While learners may have access to reference tools during second language (L2) writing, the latest developments in machine translation (MT), such as Google Translate requires examination as to how using the tool may factor into the second language learners’ writing products. To this end, the purpose of this study was to examine how MT may have an effect on L2 learners’ writing products relative to when writers wrote directly in L2, or translated a text to English from Korean. EFL university learners were asked to write for prompts that were counterbalanced for three writing modes and three writing topics. The learners’ writing products were analyzed with Coh-Metrix to provide information on text characteristics at the multilevel. The results indicate that MT could facilitate the learners to improve fluency and cohesion, produce syntactically complex sentences, and write concrete words to express their target messages. Pedagogical implications are provided for how MT can be used to improve the quality of the L2 learners’ writing products.

This study aimed to identify the continuity between 6th grade elementary school English textbooks and 1st grade middle school English textbooks using Coh-Meu'ix, an automated web-based program designed to analyze and calculate the coherence oftexts on a wide range of measures. The measured value of text types was compared and classified into the surface linguistic features (the basic count, word rrequency, readabi li ty, connective information, pronoun information, word information) and the deep linguistic features (co-referential cohesion and semantic cohesion, lexical diversity, syntactic complexity). The findings were as follows: First, the basic counts and words before the main verb had a significant different value between two levels of textbooks. The results were remarkably different in the written language. Second, FKGL (Flesch-Kincaid Grade Level) and the pronoun ratio were significantly different only in the written language. In addition, type-token ratios in written language showed more significant differences than spoken language. Third, other language features showed only a mild and gradual difference. Finally, the resu lt indicated there were no statistical differences of discourse aspects.