526 



Fishery Bulletin 102(3) 



increased our estimate of fork length by approximately 

 .'!.'!'< Applying grade-specific DCFs increased these es- 

 timates by another 19% (to 36.6 cm) for otolith length 

 and by 24% (to 33.7 cm) for a combination of otolith 

 length and width (Table 2). All six remaining struc- 

 tures in good or fair condition provided larger corrected 

 mean length estimates than did otoliths alone, but 95% 

 confidence intervals derived from otoliths did overlap 

 with other structures (Table 2). The smaller estimate 

 provided by otoliths may reflect that >83% of grade 

 corrected otoliths (n = 34) came from the inside haul- 

 out sites, where animals seem to eat smaller fish (see 

 following section). On the other hand, large pollock 

 otoliths were observed to have been regurgitated in 

 feeding studies on captive sea lions and also may be 

 more easily crushed by rocks often found in the stom- 

 achs of Steller sea lions (Tollit et ah, 2003). Regres- 

 sion formulae used in our study to predict pollock FL 

 from otolith length were similar to those of Frost and 

 Lowry (1981) for juvenile fish (<10 mm otolith length) 

 but led to smaller fish size estimates (—1—1.7 cm) over 

 the range of 30-50 cm. 



The lengths of the biggest fish (corrected mean 

 FL = 48.0 cm) were estimated from measurements of 

 INTE (Table 2), the structure with the lowest DCF. 

 Dentary bones (the most abundant structure recov- 



ered) predicted mean (Table 2) and modes (FL 44- 

 50 cm) similar to those predicted from all structures. 

 Applying DCFs increased our length estimates by be- 

 tween 9% (INTE) and 29% (DENT) (paired r- tests, all 

 P<0.001). Overall, corrected fork length estimates from 

 elements in good condition were similar to those from 

 elements in fair condition (Mann-Whitney U, P=0.47), 

 but multiple comparisons indicated a significant differ- 

 ence (P<0.05> between condition categories for INTE 

 and DENT (lengths were estimated to be longer from 

 elements graded in fair condition). 



Repeat measurements of individual elements were 

 all within 0.04 mm of the mean, and 88.9-99.5% were 

 within 0.02 mm. The highest variability was associated 

 with ANGU, HYPO, and PHAR with 88.9%, 91.7%, and 

 94.9% of their respective measurements falling within 

 0.02 mm. A 0.02-mm measurement difference corre- 

 sponded to only a 0.1-0.2 cm difference in fish length, 

 depending on the structure used. 



Small differences in estimates of fork lengths can 

 have large effects on estimated body mass (given the 

 exponential mass-length relationship, see Tollit et al., 

 2004, this issue) and can increase the mean mass of 

 fish by more than sixfold depending on which method 

 is used to estimate body length (all otoliths and no cor- 

 rection versus condition-corrected structures). The ap- 



