Bolles and Begg: Distinction between sliver hake stocks In US. waters 



453 



sex was determined by macroscopic examination of the 

 gonads, and sagittal otolith pairs were removed from each 

 fish. One otolith from each pair was sectioned and assigned 

 an age following methods in Dery ( 1988). The other otolith 

 of the pair, either left or right, was then used for morpho- 

 metric analysis. Samples were restricted to fish of ages 1 

 to 3 years, because older fish were infrequently collected 

 in the surveys. This restriction also enabled both size- and 

 age-related variation, which could have confounded dis- 

 crimination analyses, to be minimized. 



Otolith orientation was standardized by positioning the 

 otolith with its proximal side down and the rostrum was 

 used as a common starting point from which the perimeter 

 was traced in a counterclockwise direction. Broken or vis- 

 ibly damaged otoliths were not measured. Whole otolith 

 length, width, area and perimeter, and two shape indices — 

 circularity and rectangularity, were collected for each oto- 

 lith by using the OPTIMAS''"'^' (version 6.2) image analysis 

 system (OPTIMAS, 1996). Rectangularity is a measure of 

 the otolith area divided by the area of its minimum enclos- 

 ing rectangle, and circularity is the perimeter of the otolith 

 squared divided by its area (OPTIMAS, 1996). All otolith 

 measurements were performed at a magnification of 7x. 



Data analysis 



Bottom ocean temperatures measured throughout the 

 NEFSC spring and autumn bottom-trawl sui-veys from 

 1989 to 1996 were compared between the northern and 

 southern regions to determine whether different thermal 

 regimes exist between the two regions that could effect 

 growth rates and subsequent otolith morphometric char- 

 acteristics of silver hake inhabiting each of these regions. 

 A two-way, fixed-factor, unbalanced analysis of variance 

 (ANOVA) was used to compare bottom temperatures 

 between the regions and time of the survey. Following 

 a significant interaction between these factors, unpaired 

 f-tests were used to compare bottom temperatures between 

 regions for each survey, and one-way ANOVAs were used 

 to compare bottom temperatures between surveys for each 

 region. Significance levels were corrected for multiple test- 

 ing by using the Bonferroni adjustment factor Tukey's 

 honestly significant difference (HSD) tests were used for 

 a posteriori comparisons. 



Growth rates of silver hake sampled throughout the 

 NEFSC spring bottom-trawl surveys were calculated by 

 using linear regressions for samples of each sex, year, and 

 region of capture to determine if growth differences existed 

 that might be indicative of stock separation. Analysis of 

 covariance (ANCOVA) was used to compare differences in 

 growth rates of silver hake between the sexes (same year 

 and region), sampling years (same sex and region), and 

 regions (same year and sex). 



All otolith morphometric variables were first examined 

 for normality and homogeneity of variances, and were log^- 

 transformed prior to statistical analysis if these criteria 

 were not satisfied. ANCOVA was then used to determine 

 the effect of fish length on the magnitude of each otolith 

 morphometric variable. "Region" was treated as the main 

 factor and "length" was the covariate. Morphometric vari- 



ables for which "region-length" interactions were signifi- 

 cant were not included in any further analyses because 

 they could not be corrected for fish length. Variables that 

 were significantly correlated with fish length were cor- 

 rected for variable fish length by using the common within- 

 group slope (6). 



Multivariate analysis of variance (MANOVA) was used 

 to compare otoliths sampled for each sex (same region, 

 year class, and age group) by using the appropriate length- 

 corrected variables. One-way ANOVAs were then used to 

 examine individual morphometric variables to explain any 

 significant differences detected by the MANOVAs. Signifi- 

 cance levels were corrected for multiple testing with the 

 Bonferroni adjustment factor Likewise, multi- and uni- 

 variate analyses were then used to test the effects of year 

 class (samples from the same region and age group) on oto- 

 lith structure. Following these analyses, similar tests were 

 used to investigate morphological differences in otolith 

 samples from the northern and southern regions. Spatial 

 comparisons were made between samples from the differ- 

 ent regions for fish of the same age group, sex, and year 

 class. Canonical discriminant analysis was then used to 

 detect morphometric differences in otoliths of silver hake 

 sampled from the northern and southern regions. Sig- 

 nificant (P<0.05) canonical variates represented the opti- 

 mal combination of regions and morphometric variables 

 that provided the best overall discrimination between the 

 groups. Standardized coefficients provided for each sig- 



