refused for cost, RS-20260823c §8); (vii) whether Preston's lineation is his or a compositor's (page images not consulted).

7. Sources consumed

• v0.2 §7.24 (RS-20260821c), §7.25 (RS-20260822), §7.26 (RS-20260822b), §7.27 (RS-20260822c), §7.28 (RS-20260823), §7.29 (RS-20260823c), §7.30 (RS-20260824), §7.31 (RS-20260824b), §7.32 (RS-20260824c), §7.33 (RS-20260825b), §7.35 (RS-20260826): §2–§3 above.
• v0.2 §7.34 (RS-20260825c) — family K's; only its craft finding (accidental echoes, R43 vs R50) is taken here. §7.39 (RS-20260827c) — shared with K; its page-side chime ratios and Chappelow's expansion are taken here.
• v0.2 §7.44 (RS-20260830b), §7.49 (RS-20260902), corrected by RS-20260908-chime-slot (no v0.2 section: v0.2 is frozen). Hidden by the titles: RS-20260814-saj-carriage (Lane and Burton keep every colon), consumed by family L's craft report, cited here by id only; §7.32's radif sentence belongs to family C (§7.36).
• A-gulistan-hands, A-hariri-hands; regimes R41–R50, R57, R59.
• Withdrawn along the way, one line each: §7.24 item 2 (the rhyme lives among the synonyms) — withdrawn by §7.25 / RS-20260822. §7.24 item 1's rationale and §7.25 item 3 ("not lexical, it is structural"; the rhyme is in your draft) — withdrawn by §7.29 / RS-20260823c. §7.25's "nobody hears these as echoes" as a general claim — withdrawn by §7.26 / RS-20260822b, kept for its three examples. §7.26 item 1's generality for the pararhyme in prose — narrowed by §7.30 / RS-20260824. §7.27 item 3 (hands that print verse rhyme it at ~0.9) — scoped to Sa'di by §7.31 / RS-20260824b; §7.27's GENRE-CHANGED rate — unsafe by the same page. §7.28's reading prediction — withheld (saturated, note (brb)), replaced by §7.32. §7.32's "English has no slot before a repeated word to rhyme in" — corrected by §7.36 / RS-20260826b-radif (family C). §7.33's Preston exemption — withheld, untestable on his author. §7.34's both-orders preference statistic — post-hoc, not replicated at RS-20260827b (HB-telling-the-reader); direction stands. §7.39's registered P3 (both declarers below the non-declarer) and P2 (equivalence) — failed on their own run. §7.44.1's "the sense sits one word to the left of the rhyme" and §7.44.4's second sentence — withdrawn by §7.49.2 / RS-20260902. The T-hafez-bekonad-R57-v1 log's self-report on its rhyme family — corrected by §7.44.2. §7.49.1's "factor of eleven" — corrected to about 5.5 by RS-20260908-chime-slot; §7.49's by-stratum robustness claim — weakened to unresolved by the same page. RS-20260820-rhyme-bearer's closing observation (the crossing rhyme's plain words chime) — refuted by §7.24.