The subtitle that chose a word

A character in a film says four syllables. The subtitle reads 'I understand'. A different subtitler, working from the same four syllables, wrote 'I see what you have done'. Both are defensible and the film is not the same film.

Subtitling operates under constraints that translation on a page does not. A line must be readable at a given speed, which in practice caps it at about seventy characters for two lines and a few seconds on screen. It must not reveal information before the image does, which forbids some natural word orders. And it must leave room for the viewer to hear the original, since an audience that can partly follow the dialogue will notice and distrust a line that is longer than what was said.

Within those limits, the subtitler is constantly choosing which layer to keep. The literal content, the register, the implication, the joke and the rhythm rarely survive together, and the honest versions of the craft are explicit about the order of priority. Many professionals rank implication above literal content: a line that says the wrong words and produces the right understanding is usually judged better than its opposite.

This is invisible to most viewers, which is the aim, and it makes criticism of subtitles unusually unfair. A viewer who knows both languages sees the discarded layer and reports a mistake. What they are seeing is a decision, made under a character count, about which of four things to lose.

The fairest test is not whether a line matches the words. It is whether a viewer who speaks neither language leaves the cinema with the understanding that the original audience had.