Two versions of the same agreement

For two months in the autumn of 2023 I covered a bookkeeping post at a union while the person who held it was seconded elsewhere. Expense claim approvals, reporting, both in English and in French, against a collective agreement that governed what could be claimed and by whom.

It was a short placement and I think about it more than the length justifies, because it put me somewhere I had not been before. Not translating a document. Producing both versions of one, and having to approve real money against whichever version was in front of me.

The place the two versions come apart

An English clause allows a reimbursement where the expense is reasonable and adequately supported.

You can read that sentence for years and never be forced to decide anything. Reasonable is doing an enormous amount of quiet work in there. So is supported. The sentence functions perfectly well, because everybody involved has a rough shared sense of the territory and the edge cases arrive rarely enough to be handled one at a time.

Then you have to write it in French, and the vagueness does not survive the crossing. There are several available words and each one draws the boundary somewhere slightly different. One is closer to justified, one to substantiated, one carries an implication about who is doing the judging. You cannot pick neutrally, because there is no neutral option on offer. You have to decide what the English clause meant, and the English clause was written specifically so that nobody would have to.

That is the thing I did not expect. The second language is not a mirror. It is an interrogation.

Order of writing decides who gets the slack

The version written first gets to keep its ambiguity. The version written second has to resolve it.

And then the resolution leaks, because a person approving a claim reads one version, not both. Whichever language they work in is the policy, for them, that day. So a decision made by a drafter under time pressure, choosing between three imperfect words, silently becomes the operative rule for half the organisation and does not exist at all for the other half.

Nobody ratified that. There is no meeting where it happened. It is an artefact of the sequence in which two documents were produced.

What I assumed, and where the argument turns on itself

I went in assuming the job was equivalence. Two texts, the same meaning, and a good producer of them is one who minimises the gap.

That is not really the job. Often the correct output of the exercise is not a better second version, it is a rewritten first one, because the exercise has just demonstrated that the original was never precise enough to be applied consistently by two people. The translation is a test the source failed.

But I have come to be suspicious of my own enthusiasm here, because precision is not free and it is not always the good.

A clause with slack in it survives situations nobody imagined when they wrote it. That is a feature, and it is most of what makes an old agreement usable at all. Force it into two exact statements and you have converted flexibility into a set of decisions, made in advance, by people who could not see the cases the decisions would eventually be applied to. Somebody pays for that later, and it is usually the person with the unusual claim who would have been covered by the vague version and is now plainly outside the specific one.

So a bilingual requirement is not rigour applied for free. It is a machine that converts unstated flexibility into stated commitments, at a rate nobody is measuring.

I still read the French of documents I have already read in English. Not to check the translation. To find out which parts of the thing anybody actually decided.