15. 



These new principal component axes were rotated to a 

 "best fit" position, defined by the Variraax criterion (Klovan 

 and Imbrie, 1971). The principal components represent mathe- 

 matically independent end-member samples (species or groups of 

 species) , with each real sample being described to a lesser or 

 greater degree by an appropriate mixture of these end-members. 

 Rotation of the components to a "best fit" position introduces 

 no distortion or loss of information. The value of this 

 technique is that the composition of the mathematically 

 independent end-members tends to be close to real, and therefore 

 more readily interpretable, samples. Each sample was then 

 classified in terms of its faunal composition with respect 

 to the identifiable end-member assemblages. The results of 

 this analysis were then mapped in terms of the end-members 

 best describing the fauna of each area viewed. 



RESULTS AND DISCUSSION 

 Substrates 



During the course of this study considerable variation in 

 substrate type was noted, not only among the canyons, but also 

 within them. By and large, fine grained sediments (silt and 

 sand) predominate in all three canyons studied, with rock 

 outcroppings generally restricted to the canyon axes and lower 

 walls. Glacial erratics were also frequently foiind, partic- 

 ularly in the northern canyons. 



In order to determine the spatial distribution of various 



