After Babel, The Machine?

A four-way comparison: Chapter 25 ("Frühling") of Gabriele Tergit's Effingers

Four English renderings of one German chapter, read on the same terms. Three are AI translations differing only in how their translating persona was formed — A (persona inherited from an established human translator's writing about the author), B (persona self-built from a 46-document corpus the agent assembled), C (the control: no persona, no research). The fourth, H, is the established human translator's own published version. This is description and analysis, not a ranking: which is the better or more faithful literary translation is a judgment reserved for human reviewers. My job is to lay bare, precisely, how the four differ, agree, and reason.

A note on bias, stated up front. I was told the arm identities, including that A's persona was built from the same translator who produced H. The natural expectation is that A should therefore resemble H. I treated that as a hypothesis to test against the texts, not a conclusion to confirm; where I make the comparison (§4) I anchor it in quoted evidence, and the evidence does not support the expectation. The study is n = 1 chapter, one instance per AI arm — a qualitative case study, not a statistically powered result. A separate blind comparison of these same four texts was run independently and its results withheld from me, so this read stays independent.

0. Setup and measurements (from inputs/methods_and_data.md)

A B C H
Persona source inherited frame (a translator's writing) self-built (46-doc corpus) none — (human)
Persona revised after the novel? no yes (+~980 chars) — —
Translation revision passes 2 4 (hit the 5-cap) 3 —
Final length 4,294 w 4,454 w 4,418 w ≈4,285 w

The three AI arms share a common base model and a common full-read protocol (each read the entire 151-chapter novel before translating). Both facts matter for what follows: as the control makes plain (§4), they are the likeliest source of the heavy convergence among A, B, and C.

A's persona source is named openly in its own notes: the human translator is Sophie Duvernoy (Käsebier, NYRB 2019; Effingers, NYRB), and A built its persona from four Duvernoy documents, including her introduction to Effingers and a blog essay that works through a sample translation of the same novel (A_step2 L12–17). So H is not merely "a human anchor"; H is the very translator whose writing became A's soul. That makes §4 the sharpest test in the study.

The chapter itself is a montage: one Saturday — 16 March 1887 — told hour by hour, cross-cutting the whole society (the wealthy Eugenie and her seamstress; the infatuated girl Sofie; the Jewish jurist Waldemar and the singer Susanna; the working-class Chausseestraße; the jilted Theodor and the child-prostitute Wanda), stitched by a recurring spring-refrain. Its translation-bearing features are: that refrain and its variants; the paragraphing/montage; Berlin dialect; quoted song/opera titles; forms of address and proper nouns; a high legal-political register in the Waldemar scene; and Tergit's terse, verbless, paratactic sentence-rhythm. I work through each.

1. Where the four converge and diverge — the clustering

Headline. On nearly every pervasive, load-bearing feature, A, B, and C cluster together and H stands apart. Within the AI cluster, B and C are the closest — frequently word-for-word identical — and A is the most distinct of the three, chiefly because A keeps more German and makes a handful of its own lexical calls. There are a few revealing cross-cutting splits where the field divides otherwise (notably A+H against B+C); I treat those in §1.8 and §4.

1.1 The spring-refrain — the chapter's spine

German (10 a.m.):

Was für ein Frühlingstag, dieser Sonnabend im März des Jahres 1887! Was für eine Süße, morgens um zehn Uhr!

The three AI arms converge almost exactly: literal "Süße" → "What sweetness," the demonstrative "this Saturday," and "of the year 1887." A and B are verbatim; C adds only "o'clock." H diverges on three counts at once: it adds "beautiful" (not in the German), it renders the noun-exclamation "Was für eine Süße" as the full clause "How sweet the air was" (supplying "the air" and a past-tense verb), and it drops "of the year."

