338 



Fishery Bulletin 92(2). 1994 



terns, a canonical-variates analysis typically would 

 not select both for inclusion, since they provide ba- 

 sically the same information for separating blocks. 

 The second canonical variables for the two stud- 

 ies also were compared (Table 9). They showed no 

 statistical association. 



Table 10 includes results of Mantel tests, matrix 

 correlations, and product-moment correlations of in- 

 terspecific comparisons for individual morphological 

 characters. Thirteen of the 30 lvalues for Mantel 

 tests of interlocality differences for the same char- 

 acter in the two species were significant, while 15 

 of 30 product-moment correlations 

 indicated statistical associations. 

 Nine of 11 characters with positive 

 correlations were width measures. 

 Furthermore, a tenth (length of 

 antorbital process) is essentially a 

 width character as well (for illustra- 

 tion of measurement, see Schnell et 

 al. 1985). The two characters involv- 

 ing upper tooth counts, as well as 

 length of temporal fossa, exhibited 

 significant negative correlations. For 

 S. attenuata, upper tooth counts 

 tend to be higher for the western 

 blocks (but not for the Hawaiian Is- 

 land block), whereas in S. 

 longirostris, higher upper tooth 

 counts are found in the Hawaiian 

 and eastern blocks. The length of 

 temporal fossa is greater in northern 

 localities of S. attenuata (Fig. 10A), 

 whereas the shorter fossae are found 

 in northeastern blocks for S. longir- 

 ostris (see Douglas et al. 1992: fig. 11). 



Discussion 



Sexual dimorphism 



Schnell et al. (1985) conducted the 

 most recent analysis of sexual di- 

 morphism of S. attenuata in cranial 

 morphology. They found statistically 

 significant dimorphism for 23 of 36 

 characters. Our analyses used many 

 of the same specimens, with some 

 added and some deleted, and 30 of 

 the same characters. For the 30 

 characters we analyzed, Schnell et 

 al. ( 1985) found the same 22 to have 

 statistical differences between sexes 

 (one statistically significant charac- 

 ter analyzed earlier was not used in 

 our analysis). Results from the two 

 studies on sexual dimorphism are 

 essentially the same. Thus, for S. 

 attenuata, our current findings sim- 

 ply update information in Schnell et 



