Peer moderation has been used as a beneficial strategy in asynchronous online discussions to assist student learning performance. However, most studies in peer-moderated asynchronous online discussions (PMAOD) have focused only on learning effectiveness and perceptions of students rather than on students' knowledge dimensions and cognitive processing patterns. This study combined quantitative content analysis (QCA) and lag sequential analysis (LSA) to explore student knowledge dimensions and cognitive processing patterns in PMAOD. The participants were 84 students in an undergraduate blended course from University Putra Malaysia (UPM), Malaysia. The Revised Bloom Taxonomy (RBT) was used as the codification scheme to code the discussion transcripts of participants assigned the role of peer moderators in a reciprocal manner over seven weeks. Behavioural distributions and patterns of high- and low-quality discussion groups were compared. Results showed that students were primarily sharing knowledge dimensions and cognitive processes of metacognition and understanding, respectively. Additionally, it was found that there was a modest proportion of off-topic discussions. Nonetheless, by means of LSA, it was found that PMAOD exhibited a certain degree of self-sustainability in knowledge and cognitive process behaviours, with the exceptions of "procedural knowledge" and the cognitive process of "applying" and, in terms of diversity in knowledge dimension and cognitive processing, high-quality discussion groups outperformed low-quality groups.