Most of Central European rocky scrub communities formed by Cotoneaster integerrimus, Juniperus communis and Amelanchier ovalis are included in the association Cotoneastro-Amelanchieretum (= Junipero-Cotoneasteretum). However, this leads to the creation of syntaxon whose internal diversity is so great that it seems necessary to examine validity of its existence in the current form. This diversity entails species composition, habitat requirements and geographical distribution. Therefore, we posed the following objectives: i) to investigate the variability of species composition of the rocky scrub; (ii) to determine if there are ecological differences between the communities distinguished by species variability; (iii) to determine the geographical ranges of individual syntaxa. Altogether we analyzed 387 phytosociological relevés from Central Europe. Vegetation types of rocky scrubs were identified using the unsupervised K-means algorithm and detrended correspondence analysis. Mean Ellenberg's indicator values were applied to identify the environmental gradients shaping the plant communities. Obtained results confirmed the validity of dividing this broadly defined syntaxon into six distinct vegetation units. In order to present the studied communities in a broader context, we included into our analyses other rocky scrub with the occurrence of Cotoneaster sp. and A. ovalis, which formed the remaining three clusters. The observed differences in species composition were further supported by significant differences in soil reaction, temperature and continentality between the syntaxa. Moreover, the distinguished communities differed among one another in geographical range. Our study provides a new typology of the selected type of natural rocky scrub vegetation in Central Europe which involves environmental aspects, hence better reflects community-habitat relationships. This study also indicates a need for further revision of the classification of other types of natural scrub communities. Such classification should be based on modern data analysis methods and should primarily focus on lower vegetation units at pan-European scale.