20 W. A. SANDS 



analysis. Further material received later confirmed its membership of the genus. 

 The position of Ateuchotermes is shown clearly in the three-dimensional graphs of 

 the first six eigenvectors. All these were plotted as part of computer output, 

 but the diagrams have been modified in various ways to enhance their visual impact. 

 Vectors i, 2 and 3 have been converted into a 'glass box' diagram (Text-fig. 10) 

 with floors dividing the total range of vector 3 into four equal parts. Co-ordinates 

 of vectors 1 and 2 are drawn in full on the last floor through which the stem of 

 vector 3 passes before reaching the species point. From the last floor to the species 

 point the stem is twice its lower thickness. Where the terminal part of a stem is 

 hidden by the floor above it is drawn broken. The vector 2 shadow lines on the 

 base plate help in spotting the origin of points arising from clumps on vectors 1 and 2. 

 This type of diagram has been used effectively with fewer points in smaller analyses 

 but is too complicated to be of much help in interpreting these results. Accordingly 

 all other three-dimensional graphs have only been modified by adding their basal 

 co-ordinates, which reverse direction about the zero value of each vector. The 

 rest of the first six vectors have been plotted in the combinations 4, 5, 6; 1, 3, 5 

 and 2, 4, 6 in Text-figs 11, 12 & 13. This enables the relationships of species and 

 genera to be visualized more clearly than would be possible in a single diagram. 

 In the graph of vectors 1, 3, 5 the monotypic genera have been marked distinctively. 

 Some species show up as more distinct from the main cloud of points in certain 

 dimensions rather than others. An example is the monotypic genus Adynatotermes 

 (19) which is distinct on vector diagrams 1, 3, 5 and 4, 5, 6. In other dimensions 

 it is not far from the large loose group of points representing the genera Astalotermes 

 and Anenteotermes, and it falls among them in both cluster analyses. 



The distribution of points along the principal axes of the multivariate cloud 

 represented by vectors 2, 4 and 6 seems to be more even than on vectors 1, 3 and 5, 

 the latter combination producing the clearest indication of clustering. It is there- 

 fore used, with appropriate markings, in the introductory sections to successive 

 genera to indicate the positions of their respective clusters and to compare them 

 with those derived from the canonical variates analysis. In an analysis of the 

 'Q'-type such as the principal co-ordinates, the contributions of individual characters 

 to the total variance represented by successive latent roots cannot readily be 

 assessed as they can in an 'R'-type study. Thus the reasons for differences of 

 dispersion along the principal axes remain unknown. 



Other genera that stand out clearly in the cluster analyses are Aly scoter mes 

 (26-28) which is separated by a phenon gap of 6% from its relatives, and Aderito- 

 termes (56-57). Three species of Acholotermes (30, 32, 33) come out together in 

 both cluster analyses, but the fourth species (31) is variously separated from them. 

 This is because its nearest neighbours are members of Anenteotermes as can be seen 

 from the vector diagrams. It is located on the same side of the larger cluster as 

 the other members of its genus, and a conventional decision places it differently 

 from the numerical methods. It has a short mixed segment like the other Acholo- 

 termes, not a long one like Anenteotermes, and is possibly a transitional species. 



In the single linkage clustering three species of Astratotermes (22-24) form a group, 

 and median sorting adds a fourth (25), while a fifth (21) comes in at the next general 



