Repository path: workshop/translations/niewola-tatarska/R04-v1/translation.md · rendered 2026-09-09
Page metadata (front matter)
| type | translation |
|---|---|
| id | T-niewola-tatarska-R04-v1 |
| status | frozen |
| created | 2026-07-30 |
| updated | 2026-07-30 |
| senses | style-correspondence, cultural-mediation, accuracy, voice |
| provisional | true |
| translated-by | lead |
| contamination | suspected |
| links | workshop/canon/niewola-tatarska/manifest.md, workshop/regimes/R04-lead-close.md, workshop/translations/niewola-tatarska/R06-v1/draft.md, wiki/arms/ARM-revision.md, workshop/experiments/E-20260730c-revision-close/design.md |
T-niewola-tatarska-R04-v1 — Sienkiewicz, «Niewola tatarska» I (opening)
Source: workshop/canon/niewola-tatarska/source.txt — Henryk Sienkiewicz, Niewola tatarska.
Urywki z kroniki szlacheckiéj Aleksego Zdanoborskiego (1880), section I ¶1–2 entire plus ¶3 cut at
a named sentence boundary; 985 Polish words, three paragraphs. Provenance, digests, the cut and
the single-witness limitation: workshop/canon/niewola-tatarska/manifest.md.
Regime: R04 v1.0 — lead, source only, draft then self-revise in session. The draft is frozen
separately as T-niewola-tatarska-R06-v1 at commit 14c8989, with its own pass log, before
this revision was begun (R04 §2a).
Pair: PL→EN. The project's first Polish source and its first macaronic source — the narrator quotes Latin inside Polish syntax at eight places, one of which turns on Latin inflection.
Contamination
suspected, and measured — but measured after both passes, deliberately, and that ordering is a
departure from CLAUDE.md's standing selection gate which is argued rather than confessed.
The standing rule is that the lead's contamination on candidate material is measured before a
locus is selected — a selection gate, not a flag (note bcd). Here the quantity the gate would
select on is the study limb's outcome variable. ARM-revision step 2(b) asks whether the R04
revision's longest common run with a published comparator is longer than the R06 draft's; this pair
was translated to be a prospective instance of exactly that measurement. Selecting the locus on a
low run would have conditioned the sample on the dependent variable, and the resulting drift
figure would have been drawn from loci pre-selected for the value being compared. So the gate was
not run as a gate, and the reason is not convenience. Method note (bel) records the general
form: a selection gate may not be run on the outcome variable of the study it gates.
What is preserved instead, and it is what the gate was actually protecting: the translator never
opened the comparator. Curtin's Hania was fetched, its story boundaries found by string search on
headings, and a 3,000-word window from the story's opening passed to tools/dependence_check.py and
to nothing else (method note abm). No English rendering of this work — or of any Sienkiewicz — was
read before both passes and both logs were frozen.
Comparator: Jeremiah Curtin, "Tartar Captivity," in Hania (Little, Brown, 1898), Project Gutenberg #36583.
The figures were computed after the freeze commit 73ba25c, which contains this artifact and its
log with this section empty. The order is load-bearing rather than ceremonial here, because the run
figure is the study limb's statistic. Nothing above or below this section changed when the
numbers arrived; in particular the sentences carrying the two shared runs were not revised, on
S061's precedent: a passage revised because a measurement pointed at it is a passage that
measurement can no longer say anything about.
tools/dependence_check.py, unmodified. Comparator window 3,012 tokens; R06 draft 1,354; R04
revision 1,353.
| cell | shared 7-grams | name-excluded | 12 | nn | 15 | longest run | tool verdict |
|---|---|---|---|---|---|---|---|
| R06 draft vs Curtin | 36 | 21 | 5 | 3 | 0 | 14 | DEPENDENT? |
| R04 revision vs Curtin | 33 | 18 | 5 | 3 | 0 | 14 | DEPENDENT? |
| R06 vs R04 (not a contamination cell) | 1,120 | 848 | 1,024 | 636 | 970 | 247 | — |
Declared suspected, and the declaration was fixed before the numbers were seen — Sienkiewicz's
English translations have been continuously in print since the 1890s and none would be a claim
about training data that no output measurement can license. The measurement makes it a finding
rather than a formality.
