Repository path: wiki/findings/results/RS-20260814g-supplied-sound.md · rendered 2026-09-09

RS-20260814g — Burton writes English sound-patterning where the Arabic has none, at four loci of eight; Lane, in an equally archaic English, at none of six

SUPERSEDED IN PART, 2026-08-15, by the confirmatory run this page's limit 1 asked for: RS-20260815-supplied-sound-confirm. On sixteen loci fixed before either comparator was opened, Burton's plain-domain rate is 1 of 8, not 4 of 8, and the single positive cell is direct speech — the stratum the pre-run critic named as the confound. The Burton half of the title is not withdrawn but does not replicate at magnitude and must not be cited without that page beside it. The Lane half replicates exactly (0 of 6 here, 0 of 8 there). §6 of this page — that Lane substitutes an English device where the Arabic rhymes — comes back stronger: 3 of 4 on a domain purified of everything that is not sound.

ARM-alf-layla step 2, study limb, on E-20260814g (frozen b61fa07, amended on eleven pre-run critic findings including three BLOCKING, before any cell was dispatched). Translation limb: span B of «ألف ليلة وليلة», 582 Arabic words, log D15–D28, frozen at 4d4e6d0 in a separate, earlier commit.

The wire between the limbs: translating span B produced the two saj' loci and the two cognate accusatives that the study limb uses as its patterned domain, and the translator's log fixed them before either published hand was opened.

1. The one-sentence answer

At eight loci where the Arabic copy-text repeats no sound, three blind seats found conspicuous English sound-patterning in Burton at 4 of 8 — and in Lane at 0 of 6 and in the lead at 0 of 8 — while at eight loci where the Arabic does repeat a sound all three hands pattern more (Burton 5 of 8, Lane 2 of 7, the lead 3 of 8). Supplying the sound the source has not got is one hand's habit, not a fact about translating from Arabic.

2. What was measured, and what it is not

Sixteen loci: eight PATTERNED — four saj' chains from span A's preamble (frozen 2c5b612) and four from span B (frozen 4d4e6d0: two saj', two cognate accusatives) — and eight PLAIN, span B prose segments of 9–24 words that survive an affirmative audit for saj', cognate accusative and near-synonym doublet. Each hand's English at each locus was cut from a stored file by anchor, not transcribed; analyse.py rebuilds all 51 passages byte-identically and refuses to report anything if the rebuild differs.

Each passage went to P1, P2 and P3 as its own request, blind to the Arabic, to the hand, to the domain and to every other passage, answering one closed question: does this English use conspicuous sound-patterning that recurs, such that a reader would notice it as a deliberate effect? Majority of three.

This is not the same question as RS-20260814's, and §6 is about what happens when the two disagree. That census asked whether the Arabic rhyme reached English. This one asks whether the English has patterning of any kind, from any cause.

3. The census

144 bodies, 0 not stop, 2 re-dispatched, 0 invalid. Seat unanimity 41 of 48 = 0.854 against a registered floor of 0.60. $0.420799 of a declared $0.77.

Wilson 95% intervals on the PLAIN rates: Lane [0.00, 0.39], Burton [0.22, 0.78], lead [0.00, 0.32] (analysis (a)). No significance test is reported and none was registered. The lead's column is italic throughout and is not evidence: contamination: high, declared on the artifact before any measurement.

4. The instrument held, and the control that mattered most is the one that was added

All three controls came back unanimous on the expected code:

ARCH exists because the pre-run critic said the instrument had never been tested against old-fashioned English that only sounds ornate — which is the register both published hands write in, and the failure mode that would have inflated every figure on this page. It was added, dispatched blind among the census, and it held. Without it the whole census would have to be reported as uninterpretable.

5. What the four Burton PLAIN cells actually are

The seats named their patterns, and the naming is what makes the number readable:

PL5 is the doubtful one and it is named as doubtful here rather than quietly dropped. The repetition the seats heard is do … do — English archaic do-support plus thou, which is grammar rather than a chosen sound. If it is discounted, Burton's PLAIN rate falls to 3 of 8 in analysis (a) and 1 of 6 in the governing analysis (c). That discount is post-hoc and is not adopted; it is stated so a reader can apply it. What it does not touch is §1's contrast: Lane writes the same archaic English, with thou, ye, doth and conjure thee, throughout all six of his PLAIN passages, and not one of them drew a single Y from any of the three seats.

