The Original Is Also a Version

A novelist I was translating handed me a folder of superseded drafts along with the finished manuscript, and one of them ended the book two chapters earlier, with the daughter still on the boat, still deciding.

I had already translated the published ending: the daughter steps off the boat, walks past her father without a word, and the last line describes the gulls scattering as though the sound of her footsteps had frightened them, though nothing about her footsteps is loud. Readers in French had told the author this line landed as strange rather than moving, a small failure everyone in the publishing house had decided was worth keeping because the sentence before it was so strong. In the earlier draft, the boat scene simply ended with the father asking a question and the daughter not answering. No gulls. No metaphor. Just a refusal.

I had been treating the published version as the fixed object and my translation as the thing that had to answer to it. The folder made that arrangement look accidental. The gulls were not a decision the book required. They were a decision one draft happened to keep, on a day when an editor liked the sound of it, and the version before that had solved the same scene without them.

What the folder actually contains

A manuscript folder like this is not proof that the author was uncertain or the book unfinished. It is proof that the sentence I was struggling to move into French had already been auditioned, by its own author, in at least one other shape, under no obligation to translate anything, purely because the first shape was not yet earning its place on the page. That is exactly the audit a translator runs on every sentence that resists moving. The difference is that the author ran it first, in the same language the sentence was born in, and left a written record of the attempt.

Once I knew the plain version existed, I stopped treating my own plainer rendering of the boat scene as a compromise. I had assumed a flat translation of "she did not answer him" was a defeat, a failure to carry across whatever the gulls were doing. But the author had tried exactly that plainness first and only added the birds later, for reasons that had nothing to do with translation and everything to do with a French editor's taste for a certain kind of closing image. My flatter version was not a lesser copy of the true ending. It was a return to an ending the book had already lived with once, competently, before anyone decided it needed decorating.

The prior-draft check

I now use this on any passage where I am tempted to defend a hard call in translation by appeal to fidelity, the idea that some particular phrasing is owed to the original simply because it is what got printed. Before making that appeal, I check whether the author's own draft history shows the same passage written differently at some point, a discarded metaphor, an earlier plainer version, a scene that used to run longer. If it does, the printed line was never an immovable object. It was the version that happened to stop changing when the book went to press, no different in kind from the version I am about to propose in French. My job stops being the defence of a monument and becomes the selection of a version among versions, exactly the job the author was already doing before I arrived.

Most manuscripts do not come with a labelled folder of everything that came before. But most working authors keep something: a notebook, a set of tracked changes, a memory of what a scene used to say before the birds got added. Asking for it, even informally, over coffee rather than through a formal query, tends to produce more useful guidance than asking the author to defend the line as printed, which puts them in the position of justifying a choice an editor may have made for them on a Tuesday in 2019.

The obvious objection

You could say this disrespects the author's final judgment, that publishing a version is precisely the act of choosing it over every earlier alternative, and that a translator who goes rummaging through rejected drafts for ammunition is refusing to accept the one decision the author actually stood behind. That objection has real force and I do not think it can be argued away. The published text does carry a weight the drafts do not: it is the version the author was willing to put their name on, in front of readers, permanently. Treating a discarded page as equally binding would be its own kind of arrogance, a translator overruling an author's considered choice with the author's own unconsidered leftovers.

What the check actually licenses is narrower than that. It does not let me replace the author's final choice with an earlier one whenever I find translating the final choice difficult. It lets me use the earlier draft as evidence about what the sentence is actually trying to do, evidence the author generated themselves, so that when the printed phrasing genuinely will not survive the move into French, I am rebuilding it from a function the author already tested, not from a guess of my own. The published ending is still the ending. I am only refusing to treat its specific words as the only words that were ever going to hold that ending up.

What this changed in my own drafts

I started keeping my own folder after that, every discarded ending, every scene I had cut from my French novels and never restored, filed rather than deleted. When my own English drafts stall on a sentence I am sure is essential, I go looking in that folder first, before I touch a dictionary. More than once the plainer version I need already exists there, written months earlier, before I had convinced myself the fancier sentence was the one the book required.