544 



Fishery Bulletin 98(3) 



For morphometric characters, significant differences 

 between species were identified using an ANOVA of arc- 

 sine-transformed ratios of the measurement divided by 

 SL or HL (Sokal and Rohlf, 1995) and an analysis of 

 covariance (ANCOVA) with SL or HL as covariates when 

 assumptions of homogeneity of slopes were satisfied. The 

 following morphometric characters did not significantly 

 differ in variances among species pairs and were subjected 

 to an ANOVA: head length, snout length, ocular-side max- 

 illa length, blind-side maxilla length, ocular-side mandi- 

 ble length, cheek length, interorbital width, dorsal orbit 

 length, ocular-side pectoral-fin length, body depth, caudal 

 peduncle depth, and caudal-fin length. Snout length, blind- 

 side maxilla length, ocular-side mandible length, and inter- 

 orbital width were also subjected to ANCOVA. 



To aid in the discrimination and classification of spe- 

 cies, standard principal components analysis (PCA) was 

 conducted on all morphometric and meristic characters for 

 adults of all species together and on the eastern North 

 Pacific material separately. Raw morphometric data were 

 log-transformed and the covariance matrix was subjected 

 to PCA, as was the correlation matrix of raw meristics. 

 Differences between species were illustrated by separately 

 plotting principal components (PC) 2 and 3 of the morpho- 

 metric analyses, PCI and PC2 of the meristic analyses, 

 and PCI of the meristic analyses against PC2 of the mor- 

 phometric analyses (Stauffer and Hert, 1992 1. For eastern 

 North Pacific species, data points were also identified by 

 reference to 11 geographic regions: Sea of Okhotsk, western 

 Bering Sea, eastern Bering Sea, Aleutian Islands, Gulf of 

 Alaska, British Columbia, Puget Sound, Washington coast, 

 Oregon coast, California coast, and Baja California coast. 



Larval statistical analyses 



In an effort to identify additional characters to distinguish 

 larvae of the two eastern North Pacific species, morpho- 

 metric characters of yolksac preflexion, flexion, and post- 

 flexion lai-vae were further analyzed. Scatter plots for each 

 measurement versus SL were made. Morphometric data 

 were analyzed by the ANCOVA model for each parameter 

 at each developmental stage and at all stages combined, 

 which included species as a factor, standard length as a 

 covariate. and a species/SL interaction (e.g. BD = C + Spe- 

 cies + SL + (Species x SL)). A residual analysis was done 

 for each model to determine the appropriateness of the 

 model. Whenever the interaction was not significant (at 

 the 5% level), a reduced model was used, and the inter- 

 action was dropped and the slopes were forced to be the 

 same (BD = C + Species + SL ). This procedure removed the 

 effect of SL and allowed testing for significant differences 

 between species. When the species were significantly dif- 

 ferent, the P-value was reported and graphs were gener- 

 ated to illustrate differences between elevations of the two 

 regressions. Principal components analysis was also used 

 to highlight differences. 



Results 



Adults 



All three species differed significantly from each other 

 in three meristic characters (Tables 3-7): posterior ADB 

 pores (Table 5), total scales above and below the lateral 



