Repository path: workshop/translations/clos-des-ames/R04-v1/translation.md · rendered 2026-09-09

T-clos-des-ames-R04-v1 — from Paul Arène, «Le Clos des Ames», opening

Regime. R04-lead-close: one drafting pass and one self-revision against the French. No English rendering of this work, or of any work by Paul Arène, was opened before this file and its log were committed — see §Contamination, where the reason is that none exists.

Source. Paul Arène, «Le Clos des Ames», chapters I–II opening, in La Gueuse parfumée: récits provençaux (Charpentier/Fasquelle, 1907 printing; first published 1876). Public domain (Arène d. 1896). Text: Project Gutenberg eBook #60680 (French). The excerpt as given to the translator is source-fr.txt in this directory — 466 whitespace-delimited tokens, running from «Du balcon de sa chambre à coucher…» to «…les bords de la Durance par un beau coup de mistral.», SHA-256 d5780bee6238dee20ae1270897e7b0933da42080bb73731b609ecc3d2d5b9c6f.

Pair. French → English. Structurally near, culturally near-but-regional: the story is set in Arène's invented Provençal town of Canteperdrix and turns on provincial-bourgeois politics of the 1848–51 sort.

Declared purpose. A general reading edition in modern literary English, unmarked for period — the same purpose parameter the study limb that scores it declares.

Contamination: none, on an argument from absence, which is weaker than a measurement and is said so here. No English rendering of any work by Paul Arène exists on two search routes, both re-run for this file on 2026-08-03: Project Gutenberg returns seven Arène volumes, every one in French; the Internet Archive advanced search on creator:("Arène, Paul") returns 16 records, 15 French and 1 German. Both figures reproduce S089's search exactly. There is therefore no published English for the translator to have remembered, and no comparator against which a dependence_check run could be made at all. Two candidate works were gated and discarded before this one was admitted (§Gate history).

And absence of a comparator is not absence of contamination — accepted from the independent pre-run critic of E-20260803-a4-set (finding 9, amendment A9). What the gate can measure is overlap with a published English rendering, and there is none to measure against; it says nothing about whether the lead carries Arène's French from training. This artifact is therefore not described as uncontaminated anywhere: the honest statement is that no comparator exists.

Gate history — two works discarded before this one, on a rule frozen in advance

The standing rule of CLAUDE.md (note (bcd)) is that contamination is measured on the candidate material before a locus is chosen and before anything is translated, as a selection gate. The discard rule, frozen before any measurement: longest common run ≥ 12 tokens, or ≥ 1 shared 12-gram, and the work is discarded as a limb. It fired twice.

Both gate translations were frozen before their measurements ran and both are kept: workshop/experiments/E-20260803-a4-set/materials/ (fresh-gate-en.txt holds the Sand gate span; the Daudet gate span and its result are in the same directory's history and in gate-result.json). Raw figures: gate-result.json, gate2-result.json, gate-meta.json, gate2-meta.json. The rule was not revisited after it fired, which is the move this project's verification discipline exists to stop.

The translation

From the balcony of his bedroom M. Sube could see the whole of his close: first the vineyard, very old and badly kept, but which produced such good wine; then the cistern with its fountain, a strip of meadow, a square of kitchen garden, and right at the bottom, closing off both the property and the slope, a well-fed field of sainfoin where the first suns of May brought thousands of violet flowers into bloom every morning. I was forgetting: all round the close, sixteen pillars of rough sandstone which, having once carried vine arbours, must at one time have formed an agreeable cloister of greenery, and now carried nothing but ivy in place of vine-stocks, with clusters of small black berries in place of muscat grapes.

Never did schoolboy, in his dreams of playing truant, dream of a close brighter, more cheerful, more magnificently overgrown, or more deliciously untilled than M. Sube's old close. It was called the Close of the Souls. But that name, whose enigmatic look is about to produce in you, meeting it for the first time, some vague impression of superstitious terror and mystery — that name, the Close of the Souls, appeared to us at Canteperdrix as a joyful thing, green and in flower. We said Close of the Souls without knowing why, the word's original value, its power to signify, having long since worn away; and far from keeping any funereal aftertaste, those four syllables called up in our heads nothing but memories of stolen grapes, of pears eaten in the tree, of walls climbed and ditches jumped, and of subtle escapes through a gap in the hedge in cherry time.

M. Sube, thanks to his close, was — and this is not saying little — the happiest man in Canteperdrix, where there are so many happy people. The most fearful, too! But in our provincial towns, is not a little mild cowardice the obligatory seasoning of every bourgeois felicity?

That good French bourgeoisie, which one day made '89 and, a little, '93 as well, has gone on trembling ever since. Now M. Sube, a bourgeois and the son of a bourgeois, a practising Catholic, a friend of order whatever came, and respectful towards established power whatever it might be, but at bottom devoted to the elder branch for reasons he never quite explained to himself — M. Sube had trembled from birth, naturally, like a Lombardy poplar. And in the evening at the club, when all the other shivering poplars, all the alarmed men of Canteperdrix, were stirring in a group around him — to hear the whisperings and the confidences, Lyon in flames, Marseille running with blood, the terrible news poured into the ear with that harsh relish the fearful take in exasperating their own terror as soon as there are enough of them; to hear that confused noise of voices, so like the noise of leaves — one would positively have said the banks of the Durance under a fine gust of mistral.

Translator's log

Written at translation time and frozen before the study limb that scores this text was designed (charter §3, A4; continue-prompt.md §5.4). Decisions and alternatives, not quality claims. Any evaluative sentence carries internal-judgment-only, which the front matter asserts for the page.

