A mixed-data numerical approach to angiosperm taxonomy: the classification of Ericales

The taxonomic necessity for a satisfactory mixed-data numerical method, and the difficulties involved in devising one, are emphasized. Observations on 24 qualitative and 14 multi-state characters were available for 193 species of Ericaceae and Epacridaceae. The data were analysed using the programme MULTIST (agglomerative strategy, with ‘centroid’ sorting) modified to deal with missing and inapplicable observations. The resulting computer-based classification bears only a limited resemblance to the accepted taxonomic arrangement of the genera; but it provides a convenient summary of the information put into it and makes sense of observations other than those on which it was based, while the traditional scheme does neither. The results suggest (i) that most of the genera are relatively homogeneous; (ii) that the traditional distinction between Ericaceae and Epacridaceae is useful; (iii) that most of the subfamily and tribal groupings in current use are either quite worthless or in need of drastic revision; (iv) that mixed-data computational methods are likely to be of considerable assistance to taxonomists.