456 



Fishery Bulletin 98(3) 



the silver hake samples, as indicated by the multi- and 

 univariate analyses (Fig. 5). We found significant differ- 

 ences in the discriminant scores (CV Ii between those 

 samples from the northern and southern regions for both 

 1-year-old (/=-6.3075, df=87, P<0.0001) and 2-year-old fish 

 (<=4.7654, df=95, P<0.0001). However, as in the preceding 

 analyses, poor separation was observed between the sam- 

 ples of 3-year-olds (/=1.7827, df=370, not significant). This 

 was again probably due to the reduced number of vari- 

 ables in the analyses, particularly that of otolith area and 

 length. Variation in otolith area, length, and circularity 

 was mainly responsible for the north-south separation in 

 samples of 1- and 2-year-olds, whereas otolith width and 

 circularity accounted for most of the variation in the sam- 

 ples of 3-year-olds (Table 5). 



Classification success, based on each particular discrimi- 

 nant model, decreased with each age group. Overall, individ- 



ual samples of 1-, 2-, and 3-year-old silver hake were correctly 

 classified 75'7f, 69'^, and 49'7f of the times, respectively 

 (Table 6). Classification rates increased slightly when sam- 

 ples from each age group were analyzed separately by sex 

 ( 1-year-old fish, males — 78''f ; 2-year-old fish, females — 68' f, 

 males — 717f; 3-year-old fish, females — 54%, males — 56'7r) 

 and year class (3-year-old fish: 1989 year class — 54'^f ; 1991 

 year class — 64C'f-; 1993 year class — GK/f ), indicating the con- 

 founding affects that these variables can have on these types 

 of analyses when used for stock discrimination. 



Discussion 



We confirmed that two stocks of silver hake exist in U.S. 

 waters of the northwest Atlantic from differences in whole 

 otolith morphometries and growth rates. Our method of 



