Abstract. Forest patches in central Belgium were inventoried twice for the presence or absence of forest plant species to study the effects of age and distance on species composition. All forests in the study area were subdivided based on their land use history. To avoid effects of autocorrelated environmental characteristics on species composition, habitat homogeneity was indirectly investigated using a TWINSPAN classification of the vegetation data. Two major habitats (alluvial and non-alluvial forests) were distinguished and analysed separately. Patterns of species composition were investigated at the landscape level using Mantel tests. Species composition similarity measures were calculated between all pairs of fragments based on the floristic data. A highly significant correlation was found between species composition similarity and inter-patch distance. Difference in age, which we used as a measure for habitat quality, was less important in explaining species composition patterns. The effects of spatial configuration became significant when difference in age was accounted for, and a partial correlation was calculated between inter-patch distance and species composition similarity. Different results were found for alluvial and non-alluvial forest types. Alluvial forests were more influenced by the spatial configuration than the non-alluvial. For the non-alluvial forest type effects measured with the difference in age between forest fragments obscured the effects of inter-patch distance. Based on our findings we suggest that species composition is not only internally generated, but external processes such as differential colonization caused by varying degrees of isolation may be of overriding importance.