side from a few volunteer projects, my start in the translation profession was with a private school in Hiroshima where I was employed as a teacher. The English department had taken on responsibility for a visiting author who was writing a novel about the atomic bombing. She had amassed a series of transcripts taken from interviews in Japanese with atomic bomb victims, and she came to us looking for a clean English version. The project was to take 6 months. Although the department accepted the work, this was its first venture into translation and there was no one available to deal with the task on a full-time basis. Knowing of my interest in becoming a translator, I was given a portion of the responsibility, and eventually the better part of the material ended up on my desk.

The conservative emphasis on preserving as much of the structure of the Japanese as possible in the final copy yielded work that while not always aesthetically pleasing, could rarely be faulted for accuracy.

The opportunity was exciting and the material meaningful, but the pragmatics were daunting. The problem was the mismatch between my sterile textbook Japanese and that in text in front of me, with its fits and starts, colloquial grammar, incomplete sentences and emotional tone. Under pressure for time and not wanting to embarrass the school, I ended up organizing a kind of team-translation situation with several of my Japanese study partners. I would read for transcripts gist and discuss with them what I felt the speaker was trying to convey. They would correct my intuition, which tended to stray from the text particularly in the more idiomatic passages. I would then take my notes from these meetings, and through comparison with the original, arrive at an English equivalent at a later date. It could be said that the structure of the product we ultimately delivered reflected the choice to focus on conveying the main ideas as clearly as possible in colloquial English, rather than preserving the structure of the transcripts. I think of this as target language driven translation because of the distance from the source text during the process of composition, particularly during the production of the final draft.

A full-time opportunity eventually opened up at one of the larger translation agencies in Hiroshima. Accepting this offer provided my first exposure to the work methods of professional translators, and was the start of a valuable and sometimes painful exposure to a completely different perspective on dealing with text. The key aspect of the approach was what I refer to as source-based translation, because of the degree to which the terminology and rhythms of the final product were constrained by those of the original document. The conservative emphasis on preserving as much of the structure of the Japanese as possible in the final copy yielded work that while not always aesthetically pleasing, could rarely be faulted for accuracy. It was a conservative style, reflecting years of dealings with customer expectations.

It became evident in discussion with these coworkers that source-based translation is closely associated with thinking in terms of mapping—the notion that there is always an equivalent in the target domain for a given word or phrase in the source language. Less convincing was the implication in office practices that such correlations should be considered invariant and largely unaffected by context. (Hofstadter for one adopts a very different stance in discussing the need for evolutionary models if machine translation is ever to offer a realistic alternative to human expertise.)

The most contentious aspect of company policy was the dictate that the same word in Japanese should be translated with the same choice of word in English regardless of how many times the phrase appeared in the same document. Now as anyone who is familiar with the language is aware, there is a good lot of repetition in written Japanese, particularly in technical articles, and not surprisingly, following this rule produces materials that seem stilted and lacking authenticity. Yet revisions to more natural copy were inevitably dismissed as barabara (inconsistent) because of the violation of the mapping principle. In defense of the agency, assuring consistency was important given the nature of the material, which primarily consisted of patents, company standards and instruction manuals. As far as the customer was concerned, overuse of synonyms implied a nonexistent variance in the source text, something that could potentially mislead the reader. From their perspective, the artificial tone of the final product was a small price to pay for clarity.

Other aspects of the office organization including as the review process also implicitly favored production of source language driven copy. Each translated document was reviewed by at least one other person (usually Japanese) before delivery to the customer. Selection of more natural turns of phrase in English and particularly the use of colloquialisms increased the likelihood that the checker would be unfamiliar with the material and would thus flag the sentence for confirmation. The ensuing "hassle factor" was even worse in the case of client reviewers associated with certain companies, who could be counted on to reply with a list of detailed questions on word choice and syntax requiring an extensive explanation (in Japanese). The concomitant loss of time and stream of thought could be avoided by ensuring that the structure of the translation never strayed far from that of the original text.

