Central European lowland wet meadows are habitats of great conservation interest, however, their phytosociological status has been to a large extent dependent on specific phytosociological traditions in different countries. In order to bridge the gaps between different national schemes of vegetation classification, a statistical analysis of variation in species composition of these meadows in the Czech Republic, E Austria, Slovakia, Hungary and NE Croatia was performed, using a data set of 387 geographically stratified vegetation relevés sampled at altitudes < 350 m. Principal coordinates analysiswas used to identify and partial out the noise component in the variation in this data set. The relevés were classified by cluster analysis. A new method for identifying the optimal number of clusters was developed, based on species fidelity to particular clusters. This method suggested the optimum level of classification with three clusters and secondary optimum levels with five and nine clusters. Classification based on three clusters separated the traditional phytosociological alliances of Calthion palustris and Molinion caeruleae, both with a suboceanic phytogeographical affinity, and a group of flooded meadows of large river alluvia, with a continental affinity. The latter group included the traditional alliances of Agrostion albae, Alopecurion pratensis, Cnidion venosi, Deschampsion cespitosae and Veronico longifoliae-Lysimachion vulgaris; however, the internal heterogeneity of this group did not reflect putative boundaries between these alliances as proposed in the phytosociological literature. Therefore we suggest to unite these alliances in a single alliance Deschampsion cespitosae Horvatić 1930 (the oldest valid name). Classification with nine clusters was interpreted at the level of broad phytosociological associations. Particular clusters were characterized by statistically defined groups of diagnostic species and related to macroclimatic variables.