Bilingual Revision as a Diagnostic

I found the continuity error on page two hundred and eleven of my own novel only after reading it in English, a language I had not used to write a single sentence of that book, and only because the English version forced me to look up a name I already thought I knew.

The novel was drafted entirely in French. By the time I reached page two hundred and eleven I had read the opening chapters perhaps a dozen times, in French, checking rhythm, checking dialogue, checking scenes I was still revising. Somewhere around chapter four I had changed a minor character's dead brother's name for a small, sound based reason, one syllable too close to another character's name for comfort in a scene where both were spoken aloud. By chapter nineteen the old name reappeared, unchanged, in a single line I had apparently reread eleven times without noticing, because in French, fluent French, my own French, I was reading what I remembered writing rather than what was actually printed on the page.

Why fluency hides the gap

A fluent reader does not process every word of a familiar sentence individually. Skilled reading relies on prediction, the eye samples a few letters and a practiced brain fills in the rest from context and memory, which is exactly why proofreading your own manuscript in your dominant language is so unreliable, you are not reading the words on the page so much as recognising the sentence you already know is supposed to be there. A name used nine times gets predicted correctly the tenth time even when the tenth instance is wrong, because your brain is running on the pattern of the first nine, not on the actual letters in front of it.

Reading the same page in English, a language I had not drafted the book in, removed that shortcut. I do not have nine prior readings of that character's name printed in English anywhere in my memory. Every instance had to be processed as new information, which is slower, more tiring, and considerably less pleasant than reading my own French prose. It is also the only way I have found to catch the specific class of error that fluent rereading is structurally unable to catch, a fact stated once that quietly stopped holding true somewhere later, without the sentence around it ever announcing the change.

The continuity pass

I now translate roughly, by hand, on paper, the chapters most likely to carry this kind of error, the ones with recurring names, ages, dates, distances, anything stated as a fact early and depended on later. I do not aim for a publishable English version. The translation exists only to force literal processing of every word, and I read it hunting for exactly one thing, whether a fact stated once still holds, every later time it is touched. Everything else, style, rhythm, the quality of a metaphor, I ignore completely on this pass, because those checks belong to a French read and would only distract from the narrow thing an unfluent read is actually good for.

For the two hundred and eleven page novel, the rough translation surfaced four errors in three chapters, the brother's name, a character's age that had shifted by two years between an early draft and a later one without every mention being updated, a distance between two towns that changed depending on which chapter stated it, and a season, spring in one chapter and autumn in another, for what was meant to be a single unbroken week. None of these were visible on a fifth or sixth French rereading. All four surfaced on the first rough pass in English, in a single afternoon.

The obvious objection

You could argue that continuity errors are a known problem with known solutions, a character bible, a spreadsheet of names and ages and dates checked against the manuscript directly, and that translating an entire chapter by hand to catch four facts is a wildly inefficient way to do something a simple table would do in an afternoon with no second language required. I keep a table too, and it catches most of what a table can catch, facts I remembered to enter into it in the first place. What it cannot catch is the fact I never thought to log, because I never noticed it was a fact worth tracking until the moment it contradicted itself two hundred pages later. A table only protects what you already suspected was fragile. The rough translation protects against the error you did not know to suspect, because it denies your fluent memory the shortcut that let the contradiction hide in the first place.

What this diagnostic is not for

I do not run a continuity pass on early drafts, where facts are still supposed to be unstable, or on prose quality, where a French ear is the only ear that matters and an English translation would tell me nothing true about a sentence's music. It is a late tool, reserved for the draft I already believe is finished, applied to exactly the kind of error that finished belief is most likely to be wrong about. The two languages I write in were never going to agree on how a sentence should sound. On whether a fact stated on page forty still holds on page two hundred and eleven, they agree completely, which is exactly why reading in the second one is the only way I have found to catch what the first one keeps politely hiding from me.