Observations on the Use of Cluster Analysis in Botany with an Ecological Example

Abstract

ever-increasing number of accounts of techniques for the handling of very large quantities of biological data. In botanical work this flow starts with the pioneer studies of Williams & Lambert (1959). Against the undoubted refinements of modern techniques it seems prudent to balance their proliferation; increased sophistication of computer methods is not necessarily associated with greater clarity and utility of the resulting classifications. Many workers have pointed out that the ability of computers to sift data in no way relieves the biologist of his responsibility for the choice of criteria or careful judgement of the results. Computer taxonomy and ecology do not really pose the same problems, although they have often been confused (notice the frequent use of the blanket term 'numerical taxonomy'.) In taxonomy the units considered are individual organisms which must be grouped, usually hierarchically. At any level in the hierarchy the possible variations on the theme are very limited; the viability of the individual and the natural boundaries of the taxon establish its limits. Even where, in 'difficult' taxa, the individual groups merge into one another, there always develop eventually more or less clearly marked natural discontinuities which are inevitably picked out whether the techniques used are those of the classical taxonomist, the experimental taxonomist or the computer taxonomist. Within the groups defined by these discontinuities there is little room available for manoeuvering; the morphogenetic processes involved in the development of viable individuals presumably leave little to chance. The commonly drawn analogy to this of ecological community structures has generally appealed to analysts of botanical data; the analogy needs justification. It is doubtful if discontinuities of a similar sort are to be found among plant communities; however closely a plant is integrated with its environment, it seems unlikely that the integration can be as great as that of, say, a particular chromosome configuration, the development pattern of the auditory apparatus or the functioning of a multiple allele sterility system within their individual animals or plants. All of these are, or can give rise to, criteria used in the taxonomy of organisms. The occurrence of such characters in animals or plants is definitive of the organisms concerned, and leaves little room for chance inclusion or omission. The interest in the phytosociological analyses of plant communities lies in the overall facies of the communities, and not in the presence or absence of any particular species. Yet the techniques available depend essentially on such presence or absence data, or at best on some estimate of the abundance of the species