Washback research in language education aims to demonstrate, explain, and ultimately predict, the impact of tests on teaching and learning in educational contexts. A recent review in the international arena (Cheng et al., 2015) has revealed a rapidly growing field of empirical washback research, yet only two studies were identified as occurring in the Japanese context. The present article therefore sought to more fully document the washback research conducted in Japan prior to 2021 with the aim of facilitating future research in this important area. Following an extensive online search, 32 empirical washback studies in the Japanese context were identified. These studies were analyzed in terms of the following information: publication details, test (s) involved, context and participants, methodology, aspects of washback investigated, and type of consequence targeted. The review reveals a wealth of empirical literature that has adopted a variety of research methods and designs to investigate the impact of a variety of tests, notably that of university entrance exams. On the basis of these previous studies, a series of recommendations are made for future washback research in Japan.

This study examines the effects of retelling on reading comprehension at the textbase and situation-model levels and assesses whether the non-textual information appearing in retelling protocols, which differs from the literal information of a text, correlates with comprehension at both levels. In the experiment, 54 Japanese students from one of Japan’s National Institutes of Technology participated and were divided into two groups. Participants in one group read two expository texts and answered multiple-choice questions on the texts. Participants in the other group read the same texts, but before answering the questions, they retold the stories. Data from 44 participants were analyzed statistically. Contrary to the results of the previous studies, the facilitating effects of retelling on reading comprehension were not confirmed at either the textbase level or the situation-model level. Specifically, the results of the Kruskal–Wallis test indicated that comprehension at the situation-model level might significantly differ depending on the with- or without-retelling conditions and participants’ English proficiency, whereas post-hoc multiple comparisons did not report which condition yielded a significant difference. The low reproduction rates of information in the retelling protocols suggested that the participants in this study remained in the shallow processing of memorizing the literal information of a text, which did not contribute to constructing a situation model. The frequency at which non-textual information occurred did not correlate with comprehension at the textbase or situation-model levels, but the very low occurrence of non-textual information could indicate a problem with comprehension, especially among high proficiency learners.