Repository path: wiki/findings/results/RS-20260803g-distance-not-deference.md · rendered 2026-09-09

Between adults, te is not deference in this book — it is distance, and that changes what was lost

No jury, no scores, no evaluation of any translation. This is a census of a source text and a translator's reading of what English can carry. One API call did enter it, in the only direction charter §4 permits: an adversarial seat instructed to refute, whose three findings are all corrections against the lead and are all applied below (E-20260803g/critic.md). Nothing it failed to refute is cited as support. Every count below is a claim about the Finnish and is checkable by anyone re-running the search on the files named; every judgment about what English does or does not carry is the lead's own and is flagged internal-judgment-only at the page level. Nothing here rests on a panel model, and nothing here is evidence about readers.

1. The question, and where it came from

For three spans this project has been operating on a reading it never checked. register.md V13 calls the third-person address to a lady "third-person deferential address"; V14a calls te "the pronominal honorific"; RS-20260803b spent a session and $0.26 measuring a compensation for it, and D-20260725-era vocabulary has treated the T/V contrast throughout as an honorific — a marker of who is above whom.

Span 3 registered a prediction with a falsification condition (register.md §Unresolved 3):

Span 4 opens with Tiina Katri and Koposen Lopo in the room — the first sustained talk among poor women who all know each other, where the mutual-te finding of D60 predicts sinä throughout. If it is te, D60's grid is wrong about what the eight stranger-sites were marking.

Span 4 was translated this session (¶207–295, 1,800 Finnish words, T-koyhaa-kansaa-R05-v1, log D65–D83, frozen before this page was written). The prediction is the study limb's material.

2. The prediction holds, and it holds at 15 sites out of 15

Every second-person form between acquainted adults of the poor class in span 4 is T-singular. Not one te.

Six te events occur in the span and not one of them is politeness to an intimate. Three are plural by number beyond argument — ¶283 «Toin vähän teillekin» to the household, ¶279 and ¶284 to the two boys. Two are contested and are argued in §7 rather than counted past (¶222 and ¶240; the adversarial pass refuted their classification, E-20260803g/critic.md §3). The sixth is §4.

¶222 is the one the argument leans on, and the argument is a minimal pair of exactly D60's kind. Tiina Katri says teidän to Mari at ¶222 and sinä to the same woman, in the same conversation, at ¶215, ¶217, ¶261, ¶286 and ¶288. One speaker, one addressee, one scene, both forms — so the te at ¶222 is very unlikely to be politeness, and its referent (a bed belonging to six people) is what a household plural is for. An independent seat, shown the paragraph without this argument, called it "potentially polite singular" and was right that the paragraph alone does not decide it.

3. The whole-text distribution, ¶2–295

The prediction having held, the same search was run over everything translated so far — spans 1–4, 7,172 Finnish words, 53% of the novella — plus a whole-text scan for the one feature the register had made a rule about (§5). 30 sites carry a te form in ¶2–295, classified by what the relation is:

The finding is the first row, and it is about adults. (This sentence was re-worded after the adversarial pass: the claim originally read "the contrast is not status-asymmetric" tout court, and P3 broke it in one line by pointing at the child-to-parent row, which is a status relation. The corrected claim is the one the distribution actually supports.) Heikura the landlord and Mari his tenant, alone in a room, address each other with te — she says «istukaa» (¶80) and «Ettehän te ole» (¶83) and «Jos ottaisitte, isäntä» (¶97); he says «Itsekukin, näettekös, tarvitsee omansa» (¶84), a discourse particle in the second-person plural addressed to one woman standing in front of him. The form is symmetric across the sharpest status gap the scene contains, and it is symmetric again between Mari and Hanna the servant, and again between Mari and a beggar-woman poorer than herself.

A form that is symmetric across a status gap is not marking the gap. What it marks, between adults, is that the two people are not intimates — and note that it is compatible with warmth: Hanna calls Mari «Mari kulta» and says «näettekös» and «Pyytäkää» in the same breath (¶122). It is not coldness; it is not being on sinä terms. sinä appears between husband and wife, between neighbours who have lived beside each other for years, and from any adult to any child. The axis is intimacy, not rank.

And rank is marked in this book — by other means. Mari does not say te to a rouva at all: to a lady she uses the third person (V13, ¶132's scene), which is a further step out and which no one uses to her. Above te there is a form; te is the middle of a three-term system, and calling the middle term an honorific was reading a two-term system that is not there.

The exception proves the shape. The one relation in ¶2–295 where te is status-asymmetric is generational and inside one family (§4), which is why the corrected claim is about adults. A three-term system whose middle term is symmetric between adults and asymmetric between a child and her mother is not an honorific system with a gap in it; it is a distance system with one place where distance cannot be what is meant.

4. The one asymmetric te, and why the compensation argument survives

Two rows above are asymmetric, and they are the interesting ones.

¶248, new in span 4, is the cell D60 said the book had not yet supplied. Lopo quotes herself at the tax office: «kauluksenkohan otatte hameestani ensi vuonna vai kuoninginko. Taikka luuletteko rahapennin saavanne, kun nyt vihon viimeisen kiskotte» — three plural forms, upward, to officials, and no answering form. Span 3 has its mirror at ¶149: «Nuori olet, tee työtä, sanovat» — officialdom's singular, downward, to a destitute woman. Both halves of the asymmetry are in the novella, both are inside quotation inside a poor woman's own account of being dealt with, and English renders all four forms "you". The class marking that survives in English at these two sites is the content of what is quoted, not the grammar.

