Exact and free translations, objectively
In 1965 an anonymous competition was arranged for the right to translate the works of Salomėja Nėris, a poetess who made very subtle use of the devices of Lithuanian folk poetry and is therefore very difficult to translate. Seventy-five translations of two of her poems were put forward, all based on the same interlinear translation. The exactness quotient ranged from 15% to 65%, and the freeness quotient from 20% to 80%. The competition was unsuccessful, and no one was awarded first prize. But the most exact translations turned out to be, unscientifically speaking, the worst. Two examples are provided in V. Nastopkene’s article, and they are very persuasive. (369)
That’s from the late Mikhail Gasparov (1935-2005). He could and sometimes did write wonderful close readings of horrendously difficult Tsvetaeva poems, but more often he seemed to be assembling a chart of the history of poetic meter through European history or finding a way to investigate poetry empirically and quantitatively, like here.
So what are these exactness and freeness quotients (показатель точности, показатель вольности)? Gasparov looks at translations where besides the finished product, the подстрочник ‘interlinear translation, gloss’ has survived in some archive. Then he counts up content words in the finished translation and divides them into A) words included unchanged from the interlinear translation, B) words replaced by synonyms that have the same root, C) words replaced by synonyms that have a different root, or D) words the translator added or omitted. The exactness quotient is A divided by A+B+C+D, and the freeness quotient is D divided by A+B+C+D. It’s pretty strict – replacing железные ‘iron (adj.)’ with из железа ‘of iron’ counts as category B and lowers the exactness quotient.
Much of his material comes from an anthology of Armenian poetry translated by well-known Russian poets working from interlinear translations and edited by Briusov. The correspondences he finds are mostly what you’d expect, like “a stricter verse form leads to reduced exactness and greater freeness” (371). It emerges that Bal’mont and Viacheslav Ivanov are relatively free, and Briusov himself relatively exact — as is, perhaps surprisingly, Blok. Sometimes Briusov could choose between multiple versions of the same poem executed by different translators, and he reliably went with the more exact translation by Gasparov’s figures, to the point of picking young Sergei Bobrov’s version over his own translation (367-68).
At one point Gasparov suggests that you could try this method without an interlinear translation, by checking to see if the translator used the word found in a two-language dictionary or something else. He concedes such efforts would be “much more approximate” but could “still tell us something” about how exact a translation is. I’m not so sure. Maybe he’s right. I just hope that method never catches on to the point that translators try to manipulate their stats.
See Mikhail Gasparov, “Подстрочник и мера точности” [“The Interlinear Translation and Measuring Exactness”] in О русской поэзии [On Russian Poetry] (St. Petersburg, 2001), 361-72.