A syntaxonomical synthesis of calcicolous forests dominated by Fagus sylvatica (Cephalanthero-Fagenion suballiance) in the Czech Republic was carried out using the Braun-Blanquet approach. Relevés included in the analyses were selected following formalized approach by using an expert-delimited group of 38 calcicolous and/or xerothermophilous species. Only one association Cephalanthero-Fagetum was distinguished, which usually occurs on limestone, calcareous sandstone and calcareous sandy marlite; however, can be found also on base-rich siliceous bedrock (e.g. basalt, phonolite). Based on TWINSPAN analysis, three subassociations were recognized within the Cephalanthero-Fagetum: (i) Cephalanthero-Fagetum seslerietosum caeruleae on shallow rocky soils with frequent dominance of Sesleria caerulea and presence of petrophytes, (ii) Cephalanthero-Fagetum typicum on dry, shallow soils with a significant presence of light-demanding, thermophilous, and calcicolous species, and (iii) Cephalanthero-Fagetum actaeetosum spicatae on deeper, sufficiently moist soils with an abundance of mesophilous, nitrophilous and acidophilous species. The name Cephalanthero-Fagetum actaeetosum spicatae is a new nomenclatural combination. The relationships between Cephalanthero-Fagetum and similar forest vegetation types containing xerothermophilous and/or calcicolous species in the Czech Republic are discussed. The main gradients in species composition of Cephalanthero-Fagetum subassociations were revealed by gradient analysis. The Ellenberg indicator values, altitude, slope, and ‘southness’ were used to interpret these gradients. Using unconstrained ordination analysis (DCA) the syntaxonomical interpretation indicated three relatively distinct groups. Moreover, further DCA analysis revealed the well-defined position of Cephalanthero-Fagetum within Czech beech forests. The results of the above delimitation of Cephalanthero-Fagetum were compared with the results based on Cocktail-defined species groups improved by similarity-based assignment of relevés (using frequency-positive fidelity index). When the Cocktail-based formulas for beech forests were applied to the relevés selected by our 38-species diagnostic group, the correspondence between these two approaches was only 36%. However, at the lower subassociation level, the highest correspondence occurred for Cephalanthero-Fagetum seslerietosum (84%). The reason for this high correspondence is that the species composition includes many specialists (i.e. good diagnostic species) and it occurs at the end of an ecological gradient. To sum up, it is possible to define vegetation units accurately using strict formulas, as opposed to the less rigorous ‘soft’ traditional approach. However, both approaches fail when defining central units.