The other asymmetric row is a nine-year-old girl to her mother, and it is the only place in the book where te runs upward inside intimacy. That is precisely why it cannot be distance: Hellu and her mother are as intimate as two people in this novella get, they share a bed on a heap of clothes, and the form appears anyway. Where distance is excluded, deference is the only reading left.

This is the same conclusion D57/RS-20260803b reached from a single adjacent minimal pair at ¶151, and it now rests on the whole distribution instead. The argument for treating ¶175's te as deferential is stronger than it was; the vocabulary the project used to state it was wrong.

5. What the loss actually costs: 4 sites, not 30

The register's V18a says the honorific from child to parent is lost at every site and that the loss is chosen. register.md §Unresolved 2 asked spans 4–8 to count the further occurrences, because "how often a feature the translation cannot carry actually occurs is the only thing that makes the loss large or small". Span 4 has none, so the count was run over the whole novella instead of waiting four more spans for it.

It says "both known sites". There are four, all Hellu, all to her mother, all confirmed in the 1886 first edition:

No erratum: V18a mandates plain and all three translated sites already read plain. What changes is the price. The register had the count wrong by half, and the corrected figure is still small: four sites in 13,525 words, against 26 further te sites where English loses nothing anybody could point to, because at every one of those the relation is fixed by a vocative (isäntä, rouva, Mari kulta), by who is in the room, or by what is being said. (That last clause is the lead's judgment, not a count.)

So the honorific loss this project has spent two sessions on is real, is four sites wide, and is the smaller part of what the pronoun system does. The larger part — the intimacy axis, 26 sites — English loses completely and costlessly, and nobody had noticed because it had been misdescribed.

6. What this changes

1. register.md V13 and V14a are re-worded (V23): te is the non-intimate form, not the pronominal honorific; the honorific reading is available at 4 of 30 sites and is licensed by asymmetry-inside-intimacy, not by the form.
2. The count that mattered was a whole-text count, and the span-by-span rule delayed it by four sessions. V18a's own instruction — record every further occurrence, span by span, and total it in the craft report — would have produced the number at span 8 and would have produced it wrong, because the site it missed is in span 1, behind the counter. The standing lesson (V17: no rule for an unread site) does not forbid a whole-text count, and this is the case that shows why it must not.
3. RS-20260803b's finding is unaffected in substance and improved in support. Its number (+1.167, p = 0.0368) measured whether one English device recovers one girl's standing; that is untouched. Its framing — "the pronominal honorific" — is corrected here.

7. Limits

• Two classifications were refuted by the adversarial pass and are carried here openly (E-20260803g/critic.md §3; E-20260803g/design.md FC3). ¶222 «Mihinkä teidän on sänky joutunut» — the seat is right that the paragraph alone does not exclude a polite singular; the minimal-pair argument in §2 was not put to it. ¶240 «Elkäähän… ennenkuin kuulette» is the one that costs something: the addressee may be the whole room or may be Lopo alone, who has just spoken and who is not a member of the household — she is fetched from next door at ¶217. If it is Lopo alone, the form is politeness to a woman the Holpainens have just met, and this page's classification of the site as "by number" is simply wrong. Under that reading §3's thesis is supported and §2's arithmetic is wrong, which is why the site is stated rather than defended. A live tension underneath it, named rather than smoothed: at ¶249 somebody in the same room says «miehesi» — T-singular — to Lopo. Two speakers, one addressee, two forms, one scene. That is either two degrees of acquaintance or a hole in the account.
• The remaining classifications are the lead's, from referent and context: ¶29, ¶151, ¶279, ¶283 and ¶284 as plural by number (two boys, two children, a household). If any is instead a polite singular, the mutual-te rows in §3 are untouched, because the argument turns on the landlord scene, not on them.
• Spans 1–2 were not counted sinä-site by sinä-site. §3 counts te sites exhaustively over ¶2–295 and sinä sites exhaustively over spans 3 (D60) and 4 (D77) only. The claim "sinä between intimates, te between non-intimates" is therefore established as: no te between intimates anywhere in ¶2–295, and no sinä between non-intimates in spans 3–4. The second half is not checked for spans 1–2 and is not claimed for them.
• ¶154 and ¶159 are imported from D60's enumeration and were not re-derived this session.
• Chapter II (¶296–573, 47% of the novella) is unread except for the whole-text searches run for §5 and for raskita (D74). A te between intimates there would falsify §3, and nothing here forecloses it. This is the honest exposure of the result and it is named rather than hedged.
• The material was checked against a second printing before any of this was written: all 22 second-person sites named on this page read identically in the 1886 Edlund first edition, so none of the copy-text corruptions found this session (collation.md) touches the census. That check was run because the opposite outcome would have voided the page.
• No claim is made about readers. Whether an English reader recovers "distance" from a vocative is a measurable question and is not measured here.
• The adversarial pass ran with one seat, not two. openai/gpt-5.6-terra returned finish_reason: length with zero characters of content on both dispatches — 7,000 completion tokens of unreturned reasoning — and was withdrawn under the design's own FC4 rather than replaced with a fresh model after the first verdict was already in hand. One refuter is a weak instrument. Nothing the seat declined to refute is cited anywhere on this page as support.

8. Provenance

Source: workshop/translations/koyhaa-kansaa/source-fi-full.txt (Gutenberg 13976, Otava 1917) and the 1886 Edlund first edition, Doria 10024/100842 (collation.md §6). Translation and frozen log: workshop/translations/koyhaa-kansaa/R05-v1/translation.md, span 4, D65–D83. Prior grid: D60 and RS-20260803b. Adversarial pass: workshop/experiments/E-20260803g-address-axis/ (design frozen before dispatch, one seat of two, three findings, all three accepted). Cost: $0.0778522, of which $0.0444738 bought nothing; declared worst case $0.42.