Repository path: wiki/findings/results/RS-20260814-saj-carriage.md · rendered 2026-09-09

RS-20260814 — Arabic rhymed prose reaches Lane's English at 0 of 7 and Burton's at 2, and what both hands keep is the shape, not the sound

ARM-alf-layla step 1, on E-20260814-saj-carriage (design frozen b3b38a8). Translation limb: T-alf-layla-R05-v1, frozen at 2c5b612 in a separate, earlier commit, with its eight saj' loci already tabulated — neither comparator had been read at that point. That ordering is the design.

1. The one-sentence answer

At the eight rhymed-prose loci of the frame tale's opening, fixed before either published hand was opened, Lane 1839 carries the rhyme at 0 of the 7 loci his text contains and Burton 1885 carries it at 2 with 1 more partial — and neither hand loses a single colon: what falls out of English is the sound, not the structure.

2. The work, and why it is new here

«ألف ليلة وليلة», the frame tale «حكايات الملك شهريار وأخيه الملك شاه زمان» — the project's nineteenth source language and its first Semitic one, and its fifth long work. Copy-text, recension and the collation of span A against a second Arabic witness: ../../../workshop/ translations/alf-layla/collation.md. Span A is 642 Arabic words, rendered by the lead at $0 under R05 into 1,201 English words.

Saj' — rhymed, cadenced prose — is the formal signature of literary Arabic and the feature its translators most often say is unreachable. This is the first time the project has measured that claim rather than repeated it.

3. The census

Seven of eight loci are locatable in both hands, above the registered power floor of six. Locus 8 is absent in both — for a reason that is itself a finding, §6.

Every verdict exhibits its located English verbatim in coding.py, which asserts that each of the 25 quotations occurs in the stored comparator file and recomputes the tallies; it exits non-zero if any quotation has been mis-transcribed. The lead's column is printed in italics throughout and is not evidence: contamination: high, two named priming events, declared on the artifact before the census was run.

4. What the two carried loci look like

Burton at locus 5, for فيما مضى من قديم الزمان، وسالف العصر والأوان:

in tide of yore and in time long gone before

yore · before. Lane, at the same place, has "that there was, in ancient times, a King" — the doublet reduced to a single member and the rhyme gone with it.

Burton at locus 1, the doxology — and here he does something no other hand in this project's record has done:

PRAISE BE TO ALLAH * THE BENEFICENT KING * THE CREATOR OF THE UNIVERSE * LORD OF THE THREE WORLDS * WHO SET UP THE FIRMAMENT WITHOUT PILLARS IN ITS STEAD * AND WHO STRETCHED OUT THE EARTH EVEN AS A BED * AND GRACE, AND PRAYER-BLESSING BE UPON OUR LORD MOHAMMED * LORD OF APOSTOLIC MEN * AND UPON HIS FAMILY AND COMPANION TRAIN * PRAYER AND BLESSINGS ENDURING AND GRACE WHICH UNTO THE DAY OF DOOM SHALL REMAIN * AMEN! * O THOU OF THE THREE WORLDS SOVEREIGN!

Two rhyme-chains across seven of eleven cola — and an asterisk between every colon. Burton did not only carry the device; he invented an English typographic convention to make it visible, because English prose has no way of showing a reader that a sentence is divided into rhyming units. Lane's version of the same doxology is complete, dignified, and ends its cola on King, universe, pillars, bed, apostles, Family, constant, judgment — eight cola, no two of them rhyming.

5. The finding the failed prediction produced

P3 was that at locus 4 — a four-member epithet chain — at least one hand would drop members rather than rhyme them. It fails, and its failure is the most useful thing in the census.

• Lane: "but God alone is all-knowing, as well as all-wise, and almighty, and all-bountiful" — all four members present, none rhyming.
• Burton: "but Allah is All knowing of His hidden things and All ruling and All honoured and All giving and All gracious and All merciful" — six members, three of them rhyming on -ing.

Neither hand drops a colon anywhere in the census. Both keep the doublets, the quadruplets, the parallel shape, the anaphora — Lane's all-…all-…all- and Burton's All…All…All are both attempts at the chain, using a repeated word where the Arabic uses a repeated sound. The generalisation the census supports is that the structural half of saj' crosses into English intact and the phonetic half does not, and that a translator who wants any of the sound must go looking for an English device that is not the source's device — Burton's -ing participles at locus 4, his monosyllabic end-rhyme at locus 5, his asterisks on the page. This is the same shape as the project's Hungarian result, where a tense-marked attribution slot reached English as a bleached said: the mark is relocated, not lost. Here, at 5 of 7 loci in the better hand, it is simply lost.

