440 



Fishery Bulletin 92(2). 1994 



phin school abundance data as a 

 whole, with a range from 33.5% for 

 common dolphins to just 5.1% for 

 whitebelly spinner dolphins (Table 

 4). The eigenvalues for the ordina- 

 tion axes indicate that only the 

 first two are important (Table 4). 

 Axes 3 and 4 together represent 

 only 6% of the variation of the spe- 

 cies — environment data. Ter 

 Braak ( 1985) suggests that eigen- 

 values less than 0.02 be disre- 

 garded. The first axis accounts for 

 70% of the variance extracted, the 

 second accounts for 24%, together 

 they represent 94%. Further re- 

 sults and discussion therefore re- 

 late primarily to axes 1 and 2. 



The relationships between most 

 of the species categories of inter- 

 est and the first two environmen- 

 tal axes were unimodal (Fig. 4), 

 satisfying a primary assumption 

 of CCA. One exception was com- 

 mon dolphins on axis 1, where 

 there is evidence of bimodality. 



The Monte Carlo randomization 

 test resulted in rejection of the null 

 hypothesis of no relationship be- 

 tween the species encounter rates 

 and the environmental data (H : 

 X=0). One thousand random per- 

 mutations produced no ordinations 

 with a trace (eigenvalue total) larger 

 than the observed 0.443, giving a 



