The species pool concept has been used as a theoretical framework for understanding local community richness. A significant problem in putting the concept into practice is the lack of methods for determining the size of the species pool. We tested the hypothesis that species composition of recent forests is primarily determined by the species composition of neighbouring older forests against the null‐hypothesis that species are a random sample of the species occurring in the study area. Forest plant species composition of recently established fragments was significantly correlated with species composition in neighbouring older forests (i.e. the local species pool). When older forest within a neighbourhood of 1000m radius is considered, seed dispersal sources can be found for 91% of the flora in the recent forests. For an individual fragment, dispersal is a much more important determinant of species presence than the environment, with an average of 46% of the total pool excluded from local pools by dispersal limitation and only 8% excluded by environmental limitations. The species richness of recent forests is on average 23% of the local species pool. Several hypotheses are proposed for this low percentage, such as asymmetric competition due to the early successional state or the limited colonization period.

Nomenclature: Lambinon et al. (1998).

Abbreviations: TSP = All species occurring in the 241 inventoried forest fragments, which is the full forest resource in the study area; ESPSP = The subset of the TSP potentially occurring within the limits of the habitat hypervolume model of a specific forest fragment; DUPSP = The subset of the TSP occurring within a neighbourhood of 100, 500 and 1000 m around a specific forest fragment; LSP = Those species that belong to both, the ESPSP and the DUPSP (i.e. all species that can potentially reach and establish in a specific forest fragment); OCR = Those species which actually occur in the recent forest fragment; Relative richness = (OCR/LSP) × 100.