Additionally, office promotion of computer assisted translation was wholly congruent with the focus on source language structure. As those familiar with TRADOS and other such products are aware, the software provides a database for comparison of previous translations both within and across documents. Essentially, the sentence currently under scrutiny is compared with all previously translated material. Anything judged sufficiently similar is recalled for the user along with its corresponding translation. The latter can typically be modified to suit the present case with only a substitution or two of nouns, thus speeding up the handling of documents considerably.

It became evident in working daily with this software and becoming familiar with how it deconstructs the text that my "conceptual space" was contracting—from page or paragraph down to the level of the sentence—a factor which curtailed any temptation to read for gist and then paraphrase. Certainly there was less of a temptation to build up elaborate explanations in the target language because of the constraint of providing approximately one English sentence per Japanese sentence. More importantly, working with CAD also tends to promote a kind of abstract thinking with respect to the material, focusing attention squarely on the syntax of the source language. Nouns take on a disposable quality thanks to the recycling of previous sentences, and the text becomes somehow less about content and more about form—primarily the abstract pattern of particles and verbs. And this I believe is perhaps the primary benefit of source language emphasis, i.e. the ability to adapt quickly when faced with technical materials in fields outside of one's own areas of expertise. Learning to ignore the complexities of the placeholders in favor of the essential form—"A acting on B during C" for example, made it easier to visualize what the writer was attempting to convey, and to then fashion a meaningful equivalent in English.

Source/target Revisited

The hiring of a new translator by the company about a year later provided an opportunity to reexamine the question of source/target emphasis anew from a fresh perspective. As in my case, this man's background included no use of CAD; further his methodology involved an initial reading for meaning followed by an intuitive translation, guided mainly by the principle that the finished product must sound as if it had been produced originally by a native English speaker. The approach was hardly radical but friction with the other staff was readily apparent.

A frequent criticism of his copy was fidelity to content, particularly with respect to business letters and other formal documentation. In Japanese, this type of writing frequently involves long stretches of prose tied to levels of politeness not normally utilized in English correspondence. If an equivalent "western" image or tone is not readily apparent, there is a tendency to abbreviate, a habit which can belie the tone of the original if not used sparingly. In dealing with this sort of correspondence, my new friend's "authentic sounding document" rule was gradually augmented by another, less commendable rule of thumb—i.e. "if in doubt, leave it out." This approach was further reinforced by inevitable time constraints arising from an interest in subjective issues of style and phrasing.

Focusing on English form rather than Japanese content also became problematic due to the diversity of source materials. While a translator must of necessary be committed to continuous study, the ideal of operating only within one's field of expertise is rarely an economically viable option. During the course of a typical day at the agency, it was not unusual to be faced with a private letter, a financial statement, specifications for semiconductor production and an overview of the municipal water supply, all in quick succession. Providing an accurate and unambiguous text in this kind of triage situation means keeping a close eye on the syntax of the original, and assuming that the result will make sense in the specialist context in which is it will be read. My friend's insistence on finding the "authentic English tone" for each sentence was commendable. Less so was the concomitant willingness to sacrifice the phrasing and the rhythms of the original, elements which surely form a part of what the author wished to convey. On a practical level, stylistic questions also left him little time to confirm content, which meant being constantly under the gun as far as deadlines were concerned.

My circumstances have changed, and I am now translating in a freelance situation with more leeway regarding the choice of material. Reflecting back on the experience at the agency, what I take away from those two years is sense of the importance of balance. The office policies favoring source-based translation were pragmatic ones intended to maximize the throughput of material and hence profits. The results were often stilted, and stylistically there was never any comparison with text focusing on issues of presentation in English. Yet it must be said that there are times (particularly in working with unfamiliar fields) where the best assurance of correct interpretation is to internalize the grammar of the source and to give it precedence in the writing. In fact, I find myself now using computer assisted translation even with literary pieces in which there is no chance of sentence repetition. The reason is that the deconstruction of the text grounds intuition by placing attention squarely on sentence and the syntax of the sentence. Staying close to the grammar of the original provides cohesion, thereby framing the aesthetic choices that constitute the creative process in which we are all involved.