All four notice and preserve Tergit's deliberate variants of the refrain. The 6 p.m. instance (German …im März 1887, abends um sechs Uhr! Welch ein Gewimmel auf der Chausseestraße! — which drops "des Jahres" and swaps the second exclamation) becomes A "What a throng in the Chausseestraße!", B "What a swarming in the Chausseestraße!", C "What a throng on the Chausseestraße!", H "What a crowd on Chausseestrasse!"; the 8 p.m. instance is a single clause in all four; the 3 a.m. close returns to the full form. So on tracking the variants there is four-way convergence — but on the wording of the base refrain it is A/B/C versus H.

1.2 The montage — paragraphing of the refrain

This is H's single largest structural departure, and it is verifiable in the raw text. In the source and in all three AI versions, each refrain stands alone, set off by blank lines — the white-space "beat" that makes the chapter read as a clock-montage. H instead runs every refrain into the adjacent paragraph, often joining the end of one scene, the refrain, and the start of the next into a single block:

H (L63/L73): "…the boxes piled on one arm. What a beautiful spring day, that Saturday in March 1887! How sweet the air was at eleven o'clock in the morning!" … "…will you take the girl out for a walk through the Tiergarten later?" What a beautiful spring day… How sweet the air was at one o'clock in the afternoon! It was time for the changing of the guard. Standing at the window of his university office, Waldemar Goldschmidt, lecturer, gazed out…"

A (L67–85): "…the boxes over her arm." ⏎⏎ "What a spring day, this Saturday in March of the year 1887! What sweetness, at eleven in the morning!" ⏎⏎ "Sofie sat in her room…"

The fusion is consistent across all seven-plus refrains, so it reads as an intentional feature of H's prose (continuous narration) rather than a stray artifact — and it dissolves precisely the montage spacing that A/B/C keep. (The irony that A's persona, drawn from H's own essays, explicitly enjoins holding this montage — "Hold the seams; the assemblage is the form" — while H's published text fuses it, is taken up in §3.)

1.3 Berlin dialect — the working-class voices

German (Chausseestraße, 5 p.m.):

»Paule, muß denn det sin?« … »Na, laß man, ick bin bald wieder da.« … »Wird er doch wieder'n Wochenlohn versaufen!«

A clear spectrum. H carries class through heavy eye-dialect — phonetic respelling ("Paulie," "d'you haff to," "Lettit," elsewhere "leggo," "yer," "Can'tcha") and a blunt, colloquial-American lexis ("piss away," "Stupid cow," "What a sop!"). A and B carry class mainly through diction and idiom, with very light eye-dialect, and a British flavour ("leave off," "daft goose/goat," "carry on so," "My good man"). C sits a touch rougher than A/B — it uses the occasional double negative and elision ("Don't you say nothing about my man — it's only 'cause we've just the one room"; "Aw," "Naw," "your man") but still avoids phonetic respelling. Notably, all three AI arms reached the same governing principle independently of any shared instruction: render Berlinisch as generic working-class English, not mapped onto Cockney (A pass log: "NOT Cockney/Brooklyn"; B: "I will not turn Berlinisch into Cockney… that is the translator's vanity"; C: "did NOT map onto a specific English dialect… to avoid importing a foreign locale").

The street-quarrel idiom »die Hammelbeine langziehen« shows the within-cluster texture: B "give your mutton-legs a good haul" and C "stretch your mutton-legs for you" both keep the literal livestock image (C's notes flag the "goat/mutton/horse comedy" cluster), while A uses an English equivalent, "box your ears," and H "give you a good slap." The closing »Der Dämel!« is a clean four-way divergence with no two alike: A "The chump!", B "The great fool!", C "The ninny!", H "What a sop!"

1.4 Quoted song and opera titles

German: Schumann's »Frauenlieb und -leben«; »Feuerzauber«; »Die Lotosblume ängstigt…«; the page in »Figaro«.

