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



325 



al. (1986). Numerous additional specimens have 

 become available, particularly from the western 

 portion of the range and Hawaii. The repeatability 

 of 36 skull measures used in previous studies (Dou- 

 glas et al., 1984; Schnell et al., 1986) was appraised, 

 as was done previously by Douglas et al. (1992) for 

 spinner dolphins. Also, some immature specimens 

 had inadvertently been incorporated into the previ- 

 ous spotted dolphin analyses. For these reasons, we 

 have undertaken a reassessment of geographic 

 variation and sexual dimorphism of spotted dolphins 

 from the eastern tropical Pacific. This study pro- 

 vides an opportunity to re-evaluate variation pat- 

 terns previously described and to compare directly 

 patterns of variation found in spotted and spinner 

 dolphins. 



Materials and methods 



Overall, data-gathering and assessment procedures 

 outlined by Douglas et al. (1992) were used. We 

 measured 611 adult museum specimens (maturity 

 evaluated on the basis of premaxilla fusion with 

 maxilla at distal end of rostrum; Dailey and Perrin, 

 1973) of spotted dolphins (Fig. 1). These included 

 534 of 613 specimens used in earlier studies (Dou- 

 glas et al., 1984; Schnell et al., 1986; 79 specimens 

 previously used had been incorrectly aged or had 

 inadequate locality data) along with 77 new specimens. 

 As was done with spinner dolphins (Douglas et al., 

 1992), the first specimen set was measured by M. 

 E. Douglas and the new specimens by W. F Perrin. 

 In addition, Perrin remeasured 81 specimens of 

 spinner and spotted dolphins measured by Douglas 

 to determine whether measurements were repeat- 

 able. Initially, 36 characters were evaluated (illus- 

 trations and character definitions given in Schnell 

 et al., 1985). Comparisons of measurements taken 

 on the same specimens by the two investigators in- 

 dicated that 6 of the original 36 measurements (i.e. 

 width of left premaxillary [at midline of nares], 

 width of right premaxillary [at midline of nares], 

 separation of pterygoids, length of left tympanic 

 cavity, length of right tympanic cavity, and width at 

 pterygobasioocipital sutures) should be deleted, be- 

 cause we were not able consistently to repeat these 

 measurements. For some other measurements, there 

 were differences between investigators, but the dif- 

 ferences were consistent (e.g. one obtained measure- 

 ments that were smaller than those reported by the 

 other). Therefore, we calculated regression equations 

 for each of the remaining characters based on the 

 81 jointly measured specimens. These regression 

 equations were used to convert the measurements 

 from the rest of the initial specimens to appropri- 



ate values for inclusion with the measurements 

 taken by Perrin. Through these procedures, we de- 

 veloped a data set of 30 characters (see Table 1) for 

 the 611 specimens. 



Specimens were not used if, because of damaged 

 parts, we could not obtain most of the 30 measure- 

 ments. Missing values (0.50% of total) for included 

 specimens were estimated by linear regression 3 onto 

 the character that explained the greatest proportion 

 of the variance for the variable being considered. 



Animals were assigned to 5° latitude-longitude 

 blocks and each geographic block assigned a numeri- 

 cal code (see Fig. 1). These codes were modified 

 slightly from those employed by Schnell et al. ( 1986) 

 to accommodate new specimens from more westerly 

 blocks. We had specimens available from 41 blocks, 

 8 of which were represented by only a single speci- 

 men and 4 of which were inshore blocks (i.e. con- 

 tained only specimens of the inshore form; Douglas 

 et al., 1984). The 29 blocks that were not inshore 

 blocks and had more than one specimen were used 

 as the basis for most geographic variation analyses. 

 While several of the 29 blocks have relatively small 

 samples, geographic-patterning tests (described be- 

 low) suggested that, in general, sample values are 

 representative of what is expected for these blocks 

 based on their geographic positions. 



Schnell et al. (1985) showed S. attenuata to be 

 sexually dimorphic for 23 of 36 characters. Because 

 some specimens used in that analysis were removed 

 and new specimens added (see above), we conducted 

 a two-way analysis of variance (ANOVA) for block 

 and sex, based on specimens in the 11 blocks with 

 at least four of each sex (Fig. 1). Correction terms 

 were obtained to adjust measurements of the larger 

 sex downward and the smaller sex upward, thus 

 producing sex-adjusted or "zwitter" measurements 

 (method described in more detail by Schnell et al., 

 1985). As a result, we were able to combine speci- 

 mens for both sexes in an overall analysis of geo- 

 graphic variation. 



To assess whether combining specimens from dif- 

 ferent cruise sets within blocks confounded geo- 

 graphic patterns based on blocks, we performed a 

 nested ANOVA for cruise sets within the 12 blocks 

 for which two specimens were obtained from at least 

 two cruise sets. Blocks employed in this analysis 

 (with numbers of cruise sets in parentheses) were 

 the following: 0215 (4), 0216 (3), 0312 (2), 0506 (2), 

 0507 (2), 0512 (9), 0513 (10), 0515 (2), 0612(6), 0613 

 (7), 0615 (3), and 0712 (2). 



3 "Missing Data Estimator" program by Dennis M. Power. Santa 

 Barbara Mus. Nat. Hist., pers. commun. 1975. 



