BOEHLERT ET AL.: TIME SERIES OF GROWTH IN SEBASTES 



VENTRAL 



CORSAL 



S. diploproa 



S. pinniger 



Figure 1. — Schematic drawing of an otolith section from Sebastes spp. showing the axes of measurements. 

 Measurements for S. pinniger were from focus to dorsal distal increment margin (i.e., F to D2); measurements 

 for S. diploproa were from dorsal to ventral distal increment margins (i.e., V2 to D2). 



ception of those younger than 6 at the time of 

 samphng. As an example, a fish collected in 1980 

 and aged at 40 years was determined to be born 

 in 1940; therefore, growth measurements from 

 this individual were available for years 1940 

 (GKD) through 1945 (GI(6)). 



Most otoliths in our study were from males; 

 otoliths from older females were also included to 

 increase the sample size earlier in the time 

 series. Length at age for these species did not 

 differ between sexes until after sexual maturity, 

 which occurred after age 6 (Boehlert 1980; 

 Boehlert and Kappenman 1980). Still, to test for 

 differences in gi'owth between sexes, we sep- 

 arated the data by sex and then aggi'egated the 

 data for each GI(0 so that each year of birth had 

 a single, mean value. A paired f-test was used to 

 test for differences in respective gi'owth incre- 

 ments for all years in which both male and 

 female data were available. Growth did not sig- 

 nificantly differ between sexes for either species 

 (P > 0.10), so data were combined in subsequent 

 analyses. 



Certain aspects of the technicjue, associated 

 with errors in otolith increment measurement 

 and age estimation, may have led to variability 

 in results. From a methodological standpoint, 

 three errors were quickly apparent. First, small 

 changes in the location where the section was 

 removed from the otolith (Fig. 1) may have 

 resulted in slightly different increment measure- 

 ments; we expected this to introduce relatively 

 minor errors, however, since the sectioning 



technique (Boehlert and Yoklavich 1984) was 

 designed to be the same. The second source of 

 error occurred in the estimate of total age. 

 Errors in this estimate will result in the assign- 

 ment of an incorrect year of birth and, thereby, 

 incorrect years of gi'owth for each gi-owth incre- 

 ment; this type of error is jjrobably the most 

 serious, because it will tend to decrease the 

 reliability of real differences between adjacent 

 years. This type of error will increase with 

 increasing age (Boehlert and Yoklavich 1984), 

 and therefore, the reliability of between-year 

 differences in growth will decrease somewhat 

 with increasing age. Finally, errors in selection 

 of annuli (for example, selecting 2-7 instead of 

 1-6) can occur when making measurements on 

 the section. The cumulative effects of these 

 errors, averaged over many individual samples 

 for each year, should not significantly mask 

 long-term trends in the data. The errors could 

 mask correlations with high frequency environ- 

 mental features but should not affect correla- 

 tions with low frequency phenomena. 



An additional concern stemmed from the im- 

 plicit assumption that there is no hnkage be- 

 tween longevity and gi'owth; this concept, asso- 

 ciated with Lee's phenomenon, is related to 

 size-selective mortality (Ricker 1969). If long- 

 lived individuals are characterized by either 

 faster or slower gTowth rates during the first 6 

 years of life than are individuals with shorter 

 lifespans, then we can encounter problems com- 

 paring gi-owth of young and old fishes. In lightly 



793 



