Planning an adaptive numerical classification

Summary While numerical methods of classification seek to strictly formalise one part of the classificatory process, a human classifier uses a variety of different approaches to obtain a satisfactory organisation of data. In this paper some means of increasing the flexibility of numerical methods are discussed, in particular the possible uses of the taxonomic hierarchy. It is first necessary to show that higher taxonomic units contain ecologically interesting information. Then, using a two parameter classification method a means of progressing from higher to lower taxonomic units, during analysis and in a data dependent manner, is outlined. A similar procedure is possible with nested samples. However, since the value of a classification resides in its ecological interpretation, an automatic classification requires some means of ascribing meanings to classes. While such a semantic analysis seems theoretically possible, its practical attainment presents considerable difficulties.Contribution to the Symposium on Plant species and plant communities, held at Nijmegen, 11–12 November 1976, on the occasion of the 60th birthday of Professor Victor Westhoff.