40 
Fishery Bulletin 110(1) 
the /3 parameters 
and the intercept 
Si 
D, 
= estimated regression coefficients; 
the developmental stage; 
the sampling period; 
SjXD - the interaction of stage and sampling 
period; and 
e., = the normal error. 
One model was run where each maternal factor was the 
response variable: age, length, or weight. Length and 
weight met the assumptions of normality and age was 
log transformed. 
The relationship between developmental stage and the 
maternal factors was evaluated with nine linear regres- 
sions, one for each sampling period that had adequate 
sample sizes (n =24-28). Length and weight 
met the assumptions of normality and age 
was log-transformed. A significant, posi- 
tive relationship would indicate that older 
or larger fish have more developed larvae 
than younger, smaller females sampled at 
the same time. 
Three linear regressions were used to 
quantify the relationship between weight- 
specific fecundity (eggs per gram of ovary- 
free body weight) and the three maternal 
factors. 
Results 
Stage 
Figure 3 
Average oil globule diameters (OGD) of embryos from 89 gravid 
female quillback rockfish ( Sebastes maliger) (black squares) versus 
the developmental stage. Average OGD and 95% confidence intervals 
from embryos at each developmental stage are denoted by open circles. 
The solid line is a polynomial fitted to the average OGD, where each 
female’s embryos or larvae are all from one stage. 
20 
18 - 
16 - 
14 
12 
1 10 
o 
8 
6 H 
4 
2 -| 
0 
10 
20 
60 
70 
80 
Figure 4 
Distribution of ages of gravid quillback rockfish ( Sebastes maliger) 
binned by 5-year increments (n = 89). 
Oil globule diameter increased, then de- 
creased curvilinearly through stage (Fig. 
3) (7i = 89 females; age range = 10-74 years, 
Fig. 4). A polynomial fitted the data 
better than other methods (e.g., linear or 
squared) and was used to predict OGD 
(intercept = 3.51xl0 _1 , /3 1 =1.04xl0 -2 , /3 2 = 
-1.79xl0' 3 , r 2 =0.65, P«0.05). We defined 
13 developmental stages based on physical 
characteristics associated with develop- 
ment (Table 1, Fig. 2). The range of OGDs 
within a stage increased after stage 4 
(Fig. 3). 
Regressions of lengths and weights and 
OGDs, adjusted for their confounding with 
stage, were both statistically significant 
at ce=0.05, but correlations were low. The 
strongest, positive correlation was between 
weight and OGD (Fig. 5). Age and OGD 
were not significantly related (Fig. 5), but 
the linear fit had a positive slope. The in- 
teraction of sampling period and stage was 
not significant and was excluded from all 
three models. Sampling period and stage 
were significant in all three models. 
The stage of development was positively 
related to age, length, or weight in eight 
out of nine linear regressions (Fig. 6). This 
positive relation indicates that older, larger 
females have further developed embryos 
and that preparturition larvae within a 
sampling period and likely will parturi- 
ate earlier than younger, smaller ones. For 
every 35 mm increase in length (20% of 
range in lengths) the predicted increase in 
stage was 0.8, 1.0, or 1.8 stages (depending 
on the sampling period). For every 350 g 
