Studies in the Numerical Analysis of Complex Rain-Forest Communities: II. The Problem of Species-Sampling

Abstract
A set of 18 sites, previously classified by reference to all 818 species recorded, was reclassified using subsets of species; 10 ''synusial subsets were used, and each was matched by a subset containing the same number of species selected at random. The subset of all 269 large trees duplicated the original classification perfectly down to the ultimate divisions of the hierarchy, and no other subset possessed this property. Many of these larger trees are confined, or almost confined, to single sites, and such species contribute very little to the classification; it was therefore possible to duplicate the major divisions of the original classification, with only very slight modification, by the use of only 65 tree-species. The implications of these findings are discussed with reference to the structure and floristic classification of rain forests.