Repository path: workshop/regimes/R24-source-first-located-low.md · rendered 2026-09-09

R24 — source-first located low (the +I cell, written from the source)

v1.0 (frozen 2026-08-10, S155.) R24 is R23's cell A — +I −S, located idiom permitted, standard spelling — with R23's minimal-pair procedure replaced by a single pass from the source. Every rule is R23's verbatim. Only the procedure changes, and the change is the whole point.

What it is for, stated as the question it exists to answer

RS-20260810c §6 measured two procedures against each other and could not decide between two readings of the same number:

is the located idiom unavailable in Japanese narration, or is it merely unreachable by minimal revision? The instrument that would separate them writes the +I arm from the source, not from the ∅, and pays the tie-rate cost that the minimal-pair procedure was adopted to avoid.

R23's minimal-pair procedure was adopted for a good reason — RS-20260808f's independent passes produced four exact ties in eight sites and could not have reached significance whatever the data did — and it produced, at S150, an arm byte-identical to its own placeless baseline at 93% of hand-sites. A permission that is never exercised and a device that does not exist look the same under that procedure. R24 pays the tie-rate cost back in order to tell them apart.

R24 is not a claim that the minimal-pair procedure was wrong. It is the other half of a crossing: same rule set, two procedures. R23/A and R24 differ on one dimension by construction, which is what makes the pair readable, exactly as R21/R22 did.

The switches, fixed

The dictionary line is R23's and is not redrawn. A dialect word with its own dictionary spelling (nowt, summat, clout, waster) is I and is admitted. A standard word respelled to show its sound (nothin', o', 'em) is S and is refused, and the refusal is logged — under −S a form that is both located and respelled must be declined even though its I half is permitted.

The rules carried unchanged

W1 (no word above the source's own level), W2 (flatten source-culture particulars), W3 (contract, including the numerical clause), W4 (withhold what the source withholds), W6 (nothing is added — a gate, not a preference), W7 (propositional content complete and unaltered), W8 (stop short of the parodic and the incomprehensible): all R23 verbatim, which is R21/R22 verbatim.

W3 under a single pass. R23's minimal-pair carve-out does not apply — there is no baseline to stay byte-identical to — so W3's numerical clause binds as it does in R21 and R22: the finished English is no longer than the source in tokens, and a return pass is permitted only to satisfy it, with every cut named in the log.

Procedure — one pass, from the source, and the log carries a per-site availability record

1. One pass from the source alone. No published rendering consulted while translating; and, specific to R24, no previous rendering of the same span by the same translator consulted while translating, including the translator's own −I artifact. The −I half of the pair exists precisely so that the two can be compared afterwards, and a translator who reads it first is revising rather than translating.
2. A return pass permitted only for W3's numerical clause, every cut named.
3. W5+'s per-site record, and it is the rule set's own evidence. At every site where the source has a proverb, a set phrase, a figure or a popular turn, the log records: - the English chosen; - TAKEN — a located option was available and was taken, with the item named and the variety it would be placed in; or - AVAILABLE / NOT TAKEN — a located option was available and was declined, with the item named and the reason; or - NONE AVAILABLE — the translator could find no located English option at this site, with what was searched for.

R22's W5 recorded only the refused locatable option, because under −I refusal was the only outcome. R24 needs all three, because NONE AVAILABLE versus AVAILABLE / NOT TAKEN is the distinction the regime exists to make. 4. Coverage codes as R21/R22/R23: P predicted by the rule set · D deviation with reason · INT decided by interpretation · L a loss the rule set caused.

What R24 predicts about itself, written before translating

If RS-20260810c §5 is right that "where a hand does take the located option, it lands where the lead said it would — it just almost never takes it", then a source-first pass will take located options at a rate far above the minimal-pair procedure's 4 of 60, and NONE AVAILABLE will be rare. If instead the material genuinely has few purchase points for located English — which is what framework/v0.2 §7.1's third measurement suggests, reach is a property of the site's syntactic shape — then NONE AVAILABLE will dominate even under a permission a translator is actively trying to spend, and the framework's refusal stands on the material rather than on the procedure.

The rule set is indifferent between the outcomes. Both are written here before the pass so that neither can be claimed afterwards.

Roles, human involvement

As R21, R22, R23. Lead translator, labeled subject, $0 (charter §3, A4). The lead never judges the output. Where a design uses panel-written R24 arms as its primary, the lead's own R24 artifact is a secondary replication and says so — and where the lead wrote it knowing the arm's question, it enters no primary figure at all.

Relation to the other regimes