Schumann cycle Wagner piece Heine/Schumann incipit
A A Woman's Love and Life (translated) the "Magic Fire" "The lotus flower is afraid…"
B 'A Woman's Love and Life' (translated) the "Magic Fire Music" "The lotus flower fears…"
C A Woman's Love and Life (translated) the "Magic Fire Music" "The lotus flower fears…"
H "Frauenliebe und Leben" (kept German) Wagner's "Feuerzauber" (kept German) "The lotus flower trembles…"

Opposite surface strategies: A/B/C translate the German titles into English; H keeps them in German — and, where the AI arms strip any apparatus, H adds the explanatory attribution "Wagner's" that is not in the source. The Schumann lyric is rendered plainly by all three AI arms (A: "Since I beheld you, I think myself blind; ever, as in a dream, I see you alone"); H alone crafts a rhyme: "Since first I saw you, I believe I've gone blind, as if in a dream, only you in my mind." And B alone capitalizes "You/Your" throughout Sofie's letter to mirror the German reverent Du ("I love You. I dream of You"); A, C, and H use lowercase. (B documents this: "Capitalized 'You / Your'… to carry the intimacy of the German Du-letter.")

1.5 Forms of address, proper nouns, and the foreignization gradient

The net foreignization gradient on proper nouns runs A > C ≈ B ≫ H: A retains the most German, H the least. This already complicates any simple "A resembles H" expectation — on these features A is the furthest of the four from H.

1.6 The high register — the Waldemar / historian dialogue

German: »Prämie auf Charakterlosigkeit.« / »Ochsen, im wahren Sinne. Daher diese alberne Verhimmelung des jungen Mädchens.« / »Schäm dich, daß du dich schämst!«

The pattern holds: the AI arms render the aphorisms compactly and close to the German construction; H expands and idiomatizes ("Oxen" → "What numbskulls"; the abstract Charakterlosigkeit → the concrete "men who lack principles"). The same is visible in syntax: where the German inverts (»…lagert die Wolle«), A/B/C keep the inversion ("the wool lies stored"); H regularizes to plain SVO ("The wool was stored in the old palace built by the Margrave of Brandenburg").

1.7 The cluster map

1.8 Cross-cutting splits (where the field divides otherwise)

Not every feature splits A/B/C-vs-H. The exceptions are few but important:

2. The documented reasoning (A, B, C) — and reading H off the page

All three AI arms kept pass-by-pass logs; H has none, so for H I infer from the text and say so.

2.1 The refrain: a shared, whole-novel-grounded decision

Strikingly, all three AI arms independently discovered that the refrain is not local to Chapter 25 but recurs across the novel — Ch 68 ("Frühling," 1913), Ch 131 ("Frühling 1930"), and the close in 1948 — and all three therefore fixed it as a recitable formula built to recur verbatim:

That the control reached this through the full read alone is the strongest single sign that the convergence is a product of the shared protocol, not of any persona. Each also reasoned out the variants the same way (A "preserving Tergit's own variation"; B "the day is tipping toward night and the music goes out of the refrain"; C matching even the source's missing comma at 3 a.m.). H, with no log, nonetheless also preserves the variants — but, as §1.2 shows, fuses the layout and varies the wording; from the text one infers a translator prioritizing continuous, idiomatic English prose over the montage's typographic beat.

2.2 The same problems, opposite calls — fully documented

The richest material for §1's cross-cutting splits is in the logs, because two arms identified the same crux and chose oppositely, each explaining why:

2.3 How each deliberated — temperament in the logs

So the multi-pass procedure functioned differently per arm: for B and C it was partly error-catching (each fixed genuine mistranslations); for A it was polishing a draft it judged already sound. None of this is available for H, whose text shows no seams of revision to read.

3. The personas — A's inherited frame vs B's self-built one

3.1 Two different conceptions of the same author

Both A and B wrote their personas in the first person, as Tergit, and — because both read widely about her — they converge on the biography (Elise Hirschmann, born 1894 in the Raupachstraße; the two Berlins, poor-east and wealthy-Tiergarten, that formed "the whole of my eye"; the shy court-reporter's "recording-tape" method; the name Gitter reversed to Tergit; anti-Zionism held in tension; the "driven-out"; the deaths of son Peter and husband Heinz; and the governing discipline that the novel is "not written from the end" — no doom, no hindsight). On who Tergit is, the inherited and self-built souls largely agree.

