|Oak-hornbeam forests represent one of the most frequent mesophilous deciduous forest types within Europe and adjacent areas. In their tree layer, Carpinus betulus and oaks play an eminent role. Forest mesophytes prevail in their herb layer. Regarding the EUNIS habitat classification, they are unified in the unit T1-E (Carpinus and Quercus mesic deciduous forests) and evaluated as near threatened. A large majority of them are traditionally assigned to the order Carpinetalia. For the Caucasian types, an individual order Lathyro-Carpinetalia was designated, however, it is sometimes merged with the former one. In the recently released EuroVeg checklist, ten geographical alliances of these forests were recognized. This traditional geographical approach to alliance delimitation mirrors the existence of numerous geographically distinctive groups of forest mesophytes (e.g. Caucasian, Euxinian and Illyrian) likely reflexing various glacial refugia of deciduous forest vegetation. Therefore the alliances are characterized by their own combination of such species groups. Nevertheless, alliances representing northern types of oak-hornbeam forests in the context of their whole distribution range (i.e. Carpinion betuli for Central Europe and Querco-Tilion for subboreal Europe) are usually poor in geographically distinctive species and thus delimited mainly negatively. Although oak-hornbeam forests are broadly distributed and mostly well-represented by phytosociological relevés, differences in species composition among the particular alliances based on numerical data analyses have not been provided so far. Therefore the main aim of the study is to recognize main patterns in their species composition variation and subsequently to assess how well they are expressed by the alliances in the currently accepted syntaxonomical system. We used mainly phytosociological data (more than 10,000 relevés) extracted from the databases stored in the European Vegetation Archive (EVA). Vascular plant nomenclature was harmonised according to the Euro+Med Plantbase, non-vascular cryptogams were omitted. To obtain more representative dataset, we performed heterogeneity-constrained resampling. Afterwards, we applied various strategies including unsupervised and supervised classification analyses coupled with ordination analyses. The final results will be presented during the meeting.