32 W. A. SANDS 



canonical variate analyses are shown graphically for the imago and worker castes 

 in Text-figs 14 & 15 respectively. These diagrams are intended to be self-explanatory 

 and avoid the necessity for tables of figures. In the principal component analysis 

 the weighting coefficients of the first eigenvector all tend to be similar in size, repre- 

 senting what in factor analysis is sometimes called a general factor, apparently 

 related to the general size of the insect. All measurement characters will naturally 

 contain a large element of closely correlated size variation and the first vector is 

 often disregarded except to take one convenient measurement as a size indicator. 

 In the present work the width of the head is used. In the worker diagram it is 

 worth noting that character 3 carried a much lower weighting on the first vector, 

 indicating an unusually large proportion of the variation independent of size. 

 This reflects its importance as a taxonomic character and its interest as an adaptive 

 feature. 



The correlation matrices, the numerical values of the vectors, the component 

 scores, and the canonical variates are not published in full to save space. Two- 

 dimensional graphs of principal component scores are shown under the headings of 

 individual genera where they are of interest in separating pairs of closely related 

 species or indicating misidentifications. The first three canonical variates of imago 

 and worker castes are used in the introduction of each genus to illustrate the cluster- 

 ing of species based on numerical characters for comparison with that of the principal 

 co-ordinates. The positions of the species of each genus are shown by solid spots 

 instead of circles. The monotypic genera are shown together in Text-figs 16 & 17. 



Text-figures 



The termites have all been drawn with the aid of a camera lucida, from ethanol- 

 preserved specimens immersed in ethanol. Two scales are used, the larger being 

 twice the smaller. All illustrations of entire head capsules are on the smaller scale 

 and all mandibles on the larger. Parts of the intestines of the worker caste are on 

 either scale according to their relative size but the scale of any particular part is 

 not altered between species on one page of illustrations. The only exceptions to 

 the above are illustrations of the entire gut, which are on a third, lower scale. 



PHYLOGENY 



The description here of numerous new species and genera throws the existing 

 classification somewhat out of balance. Similar undescribed complexes exist in 

 other zoogeographical regions, and will, when described, add to this difficulty. 

 The previously described species included in this study have hitherto been placed 

 in the genus Anoplotermes of the subfamily Amitermitinae. This subfamily has 

 included a considerable variety of forms, and habits ranging from soil-feeding, 

 through wood- and detritus-feeding to harvesting grasses and foliage. Ahmad 

 (1950) recognized four evolutionary series based primarily on the dentition of the 

 imago and worker mandibles within the subfamily. He regarded the group that 

 included Anoplotermes and Speculitermes as the most primitive, and in his hypo- 

 thetical phylogenetic tree showed it branching off near the base of the main stem. 

 The remaining groups he regarded as more closely related to one another as side 