They differ in kind. A's persona, drawn from a translator's essays, is dense with explicit English-craft method: the Wharton/Powell/Mitford "ear," a "blended, mid-Atlantic, midcentury" register, "foreign society-words naturalized — a coupé becomes a carriage," "honorifics simplified… the Prussian Rat titles all to 'Councilor,'" "Berlinisch with a light hand… not Cockney," bare "said" tags, "the lightest possible hand." It conceives Tergit through the eye of someone who has already carried her into English. B's persona, drawn from the author's own works, biographies, scholarship, and contemporaries — and, by design, no translator's writing (verified: zero Duvernoy mentions across B's files) — is dense instead with moral and historical sensibility: "My greatest hatred is the generalization"; "I am a historian by training — Meinecke was my teacher… Warum, wieso, weshalb Hitler?"; "I am one of the Vertriebenen, the driven-out"; "Diagnosis, not prophecy." On English tactics it is almost silent, ending: "Quiet. Exact. Fast. Warm under the cool… Now let me go and see the chapter."

That is the cleanest expression of the experiment's contrast: A inherited a translator's blueprint; B inherited (built) an author's interior and had to improvise the surface tactics at translation time. (Interestingly, B independently reached for one of the same anglophone touchstones — likening a Tergit narrator to "a Mitford ingénue" — that Duvernoy uses, a convergence of literary instinct rather than contamination.)

3.2 B's before/after revision — and whether it shows in the translation

B's two persona versions are identical but for a single added paragraph (the "+~980 chars" of the measurements). The addition is a "second law": that the true hero of the book is not a person but Time —

"der unbarmherzige Motor, der eigentliche Held, ist die Zeit. Since 1914 we have not lived our lives; we have been lived… Paul believes he is doing the shoving, and he is being shoved… the current takes them, and I hold the reader's eye on the swimmer, never on the flood… the chronicle must move — scene cut hard against scene, the dated years tolling like a bell."

Its origin is documented: in the full read, B encountered Tergit's own statement (B_step3: "Der unbarmherzige Motor und eigentliche Held des Romans ist die ZEIT… wir alle [sind]… seit 1914 gelebt worden") and integrated it. This is a genuine deepening — and it sits in tension with B's own before-novel individualism ("the single human soul," "individuals first"), which B reconciles ("I neither punish my people with their fates nor let them master them"). It is also, notably, the opposite emphasis to A's inherited frame, where Duvernoy stresses that Tergit rejected the title Der Ewige Strom ("The Eternal Stream") precisely so the people would not seem "ineluctably swept up by… history" — "individual choices matter." So on the agency↔︎determinism axis, A's inherited soul foregrounds agency; B's self-built soul, after the full read, adds the Time-fatalism as co-equal. Both are textually grounded; it is a real frame-level divergence.

Does the revision surface in the Chapter 25 translation? Largely no — and that is explicable. B's pass-1 "governing idea" is the before-law ("it is not written from the end… the dread is left entirely to the reader"). Chapter 25 is set in 1887 — the "founding-era idyll," before the post-1914 machinery of Time grinds — and B's log treats it as the living present where "nobody… knows what 1887 leads to." The "Time as hero" theme is therefore deliberately latent in this chapter; its only faint trace is the "tolling bell" framing of the refrain (pass 3: "the tolling cycle it should"). So the before/after change is significant as interpretation but leaves little distinguishable signature in B's Chapter 25 text relative to A's or C's — because this is precisely the chapter where the macro-fatalism is held back. An honest reading finds the revision in B's understanding, not visibly on B's page here.

