Definitions of biogeographical regions in Switzerland are traditionally based on expert knowledge. Six different sets of systematically collected floristic data, including 129 to 2280 species, were analysed by means of minimum variance analysis, using two similarity measures, the correlation coefficient of standardized data and the Jaccard coefficient. The resulting groupings are compared with some of the traditional classifications using kappa statistics. The main divisions obtained by the statistical procedure are roughly corresponding to those based on expert knowledge. There is evidence that standardized and correlated distribution data tend to build more coherent groups. Datasets with a high number of species are divided into more coherent groups than datasets with a low number of species. Biogeographical divisions of Switzerland into 5, 11, 19 and 30 groups are suggested. Possible causes for these patterns are discussed.