ABSTRACT
Increasingly used in qualitative research, intercoder reliability (ICR) remains contested. Critics argue that it imports positivist assumptions and risks flattening the polysemic, context-dependent nature of meaning. This chapter contributes to these debates by examining how ICR can be operationalised in qualitative discourse analysis, drawing on the annotation of ‘populism’ and its derivatives in Belgian, French, and Spanish public discourse as an illustrative case. We identify three challenges specific to sociopolitical keywords: the need for context-rich segmentation, the interpretive work required to reconstruct implicit meanings, and the difficulty of assessing agreement in multi-label and open-ended coding. We then present a reliability protocol tailored to a heterogeneous coding grid that includes closed single-choice and multi-choice variables, alongside open and mixed formats. The initial ICR assessment reveals a consistent pattern: strong convergence on formal, directly observable descriptors, and weaker convergence on inferential or highly polysemic dimensions. Ultimately, this chapter shows that ICR can be integrated into discourse analysis as both a device for coordinating interpretive coding and an analytic lever, making coding decisions traceable, supporting iterative codebook refinement, and identifying zones of semantic density where disagreement is theoretically meaningful.
