Introduction

It was in 1960 that J.C.G. initially became interested in problems of classification in the sense of constructing classes of individuals. This type of problem contrasts with the use of the term ‘classification’ in classical statistics which is concerned with discrimination, that is the problem of assigning an individual to one of several a priori classes. Sneath (1957) had written a paper in the Journal of General Microbiology about a computer program that he had developed for generating a hierarchical classification of bacterial strains. His approach was to compute coefficients (the simple matching and related coefficients) that gave a measure of the similarity between each pair of strains as judged by their abilities, either ‘present’ or ‘absent’, to thrive on a selection of nutritive bases. Sneath used a simple hierarchical classification algorithm, now termed the single-linkage algorithm, in which strains are successively grouped, the pair of groups fusing at each stage which have the biggest similarity between any pair of strains, one from each group. Similar work was reported from the USA by Sokal and Michener (1958) (classifying bees) and Rogers and Tanimoto (1960) (botanical classification) while Williams and Lambert (1959), in Southampton, were concerned with the ecological classification of quadrats based on the presence/absence of plant species within quadrats sampled from a region. It was the newly available computers that stimulated these and other scientists to investigate problems that had not previously been viewed as amenable to a numerical approach.