This study aimed to theoretically explore new implicit metadiscourse devices of text coherence which have not been touched upon in metadiscourse literature, labeling them ''macro-level metadiscourse coherence devices'' (henceforth, macro-level MCDs); and empirically identify the level of using macro-level MCDs in written texts of the second year students of English, Faculty of Education, Ibb University, Yemen, in the second semester of the academic year 2021-2022. To achieve the objectives of the study, content and descriptive analysis approaches were used, respectively. To collect the required data, a writing test was administered to all 60 students where a sample of 30 texts was selected randomly for analysis. The study revealed a number of findings; most notably, the overall level of using macro-level MCDs in written texts by the students was low (M = 7.93, SD = 2.44); and Sequencing Ideas scored the most macro-level MCD achieved by the students (M = 2.43, SD = 0.82) while Conclusion was the most problematic one encountered by them (M = 0.57, SD = 0.57). Accordingly, the study recommends curricula/syllabi designers and language teachers to consider macro-level MCDs to be introduced explicitly to students in the classroom.