Repository path: workshop/experiments/E-20260727c-sense-boundary/contamination-preprobe.md · rendered 2026-09-09

Contamination pre-probe — 「最後の一句」, frozen before translating

The standing gate could not be run, and this file says exactly why and what was done instead. CLAUDE.md requires the lead's contamination on candidate material to be measured before the experiment on it is designed — one longest-common-run call against a published English rendering. That call needs a rendering the project can obtain.

What exists

One English translation. David Dilworth and J. Thomas Rimer, "Saigo no ikku: The Last Phrase", in The Historical Literature of Mori Ōgai, Volume 1: The Incident at Sakai and Other Stories, University of Hawai'i Press, 1977. In copyright, not freely reachable, and not stored by this project. Other languages: Hungarian (Teleki 1983), German (Schamoni 1989), Russian (Ivanova 1990/2002), Romanian (Săndulescu 1996), Spanish (Gallego Andrada 2000). Source: the Ōgai-Portal translation bibliography, Humboldt-Universität zu Berlin, read 2026-07-27.

So the measurement is impossible, and the declaration on the translation is contamination: none — UNMEASURED. It joins the five unmeasured declarations CLAUDE.md already flags. Recording it as measured would be false.

What the gate's second clause says, and it clears

The binding prohibition is that "the lead cannot serve as the independent third translator for any work whose standard English translation is canonical." One 1977 academic-press translation of a rarely anthologised Ōgai story is not canonical in the sense Garnett's Turgenev is — there is no ambient English text of 「最後の一句」 the way there is of «Свидание». The measured 21-token Turgenev result exists because Garnett is unavoidable; nothing here is unavoidable. The gate clears on its own terms; what is missing is the number, not the safety.

That distinction is worth keeping straight, because CLAUDE.md warns that an unmeasured declaration can hide either end of the 0–21 token range. The reason to think this one sits at the low end is bibliographic — a single low-circulation rendering — and bibliographic evidence is weaker than a measurement. It is not nothing, and it is not a measurement.

The free substitute, pre-registered here before translating

A self-report, and its weaknesses are the point of writing it down in advance rather than afterwards.

Asked, before translating, to produce whatever English wording I associate with this text, I can produce: the title, as "The Last Phrase" — which is Dilworth and Rimer's, and which I hold because it appears in the bibliography I just read, not from the translation. I cannot produce a rendering of the story's final spoken line 「お上の事には間違いはございますまいから」, of the 「マルチリウム」 sentence, or of any other sentence in §4. No candidate English wording for any of them is available to me to write down here.

What this is worth: very little, and in one direction only. It is the lead reporting on its own memory, which is internal-judgment-only by nature and unfalsifiable in the direction that matters — an absence of recall is not an absence of influence, and a rendering can be reproduced without being recognised, which is precisely what the Turgenev measurement showed. It would have been informative had it come out the other way. It did not, so it is recorded and not relied on.

Frozen before translation begins. Nothing below this line was written after §4 was rendered.