260 



Fishery Bulletin 97(2), 1999 



This study generally confirmed the criteria used 

 for ageing young sablefish. Even though a high pro- 

 portion offish were misaged, most fish were misaged 

 by only one year. In addition, after reexamining the 

 otoliths after the known ages had been revealed, the 

 primary reader could reconcile the differences be- 

 tween the known and assigned ages in most cases 

 (25 of 31). Anderl and Heifetz- have described three 

 types of misinterpretation of otolith patterns that 

 resulted in most misages: 1) misinterpretation of an 

 ambiguous check (i.e. false annulus) immediately 

 following the first annulus; 2) misinterpretation in 

 assessing whether the most recent years annulus 

 had been formed; and 3) misinterpretation of mul- 

 tiple checks on parts of the otolith (a check is a mark, 

 growth zone, or part of a growth zone on an otolith 

 that does not form annually but reflects various en- 

 vironmental or physiological changes [Chilton and 

 Beamish, 1982]). 



Analysis of ageing errors 



Table 2 summarizes the parameter estimates for 

 various model specifications from the data in Figure 

 2. The estimates based on between-reader variabil- 

 ity include only data where both primary reader and 

 tester readings were available. The data consisted 

 of/ = 44 fish and J = 2 readings per fish resulting in 

 ;i := 88 observations. The estimates based on reader 

 and known ages were derived from data where at 

 least one reader determined an age. The data con- 

 sisted of / = 44 fish with J = 2 readings per fish and 

 7=4 fish with J = 1 reading per fish resulting in n = 

 92 observations. 



For the estimates that included only between- 

 reader variability, a model with all three parameters 



^ Anderl, D., and J. Heifetz. 1999. An evaluation of ageing crite- 

 ria for sablefish ba.sed on known age specimens. In prep. 



estimated (case 1 ) provided the best fit with the few- 

 est parameters (i.e. lowest AIC values ). The AIC value 

 for a model with or constrained to (case 2) was nearly 



