Repository path: workshop/regimes/R26-light-ennoblement.md · rendered 2026-09-09

R26 — light ennoblement (the upward pole at reduced dose)

v1.0 (frozen 2026-08-14, S183.) An R06 single pass with a numbered rule set frozen before translating and a coverage code on every logged decision, in the shape R07/R08/R21/R25 established.

What it is for, and it is an instrument requirement that turned out to name a real translator. R25 was built at S174 to give the register axis its upper pole, and it worked too well: on a −1 / 0 / +1 height scale the seats coded the plain arm 0 and the ennobled arm +1 in every one of 56 cells, zero variance, while the quantities the run existed to measure — three published hands — sat at +0.125 to +0.70, entirely in the interior the gate had said nothing about. That is method note (bnb): a calibration gate that passes at ceiling has not calibrated anything in the range the primaries occupy. Its registered remedy is a half-dose arm — the same rule set applied at reduced strength — so the gate has to separate something smaller than a maximal manipulation.

But a reduced dose of ennoblement is not an artificial object. R25 is the translator Berman is complaining about, executed. R26 is the translator almost everybody actually is: the one who will not write slang and will not write Victorian either, who takes the source's colloquial cry and makes an ordinary formed English exclamation of it, and who stops there. Naming that policy and writing it down is worth doing whether or not a gate needs it.

It is R25 at half height with R25's two amplitude rules struck. R25's known-limitations section says in terms that E3 (expand) and E4 (clarify) make it a compound arm and that "a finding that R25 moves a register cannot attribute the movement to E1/E6 rather than to E3/E4, and any run wanting that attribution must build a rule-subset arm." R26 is that arm as well: it keeps the lexical-register rules at reduced height and forbids expansion and clarification outright. The reduction is therefore compound — height down and amplitude removed — and §Known limitations says what that costs.

Roles, human involvement

As R25. Lead translator, labeled subject, $0 (charter §3, A4). The lead never judges the output. No human involvement.

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

N1 — One step above the source, and never two; the neutral is left alone. Every lexical choice is pitched at the source's register or one step above it. Nothing colloquial: no slang, no contraction in narration, no bare cry left bare. But the ceiling binds as hard as the floor: nothing literary, latinate, archaic, poetic or decorous is admitted, and no word is chosen that a reader would notice as a choice. The target is the ordinary careful written English of a general publisher. Where the source is neutral, the English is neutral — R25 E1's second clause ("where the source is neutral, the English is literary") is struck, and that deletion is the single largest difference between the two regimes.

N2 — Idiom takes the ordinary standard equivalent. Where the source has a proverb, a set phrase or a figure, the English takes the plain current standard English equivalent — not a calque, and not the dignified or literary phrase R25 E5 requires. Where the source's figure is homely, the English figure is ordinary.

N3 — Exclamations are formed, not decorated. A bare colloquial cry or noun-shout becomes an ordinary complete English exclamation — the minimum that stops it reading as a bark. It does not become a decorous or literary formula (R25 E6), and it does not stay bare (R06, and R21 V6 in the other direction). This is the rule that does most of the visible work on a source built out of shouted single words.

N4 — Do not expand. R25 E3 is forbidden. Sentence division, coordination, connectives and clause order follow the source; no periodic sentence is built where the source runs short ones, no coordinated pair replaces a bare word, no connective is supplied for smoothness. The finished English is expected to sit at or near the source's own length, and the ratio is reported — this is a numerical rule and it is checked, exactly as E3 is checked in R25.

N5 — Do not clarify. R25 E4 is forbidden. What the source leaves indefinite stays indefinite: no antecedent, subject, reason or connective is supplied that the source withholds, and no reader is helped over a place the source makes them work at.

N6 — The propositional content is complete and unaltered. R25 E7 verbatim. Every event, speech act, referent, participant and causal relation in the source is present in the English. This is the rule that makes the output a translation and an admissible experimental arm rather than a rewriting, and it is checked, not assumed (§Procedure).

N7 — Stop short of the parodic, and equally short of the flat. R25 E8's constraint plus its mirror. The English must read as a serious rendering by a translator with an unremarkable register policy — a competent modern trade translation — neither a send-up nor a deliberately plain or stripped text. Where a choice would raise a laugh the source does not raise, the next-lowest choice that does not is taken; where a choice would read as pointedly ordinary, the next choice up is taken.

Procedure

As R25, with R26 substituted throughout:

• §5b — Every logged decision carries a coverage code: P (the rule set predicted it), D (deviation from the rule set, with reason), INT (decided by an interpretation of what the source means, not by a rule), L (a loss the rule set caused, recorded as a loss).
• §5c — N6 is checked, not assumed. Before the log is frozen, the translator walks the source and the target clause by clause and records any place where a fact, participant or event could be said to have moved.
• §5d — N7 is checked in both directions, and the check is recorded: the translator names every site where the rendering came near burlesque and every site where it came near flatness, and says what was taken instead.
• §5e — N4 is checked numerically. The token ratio against the source is computed and reported beside R06's and R25's on the same tale, since N4 is the rule most easily broken without noticing.

Outputs

As R25. Additionally: the token ratio against the source (N4), and the §5d flatness check, which R25 does not have.

Known limitations (by design)

• The reduction is compound, and no run using R26 may attribute a difference from R25 to height alone. Two things change together: lexical height falls from as high as English admits to one step, and E3/E4 (expansion, clarification) are removed entirely. A measured gap between the two arms is a gap between policies, not between doses of one rule. Decomposing it further needs a third arm and this regime does not supply one.
• "Half dose" is a description, not a metric. Nothing here guarantees that R26 lands between R06 and R25 on any scale, and a design that assumes it does has assumed its own gate. That R26 may be indistinguishable from an unruled pass is a live outcome — an unruled lead pass of a colloquial source is already fairly plain-but-not-slangy, which is most of what N1 asks for. Any run using R26 as a resolution gate must be willing to report that null.
• One translator's idea of "one step" is one idiolect, and it is the same translator who wrote R06 and R25 of whatever tale this is applied to. Note (bhb): the lead is not an independent sample of itself. Self-overlap against the sibling arms is measured on the artifact, not assumed.
• R26 is the lead's assembly from Berman's description of ennoblissement reduced by the lead's own judgement, not Berman's protocol and not any published translator's declared policy. Any finding about "the light upward pole" is a finding about this rule set. R21's and R25's version of this limitation, unweakened.
• N1's ceiling is period-blind. "Ordinary careful written English" means ordinary now. Where a run compares R26 against a Victorian published hand, the comparison is with a constructed contemporary pole and not with a reconstruction of that hand's norms — R25's limitation, and it bites R26 harder, because a Victorian hand's ordinary is above a modern one's.

Changelog

• 2026-08-14 v1.0 (frozen, S183) — created for ARM-elevation-resolution step 1 and E-20260814d-elevation-resolution, before any word of the translation was written, and against method note (bnb).