HEDGECOCK ET AL.: GENETIC VARIATION IN PACIFIC SARDINES 



Morphological Variation 



In contrast to the similarity of age class com- 

 positions among population samples, size distri- 

 butions among sites are grossly different, with 

 sardines from California being much larger than 

 those from Mexico (Fig. 2). Stepwise discrimi- 

 nant function analysis (DFA) of the 12 log- 

 transformed morphometric variates reveals, 

 after 10 steps, significant differences among the 

 five population samples (approximate F = 

 22.085 with 32, 499 df, P « 0.001). However, 

 discrimination is based primarily on log of stan- 

 dard length which enters the discriminant func- 

 tion first with F = 243.49 (P << 0.001; 4, 142 

 df). A principal component analysis (PC A) of 

 those traits contributing to between-gi'oup vari- 

 ance in the DFA produces a single factor, 

 heavily and positively loaded by all traits and 

 accounting for 97% of the variance in data space 

 (minimum factor eigenvalue set to 1.0); such a 

 factor is generally interpreted to represent 

 variance in size (Humphries et al. 1981). The 

 evident geogi'aphic cline in size apparently re- 

 flects a cUne in gi'owth rate; at the extremes, 



fish of the same age from central California and 

 from the Gulf of California can differ in stan- 

 dard length by nearly 100 mm (see Figure 2). 



Two further analyses do indicate minor but 

 significant morphological variation attributable 

 to shape differences among sardines from differ- 

 ent geographic areas. First, two factors ex- 

 tracted in a PCA of log-transformed variates 

 standardized by subtraction of the log of stan- 

 dard length show complementary patterns of 

 factor loadings suggestive of different allo- 

 metries of head size relative to body size among 

 populations (Fig. 3A). The separation of popula- 

 tions along the factor 1 axis (Fig. 3B) is inversely 

 related to standard length; i.e, larger California 

 fish are on the left of smaller Mexican fish, owing 

 to the negative allometry of head dimensions 

 relative to standard length. This result is consis- 

 tent with observations on postlarval Pacific sar- 

 dines (McHugh 1950) as well as with a general 

 geographical pattern in fish morphology (Jensen 

 1944; Martin 1949). 



Second, comparisons of pairs of population 

 samples for which there is considerable overlap 

 in the sizes of specimens (Southern California 



1 2 3 



1-3 



Age 



12 3 4 12 3 



2 3 



250 



O) 



c 



0) 

 -O 200 



■o 



c 



CO 



150- 



10 



If 



GUAYM 



MAGDA 



1 



TOMAL 



SOCAL 



MONTE 



Figure 2. — Histograms of standard lengths for various ages of Pacific sardines in five population 

 samples. Open arrowheads along the baselines of the Magdalena Bay and Southern California Bight 

 population samples indicate the mean sizes of two- and three-year-old fish in the 1961-62 sardine catches 

 of Baja Cahfornia and California, respectively (fi'om Vrooman 1964). 



659 