6. Three things the census turned up that it was not looking for

(a) The copy-text is shorter than the Arabic of both hands. Locus 8 is absent in Lane and in Burton because both of their texts carry a vizier-and-letter episode — a written invitation, gifts of horses and slaves and stuffs, a three-day guest rite, a journey — that the Hindawi 2022 copy-text compresses into eleven words. Lane abridged and Burton expanded, but neither can explain material that is present in both and absent from the copy-text. This is a recension fact, not a translation fact, it was foreseen in the design's failure criteria, and it means absent verdicts here stay ambiguous. collation.md is amended to record it.

(b) Neither published hand renders the one obscenity, and Burton euphemises it by ornamenting. The Arabic ولم يزالوا في بوس وعناق ونيك ونحو ذلك puts an ordinary vulgar verb third in a list of three neutral verbal nouns, which do not rhyme.

• Lane: "and all of them continued revelling together until the close of the day" — the list collapsed to one word.
• Burton: "they ceased not from kissing and clipping, coupling and carousing till day began to wane" — an alliterating quadruplet where the Arabic has an unpatterned triplet.

So P4 splits: Lane does not omit (he euphemises by compression) and Burton does not render (he euphemises by expansion). The observation worth carrying forward is the second one: at the one locus in span A where the Arabic is plain and coarse, Burton supplies sound-patterning the source does not have — while at five of seven loci where the Arabic is patterned, he supplies none. This is one instance and is unregistered; it is handed to ARM-alf-layla step 2 as a question, not offered as a finding.

(c) An independent check on a textual crux. The lead's log flagged D6: الخرزة التي أعطيتها لك — the bead which I gave you — is illogical, since the speaker is going back to fetch it. Lane renders the illogic straight ("the jewel that I have given thee"); Burton repairs it ("a string of jewels intended as a gift to thee"), which is the solution the lead reached independently. The crux is in the Arabic, not in the copy-text, and two of three hands silently mend it.

7. Extent, and the dependence check

Over the same stretch: lead 1,201 words · Lane 1,522 · Burton 2,810. Burton is 2.3× the lead and 1.8× Lane, part recension (§6a) and part his own expansion.

tools/dependence_check.py on all three pairs — every pair clean: 0 shared 12-grams throughout; longest common runs lead~Lane 9, lead~Burton 8, Lane~Burton 8. P5 holds. The lead's longest run against Lane is "a black slave came to her and embraced her" and against Burton "and out of it came twenty slave girls" — stretches where the Arabic leaves English almost no choice. The two published hands' own longest shared run is the doxology's "the beneficent king the creator of the universe". contamination: high was declared on the artifact on grounds of the work's fame, and the measurement does not support dependence at span A — which is what the standing rule predicts about run length as a proxy and is recorded, not interpreted further.

8. Registered predictions, resolved

Three of seven registered predictions failed or half-failed. They are reported as they stood.

9. Limits

1. The coding is the lead's, single-handed, after having translated the same span. This is the census's largest weakness and it is not remediable inside one session. It is bounded, not cured, by the rule that every verdict exhibits its located English: coding.py makes the coding auditable by a later session or a panel seat without re-locating anything. Every claim on this page carries internal-judgment-only.
2. n = 7. Eight loci in eighty words of preamble. Lane's 0 of 7 is a clean zero on a small sample, not a law; the span-B census extends it.
3. absent is ambiguous between translator's omission and recension, per §6a, and is never counted as lost.
4. "Rhyme in English" is a judgement, made orthographically-plus-phonetically by one reader. The two carried verdicts on Burton would survive any reasonable reading (yore·before; STEAD·BED·MOHAMMED); the partial at locus 4 rests on three -ing participles among six cola and is the verdict a second coder is most likely to move.
5. Payne 1882 is missing and would be the natural third hand — the one that sits between Lane's programme and Burton's. Not freely reachable; searched and recorded in collation.md §5.
6. Nothing here licenses a claim about good translation. The census says what happened to a device, not what should. senses: [style-correspondence]; no anchor is cited and none is claimed.