io W. A. SANDS 



absolutely characteristic of the soldierless termites in Africa. The mixed segment, 

 when present, always consists of a single extension of the mid-gut around the inner 

 curve of the loop, with a corresponding extension of hind-gut round the outer curve. 

 This is the reverse of all other groups, and will be discussed further in the section 

 on phylogeny. 



The dissection of the worker caste to obtain a clearer view of these structures 

 is a simple matter. All that is needed is to pinch up the body wall of the ventral 

 side near the hind coxae with fine forceps, and to tear it back towards the rectum. 

 The pleural membranes rupture more or less evenly leaving the gut exposed. The 

 ventral nerve cord and any residual muscle coats can then be picked off very quickly. 



The other structure requiring more detailed description is the enteric valve. Its 

 position has already been indicated. It can be dissected out by gripping the pouch 

 beyond the valve seating, and the first part of the proctodeum anterior to the valve, 

 with forceps. The valve can then be gently pulled out of its seating and detached 

 from the proctodeum as a roughly conical stump with the valve armature inside 

 its truncated apex. Repeated gentle pressure will expel the contents, and the 

 valve can be mounted whole and cleared in one operation. For this purpose the 

 variant of Gum chloral commonly known as 'Swan's Berlese' has proved com- 

 pletely satisfactory. The armature is easier to see if the valve is slit down one side 

 with a fine blade and opened out, as shown in many of the plates. The valve, as 

 described by Grasse and Noirot (loc. cit.) consists of longitudinal rounded ridges 

 or cushions fbourrelets') which carry a variety of spines and sclerotizations. In 

 some genera these are such as to render the term 'cushion' inappropriate, but it is 

 used throughout this paper to indicate the homology of the parts and avoid con- 

 fusing circumlocutions. In a few species the thin membranous wall between the 

 cushions also shows small scales or carunculations that bear fringes of spicules of 

 varying size. These can usually only be seen at high magnifications with phase 

 contrast illumination, and are not therefore intended as key or diagnostic characters 

 except when very conspicuous. However they are of phylogenetic interest and so 

 are included in descriptions and used for multivariate similarity studies. For the 

 latter purpose, it was necessary to express the variation of each character in coded 

 form, as dichotomies, multi-state qualitative characters, or as ranked or continuous 

 quantitatives (measurements). The similarity analysis utilized species as O.T.U's 

 with the objective of clarifying generic groupings. In theory, the characters are 

 used without prior weighting, but it is inevitable in the process of coding that the 

 logical structure of certain 'gross' characters necessitates their breakdown into a 

 number of smaller units. This results in some centres of anatomical variation 

 being represented by many more or less independent characters and so being weighted 

 in the overall consideration of phenetic affinity. Such weighting seems intuitively 

 to be natural, and might even to some extent be thought to reflect genetic complexity 

 though there is of course no way of assessing this at present. An example of such 

 a feature is the enteric valve in which 34 coded characters were required to express 

 its variations in form. On the other hand some characters, such as the length of 

 the mixed segment, appear to be under-represented because they do not lend them- 

 selves readily to logical breakdown. This feature is probably of considerable 



