Second language (L2) writing is a subfield within the field of applied linguistics concerned with applying knowledge and insights from linguistics, psychology, and education to develop teaching approaches for those who need to acquire L2 writing skills for academic or work purposes (Belcher, 2012). Much research on L2 writers over the past five decades has been focused on students in university contexts (Ferris, 2018) in part because universities are often not equipped to meet students’ needs for academic writing support in writing courses (Cimasko & Reichelt, 2011; Kubota & Abels, 2006). One student need pertains to the learning of discourse organization to create coherent text. This issue is of particular concern to EFL undergraduates who often experience difficulty with organizational patterns when writing in English (Tang, 2012).In response to this student need, the current study investigated the effectiveness of the explicit teaching of coherence in EFL undergraduates’ research writing in English with a pre- and post-intervention embedded mixed-methods design of three study groups of Japanese undergraduate students. More specifically, the study involved two experimental groups—a process-genre group and process-writing group—and a comparison group. The teaching program for these three groups differed in the combination of classroom instruction (i.e., coherence-focused process-genre approach or process writing) and written teacher feedback (i.e., coherence-focused or meaning-focused) they received. A total of 36 third-year female undergraduate English majors participated in this study. Writing samples were collected at three points throughout one semester and analyzed based on three measures of writing coherence: reader-based logical development, reader-based cohesion, and text-based coherence. The reader-based measures of an analytic rubric were used by human raters in evaluating reader-based coherence in the writing samples. Rasch measurement was used to assess the rubric’s functioning via a Rasch principal components analysis (Linacre, 2019). In addition, the Rasch model was used to identify raters who were too lenient or too severe and calculate fair average measures of the ratings using Many-Facets Rasch analysis (Linacre, 2014). These ratings were then analyzed and investigated for changes across time and between groups. The text-based coherence measure for each writing sample was obtained via a form of textual analysis called topical structure analysis. Mixed-design ANOVAs were used to analyze the three measures to investigate statistical differences within-group and between-groups differences. In addition, a Pearson’s correlation analysis was conducted to investigate if raters’ assessment of logicality correlates with their assessment of cohesion usage in student samples of research writing. Results from the statistical analyses revealed that the process-genre group was the only group out of all three groups in this study to have made statistically significant improvements on coherence in their research writing during the course of study. To help explain the results of the statistical analyses, qualitative data collected from background questionnaires, rater’s questionnaires, reflective journals, and student interviews were coded and analyzed. The findings indicated that the process-genre group was able to develop coherence at the sentence, paragraph, and discourse levels. Further, a comparison of the results from both the reader-based and text-based perspectives of coherence suggested that coherence development achieved by the process-genre group (i.e., improvement in both logical development and cohesion) was due to the treatment they received as observations from their background questionnaire reflective journal responses and interview data suggested that they appeared to be unaware of the concept of coherence prior to the study. However, as the treatment started, they gradually acquired knowledge and skills for creating coherence, first at the sentence level, then at the paragraph level, and toward the end of the study, at the discourse level. The improvements made by the process-genre group appeared to be related to the changes observed in their perception of coherence throughout the treatment period. The qualitative findings indicated that their perceptions changed from a focus on the relevance of information included in their writing in the beginning of the treatment period to an expanded understanding of coherence as a genre-specific concept that is important in making their writing reader-friendly by using both local and global cohesion and coherence devices. As to the other two groups in this study, the qualitative findings from the background questionnaire responses and interview data suggested that unlike the process-genre group, the process writing group’s coherence development was limited to the sentence and paragraph levels, and that of the comparison group only at the sentence level. The fact that these two groups failed to develop their knowledge and skills in writing coherently at the discourse level might explain the non-significant statistical results for the within-group and between-group analyses conducted with those groups. In sum, the findings showed that the development of coherence in EFL undergraduate research writing is influenced by writing program design. Particularly, the program needs to explicitly teach coherence through a systematically designed curriculum that includes the teaching of both useful genre knowledge and skills for writing coherently. In addition to teaching the textual construction of coherence, because coherence is a reader-based concept, its crucial role in producing research writing that is logical in the eyes of readers also needs to be reinforced in the teacher feedback given to learners. Such a program where both the instruction and the teacher feedback are focused on form (i.e., coherence in research writing) enables learners to improve coherence in their writing as they progress through the drafting cycle of writing and revising in the program. / Teaching & Learning