Letter 57, to Pammachius, On the Best Method of Translating
What it is about
Jerome had translated a Greek letter of Epiphanius into Latin for a monk who knew no Greek, and was accused of falsifying it because he did not render it word for word. He defends his method: outside scripture, he gives "sense for sense and not word for word" (§5), with Cicero, Horace and Terence as witnesses. Scripture is the exception, because there "even the order of the words is a mystery" (§5). He then shows that the evangelists and apostles themselves quote the Hebrew Bible by sense: Matthew, the Septuagint and the Hebrew differ in words, order, even the speaker (Micah 5:2, Zechariah 11:12–13, Isaiah 7:14), yet agree in meaning. Mark adds "I say unto thee" to talitha cumi for emphasis (§7). He rejects Aquila, who translated the etymologies of words and rendered every Hebrew prefix, the object marker included, as a separate Greek word (§11). He ends by setting his critic's literal rendering of the letter's first line next to his own (§12).
Ideas
- I1. Literalness is a property of the text type, not of the translator. Give each author delta (
tools/translate) a declared mode, sense or literal. A literal author is scored on the order and function classes of the residual, a sense author only on the world level. KJV and ASV are the literal Bibles, WEB and the school are sense authors. This gives one measure per author. - I2. Aquila is the baseline Mova already has: MACULA's glosses read word by word, which beat the Hebrew reader today. Add a stricter "Aquila" baseline that also writes a word for every prefix and for אֵת, and report both. If the reader cannot beat glosses, it should at least beat Aquila on the function-word class. That gives the function-word residual a floor and a ceiling.
- I3. A natural paraphrase test set: the New Testament quotations of Hebrew Bible verses, in the same English Bible (KJV Matthew 2:6 against KJV Micah 5:2). The same concepts appear with a different order, speaker and wording. Measure meta-world agreement (
measure.rs, reader stage) on these pairs against shuffled pairs as the control. They test the reader on sense-level divergence, which verse-aligned Bibles never show. - I4. Jerome's §7 example is an author adding emphasis words that have no source (I say unto thee). Count inserts with no source concept as their own sub-class of the reader residual, apart from lexical substitutions. Additions are the signature of a sense translator and should land in the delta, not in the lexicon.
- I5. §12 contrasts a word-for-word rendering that is correct word by word with a fluent rendering that keeps the sense. In normal mode, report chrF against the world-level agreement (events/entities P/R) for each system. A word-by-word table should win chrF and lose the world; if it wins both, the world measure is too coarse.
For Mova
This is the oldest statement of the split tools/translate makes between the meta-world (sense) and the author delta (the form an author chooses), and of the exception exact mode keeps: scripture whose order matters. I2 and I4 bear on the Hebrew reader (hebrew.rs) and on function words as features (open problem 4). I3 needs only verse references, already in the Bible pools.
Cite
@book{jerome1893letter,
title = {{Letter 57, to Pammachius, On the Best Method of Translating}},
author = {Jerome and Fremantle, W. H.},
year = {1893},
url = {https://en.wikisource.org/wiki/Nicene_and_Post-Nicene_Fathers:_Series_II/Volume_VI/The_Letters_of_St._Jerome/Letter_57}
}This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/jerome-1893-letter-57-pammachius-best/