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



337 



association with environmental 

 component III. Table 8 indicates 

 that the pattern for length of ros- 

 trum (from pterygoid) is associ- 

 ated statistically with those for 

 water depth and thermocline 

 depth (summer). This morph- 

 ologic character also is shown to 

 have geographic patterning statis- 

 tically similar to that exhibited by 

 environmental components II and 

 III. For length of braincase, the 

 Mantel tests were significant (but 

 weak) only for sea current (N., 

 winter) and solar insolation 

 (Jan.). The strongest association 

 of length of braincase is with en- 

 vironmental component III; its 

 pattern also is linked statistically 

 to the second environmental com- 

 ponent. 



Interspecific comparisons 



The study by Douglas et al. (1992) 

 reported comparable statistical 

 analyses on skulls of S. longiro- 

 stris, a dolphin species that over- 

 laps broadly with S. attenuata in 

 the eastern tropical Pacific. The 

 projections onto the first two prin- 

 cipal components for S. attenuata 

 were evaluated against projec- 

 tions on the two components ob- 

 tained for S. longirostis (for sum- 

 mary information on these compo- 

 nents, see Fig. 3 and Table 4 for 

 offshore S. attenuata, and fig. 3 

 and table 3 of Douglas et al. 

 [1992] for S. longirostris). A 

 strong correspondence exists be- 

 tween the first principal compo- 

 nents for the two species, as indi- 

 cated by product-moment correlations, Mantel t- 

 tests, and matrix correlations comparing the com- 

 ponent projections. The second principal components 

 for the two studies are not similar (Table 9); they 

 summarize different general trends in variation. 



A similar interspecific comparison was made of 

 projections of the 16 blocks onto canonical variables 

 (Table 9). The first canonical variable for S. 

 attenuata and that for S. longirostris are virtually 

 identical, reflecting highly concordant geographic 

 patterns for the two species (Table 9). While block 

 projections on the first canonical variables for the 



two species were very similar, the actual characters 

 incorporated into the canonical variables are not the 

 same (see our Table 5 and table 4 of Douglas et al. 

 1992). Of the five characters entered for each spe- 

 cies, only length of rostrum (from pterygoid) was 

 present in both character sets. However, the first 

 and most important character entered into the 

 analyses — preorbital width for S. attenuata and pos- 

 torbital width for S. longirostris are highly corre- 

 lated (see close association of these two characters 

 in S. attenuata indicated in Fig. 2); because these 

 two characters exhibit very similar variation pat- 



