Frable et al: A new species of Sebastes 
359 
was assigned a priori on the basis of trunk color pat- 
tern (blotched versus solid). To allow for the inclusion 
of juveniles, which lack distinguishing color patterns, 
and for faded long-preserved specimens in the morpho- 
metric analysis, we assigned specimens from extreme 
ends of the known geographic ranges to the only spe- 
cies known to occur in such regions. No ambiguous 
juveniles or faded specimens from the known region 
of geographic overlap were included in morphometric 
analysis. 
The overall morphometric distinctiveness of the 2 
putative species was tested by inputting the eigen- 
vectors that explained the greatest percent variance 
from the PCA into a multivariate analysis of variance 
(MAN OVA) implemented in PAST software. The sig- 
nificances of differences between the means of putative 
species (P<0.05) on each axis were tested with pairwise 
post-hoc comparisons, by using Tukey’s honestly sig- 
nificant difference (HSD) test. A discriminant function 
analysis (DFA, also implemented in PAST) was used to 
further determine which morphometric measurements 
best separated the species. The robustness of the DFA 
to assign the data to a priori groupings was evaluated 
with a leave-one-out cross-validation in PAST. Results 
of this procedure are reported in percentages assigned 
to the same a priori grouping. The untransformed 
versions of the 32 variables that explained the high- 
est proportion of variance in the DFA were regressed 
against length (standard length [SL] ) for each species 
to determine whether the traits differed allometrically 
or isometrically. This comparison of allometric trajec- 
tories was performed in the smatr package (Warton et 
ah, 2012) in R, vers. 3.0.1 (R Core Team, 2013). Linear 
regressions were plotted with the ggplot2 package in R 
(Wickham, 2009). Some individual linear morphometric 
measurements were compared to determine statistical 
significance with a 2-tailed Student’s t-test for unequal 
frequencies implemented in R. 
Results 
Genetic analysis 
When K was 2, all of the individuals genotyped from 
this study were assigned high ancestry (>90%) with 
1 of the 2 clusters (Fig. 2). The specimens identified 
as S. diaconus or blue-sided sensu Love (2011) clus- 
tered unambiguously with the type-1 blue rockfish and 
the S. mystinus or blue-blotched individuals clustered 
with the type-2 blue rockfish of Burford and Bernardi 
(2008) (Fig. 2), thereby confirming that the color poly- 
morphism reliably separates the genetic types. 
Morphometric analysis 
Principal component analysis of size-standardized lin- 
ear morphometric variables for specimens with clear 
a priori classification (see the Materials and methods 
section) revealed overlap between the 2 putative spe- 
S. diaconus S. mystinus Type 1 Type 2 
Figure 2 
Structure plot derived from microsatellite data show- 
ing genetic group assignments of specimens of Sebastes 
diaconus (deacon rockfish; n= 20) and S. mystinus (blue 
rockfish; n=15) from this study (on left) compared with 
specimens previously genotyped by Burford and Ber- 
nardi (2008) (on right), type 1 (n= 8) and type 2 (n= 8) 
with a K value of 2. Each bar represents one individual 
with color corresponding to population assignment. 
cies (Fig. 3) but a clear difference in mean phenotype 
(P<0.0001, MANOVA on first 4 axes explaining 51.0% 
of total variance). Percent variance explained dropped 
after PC4 (Fig. 4A). The most discrimination was of- 
fered by PCI (17.1% of variance; Table 1), with blue- 
sided individuals, or S. diaconus , having on average 
more positive scores than members of S. mystinus, or 
the blue-blotched phenotype (Fig. 3). The variables 
with the highest loadings along PCI were symphyseal 
knob length, the lengths of the first 2 anal-fin spines, 
and the length of the first dorsal-fin spine (Table 1). 
Explaining 12.1% of variance, PC2 primarily indexes 
variation in suborbital depth, anal-fin spine I length, 
and the dorsal and ventral lengths of the caudal pe- 
duncle (Table 1). Pairwise Tukey’s HSD tests recovered 
significant differences between the 2 species on PCI 
(P=0.0001) but not on the subsequent axes. 
When individuals under 150 mm SL were removed 
from the data set, the percent variance explained by 
the first 2 principal components (PCs) increased (Fig. 
4B). For PCI (23.1% of total variance), the main mea- 
surements of variation were symphyseal knob length, 
anal-fin spines I and II lengths, and ventral caudal 
peduncle length (Table 1), and, for PC2 (13.4% of to- 
tal variance), the largest loadings were for symphyseal 
knob length and lengths of anal-fin spines I, II and III 
(Table 1). As with the analysis of the complete data 
set, the multivariate means of the 2 species differed 
significantly (MANOVA, PcO.001): only the analysis of 
PCI revealed a significant morphometric difference in 
the pairwise Tukey’s HSD tests (P=0.001). 
The depth at dorsal-fin origin, pelvic-fin ray length, 
preanal fin length, and dorsal-fin origin to anal-fin ori- 
gin received the highest weight in the linear discrimi- 
nant equation of the variation between the 2 species 
(Table 2) in the size-standardized morphospace. This 
discriminant function correctly reclassified all but 2 
individuals in a leave-one-out cross-validation (98.3% 
correct reclassification; Fig. 5A). The exclusion of juve- 
niles under 150 mm SL improved the performance of 
