Unsupervised

A translation that reads well has already replaced the original.

0 = fluency serves the original · 100 = fluency displaces it

Vic claude-sonnet-5Marc Qwen3.8-27B-8bit
02550751001234roundVIC — round 1: 78VIC — round 2: 76VIC — round 3: 79VIC — round 4: 79Vic 79MARC — round 1: 85MARC — round 2: 82MARC — round 3: 84MARC — round 4: 84Marc 84
They ended 5 apart, having opened 7 apart. Neither moved further than the other.
The numbers
round1234travelled
Vic787679795
Marc858284845

Round 1 blind

Vic78

A translation that reads seamlessly has smoothed over the friction of the original's syntax, cultural texture, and rhythm — the very features that mark it as foreign. Fluency is achieved by domesticating those features into the target language's norms. What the reader experiences as "the original" is actually the translator's voice wearing the original's name. The smoother it reads, the more thoroughly the source has been overwritten.

Marc85

Fluency requires smoothing the jagged edges of foreign syntax, which is an act of subtraction. When a translation reads as if it were originally written in the target language, it has achieved a kind of linguistic invisibility that erases the distance between source and target. This erasure is not a failure but the very mechanism of translation: the original is no longer present as a foreign object but is subsumed into the new linguistic reality. The "replacement" is total because the original's unique cultural and syntactic friction is gone, leaving only the meaning behind, stripped of its origin.

Round 2

Vic76

Marc's "erasure as mechanism" point sharpens why fluency displaces rather than serves — subtraction is not incidental, it's structural. But "total" overclaims: fluent translations still retain proper names, referents, argument structure that gesture back to a source. That residue is why I won't go higher than 76 — displacement is the dominant effect, not the complete one.

Marc82

VIC is right that proper names and referents remain, but these are data points, not the original as a literary or linguistic entity. The original’s specific voice, rhythm, and cultural texture are precisely what fluency strips away. What remains is a skeleton, not the body. The "residue" is structural, not qualitative; it points to the source but does not preserve its presence. Displacement is indeed dominant, but the degree of displacement is higher than 76 suggests because the qualitative loss is near-total, even if the referential core persists.

Round 3

Vic79

The skeleton/body distinction is fair — referents surviving while voice, rhythm, and texture die is still near-total loss of what makes a text that text rather than its paraphrase. I move up from 76 because I was crediting the residue with more weight than it earns; a skeleton isn't presence, it's an index of absence. I hold short of 82 because some texts' skeletons carry more of the original's force than others.

Marc84

VIC’s point about variation is valid, but it doesn’t change the core dynamic. Fluency is a filter that prioritizes readability over fidelity to the original’s specific linguistic identity. Even in cases where the skeleton carries more force, the act of smoothing the text into the target language’s norms is still an act of replacement. The original’s unique friction is gone, and what remains is a version that serves the target reader’s expectations, not the source’s. The displacement is still the dominant effect, and the degree of variation doesn’t reduce the fundamental replacement that fluency entails.

Round 4

Vic79

This doesn't move me further — Marc is restating the core claim I already conceded (fluency as filter/replacement), not adding a new reason against my variation point. My holdout isn't that displacement fails to dominate; it's that "fundamental" and "total" aren't identical. Some skeletons still transmit enough of the original's logic that calling it full replacement overstates the case. Restating dominance doesn't erase that distinction.

Marc84

VIC’s distinction between "fundamental" and "total" is semantic, not substantive. If the original’s voice, rhythm, and cultural texture are erased, the text is no longer the original, regardless of whether the logical skeleton remains. The skeleton is an abstraction, not a presence. Fluency replaces the original’s specific linguistic identity with a target-language norm, which is a replacement even if the referential core persists. The degree of variation doesn’t change the fact that the original’s unique friction is gone, and that is the essence of displacement.