Fuzzy Boundaries in the Research into Talk-in-Interaction

We aim to explore the nature of data analysis in the investigation of talk-in-interaction, concentrating on methodological fuzziness. On the basis of a case study on the phenomenon of repair, we argue that even in a traditionally corpus-based field like conversation analysis, the source of corpus data is of an inherently integrated nature. The actual research practice shows that the methodology of conversation analysis cannot be clearly delimited but has fuzzy boundaries. In order to be able to use occurrences found in corpora as data, corpus as data source should be complemented with further pieces of information from other data sources, for instance, the researcher’s intuition, the results of earlier investigation, the theoretical framework we work in and inferences.