A numerical taxonomy of the Pterygote insects.

A data set consisting of 196 dichotomous, morphological characters, scored on 38 exemplars from different Pterygote insect suborders, was assembled and analysed using cluster analysis and ordination. The resulting phenetic patterns show a largely plausible arrangement of the insects, when compared with more traditional, cladistic taxonomies. In most cases, different suborders from the same order cluster together and three large groupings are evident, corresponding to the Orthopteroid, Hemipteroid and Holometabolous groups of orders.

When subsets of the data, from different body regions (head, thorax and abdomen) were analysed separately, considerable differences were found in the resulting dendrograms and ordinations. Novel statistics were developed for analysing the incongruence due to individual exemplars. A few exemplars were congruent between all three data sets but some were highly incongruent It was possible to relate some of these differences to a priori differences in the data sets and to the small numbers of characters in each case. Similarly, the data for the Orthopteroid insects were compared with data sets from two other authors for the same orders. Again, large differences were found in the phenetic patterns, especially for some exemplars. When the three Orthopteroid data sets were combined, a highly plausible and stable arrangement of the orders was found.

A bootstrapping technique was used to estimate the stability of the dendrogram from the full data set. Very few statistically significant clusters were found and these were all subordinal groupings. The association coefficient, used for the phenetic analysis, does not treat two character states (present/absent) symmetrically. To test the effect of using this coefficient, the data were reversed (all presences converted to absent, all absences converted to present) and analysed to give a dendrogram that was almost identical to the original one.

The main conclusion is that, while the basic phenetic patterns in this data set appear interesting and useful, more characters than the 196, used here, will be needed to fully analyse the phenetic affinities of the orders.