In this paper, we will discuss the importance of (socio)lectometric research in attempts to empirically map the range and internal structure of language repertoires, on both the level of the individual and the level of the speech community. Spoken data from to East Flemish women (city of Ghent) in s speech settings are analysed using 3 multivariate statistical techniques: correspondence regression, multidimensional scaling, and cluster analysis. These analyses show that on a community level, the language repertoire in Ghent can be called following AUER (zoos) - "diaglossic; with a continuum stretching from dialect to standard language with intermediate variations in between. In this continuum, focal points or prototypes can be identified, such as dialect, tussentaal and Standard Dutch. On the level of the individual language user, however, the repertoires are not necessarily diaglossic. Methodologically, we will argue that different multivariate techniques are ideally combined and complemented, as each technique has both advantages and drawbacks, and only structures emerging in multiple analyses can be considered robust. The Flemish case-study will moreover demonstrate that the techniques can only be usefully implemented in repertoire structure studies when stylistically stratified datasets are available and a diverse set of language features is observed.