Variety, as a linguistic concept, is of central theoretical importance for linguistics, especially for sociolinguistics, variational linguistics and dialectology. This is because it – in all its manifestations – refers to the systemacity and thus to the “reliability” of variation patterns. For social agents, who orient their actions towards such patterns, this reliability or accountability, using a term from conversation analysis (c.f. Robinson 2016), is of central practical importance. From a praxeological point of view (cf. Meyer 2015), varieties can be described as a means for designing social actions in a way that they are recognizable and understandable for others. The typification and understanding of utterances in verbal interaction always depends on the classification of stretches of sound as linguistic units such as words, phrases and syntagms. These unit-types at the same time exemplify formal characteristics of linguistic variants within a space of structural alternatives of saying the “same” thing (cf. Lanwer in prep.). This space of alternatives can be grasped with Gumperz’ (1971 et al.) concept of a verbal repertoire of a speech community. According to Gumperz (1982:33), the patterning of such a repertoire is grounded in regularities of language use with respect to the “sequential co-occurrences among [linguistic] features”. In our talk we will present a methodological framework for (a) the reconstruction of such regularities as recurrent co-occurrence patterns in natural conversations and (b) the interpretation of such patterns as contrastive configurations of linguistic variants that are used by interlocutors as a resource for the coordination of social action (cf. Lanwer 2015; in prep.). The operationalization of this approach, including transcription, annotation and statistical modelling, also will be presented (cf. Schürmann in prep).