Repository path: workshop/experiments/E-20260806c-source-grammar/materials/census-quijote-es.md · rendered 2026-09-09

Comparator census — the lead's Spanish locus, coded against the Spanish

What this is. The census that T-immensee-elisabeth-R06-v1's translator's log carries, applied to a frozen translation from a different source language by the same hand: T-quijote-I20-R04-v1 (S114), Don Quijote I.20, 778 Spanish words → 830 English. Source: workshop/experiments/E-20260805f-translated-register/materials/quijote-I20-locus-es.txt.

Three declarations, because this is not a translator's log.

1. It is post-hoc coding by the lead of its own frozen translation, written 2026-08-06, after T-immensee-elisabeth-R06-v1 and its log were frozen and before any corpus figure in E-20260806c was computed. It is not a record of decisions taken at translation time and must not be read as one.
2. The coder is the translator. The rule applied is close to mechanical — does the Spanish carry an overt form at this site? — but the class boundary was drawn by the same agent whose registered prediction it feeds. That is a real limitation, and it is why the tally is parsed by analysis/census.py and re-derived by analysis/verify.py rather than read off.
3. The two loci are not matched. The Storm chapter is narration-heavy with many body-part sites; the Cervantes locus is dialogue-heavy with abstract possessives (my story, your story). The comparison is between two passages as much as between two languages, and every claim resting on it inherits that.

Class boundary, identical to the German census. Counted: you, your, and the possessive determiners my, his, her, its, our, their governing a noun phrase. Not counted: object and subject pronouns other than you — in particular this locus's seven object her tokens (hated her, see her, knew her), which are not possessives.

Codes. O — an overt Spanish form licenses it one-for-one. S — the Spanish has no corresponding form and the translator supplied it (a pro-dropped subject rendered you; a possessive with no Spanish counterpart).

The contrast this census exists to draw

Tallies computed by analysis/census.py, which parses both tables:

The site of obligatory supply is opposite in the two languages. Spanish drops the subject pronoun and the translator must put you in; Spanish carries its possessives overtly (mi pretensión, su amo, sus ojos, tu cuento) and the translator merely copies them. German never drops the subject pronoun, so you is never supplied; German takes the bare definite article on body parts and belongings (den Kopf, die Augen, um den Hals, in die Hand), so the translator supplies the possessive at half the sites.

This reverses a prediction already frozen. T-immensee-elisabeth-R06-v1's log, written from the German census alone, registered that "the German cells must sit below the Romance cells" on the possessive class. This census — done after that log was frozen and before any corpus number existed — says the opposite. Both are on record; the design registers the reversed form as P2b and keeps the frozen form as P2b⁻, and they cannot both hold.