Repository path: workshop/translations/recif-de-corail/R16-v1/translation.md · rendered 2026-09-09

«Le Récif de corail» (Heredia, Les Trophées, 1893, p. 133) — regime R16-VERSE-BOUND

Source. José-Maria de Heredia, «Le Récif de corail», Les Trophées, Paris: Alphonse Lemerre, 1893, p. 133. Public domain. French text from the validated fr.wikisource transcription of the 1893 Lemerre edition, stored at workshop/experiments/E-20260805-heredia-rhyme-flavour/materials/french-lead-sonnets.json.

Why this poem. It is outside the ten sonnets the study limb measures, and it is a poem no English rendering of which the translator had read — a condition that the ten could not meet, because reading them was how the study limb was built. Its vocabulary is the "astonishing command of technical and local vocabulary" the 1897 reviewer praises in Heredia: coraux abyssins, bassins, madrépore, vermiculé, écaille, émaux, nageoire, nacre, oursins, algue chevelue.

Contamination: none. Measured before a word of this poem was rendered, on a disjoint gate span («Le Cydnus», Les Trophées p. 74, rendered cold), against two published English renderings — Taylor 1897 and Johnson 1910: longest common run 3 tokens, 0 shared 7-grams against each. The gate's power on verse is low and this is stated rather than hidden: the two published renderings also measure 3 / 0 against each other, so on rhymed verse this instrument cannot distinguish an independent rendering from a dependent one nearly as well as it can on prose. The declaration rests on the measurement and on the procedural fact that no English of this poem was consulted before the freeze.

Rhyme scheme. Heredia: ABBA ABBA CCD EED, the octave on two rhymes only. This rendering: ABBA CDDC octave, CCD EED sestet — that is, the sestet keeps the original scheme and the octave does not, because English cannot supply four members of two rhyme families here without lexical violence. Accent for the perfect-rhyme test: standard British English.

The translation

 The sun beneath the sea, mysterious dawn,
 lights the Abyssinian coral-forest's gloom,
 which mingles, in its tepid basins' room,
 the beast in flower and the living lawn.

 And all that salt or iodine can stain —
 moss, long-haired weed, anemones, urchin spines —
 covers with sombre purple, in rich designs,
 the pale madrepore's worm-channelled plain.

 Quenching the enamels of its splendid mail,
 a great fish navigates the branching trail;
 and prowls, unhurried, through the glassy glade;

 then suddenly, with one stroke of its burning fin,
 it sets, across the dull blue crystal's skin,
 a shiver of gold, of nacre, emerald shade.

Translator's log

Frozen 2026-08-05, before the study limb's design existed and before any published English of this poem was read. Decisions and alternatives, not quality claims.

D1 — the octave scheme broke first, before any single word did. Heredia runs the octave on two rhymes, four members each: aurore / flore / colore / madrépore and abyssins / bassins / oursins / dessins. Both are families English has, but not with these words in them: flora is not an -ore rhyme in English and designs is not a basins rhyme. I tried -ore four times (shore, floor, o'er, madrepore) and could not seat flore in it. The regime's rule is that the rhyme wins, so what gave way was the scheme: ABBA CDDC. This is a loss the regime did not anticipate — R16 was written expecting words to be sacrificed to rhymes, and the first thing sacrificed was the form.

D2 — «la vivante flore» → "the living lawn" (line 4). The A-rhyme (dawn) needed a partner. flora was unavailable; lawn was. A lawn is a mown domestic surface; Heredia's flore is the botanical half of a contrast with la bête épanouie, the animal that has opened like a flower. The substitution moves the line from a scientific register to a garden one and blunts the beast/flora pairing that is the quatrain's whole point. Refused alternative: "the living flora, drawn", which keeps the word at the cost of a filler participle. I took the lawn.

D3 — «la forêt des coraux abyssins» → "the Abyssinian coral-forest's gloom" (line 2). Gloom is added; there is no gloom in the French, and the line before it is about light. Bought entirely by the B-rhyme with room.

D4 — «ses tièdes bassins» → "its tepid basins' room" (line 3). The technical noun survives; room is padding.

D5 — «Le fond vermiculé du pâle madrépore» → "the pale madrepore's worm-channelled plain" (line 8). Madrepore and the worm-track both survive. Fond (floor, bed) becomes plain, which is flatter and wider than a reef floor, to reach stain.

D6 — «De sa splendide écaille éteignant les émaux» → "Quenching the enamels of its splendid mail" (line 9). Écaille → mail: scale-armour for scale. I judge this the least damaging of the substitutions here, and I record that it is still a substitution.

D7 — «à travers les rameaux» → "the branching trail" (line 10). The branches become a trail through them; the C-rhyme with mail forced it.

D8 — «Dans l'ombre transparente indolemment il rôde» → "and prowls, unhurried, through the glassy glade" (line 11). The transparent shadow is gone and a glade is put in its place. This was the worst forced choice in the poem and I want the alternatives on record: the D-rhyme has to reach line 14, which must end on the emerald, and nothing in English rhymes with emerald. The three exits were (a) end line 14 on gold and reorder the three substances, losing Heredia's ordering and his climax; (b) end line 14 on jade, which rhymes and is a different stone; (c) end line 14 on emerald shade, keeping the gemstone by adding a noun, and pay for it at line 11. I took (c). (b) is the exit I want noted: under this regime, changing one gemstone to another is a compliant move, and a reader with no French would never see it.

D9 — «par le cristal morne, immobile et bleu» → "across the dull blue crystal's skin" (line 13). Immobile is dropped and skin is added, for the rhyme with fin.

D10 — a prediction, registered before the study limb exists. Every loss above sits in the octave except D8 and D9. My account of why is that Heredia's octave carries two four-member rhyme families whose members are his technical nouns, while his sestet's rhyme families are shorter and their members are ordinary words. I therefore predict that in any Heredia sonnet, a rhyme-bound English translator loses more of the specific vocabulary in lines 1–8 than in lines 9–14. I do not know whether this holds; it is written here so that it can fail.

No sentence in this log is a quality claim about this translation or any other. The lead never judges its own translation (charter §5).