Many teachers say, "Don't write in Japanese first. Think in English." But a 1992 study found something different. Forty-eight Japanese university students wrote essays in two ways. First, they wrote in Japanese and then translated into English.

Second, they wrote directly in English. Students with lower English levels wrote better essays when they translated. But students with higher levels did not get much better by translating.

Teachers often tell students not to write in Japanese and then translate, but to think in English from the start. In 1992, Kobayashi and Rinnert tested this advice with 48 Japanese university students, who wrote essays both by translating from Japanese and by writing directly in English. For lower-level writers, translation tended to produce better content, organization and style, probably because they could plan their ideas fully in Japanese.

Higher-level writers, however, gained little from translating and actually made more errors that confused the meaning. Translation may also create learning chances. When writers notice that they cannot express something in English, they start thinking about the form they need. So lower-level writers can use Japanese as a bridge, while higher-level writers can try writing directly.

"Don't translate; think in English" is standard advice in writing classes, but evidence from Japanese learners suggests it may not suit everyone. Kobayashi and Rinnert asked 48 Japanese university students to write under two conditions, composing first in Japanese and translating, and composing directly in English. Lower-proficiency writers tended to produce essays with better content, organization and style when translating, and translated essays overall showed greater syntactic complexity, presumably because ideas developed in Japanese pushed writers to stretch their English. Higher-proficiency writers, by contrast, benefited little and made more errors that obscured meaning when translating, perhaps because Japanese expressions pulled them toward unnatural English.

The quality of directly composed essays was more closely related to speaking and listening skills than to grammar knowledge. Research on noticing adds another angle: Swain and Lapkin showed that encountering a gap between what one wants to say and what one can say triggers thinking about form, and Hanaoka found that Japanese students who compared their writing with model texts mainly took in expressions they had noticed lacking. Given the small sample and limited topics, the findings are tentative. Even so, they suggest adjusting writing methods to proficiency.