This study examined how students of English as a foreign language (EFL) with different first language (L1) backgrounds use interactional metadiscourse markers in argumentative writing. Specifically, to explore unique patterns of metadiscourse features that reflect context and development, the essays written by Chinese, Japanese, and Korean EFL students at three proficiency levels were analyzed for topic, L1 background, and L2 proficiency. For a comprehensive analysis of 1986 essays, I used a natural language processing tool that generates quantity scores for Hyland's (2005) metadiscourse categories (i.e., hedges, boosters, attitude markers, self-mentions, reader pronouns, and directives). The results showed notable differences in the students' use of metadiscourse features across topics, and significant variation was also found across different L1 groups. However, their use of interactional metadiscourse did not differ by L2 proficiency. A post hoc analysis of a parallel native-speaker corpus further revealed EFL students' underuse of hedges and overuse of reader pronouns. Findings are discussed in terms of academic writing instruction, writing prompt development, and L2 learner categorization.