668 



Fishery Bulletin 98(4) 



had no obvious latitudinal pattern and were relatively 

 similar (0.16 to 0.21) across areas. Overall, median age at 

 maturity of both male and female Dover sole appeared to 

 decrease with increasing latitude. 



Length-weight curves 



Length-weight cui-ves for all areas combined were highly 

 significant (P<0.001 ) for pooled-sex, male, and female sam- 

 ples (Table 6). Males had a larger length exponent than 

 females but there was little apparent difference between 

 male and female curves. For pooled-sex samples, the length 

 exponent exceeded 3 and showed that Dover sole weight in 

 autumn was more than proportional to the cube of length. 

 CVs of intercept parameters ranged from 3'7f to 5'~l for 



15 



25 



35 



45 



pooled-se.xes, males, and females, whereas CVs of length 

 exponents were all less than 1%. Bootstrap estimates of 

 parameter bias were low compared with standard errors 

 and were less than O.S^f of point estimate in all cases. For 

 the intercept parameter, ratios of bias to standard error 

 were -1%, 5%, and 6% in pooled-sex, male, and female 

 samples. Corresponding ratios for the length exponent 

 were 2%, -2'7f, and -4%. 



Geographic variation in length-weight relationships 

 of Dover sole was suggested by the generalized linear 

 model analyses. The best model contained the terms for 

 \og{length), area, and \og{length)xa>-ea . Although length- 

 weight curves by INPFC area exhibited some geographic 

 variation (Fig. 5), there was no detectable difference 

 between male and female weight at length. Predicted 

 weight at length was smallest for the Vancouver area 

 and greatest for the Monterey area, whereas 

 length-weight curves from the Columbia and 

 Eureka areas were very similar Length expo- 

 nents exceeded 3 for all areas except Van- 

 couver, where few length-weight data were 

 available. As above, the length exponent (with 

 CVs less 29c) was more precisely estimated 

 than the intercept (with CVs of 5-19%). 

 Estimates of parameter bias were less than 

 1.6% of the parameter estimate in all cases. 

 For each area, the ratio of estimated bias 

 to standard error for the intercept para- 

 meter was also low: Vancouver (9%), Colum- 

 bia (0%), Eureka (0%), and Monterey (5%). 

 Corresponding ratios for the length exponent 

 were also low: 1%, 2%, 7%, and -2%. Thus, 

 parameter bias for the length-weight cui-ves 

 was inconsequential. 



55 



15 25 35 45 55 



Length (cm) 



Figure 3 



Maturity-at-length curves for female lA) and malo (Bi Dover sole by 

 INPFC area. 



Discussion 



We found that the use of the alternative form 

 of the von Bertalanffy growth curve gave satis- 

 factory parameter estimates that had greater 

 precision than those of the standard form. 

 Bootstrap estimates of standard errors and cor- 

 responding coefficients of variation were lower 

 with the alternative form. Ratkowsky (19831 

 recommended the alternative form used in our 

 study because its parameters were more read- 

 ily interpreted, it had close-to-linear behavior 

 that facilitated comparisons among data sets, 

 and it was insensitive to the choice of initial 

 parameter values. Our results provide further 

 support for its use. 



Variance in size at age of females was 

 40% greater than that of males. This may 

 result from differences in seasonal patterns 

 of movement between females and males. In 

 particular, Hagerman (1952) noted that com- 

 mercial Dover sole catches were often segre- 

 gated by sex and that Dover sole appeared 

 to undertake an inshore feeding migration 



