FISHERY BULLETIN: VOL. 81. NO. 1 



o 



O 



a 



03 

 < 



to 

 >> 



ccj o 



<D © 



c V 



« !2 E en 



«- o Q- c 



(0 



co o> £ 



■o 2 



Q. C 



■O <"i> 



mi? 



< _ 



co Q) 



I CD 



CO -D ' 

 O 1> n 

 Q. 



il t 1) 

 i- CO *~ 



-So* 



2 <= o) 



o — c 



CD 0) 



CL — 



CD 

 CO 



CO 

 CO 



t- o o 

 ■* ^ m 



C\J C\J OJ 



CD 



O 

 CO 



O) in 



O CM 



o og 

 CO o 



t- CO 



t- d 



en t- o) 

 CD CD o 



CM CO 



in o 



CO 



co o 



1 O) 



in o 



The nature of the variation in the morphomet- 

 ric characters examined varied between sexes 

 and locations (Tables 3, 4). For several characters 

 the least square mean values appeared to vary 

 clinally. This was most evident for male adipose 

 flap height and orbit diameter as seen in plots of 

 raw data (Figs. 4, 5), female interorbital width 

 and male head length. The values for other char- 

 acters showed less consistent patterns and in 

 some cases could be interpreted to suggest two 

 distinct groups with the South Carolina samples 

 most similar to Mid-Atlantic Bight groups 

 (Tables 3, 4). This was most obvious for male pec- 

 toral fin length and female pectoral fin length, 

 caudal peduncle depth, and head length. Clinal 

 variation was also suggested by the increasing 

 number of significantly different morphological 

 characters with increasing geographic distance 

 between compared samples. 



The discriminant function analysis was con- 

 ducted with both raw and size-corrected data. In 

 each case the results were virtually identical 

 with two exceptions (males, east Hudson Can- 

 yon - 60% correct classification with size correct- 

 ed vs. 23% raw data, and Campeche - 86% correct 

 classification vs. 43% raw data). We believe 

 neither of these significantly affects the overall 

 interpretation of the results, and we report the 

 raw data results here (Tables 6, 7). 



The discriminant function analysis suggests a 

 similar clinal pattern of variation for both males 

 and females (Tables 6, 7). There was generally 

 low differentiation within the Mid-Atlantic 

 Bight samples, and where misidentification oc- 

 curred it was to other Mid-Atlantic Bight or 

 South Carolina samples and infrequently to west 

 Florida and the Gulf of Mexico off Texas. Gulf of 

 Mexico samples naturally had higher percent- 

 age correct classification (sample locations were 

 more widely separated geographically) and in- 

 correct classifications were usually to other Gulf 

 of Mexico samples. Classifications for South 

 Carolina samples had a high correct classifica- 

 tion, and where misclassification occurred it was 

 to both Mid-Atlantic Bight and Gulf of Mexico 

 locations. 



DISCUSSION 



For purposes of interpreting the significance 

 in allelic frequencies observed for IDH and EST 

 we are assuming that the genetic variation ob- 

 served is neutral (Allendorf and Phelps 1981; 

 Ihssen et al. 1981). Thus, based on the patterns 



46 