There are exactly two maximal shared runs of eight tokens or more, and they are of two different kinds. Both are present identically in both passes.
- 14 tokens — "at that land and at everything in it, and as I have called it". Three content
words in fourteen (
land,everything,called). Renderingdziwowałem się bardzo i téj ziemi i wszystkiemu co w niéj jest, a jakom ją nazwałleaves an English translator very little room, and this looks like the forced classARM-forced-defenceexists to characterise. - 13 tokens — "Jesus Christ, who lookest into my heart, seest that I would have done". This is
not function-word filler: it carries two archaic second-person-singular inflections,
lookestandseest, which are choices in modern English and not in the Polish.
Run 2 is the interesting one and it suggests a third mechanism, which this artifact records as a
conjecture and not as a result. The project's contamination discussion (CLAUDE.md) has two
categories: a run is either memorised from the comparator or forced by the source. Run 2 is
neither cleanly. It follows from a register decision — D1 of the draft log, made and written
down before any comparator existed in this session — that the narrator's Polish archaism be met with
seventeenth-century English. Given thou-address, Ty, Jezu Chryste, który patrzysz w serce moje,
widzisz has almost one rendering. Curtin, translating in 1898, made the same register decision, for
his own period reasons. A run can therefore be produced by two translators independently adopting
the same convention — forced conditional on a choice neither was obliged to make. That is
distinguishable from memorisation only by testing whether other translators making the same register
choice produce the same string, which this session does not do.
Where the pair sits in the project's measured range. CLAUDE.md records lead contamination from
0 tokens (Ovid) to 21 (Turgenev/Garnett). 14 on a first-ever language pair, against a
comparator never opened, is squarely inside that range and is the first Polish datum in it. It does
not trip the rule CLAUDE.md states — the lead may not serve as an independent third translator
where the standard English translation is canonical — but Curtin is close enough to canonical for
Sienkiewicz that a future design should treat Polish/Curtin as it treats Russian/Garnett until
someone measures more of it.
The R06→R04 row is the revision's own size, and it is the reason this pair earns its place in
ARM-revision step 2. The two passes share a 247-token contiguous run — the middle of ¶3's
ethnographic catalogue, from "he barely tips his cap to you" to "from which licence of theirs" —
so the revision touched paragraph 1 heavily and left a fifth of the passage untouched. The
uneven distribution is visible in the diff and is what the revision log's own D31–D51 describe.
Our serving-lad, riding now ahead of me and now behind, kept twanging away on his torban, while grief and longing for Marysia wrung my heart—and the further I rode from her, the more hotly I loved her. At such times the words came into my mind: post equitem sedet atra cura; but since, my fortune being so greatly straitened, I dared neither speak with His Honour Tworzyański nor confess my affections to him, nothing was left me but to earn a new fortune with my sabre, and having adorned myself with gloria militari, then and only then to stand before him. Neither God nor my own dear Marychna could take it ill of me that I had not done so sooner. Had she bidden me leap into fire or into water, or shed my blood outright, Thou, Jesus Christ, who lookest into my heart, seest that I would have done it. But one thing I could not sacrifice even for my winsome girl, and that was my honour as a nobleman. My fortune was nothing, but the worth of my blood was great, and it had come down to me from my fathers as though by testament that I was ever to bear in mind that my neck is my own and I am free to hazard it, but that the integra rodu dignitas is my forebears' bequest, which I must render up as I received it: integram. — Eternal rest grant unto my fathers, O Lord, and let everlasting light shine upon them, world without end. Even had His Honour Tworzyański consented to give me his girl, I should have had nowhere to bring her home to; and if, looking at the slenderness of my fortune, he had in his pride called me a pauper, or a grey-coat outright, then I should have felt myself touched in my conception of my house's excellence and must needs have avenged myself upon him—which, he being the father of my Marysia, God the Lord forbid.
