Not only does college-level critical writing aim to boost students’ causal thinking and reasoning skills, but it also eventually reflects the cognitive change in their assignments (Carroll, 2007; Nosich, 2022). However, to our knowledge, students’ process of developing these skills has not been studied extensively. Therefore, this study examines the language of critical thinking through interactional metadiscourse markers in Hyland (2005) and phraseological items commonly used to express evaluation in the critical response papers written by preservice EFL teachers over the course of a semester in a Turkish public university. Research questions are as follows: (1) What are the commonly used interactional metadiscourse markers as listed in Hyland (2005), as well as linguistic items used only in the critical response papers by pre-service EFL teachers ?; (2) How does the use of these markers differ across tasks?; and (3) How does the use of these markers differ across students? The Critical Response Corpus, henceforth CRP, includes 198 response papers written by 22 students. Of the 81 students enrolled in “Critical Reading and Writing,” an undergraduate course offered primarily to sophomore-level students in a TEFL program, texts on the current issues in ELT such as translanguaging, use of digital platforms in ELT classes, advocacy, mindfulness, and social-emotional learning, by 22 students who submitted all nine papers were included. The submissions were based on critical responses to the articles read and discussed in class every week. The frequencies of metadiscourse markers listed in Hyland (2005) were extracted from the corpus using AntConc 3.5.9 (Anthony, 2020). Additionally, following the manual annotation of the papers by five students, that is, over 20 percent of the data, multi-word units including interactional metadiscourse markers were closely examined. Preliminary findings suggest that students made frequent use of engagement markers (e.g., should, parenthetical information, and we) and boosters (e.g., think, must, and know), while the use of self-mention (e.g., I, we, and my) and attitude markers (e.g., important, even, and agree) decreased from the first to the last task. These results align with Caroll (2007) in that the students moved from sharing their personal experiences to using research evidence to support their arguments. However, the frequency of hedging devices (e.g., should, may, and would) was found to be closely linked to the topics of the readings. Students were found to use hedging devices more frequently when critiquing the readings about instructional practices, which, in a way, supports Liu and Stapleton’s (2018) finding that lack of familiarity with the topic affects students’ writing negatively. Overall, the findings shed light on the important linguistic qualities of critical writing (Woodward-Kron, 2002), and the key role of topic selection in assigning critical writing tasks.

Keywords: L2 academic writing; critical thinking; metadiscourse; corpus linguistics; pre-service teacher education