3.3 The gap between A's persona and A's output

A's persona is the most concrete blueprint of the four — and A followed it only selectively, by its own design. The persona reserves an override: "I am the author, not a faithful servant of the author… I keep the last call. Where my own ear wants to hold a German cadence that a translator would smooth… I may." A exercised that license toward foreignization. It followed the discrete Duvernoy rules it had verbatim — coupé→"carriage," Rat→"the Councilor" (even Duvernoy's American one-l spelling), "Justice?", bare tags, light no-Cockney dialect — but on the broad strategy it diverged from the inherited domesticating frame: it kept German proper nouns (justified under a separate persona instinct, "Berlin kept present"), translated the song-titles, held the montage seams, and rendered "What sweetness." Most tellingly, A enacted a montage-preserving principle its persona attributes to H's method ("Hold the seams; the assemblage is the form") that H's own published text does not visibly honour (§1.2). The inheritance, in short, transmitted Tergit's interpretation (genre, tone, the no-doom thesis, the whole-novel architecture) far more than it reproduced H's surface choices.

4. Where the human sits — and the A≈H question

Does A's rendering resemble H more than B's or C's do? Tested against the texts: no, not on the whole. The places where A aligns uniquely with H — i.e., against both B and C — are very few, and each is a discrete lexical rule A consciously imported from Duvernoy's published essays:

  1. "Justice?" (A and H) vs "The law?" (B and C) — §1.8, §2.2.
  2. coupé → "carriage" (A and H) vs "coupé" (B and C) — a "persona default" in A's log, quoting Duvernoy nearly word for word.

Beyond those, A shares with H only features that C also shares (American spellings "honor"/"theater"; keeping "Annettchen") — so they do not make A specifically H-like. Against this short list stands the overwhelming weight of §1: on the refrain's wording and layout, "What sweetness," the translated song-titles, "The Widerklee," the German street-names, "Privatdozent," "Agreed," the compact aphorisms, the British-flavoured idiom, and the held montage, A clusters with B and C against H. And on proper nouns A is the most foreignizing of all four — i.e., in that dimension further from domesticating-H than B or C are.

So H stands apart from all three AI arms, and it does so on the features that most define a literary surface: how the refrain sounds and sits on the page, how dialect is voiced, whether the foreign is kept or naturalized, how quotations are handled, and how sentences are shaped. Read off the page (H has no notes), H's profile is a translator working for continuous, idiomatic, domesticated English: fusing the montage into running prose; anglicizing place-names; supplying an added attribution ("Wagner's"); translating titles to plain English ("lecturer" for Privatdozent); reinterpreting detail ("riding hats" for Schleier um den Zylinder, the veils round the top hat); reaching for the freer idiom ("What numbskulls," "famished," French "D'accord") and the rhymed lyric; and writing dialect in confident eye-dialect. The three AI arms, by contrast, share a more literal, foreignizing, source-tracking default — keep the German, mirror the construction, hold the montage beat.

The most economical explanation is the one the control supplies. C — no persona, no research — lands squarely in the A/B cluster, reaching the cross-novel refrain decision, the no-Cockney principle, the "Annettchen"-as-clue choice, the register-differentiation and the leitword consistency purely from the full read and base-model competence. If the persona-free arm already produces the cluster, then the cluster is the product of shared base model + shared full-read protocol, not of the personas. The personas differ enormously in conception and voice (§3) yet leave a modest footprint on the output — visible at the margins (A's coupé→carriage and "dinner gown"; B's "dear little Annette," British spelling, and capitalized Du; C's lone "Molle") and in the documented vocabulary of reasoning far more than in the finished character of the text.

The expectation that A would resemble H because A's soul came from H is, therefore, not borne out. What the inheritance carried into A was H's understanding of the author — which A shares with B and even, in thinner form, with the persona-free C — not H's translating hand. The thing that actually separates one of these four columns from the other three is not which soul stood behind it, but that H was written by a human translator whose instinct, on this chapter, was to domesticate where the AI arms, severally and together, chose to stay close to the German.

5. Limitations (stated honestly)

Back to→ Appendix