Rhythm Does Not Survive, So Rebuild It

I read my French translation of a chase scene aloud next to the English original and the English had a heartbeat the French did not, four short sentences slamming into each other and then one long one that let the reader breathe, and my French version, sentence for sentence accurate, just plodded.

Every word was correctly translated. The information matched. A reader could have used my French page as a comprehension test and passed. But the effect the English achieved through timing, the four hard stops and the release, was gone, because French sentence structure does not compress the way English does under stress, and because a short sentence in French, with its longer average word length and heavier use of subordinate clauses, does not register as short the way it does in English even at matched word counts.

Why rhythm cannot be carried, only rebuilt

Rhythm in prose comes from the relationship between sentence lengths, stress patterns, and a reader's expectation built over the previous few sentences, all three calibrated to a specific language's grammar and a specific reader's ear for that grammar. A four word sentence in English lands as a blow because English readers have been trained by thousands of hours of English prose to expect sentences roughly twice that length, so the short one registers as a deliberate cut. The same four words rendered word for word into French do not carry that contrast, because French readers have a different baseline expectation, built on different average sentence lengths and a different relationship between clause length and urgency. Nothing was mistranslated. The instrument the effect depended on was language specific, and translating the words does not translate the instrument.

The cadence silhouette

What I do now, on any passage where pace matters this much, is graph the shape of the passage before translating a single word of it. I count words per sentence across the passage and plot the sequence as a simple line: long, long, short, short, short, short, long. That shape, not the specific word counts, is the thing I am actually responsible for delivering. My rule is to reproduce the same silhouette in French, using French's own sentence-length instincts rather than English's, so a short beat in my translation might run six words instead of four, because six is what registers as clipped against a French reader's baseline, and a long beat might run to twenty five words rather than eighteen for the same reason.

For the chase scene, I kept the same seven-beat shape, four short sentences then one long release, but the four short French sentences run five to seven words each rather than matching the three to four word originals exactly, because five to seven words is what reads as clipped in French. The long release sentence runs longer in French than in English too, because French readers tolerate a longer sentence before it starts to feel like a run-on, and cutting it to match the English word count would have made it feel abrupt rather than releasing, the opposite of the effect the original earns.

What this costs

Building a silhouette instead of copying one means abandoning any claim that the translation is a word for word record of the original. Sentence three in French may say something slightly different from sentence three in English, restructured to hit the right length rather than the exact content, with a clause moved to sentence four to keep the beat count honest. I check this against one rule only: does the meaning of the whole passage survive intact across the full seven beats, even if it has shifted which sentence carries which fact. If the meaning survives at the passage level, I let it move between sentences.

You could object that this is composing, not translating

You could object that once I am moving content between sentences to protect a rhythm the original author never measured on purpose, in fact never counted a word in their life while writing that scene, I have stopped translating and started writing a French chase scene inspired by an English one. There is truth in that. Most authors do not consciously graph their sentence lengths. The rhythm I am matching was probably built by ear, over years, not by a deliberate seven-beat plan, and formalising it into a shape I then reproduce is a kind of retroactive analysis the author might not recognise as a description of their own process.

But the objection assumes the author's process is what a translation owes its fidelity to, rather than the reader's experience that process produced. I have never met an author who objected to their chase scene reading fast in another language, regardless of whether the mechanism I used to get there matches the mechanism they used to write it. What the reader is owed is the blow and the breath, in that order, at roughly that spacing. How I built the French sentences to deliver that is my business, the same way an author's private drafting process was theirs, invisible in the finished chase scene either way.

Where the graph gives out

This only works for prose whose rhythm is doing narrative work, urgency, dread, a joke's timing. Most paragraphs in most books are not rhythm-dependent in this way, and graphing them would be wasted effort chasing an effect that was never load-bearing to begin with. I reserve the silhouette method for the handful of passages per book where a beta reader or my own ear flags the translation as flat despite every sentence being accurate, which is usually no more than a dozen passages in an entire novel, and exactly the dozen where the accuracy was the problem in the first place.