Van Waerebeek: Variation in skull morphology of Lagenorhynchus obscurus 



759 



mality were encountered in any of the characters (Kl 

 S,P>0.5). 



Sexual dimorphism The overall size of the adult 

 dusky dolphin skull, based on condylobasal length and 

 zygomatic width, is equal in males and females it- 

 tests, P>0.50; Tables 3 and 4). Highly significant sexual 

 dimorphism (/-tests, P<0.005) was found in the width 

 of the rostrum over most of its length (RW60, RW1/4, 

 RW1/2, PMX1/2, RW3/4) and in the length of the tem- 

 poral fossa; males had consistently higher mean widths. 

 Analyses of covariance showed sexual dimorphism in 

 exactly the same variables and with very similar sig- 

 nificance levels as in the two-sided /-tests, indicating 

 proportional differences (shape). However, it is uncer- 

 tain whether the observed, small dissimilarities have 

 biological meaning, particularly as there is wide over- 

 lap in ranges. I was unable to produce a satisfactory 

 model to predict the sex of an unknown animal through 

 stepwise variable selection in a ^multiple regression 

 (see Douglas et al, 1986) exactly because of those subtle 

 deviations. 



Twenty-four of 36 characters showed significant cor- 

 relations with the condylobased length (CBL) covariate, 

 including five of the six dimorphic characters (Table 

 4), implying that much of the variability is explained 

 by variation in CBL. Eliminating dimorphic charac- 

 ters from the geographic analysis therefore entails little 

 loss of information. 



Seasonal variation An analysis of seasonal variation 

 was feasible only for Peruvian specimens. Skull size 

 (CBL) was not seasonally dependent (MANCOVA, 

 F=1.45, df=3, P=0.23) but six other skull characters 

 (RW60, RW1/4, APL, RAL, TCUL, and TCUR) demon- 



strated seasonal variation at a 

 low level of significance 

 (MANCOVA, 0.05 < P < 0.02). 

 However, at least in the case of 

 RW60, RW1/4, and RAL, signifi- 

 cant interactions due to sexual 

 dimorphism in these measure- 

 ments seemed to be the cause (P- 

 values for 2-factor interactions 

 were respectively 0.039, 0.047, 

 and 0.037). In APL, TCUL, and 

 TCUR, which are sexually mono- 

 morphic characters (see above), 

 there were no sex/season inter- 

 actions (P>0.50). The significance 

 in these few cases could be due 

 to chance in view of the large 

 number of /-tests. This is sup- 

 ported by the results of a series 

 of Runs tests (Runs above and 

 below median, Runs up and down, STSC Inc., 1989) 

 on chronologically ordered CBL, tooth counts and tooth 

 widths. These results did not suggest any non-ran- 

 domness (Z>0.05 for all tests). 



Geographic variation Exclusion of the few sexually 

 dimorphic characters allowed combining samples of 

 males and females and the use of cranially mature 

 skulls of unknown sex. Highly significant differences 

 between geographic regions were confirmed for 29 of 

 31 cranial characters (for all ANOVA, P<0.005 and 

 mostly P<0.0001; Table 5). Tympanic bulla width (BW) 

 proved divergent between groups but with less confi- 

 dence (ANOVA, F=4.1, df=3,117, P=0.019) and no dif- 

 ference could be detected for periotic length (ANOVA, 

 F=0.24, df=3,201, P=0.87). Variance generally was ho- 

 mogeneous (Cohran's C-test and Bartlett's test, P>0.01 

 and mostly P >0.05); exceptions include TCLL 

 (Cohran's, P=0.006) and PEL (Cohran's, P=0.003; 

 Bartlett's, P=0.009). 



ANCOVA analyses revealed discernible variation in 

 16 of 30 skull characters which can not be attributed 

 to divergences in skull size alone but are due to pro- 

 portional differences (i.e., shape); 11 characters were 

 highly significantly different between geographic groups 

 (P<0.01, Table 5). Means and 95^f confidence intervals 

 are plotted for each variable against geographic region 

 (Fig. 2). The most striking element is the overwhelm- 

 ing difference in skull size, reflected in most measure- 

 ments, grouping skulls from Peru and Chile and speci- 

 mens from New Zealand and Southwest Africa. Skulls 

 of dusky dolphins from the latter pair-group have, on 

 average, a CBL about 3 cm shorter than skulls from 

 Peru and Chile. 