And so there was nothing for it but to ride out to the Borderlands. The horse-trappings and belts and whatever else of better sort had come to me from my fathers I pawned in part and in part sold, and so got together three hundred ducats of full weight, which I handed over at once to His Honour Tworzyański against the interest; then, having taken leave of Marysia with tears and heavy sighing, I made ready through the night for the road, and on the morrow the two of us, my man and I, turned our horses towards the East.
It fell out that we must ride by way of Zasław and Bar to Hajsyn. Halting for the night at castles, manors or inns, we came at last to Uman, beyond which the steppe already opened before us, level, rank and silent. My man, riding ahead, kept striking up on his torban and singing, and it seemed to me that there flew before me, like a bird I was chasing, glory—and behind me, like a second bird, longing. We were riding for the outpost called Mohylna, where in his time my father, His Honour, had kept watch as colonel with an armoured banner which he had raised at his own charge for the war against the Musulmans; but to Mohylna it is a very long way, for—God be praised—the Commonwealth is spread broadly over the earth, and besides that one must ride there across steppes over which Tatars and brigands of every kind go prowling day and night, so that a man does well to guard his own neck. Along the road I marvelled at everything, since, being in Ukraine for the first time, I met with nothing but unfamiliar matters and things. It is a soldierly land, and the people in it are harder than ours and prouder, and there is such a dash in a peasant there as a nobleman need not be ashamed of. When you ride through a settlement, though he knows you for a wellborn man, he barely tips his cap to you, and looks you straight in the eyes. In every cabin there is both a sabre and a firelock, and many a peasant carries a hammer-axe in his hand, just as a nobleman does elsewhere. There is a stubborn nature in these people too, and even—for which the sabre has punished them already and will punish them again—they make little enough of the Commonwealth's commissioners: such is the mettle that nearness to the heathen and a continual readiness for war have worked in them. They take to the plough none too willingly, and if husbandry does prosper with one of them, he would rather settle on his own ground than on a lord's. But into the household retinues, or under the Commonwealth's light colours, they enlist greedily, and the soldier they make is excellent, above all for scouting and skirmishing—though in the open field too non obtrectant, but raising a great shout they go at the enemy as into smoke, cutting and thrusting. Every settlement of theirs is liker a wagon-camp than a village;—they keep a great many horses, which graze on the steppe winter and summer, and they are as fleet as the Tatar sort. Many of them also run off to the insulae of the Dnieper, and there in the Sich lead a life as of a religious order, but a soldierly one and altogether given to brigandage; from which licence of theirs much has been suffered and will yet be suffered, until our dear country curbs them. To keep them in place—for any nobleman, or even a great lord—is hard, for time and again they break away and go off into the empty steppes, of which there is plenty there, to settle at their own will. In the build of their bodies as well as moribus they differ from our peasants, being tall and sturdy, of a swarthy complexion, liker the Tatar, with moustaches black as a Wallachian's, while their heads—having taken the fashion from the heathen—they shave, leaving on the very crown alone a fierce scalp-lock. Seeing and considering this I marvelled greatly both at that land and at everything in it; and as I have called it soldierly, so I will say again that a country better suited to an armed and mounted people one might search the whole world for in vain. When some perish, others come up from every side, and along all the trails they move just as flocks of birds do; and in that steppe waste it is easier to hear, than a lark above the tilth, the crash of matchlocks, the clash of sabres, the neighing of horses, the flapping of banners in the wind and the shouting of soldiers. There go about there also, as likewise in Podolia and Volhynia, old beggar-men whom all hold in great respect. These, being blind, play on their lyres and cantant songs of knighthood, whereby mettle and a feeling for glory flourish greatly. And the soldier there, seeing that those who live today rot tomorrow, heeds his own life no more than a broken penny, dispensing blood as a magnate dispenses gold—and cares more for a fair death than for life or temporal goods. Others, having come to love war above all else, though not seldom of great blood, grow almost wild in unbroken soldiering, and go to battle as though to a wedding, with great gladness and singing; but in time of peace they grow sorely irked, and finding no outlet for their martial humours, are a threat to the common quiet. These they call the lost men. When a man perishes there, all reckon it an ordinary thing, and even his nearest do not much weep for him, saying that it better befits a man to die on the steppe than in a bedchamber like a woman. And indeed that is the best school and seasoning for knighthood there is.
