Angiosperm taxonomy: a comparative study of some novel numerical techniques

SUMMARY A critical comparative study is made of the results from 9 different numerical analyses (involving combinations of ‘nearest neighbour’ and ‘centroid’ sorting with 5 different similarity coefficients: correlation coefficient, Euclidean distance, Euclidean distance standardized, non-metric coefficient and information statistic) of data for 24 genera of Epacridaceae. All the methods point to essentially the same pattern of similarities among the genera, suggesting that ‘good’ taxonomic groups tend to emerge regardless of crudity and variation in the numerical approach. A classification based on these analyses seems, judged by external criteria, to represent an improvement on that of Bentham, whose scheme appears in turn to be superior to that of Drude.