MA students often find writing the discussion section of a thesis (hereafter discussion writing) a difficult task (Bitchener & Basturkmen, 2006). Thesis advisors, research writing instructors, and graduate students are able to recognize the strength and weakness of discussion writing through established levels of quality. Therefore, grading discussion writing has the potential to contribute to instructional quality and learning enhancement. This study distinguished the qualities of discussion writing and identified the factor(s) that affect differences in such quality. To achieve this, 30 discussions in applied linguistics MA theses were classified into four quality levels based on raters' judgments. The discussion writings were then coded using the move structure of discussion sections (Basktukmen, 2009). In these sections, it was found that move 3 [reporting a result] and move 4 [commenting on results] were obligatory across only three groups (marginally adequate, competent and mastery groups) and a move cycle was identified as the sequence from move 3 [reporting a result] to move 4 [commenting on results]. The combinations of move 3 and move 4 were found to be a possible factor in determining the quality of discussion writing for competent and mastery levels works. It was also found that the move combinations were single move, two-move combination, three-move combination, and multiple move combinations. The more complex the structure of move 3 and move 4, the more sophisticated and critical the discussion writing. The pedagogical implications of this study are considered and suggestions for future research presented.