326 



Fishery Bulletin 92(2). 1994 



09 



08 



07 



Lh 



05 



04 



03 



02 



01 



01 



03 



05 



07 



09 



15 



1S 



Figure 1 



Known range of Stenella attenuata in eastern tropical Pacific Ocean (based on Perrin et al., 1983), 

 with numbers of males (above) and females (below) available for each 5° latitude-longitude block 

 (total of 611 specimens). Each block is identified by numerical code (numbers on left and bottom mar- 

 gins are combined; e.g. block 0802 encompasses the northern Hawaiian Islands). Asterisks indicate 

 the 11 blocks included in analysis of sexual dimorphism. Initial analyses indicated that specimens in 

 four blocks (0516, 0517, 0518, 0812) represented a well differentiated inshore form (see Douglas et 

 al., 1984); thus, these were not used in subsequent evaluations. The remaining 29 blocks with two or 

 more specimens were used as the basis for analyses of geographic variation; for some evaluations, the 

 eight blocks with single specimens were projected onto axes based on the other 29 blocks. For the 16 

 blocks marked with a dot in lower-right corner, sufficient numbers of S. attenuata and S. longirostris 

 (i.e. at least two of each species) were available for interspecific comparison of geographic trends. 



Correlation, ordination, and clustering 



After conversion to zwitters, characters were stan- 

 dardized (means=0, standard deviations=l). Prod- 

 uct-moment correlations were computed among 

 characters, and associations among characters were 

 summarized by clustering characters (unweighted 

 pair-group method with arithmetic averages; 

 UPGMA). This technique is a type of hierarchical 

 cluster analysis that also was used to summarize 

 average distance coefficients (Sneath and Sokal, 

 1973) calculated for all pairs of blocks based on stan- 



dardized data. Cophenetic correlation coefficients 

 indicate the extent to which distances in resulting 

 dendrograms accurately represented original inter- 

 block morphologic distances. 



Standardized data also were summarized by us- 

 ing a nonhierarchical if -group method (function- 

 point cluster analysis; described in Katz and Rohlf 

 [1973] and Rohlf et al. [1979]). Through use of this 

 technique, blocks are assigned to subgroups at a spec- 

 ified level. A w-parameter value used in function- 

 point clustering was varied. An hierarchical, but not 

 necessarily nonoverlapping, system of clusters was 



