This study aimed to explore the etfect of translated writing on English writing performance of Korean learners of English and their perceptions on translated writing in comparison with direct writing depending on their English proficiency level. A total of 75 male high school students were divided into two groups according to their L2 writing expertise and were asked to perform two ditferent writing tasks: 1) write directly in English (direct writing), and 2) write in Korean first and then translate the draft into English (translated writing). 136 written compositions (68 direct writing samples and 68 translated writing samples each) from 68 students were selected for analytical scoring and T-Unit analysis. The 68 students also filled a survey questionnaire which asked which type of writing was preferred for their English writing. The results showed that the performance of the direct writing was significantly better than that of the translated writing for hjghlevel writers, while there were no great ditferences among low-Ievel writers between the two writing tasks. T-Unit analysis revealed that syntactic complexity between direct and translated writings by two proficiency groups did not yield any statistically significant ditferences, and more students preferred translated writing to di rect writi ng in general. 8ased on the results, pedagogical implications are suggested.