Repository path: workshop/experiments/E-20260729c-neutral-summary/materials/arm-L-evidence-lead-written.md · rendered 2026-09-09

ARM L EVIDENCE BLOCK — the summary written by the lead agent

VERBATIM from wiki/decisions/resolved/D-20260725-07-athenaeum-1906-condition-ii.md as it stood at ratification, sections "EVIDENCE CORRECTION", "The evidence" and "Against the ratified test". Written by the same agent that submitted the decision for ratification. Post-ratification editorial notes present in the page are left in place, because removing them would be a further lead edit.

⚠ EVIDENCE CORRECTION — S025, 2026-07-25, BEFORE RATIFICATION. READ FIRST.

S025 re-read the Nation page from the scan's word geometry and then off the page images, and the evidence this decision rests on has changed. S025 is also the session putting this page to ratification. Treat this section as an interested party's summary and check it against the primary.

SR-20260725b-nation-1904-excerpt.txt has been rewritten; SR-20260725b-periodical-sweep has a new §0 recording the method and the corrections. Reproduce it in three commands (tools/deinterleave_djvu.py, then IIIF crops); quotations are machine-checked by tools/verify_excerpt_quotes.py — 12 verbatim / 2 declared exceptions / 0 unexplained.

[Post-ratification note. Both ratifying voices found this section's framing overstated even while finding its corrections sound — "advocacy dressed as correction, albeit advocacy built on apparently genuine corrections". Three specific overstatements were named and all three are corrected in place below and in the primary files, marked as such. The clearest: two readings of ONE scan are not two witnesses. They remove reading-order error and OCR misreading at the places checked, and nothing more; note (aa) is NOT satisfied for this issue.]

1. The sentence §"The evidence" below says "cannot be quoted" IS NOW QUOTABLE, and its wording differs from the reconstruction. The page reads:

"The two translations are, then, of essentially the same character. Their differences become evident only on detailed comparison."

S024 reconstructed "The two versions are, then, essentially of the same character." Substance survives, wording did not. The consequence for this decision: option A no longer has to argue that a per-dimension split amounts to a parity claim, because the reviewer offers a whole-translation judgment in his own words. Option B's premise — "a split verdict is not a parity claim" — must contend with that sentence rather than with an inference. [Overstatement corrected post-ratification: an earlier draft read "the reviewer states essential parity in his own words". "Of essentially the same character" supports comparability; it is not a declaration of equality. Both voices named this as advocacy.]

2. The 40:12 count was quoted without its scope or its hedge, and both bear on the audited pair. In full:

"we have noted in our reading of 'A Nobleman's Nest' some forty passages in which Miss Hapgood's rendering seemed less happy than that of her predecessor, and only a dozen in which the reverse was true. The 'Memoirs of a Sportsman' gave evidence pointing in the same direction, though by no means so emphatic."

So the counted asymmetry is about the novel; on Memoirs of a Sportsman, the cycle this project has audited twice and would build the arm on, the reviewer's own English-style evidence is explicitly weaker. Option B's strongest number — "40 to 12 is not a small gap" — is not a number about the candidate materials. Option C's sense-narrowing rests on the same misscoped figure and should be re-read in this light. Neither option is thereby refuted; both were argued from a figure that does not say what the page said it said.

3. Two sentences never recorded, both cutting against a ranking reading:

"Both usually write simple and idiomatic English, but neither produces work of marked literary excellence."

"One should not lay too much stress on these errors of detail. They can be entirely avoided only by the coöperation of several scholars…"

The first is a level judgment on literary-quality — the dimension the split favours Garnett on — sitting in the same paragraph as the parity sentence. The review therefore both levels the pair overall and splits it by dimension, and a reader who takes only one of those is taking half the document. The second is the reviewer moderating the weight of his own accuracy columns. [Overstatement corrected post-ratification: an earlier draft read "discounting". Both voices noted that "not too much stress" is moderation, not repudiation.]

4. A whole argument the project had not seen: the "thou" section (excerpt §7). The reviewer's leading stylistic objection to Hapgood is her rendering of the Russian familiar 2sg by "thou", which he argues is a dead idiom in English. That is a foreignisation choice, the project's central axis, criticised by name in 1904. It does not decide this page, but a reviewer weighing "is the English-style gap a quality gap or a strategy difference?" now has the reviewer's own answer to read.

What has NOT changed: the Athenaeum corroboration; conditions (i) and (iii); that Tier D remains NOT CALIBRATED whichever way this goes; and that no artifact is built on any option.

The direction of the correction, declared. On balance it makes the Nation look more like a parity document and makes the strongest anti-parity number smaller and off-target. That is the direction convenient to the lead's recorded preference and to S025's own interest in a permissive answer. It is also simply what the page says. The remedy offered is not assurance but reproducibility: the primary is quoted verbatim, the pipeline is two scripts in tools/, and a reviewer who disagrees can re-derive every word.

The evidence

1. The Nation (New York), Vol. 78 No. 2014, 4 February 1904, pp. 93–94 — the stronger

A joint review of both editions, headed "TURGENEFF AND HIS TRANSLATORS", with both title-pages printed beneath. Not two notices to be synthesised: one critic assessing both at once. Excerpt at SR-20260725b-nation-1904-excerpt.txt.

The reviewer's method is the project's own: he read A Nobleman's Nest and three sketches of Memoirs of a Sportsman in the Russian, compared both renderings against the source, and printed parallel columns of errors from each. "Mistakes of the sense are not hard to find in either version."

The verdict is split by dimension, with counts:

[SUPERSEDED by the S025 evidence correction above — the sentence IS quotable, and the reconstruction below was wrong in three words. Left in place, not edited away, because it is the record of a discipline that worked.] One sentence that would settle this outright cannot be quoted, and that is declared rather than glossed. The passage summarising the two versions falls where the OCR interleaves the page's two columns. The lead reconstructs it as "The two versions are, then, essentially of the same character. Their differences become evident only on detailed comparison", and the raw tokens are printed in the excerpt file. Only one scan of this issue exists on archive.org, so note (aa)'s two-scan precondition cannot be met. A reviewer should treat that sentence as unavailable and decide on the quoted, continuous passages above, which is how this page argues.

2. The Athenaeum (London), No. 4082, 20 January 1906, p. 70 — the corroboration

Unsigned. Excerpt at SR-20260725b-athenaeum-1906-excerpt.txt. Frames Hapgood as "entering into competition with the version of Mrs. Garnett, which first occupied the field", names The Memoirs of a Sportsman, and splits the same way:

"Her version is in elegant English, and perhaps in this respect superior to that of Miss Hapgood, who indulges in an occasional Americanism. But on the whole the translation of the latter is distinctly good, and she has the advantage of giving more notes than her English rival."

Two independent venues, two countries, two years apart, agreeing on the direction of every dimension they both address: Garnett better on English prose, Hapgood at least as good on accuracy and better on apparatus, neither better overall.

Against the ratified test

Both name both (✅), both address the same work with the audited cycle named (✅), neither is a lead-authored synthesis (✅). The contested element is whether they support comparability. What they support is a per-dimension split in which the two trade the lead, corroborated across sources — not an assertion of equal quality simpliciter.