Philosophy and Method in Biological Classification

Abstract

Classifications are designed for various purposes. In most of science they are pragmatic, with general purpose classifications (in philosophy of science termed natural classifications, and phenetic in outlook) used as a basis. These are supplemented by others for special purposes. In biology there is debate on the type of classification that is proper as a basis; the main contendants for this are phenetic and phylogenetic classifications.

The difference between falsifiability and verifiability is discussed. Phenetic classifications (using agreed data and algorithms) are verifiable. Phylogenetic classifications are not verifiable because actual phylogenies are unknown. A phylogenetic reconstruction depends critically on assumptions on how evolution operates, and on whether evolution did operate thus in that particular instance. These assumptions, and whether they operated, cannot be tested directly. It is not desirable to base science on the unverifiable.

Distinction is drawn between classifications derived from a small sample of data and those from a large or complete data set. Analogies are drawn between these and the statistical concepts of samples and populations. Difficulties in these concepts are discussed.

The logic of phylogenetic reconstruction is briefly discussed. Evolutionary homology cannot be a basis for classification because evolutionary homologies are unknown and must be deduced from the classification itself. What can be known is isology or high one-to-one correspondence in complex structures or systems. Phylogenetic work faces the same kind of methodological problems with character weighting, similarity coefficients and the like, as phenetics. Monophyletic groups that depart too much from phenetic groups are not useful for general taxonomy.

It is concluded that phenetic classifications are preferable as general purpose classifications, and their aim is to maximize the number of predictive statements about members of their constituent classes. This can be measured in terms of information content in a number of ways. Those most appropriate for biological systematics are discussed.