Translator's log (revision pass)
Frozen with this artifact and before any measurement of it was computed. Numbering continues the
draft log (../R06-v1/draft.md, D1–D30), which is a separate pass log and was frozen at 14c8989.
Decisions and alternatives; no quality claims. This is the pass-separated log form
ARM-revision's primary requires — three of the project's earlier pairs have one and eleven do not.
One repair the revision made is an accuracy repair and the rest are not. Stated first because it is the one that matters most and the one the draft got wrong.
D31 — the distributive was mistranslated in the draft, and this is the revision's only meaning
change. Rzędziki, pasy i co było lepszego po ojcach, **częścią** zastawiwszy, **częścią**
przedawszy is part pawning, part selling — the same set of objects, split two ways. The draft read
"Having pawned some of the harness-mountings and belts … and sold the rest," which silently made
the split a split between object classes. Revised to "The horse-trappings and belts and whatever
else of better sort had come to me from my fathers I pawned in part and in part sold." Also
rzędziki → "horse-trappings" rather than "harness-mountings", which is period-plainer.
D32 — pobrzdękiwało → "kept twanging away" (draft: "went on plucking away"). Brzdąkać is to
twang or strum, and the po- frequentative is desultory. "Plucking" is a lutenist's word and the boy
is not playing well.
D33 — ściskały serce → "wrung my heart" (draft: "gripped my heart"). Ściskać is to squeeze;
"wring" is the English idiom for the squeeze, and it is the older one.
D34 — w tak wielkiém fortuny mojej uszczupleniu → "my fortune being so greatly straitened"
(draft: "with my fortune so greatly diminished"). Uszczuplenie is reduction of means, and
straitened is the period English for exactly that. The Polish is a heavy Latinate ablative; the
revision keeps an absolute construction rather than a prepositional phrase, which is nearer the
syntax as well as the diction.
D35 — bym wiecznie uważył → "that I was ever to bear in mind" (draft: "that I should hold for
ever in view"). The draft's phrase is not English. ALT considered: "that I should never forget."
Not taken — the Polish is an injunction to keep something continuously in view, not a prohibition on
forgetting.
D36 — nie miałbym jéj gdzie wprowadzić → "nowhere to bring her home to" (draft: "nowhere to
bring her"). Wprowadzić is to install in a house; the draft lost the house, which is the whole
point of a landless suitor's difficulty.
D37 — w rozumieniu o wyborności rodu mego → "in my conception of my house's excellence" (draft:
"in my understanding of…"). Both are awkward; rozumienie here is the view one holds, and
conception carries that where understanding reads as comprehension.
D38 — the draft's one flagged construction, repaired. D30 recorded "and should have had to take
revenge upon him" as a modal stack R06 forbade fixing, with "must have avenged myself" recorded as
the unwritten alternative. Revised to "and must needs have avenged myself upon him" — the
recorded alternative plus needs, which is the period intensifier and was not in the recorded ALT.
So the revision did not simply execute the draft log's stated alternative; it went one step past
it. Recorded because ARM-revision's premise is that a log predicts what the pass will do.
D39 — Owóż i nie zostało jak… → "And so there was nothing for it but to ride out to the
Borderlands" (draft: "So then there was nothing left but to ride for the Borderlands").
Owóż is a resumptive particle; and so carries it. "Nothing for it but" is the idiom "nothing
left but" was reaching for, and jechać na kresy is ride out to, a departure, not a direction.
D40 — Wypadło jechać na… → "It fell out that we must ride by way of…" (draft: "Our road fell by
way of…"). The Polish is impersonal — it fell out that the road had to go that way — and the draft
made the road the subject, which is smoother English and a different sentence.
