Repository path: workshop/regimes/R09-futabatei-form.md · rendered 2026-09-09

R09 — Futabatei's form-preserving standard, executed

v1.0 (frozen 2026-07-30, S066.) An R06 (lead, source-only, single pass) with one addition, on the R07/R08 pattern: a numbered rule set frozen before the source is read for translation, consulted at every choice and cited by number in the log.

The rule set is not the project's invention. It is 二葉亭四迷's own account of his own method, in 「余が翻訳の標準」 (Waga hon'yaku no hyōjun, 「成功」, 明治39年1月 = January 1906), read in the original Japanese and held in the project at workshop/translations/futabatei-honyaku-hyojun/R04-v1/. Every rule below quotes the sentence it comes from.

Why it exists, and why it is not another R07

R07 and R08 were built from Venuti describing other people's translations and turned round into instructions. Tymoczko's objection — that no operational criteria are supplied — was the thing being tested, and the turn from what a reviewer rewards to what a translator does was the project's own move and is contestable.

R09 needs no turn. Futabatei states his rule in the first person, as something he did, and his execution of it survives: 「あいびき」 (1888), the whole of it, freely readable. This is the only regime in this project whose author also supplies a translation made under it, by himself, which can be scored on the regime's own terms.

Two of its rules are countable. J1 and J2 below are not judgements about register or resistancy; they are integers. That is unique here and it is the whole reason this regime was worth writing.

Roles

• Translator — the lead agent, as a labeled subject. No panel call, no API cost (charter §3, A4).
• Judge — never the lead (charter §5). R09 makes no quality claim at all; what it produces is a compliance count.

Human involvement

None.

The rule set (frozen; cited by number in the log)

J1 — Match the sentence count. Each source sentence becomes exactly one target sentence, with the same sentence-final mark where the target language has one. No splitting; no joining.

「原文にコンマが三つ、ピリオドが一つあれば、訳文にも亦ピリオドが一つ、コンマが三つという風にして、原文の調子を移そうとした。」

J2 — Match the comma count. Within each sentence, the target carries the same number of commas as the source.

ibid.

J3 — Discard no mark. No source comma or sentence-final mark may be dropped, and none may be added, even where the target language's own punctuation conventions would want it otherwise. J3 is J1 and J2 stated as a prohibition, because Futabatei states it that way and because the prohibition is what bites.

「コンマ、ピリオドの一つをも濫りに棄てず」

J4 — Match the word count as closely as the language allows. Not a hard count; a standing pressure against expansion and against compression.

「殊に翻訳を為始めた頃は、語数も原文と同じくし」

J5 — Do not break the form. Clause order, phrase order and the position of every mark follow the source unless J6 overrides. Where English would ordinarily reorder, do not.

「形をも崩すことなく、偏えに原文の音調を移すのを目的として」

J6 — 詩想 overrides form, and only 詩想 does. Where holding J1–J5 would lose the author's 詩想 — his cast of imagination, the thing Futabatei describes for Turgenev as late spring, cherry blossom just beginning to scatter, loveliness with something desolate in it — J6 wins. Every J6 departure is logged with the mark it cost and the reason. Nothing else overrides J1–J5: not idiom, not readability, not English convention.

「此の際に在ては、徒らにコンマやピリオド、又は其の他の形にばかり拘泥していてはいけない、先ず根本たる詩想をよく呑み込んで、然る後、詩形を崩さずに翻訳するようにせなければならぬ。」

J7 — The author's 詩想, not a general manner. Turgenev is written as Turgenev.

「ツルゲーネフはツルゲーネフ、ゴルキーはゴルキーと、各別にその詩想を会得して」

J8 — Do not smooth the result. The regime's own author reports that the output came out 佶倔聱牙 — knotty, jarring — and was badly received, and he kept the method for years anyway. A translator working under R09 does not repair awkwardness that J1–J5 produce. Smoothing it is a departure and is logged as one.

「出来上った結果はどうか、自分の訳文を取って見ると、いや実に読みづらい、佶倔聱牙だ、ぎくしゃくして如何にとも出来栄えが悪い。」

What the regime does NOT contain

Futabatei's essay spends its second half on Zhukovsky's opposite method — break the form entirely, carry only the 詩想 — and says he thought it better and was too timid to adopt it (「自分は臆病だ」). That is not in R09, because it is not what he did. A regime built from what a translator says he admires rather than from what he says he practised would be the same defect R07 was built to avoid.

Procedure

1. The rule set is frozen in git before the source text is read for translation.
2. The source is translated in one pass, source only, no comparator opened.
3. Every choice where a rule bound, or where J6 overrode one, is entered in the log with the rule number.
4. The compliance counts (J1, J2, J4) are computed after the log is frozen, by a program, not by the translator.
5. The translator makes no quality claim. Compliance is not quality and the regime does not assert that it is.

Status

frozen 2026-07-30 (S066), before workshop/canon/svidanie/source-ru.txt ¶1–¶19 was read for translation. First and so far only artifact: T-svidanie-R09-v1.

Unratified. No independent vote has been routed through this page. It binds E-20260730g by being cited, and that experiment says so in its own text (framework/control-arm-spec.md §Status, the standing condition adopted at S061).