Perrin et al.: Geographic variation in cranial morphology of Stenella attenuata 



343 



in our expanded study. Likewise, strong associa- 

 tions with measures of solar insolation were sub- 

 stantially reduced for all but a few characters (i.e. 

 length of braincase, length of temporal fossa, and 

 width of temporal fossa; Table 7). Some charac- 

 ters, such as sea current (W., winter) and oxygen 

 minimum layer (depth), did not have variation 

 patterns in either study that corresponded to 

 those for morphologic variables. Water depth, 

 however, has significant correlations with more 

 characters now that western blocks have been 

 added. At least three environmental measures — 

 sea surface temperature (Jan.), sea surface tem- 

 perature (July) (Fig. 10B), and surface salinity — 

 had relatively strong covariation patterns that 

 stayed relatively constant through the two studies. 



Availability in the future of bet- 

 ter information on environmental 

 variation and, possibly, on other 

 relevant parameters reflecting 

 environmental heterogeneity will 

 allow researchers to analyze dol- 

 phin-environmental covariation in 

 a more meaningful way. For ex- 

 ample, better environmental data 

 and more comprehensive informa- 

 tion on dolphin feeding ecology 

 could provide a basis for testable 

 predictions concerning why cer- 

 tain morphological characters 

 covary with specific environmen- 

 tal variables. Our admittedly de- 

 scriptive analyses demonstrate 

 some striking cases of dolphin- 

 environmental covariation and, 

 thus, provide initial guidance in 

 terms of possible causal relation- 

 ships that may be examined with 

 greater sophistication by future 

 investigators. 



Interspecific covariation 



In this paper we have presented 

 statistical data for trends in geo- 

 graphic covariation of S. longiro- 

 stris and S. attenuata skulls from 

 16 blocks for which samples of 

 both species were available (brief 

 comments on covariation were 

 included by Douglas et al. 1992). 

 Geographic patterns in 13 of the 

 30 morphological characters 

 showed statistical correspondence 

 based on Mantel tests, whereas 



Table 9 



Comparisons of principal component and canonical 

 variable projections for offshore Stenella attenuata and 

 S. longirostris based on corresponding data for 16 lati- 

 tude-longitude blocks. 



Principal 

 component 



Canonical 

 variable 



Statistic 



I 



(I 



Product-moment 



correlation 

 Mantel lvalue 

 Matrix correlation 



0.745*** 



3.46** 



0.528 



0.289 



0.32 



0.039 



0.896*** 



6.65*** 



0.741 



-0.328 



1.62 



0.204 



*P<0.05; **P<0.01; ***P<0.001. P-values for Mantel tests based on 

 number of times permutational Z-values less than or equal to observed 



Z-values (one-tailed test). 



Table 10 



Results of Mantel tests, matrix correlations, and product-moment cor- 

 relations for 16 overlapping latitude-longitude blocks of offshore 

 Stenella attenuata and S. longirostris. Comparison of interlocality dif- 

 ferences for 30 morphological variables. 



P<0.05; **P<0.01; ***P<0.001. P-values for Mantel tests based on number of times per- 

 mutational Z-values less than or equal to observed Z-values (one-tailed test). 

 Abbreviations identified in Footnote 1 of Table 1. 



