Should Translators Favor the Source Language or the Target Language?
My first big break as a translator came while working at an English school in Hiroshima. We were contacted by an author who needed transcripts of conversations with A-bomb survivors translated into English, and I was asked to take part. What struck me immediately was the difference between my textbook Japanese and these real-world samples with their stops and starts, emotional tone, and unfamiliar expressions from a bygone era. I quickly realized that the best I could do was to read for gist, try to express this clearly in English, get a Japanese native speaker to correct any omissions or misunderstandings, and then try again, striving for what the speaker intended through a process of successive approximations. I like to think of this as target-language bias, because the intelligibility of the final product took precedence over any concerns about retaining the original phrasing.
I later accepted a position at a translation agency, where I learned that reading for gist is not the way to please clients who are generally expecting to see the rhythms and structure of their text mirrored in the English they are purchasing. I call this expectation that everything from the original should somehow map to the target document a source-language bias.
To my mind, this insistence on a one-to-one correspondence between Japanese and English is sometimes carried too far, such as when the same word in the original is always translated using the same word in English, lest the addition of synonyms imply a nonexistent variability in the source. To be fair, when translating in technical fields, especially when deadlines are looming and one has limited experience with the topic, consistency should trump any concerns over the aesthetic appeal of the final product.
Other aspects of the profession also encourage this source-language bias. At many agencies, once a text had been translated by a native speaker, it is passed to a Japanese checker, who then typically flags anything out of the ordinary (read idiomatic) for further discussion. Given real-world time constraints, one quickly becomes conditioned by this reality to avoid creative flourishes and the concomitant hassle factor, however much they would increase readability.
Source-language bias is also reinforced by the now ubiquitous use of computer aided translation (CAT) tools. These programs parse the text into individual sentences, and create a database (translation memory or TM) of corresponding translations as one works through the document. Such TMs can then be applied to other projects to automatically search for and replace previously translated elements, thus ensuring consistency and reducing the workload, which is advantageous even if the final product can seem lifeless due to the focus on sentence-level parsing.
Reflecting on all this after decades in the business, I find that after years of being structure obsessed, I am gradually circling back to my roots and a more target-language based style. Part of this is a shift in the projects that now interest me, such as releasing previously untranslated essays and poetry by a-bomb survivors. But another element is the unstoppable rise of machine translation, which has created a niche in the business for people who can take text that varies between gibberish and merely stilted, and polish it quickly until it rings true. As in so many fields, I think our future as wetware in an AI world can only be ensured by adding unique value to increasingly automated products, In our case, this means drawing on our creative writing skills as well as our aesthetic sensitivity borne of a lifetime as native speakers.
