This paper considers issues relating to the identification and categorisation of interruptive acts for cross- cultural study, as revealed by the conflicting methodological requirements of a medium-scale project involving contrastive analysis of confrontational native speaker and non-native speaker talk in French and English. The paper opens with a brief introduction to the project, followed by a review of issues from the conflicting ends of corpus annotation and Conversation Analysis, the main locus of information about, and research into, sequential aspects of talk and interruptive phenomena. It then uses two examples from the project data for native English and French respectively to reveal and discuss tensions between diverging requirements in the categorisation of interruptive acts. It shows that, while categorising interruptive phenomena inevitably entails a degree of arbitrariness - minimised in either very large corpora or small scale situated analysis -, medium-size data are peculiarly vulnerable to issues of empirical validity, but that their function and the options they create to derive critical findings from the tensions between approaches make them an important tool for research, notably cross-cultural research.