Teacher collaborative discourse is a process at the heart of many existing models of professional learning, yet there are gaps in understanding the process’ complex and contingent nature, especially with research from ‘non-Western’ contexts and using relatively ‘naturalistic data’ without significant external or researcher intervention/facilitation. To address this problematic, this thesis uses relatively naturalistic data to examine how teachers learn through collaborative discourse within professional learning communities, and how their discourse practices such as explicit reasoning, storytelling and discussing external expectations make a difference to their learning. Utilising a conceptualisation of learning based on cultural models (Gee, 2015) and cognitive dissonance (Festinger, 1957; Beswick, 2017), teacher learning is examined in the context of two English-language teacher PLCs in urban Malaysian secondary schools, thereby making novel contributions to the theory and practice of teacher learning while advancing methodological tools for the field. The study’s primary data comprises verbatim transcripts of 13 teacher meeting videos, generated over a nine-month period in 2019, segmented into 112 discrete episodes of discourse. Various descriptive and analytical procedures were conducted to generate a theoretically-informed gloss of the data, from which relevant samples were identified to ask three analytical research questions: [RQ1] ‘What are the different outcomes to teachers’ shared cultural models about their students when they (the teachers) discuss discrepant evidence?’; [RQ2] ‘What observable features appear to make a difference with respect to the outcomes found in RQ1?’ and [RQ3] ‘What are the observable nuances and contingencies to those ‘difference-makers’?’ With respect to RQ1, this thesis reports three distinct outcomes when teachers discussed discrepant evidence about their students: the teachers’ initial cultural models were either ‘maintained’, ‘revised’ or ‘rejected’, with implications for learning and practice. As for RQ2 and RQ3, three discursive ‘difference-makers’ were found: (1) explicit reasoning, (2) storytelling, or narrative representations of practice (‘replays’, ‘small stories’) and (3) discourses about external expectations. Through systematic comparison of relevant episodes of data, I make arguments about the uses, contingencies and caveats of these ‘difference makers’, some of which constitute novel contributions to existing scholarship. All in all, the findings offer insights into the situatedness of learning as a dynamic interplay between individual and professional learning community, contributing to the theory of professional learning, and learning more generally. Methodologically, the study breaks new ground by being the first to present evidence of this nature from the Malaysian context. Moreover, a detailed 10-step research process is proposed, with corresponding rationales, that would be instructive for the study of teacher collaborative discourse, especially using more unstructured/naturalistic data. For educational practitioners, the findings constitute a form of ‘cultural knowledge’ (Biesta, 2020), functioning as interpretive resources or tools to support conscious engagement with and facilitation of collaborative discourse, creating possibilities of ‘doing dialogue differently’. Specific reflection and action for various categories of educational practitioners (teachers, school leaders, policymakers, professional development facilitators and researchers) are discussed. Through the aforementioned contributions, this thesis deepens our understanding of the nascent yet highly consequential field of teacher collaborative discourse in professional learning communities.