1. clos → "close". The word has to survive into the proper name, «le clos des Ames», so a phrase ("walled garden", "enclosed ground", "the enclosure") was out: the title would have gone to four or five words. English close in the sense of an enclosed piece of ground is available, slightly regional-archaic in feel, and it lets the Close of the Souls stand as a name. What it costs: an English reader may first hear the adjective, and the sentence has to do the disambiguating work.
2. ces trois syllabes → "those four syllables", and this is a deliberate alteration of a stated numeral. Arène counts the syllables of clos des Ames, which is three in French. The Close of the Souls is four. Keeping "three" would make a false statement about the words the English reader is holding; keeping "four" makes a statement the French does not contain. I took the second. The third option — dropping the count ("those syllables", "those few syllables") — loses the joke's precision and was refused. Recorded as an unlicensed alteration under accuracy, knowingly made.
3. la branche aînée → "the elder branch", copy-opaque. The legitimist Bourbon line. A gloss ("the legitimist Bourbons", "the senior Bourbon line") would explain a joke that works on M. Sube precisely because he cannot explain it himself — the sentence says he was devoted to it for reasons he never quite explained. Opacity here is the content. Cost: a reader with no French nineteenth-century politics gets an unresolved phrase.
4. Canteperdrix, Durance, mistral, sainfoin retained. Three are proper nouns; sainfoin is in English dictionaries and is the plant's English name, so it is not a foreignism at all — the only one of the four where the retention costs nothing.
5. un peuplier d'Italie → "a Lombardy poplar". The English name of Populus nigra var. italica. Denotationally exact and a substitution of the culture-bound label: the French names Italy, the English names Lombardy. The tree's whole function here is that it shakes, and both names carry that equally.
6. 89 / 93 → "'89" / "'93". The apostrophe is an addition; without it the English reads as two bare numbers. The alternative — "1789" and "1793" — makes explicit what the French leaves as the shorthand of a class that does not need to be told.
7. ce qui n'est pas peu dire → "and this is not saying little". Drafted as "and this is saying something", revised back on the second pass: the litotes is the figure, and the positive version flattens it. This is one of the two places where the revision pass changed the register rather than the wording.
8. douce couardise → "mild cowardice", drafted "gentle cowardice". Douce here is the temperature of the cowardice, not its kindness.
9. cette âpre volupté → "that harsh relish", drafted "that harsh pleasure". Volupté is stronger than pleasure and English has no one word for it at this register; relish keeps the appetite and loses the sensuality. A named loss.
10. s'agitaient en groupe → "were stirring in a group", drafted "were milling". The poplar figure runs through the whole sentence and stirring is what leaves do; milling is what people do and would have killed the figure two clauses before its punchline.
11. un beau coup de mistral → "a fine gust of mistral", drafted "a good gust". Beau is the Provençal admiring usage; fine carries the admiration, good reads as measurement.
12. Sentence shape kept long. The last paragraph is one French sentence of ~120 words held together by two d'entendre infinitives and released by quelqu'un eût dit. I kept it as one English sentence with em-dashes standing in for the suspension, rather than splitting at the semicolon, because the suspension is the joke: the reader must hold the noise of the frightened bourgeois until it turns into the noise of leaves. Splitting it was the first thing I tried and it delivers the punchline early.
13. What I know I did not solve. Le clos des Ames is a name whose original meaning has worn away, and the paragraph's whole subject is that wearing-away. In French Ames still says souls while the name says nothing; in English the Close of the Souls is, if anything, more legible than the French, so the reader is told the name has lost its meaning while being handed a name that has not lost it. I have no solution and did not invent one.

Evaluation — first blind panel scores (RS-20260803-a4-set, S094)

provisional and internal-judgment-only: Tier D is NOT PASSED (RS-20260802-tierD-verdict), so these scores carry no evidential weight and may not support a framework recommendation. Three non-Anthropic jurors (P1, P2, P5), blind, authorship-stripped, rating this text alone against its source on a 1–7 scale, twice each; the figure is the mean over three jurors and two passes. Scored span: the whole filed span.

Instrument floor for comparison: 0.233 scale points (same juror, same text, independent repeat). Two of the three jurors sit at the top of the scale on ~72% of all cells, so these figures are compressed upward and the compression is unequal across jurors.

From this page's frozen translator's log alone, an independent seat that had seen neither the source nor the translation predicted the lowest-scoring sense would be accuracy; the sense that actually scored lowest was accuracy. The full test, and the benchmark that deflates it, are in RS-20260803-a4-set §5.

This text was also the base of the run's null control. A ten-edit meaning-preserving micro-paraphrase of it (CLO-P, edit list frozen in workshop/experiments/E-20260803-a4-set/materials/edits.json) was scored as a separate blind item and came out 0.139 scale points away, pooled — below the instrument's own retest noise of 0.233. That is the resolution floor the project did not have before. This translation was made in knowledge of the design that scored it, and is outside every primary aggregate.

Re-judged — cross-session stability check (RS-20260907-panel-judging-2, S253)

provisional and internal-judgment-only: Tier D is EXHAUSTED, permanently (config/models.md, S250) — these scores carry no evidential weight. Re-judged 159 sessions after the first scoring above, on the byte-identical text, same three non-Anthropic jurors, blind, order-swapped, twice each.

Cross-session shift from the S094 scores above: mean absolute 0.127 scale points overall (this item's own deltas are in RS-20260907-panel-judging-2 §2), against a within-session retest floor of 0.222 — the panel reads this text the same way five weeks and 159 sessions later. One juror (P2) was flagged degenerate this run (near-ceiling on 90% of its cells, SD > 0.30 on only 2 of 6 senses) — see RS-20260907-panel-judging-2 §3 for what that does and does not affect.