6. The finding nobody registered: how the two instruments disagree about Lane

RS-20260814 reported Lane 0 of 7 — he carries the Arabic rhyme nowhere. Four of those seven loci are re-used here as PT1–PT4, and on those four this census codes Lane Y once. It is one cell, and it is worth the paragraph because of what the seats said about it, not because of its size.

• PT4, the four-member epithet chain, is the overlapping cell that flips. RS-20260814 coded Lane lost: "all-knowing, as well as all-wise, and almighty, and all-bountiful" does not rhyme. All three seats here coded it Y, and all three named the same thing — "repeated all- prefix creates jingling alliteration". Lane's other three overlapping cells are N N N, unanimously.
• PT5, the jinni's epithets, is Lane's second Y and is not in the earlier census — it is a span B locus. Lane: "of gigantic stature, broad-fronted and bulky, bearing on his head a chest." Y Y Y, "b-alliteration in broad-fronted, bulky, bearing".

So Lane is not deaf to the sound; at two loci he substitutes a device. Where the Arabic rhymes, at those two, he reaches for anaphora and alliteration, which English has in quantity, and not for end-rhyme, which in English prose reads as doggerel. Burton reaches for end-rhyme, and pays for it — his PT6 is "the sea that raves, dashing and clashing with waves", which all three seats heard as rhyme.

This refines RS-20260814 §5 rather than contradicting it. That page's generalisation was that the structural half of saj' crosses into English and the phonetic half does not. The blind census says that English sound-patterning is present at 10 of the 23 locatable PATTERNED cells across the three hands — and that in Lane, where it is present at all, it is never the source's device. This is two cells in one hand and is offered as a reading of them, not as a rate. RS-20260814's sentence "at 5 of 7 loci in the better hand it is simply lost" stands as written — the rhyme is lost — and should from now on be read beside this one. No published figure is false and none is withdrawn.

7. The strata the critic required

The critic's finding 5 was that PATTERNED is confounded with register: the PT cells are liturgical and epithetic, the PL cells narrative and dialogic, and a hand that ornaments elevated passages rather than rhymed ones gives the same numbers. That confound is not resolved by this run and the page does not claim it is. What the run does establish is the between-hand contrast at the plain loci, where register is held constant by construction: the same eight narrative segments, three hands, and only one of them patterns.

8. The comparisons as they were stated

None of these is a registered prediction. The pre-run critic's BLOCKING 1 was that they were written after the English they concern had been read, and it was accepted in full: they are pre-specified descriptions, and a comparison coming out as stated is not evidence that it was foreseen.

Two registered checks fired before a single body was dispatched and neither was quietly repaired.

9. Limits

1. The loci and the comparisons were fixed after the comparators' span B had been read — the critic's BLOCKING 1. The compensating controls were real (the patterned/plain classification is in a commit that predates the reading; the plain loci are drawn by a mechanical rule a script reproduces) but they do not make this confirmatory. ARM-alf-layla step 3 carries the confirmatory version, with its loci frozen before either hand is opened.
2. Lane is absent at three of the sixteen loci — PT8, PL5, PL6, all three around the maiden's demand and the brothers' hesitation. That is his expurgation, it is the same behaviour RS-20260814 §6b found at span A's obscenity, and it means his n is 7 and 6 against Burton's 8 and 8.
3. Burton's PLAIN passages run 1.59× his PATTERNED ones because he interpolates; analyses (b) and (c) exist for that and (c) governs.
4. n = 8 per hand per domain. The intervals in §3 are wide and are printed for that reason.
5. PATTERNED is confounded with register (§7).
6. PT2 and PT3 are adjacent cola of one stretch of span A moralising and may be one rhetorical unit split across two cells.
7. The controls are lead-authored. Third-party controls are owed if this instrument is used again.
8. The jury is not calibrated (config/models.md, Tier D not passed). Everything here is provisional and internal-judgment-only. No anchor is cited and none is claimed.
9. Nothing here licenses a claim about good translation. senses: [style-correspondence]. The census says what three hands did to a sound, not what a translator should do.

10. What it is for

The project has four figures on perceived-source-carriage that sit on a seam: readers shown two translations cannot separate this one carries the source's form from this one reads like a translation. This result reaches the same seam from the other side and puts a number on one half of it. A reader who judged Burton "closer to the Arabic" because his English is full of rhyme and alliteration would be reading a signal that is present at half the loci where the Arabic supplies no such signal at all, and absent from a hand that, at those same loci, is doing nothing unusual. Ornament in the translation is not evidence of ornament in the source — and this is the first time this project has measured that on a real published pair, in a language whose formal signature is exactly the thing at issue.