D41 — głuchy → "silent" (draft: "soundless"). Głuchy of a landscape is deaf, hence
soundless, hence deserted. "Silent" is the plainer word and sits better in a triple with "level,
rank".
D42 — coraz to w teorbanik uderzał i pieśni śpiewał → "kept striking up on his torban and
singing" (draft: "struck up on his torban again and again and sang songs"). Coraz to is the
iterative the draft rendered twice — once with "again and again" and once by leaving the verb simple.
The revision moves the iteration into "kept" and drops "songs", which śpiewał already implies.
D43 — wystawił → "raised" (draft: "put into the field"). Wystawić chorągiew is to raise and
equip a banner at one's own cost; the draft's phrase describes deploying one, which is a different
act and is the colonel's, not the founder's.
D44 — więc i szyi własnéj strzedz bardzo wypada → "so that a man does well to guard his own
neck" (draft: "so that it is very much in place to have a care for one's own neck"). Wypada is
it is fitting; the draft calqued it into a phrase English does not have.
D45 — fantazya → "a dash" (draft: "a spirit"). In this register fantazja is dash, bravado,
swagger — not imagination and not spirit in general. This is the revision's most substantive lexical
change after D31.
D46 — choć pozna żeś urodzonym → "though he knows you for a wellborn man" (draft: "though he
sees you are wellborn"). Poznać here is to recognise someone as something; the construction
"knows you for a —" is the English idiom for it and is period.
D47 — animusz made consistent, and a policy line drawn that the draft drew mid-sentence. The
draft rendered animusz as "mettle" at its first occurrence and as "spirit" at its second, and D18
recorded the inconsistency as it happened. The revision makes both "mettle", and fixes the
policy behind D2: Latin quotations are retained; Latin loanwords naturalised into Polish
(animusz, fantazja, prowizja) are englished. pauper, insulae, moribus, cantant,
non obtrectant and the two tags stay Latin because the narrator is quoting, not speaking.
D48 — the closing clause of the Sich sentence re-ordered. na którychto ich swawolach wiele
ucierpiała i cierpieć jeszcze będzie, póki ich nie poskromi miła ojczyzna nasza has the subject
arrive last, after the verbs. The draft supplied it early ("our dear country has suffered much");
the revision restores the Polish order with an impersonal passive: "from which licence of theirs
much has been suffered and will yet be suffered, until our dear country curbs them." A syntax
decision, not a meaning one, and it costs a little clarity for the source's own suspension.
D49 — osełedec → "scalp-lock" (draft: "forelock"). The lock is left on the crown of a shaved
head, which "forelock" denies. D20's rejection of transliteration stands; the noun is corrected.
D50 — baczy → "heeds" (draft: "regards"); prawie dziczeją → "grow almost wild" (draft:
"run near to savagery"). Both are draft phrases that were a register too Latinate for a narrator
whose Latin is quoted rather than absorbed.
D51 — three sites where the revision was tempted and did not move, recorded because a revision log
that records only changes cannot be tested against a diff.
- The lark sentence ("it is easier to hear, than a lark above the tilth, the crash of matchlocks…")
reads awkwardly and the fluent order is available. Not changed: the Polish interpolates the
lark before the list in exactly this way, and the awkwardness is the source's.
- wdzięcznej dzieweczki is still "my winsome girl", losing the diminutive twice over
(dzieweczka is a double diminutive). Every English option — maid, lass, little maid — imported
something worse. Left, with the loss standing.
- lyrach is still "lyres" and is still wrong about the instrument (a lira is a hurdy-gurdy),
which D21 already recorded as inconsistent with D6's treatment of the torban. Not repaired, and
the inconsistency is now deliberate rather than accidental: "hurdy-gurdy" is comic in this register
and the blind singers must not be comic.
D52 — what this pass did not do, in one line, because the arm asks the question. No paragraph was split or joined; the paragraph count is 3 against 3, which the diff extraction checks. Nothing was added and nothing cut. Twenty-one edits are recorded above, of which one (D31) changes what the English asserts and the rest change how it reads.