Translation is usually discussed as a problem of accuracy, which is the least interesting thing about it.
Any competent translator can render the of a sentence.
What resists transfer is everything that signals who is speaking and from where.
, , period and are carried by choices that have no equivalent in the receiving language.
A character marked as rural and uneducated in one language cannot simply be made rural and uneducated in another.
Every available equivalent carries its own history, and the translator inherits that history whether or not it fits.
The two classic strategies are conventionally described as and .
Domestication makes the text read as though written in the target language, which conceals the labour and the distance.
Foreignisation keeps the strangeness visible, which respects the original and risks producing an unreadable book.
Neither is a rule, and every serious translator alternates between them within a single page.
The economics of the profession shape the result as much as any theory.
Literary translators are paid by the word or by a and rarely receive a share of sales.
A translation that takes three years to do well earns less per hour than one done in eight months.
Naming the translator on the cover has become a campaign for exactly this reason.
is the precondition for a reputation, and reputation is what eventually permits a better contract.
Machine translation has changed the lower end of the market and left the literary end largely untouched.
A model trained on average usage produces average prose, which is precisely the failure that matters here.
The translator's task was never to find the usual word, but to decide when the unusual one is the honest choice.