Sogard and Berkeley: Movement, growth, and survival of Anoplopoma fimbria off Oregon 
237 
fish by Kimura et al. (1998). All produced comparable 
results that defined the roles of initial size and time at 
large on growth. Because our objective was to assess 
potential effects of different environmental factors, we 
elected to use the nonlinear regression approach, which 
provided a simple means of incorporating additional 
parameters into the model and evaluating their infiu- 
ence. We anticipated that depth would infiuence growth 
rates, with reduced growth at the deepest depths sam¬ 
pled due to low temperatures, low oxygen levels, and 
low food availability. We also anticipated that recap¬ 
ture gear could affect growth because Kimura et al. 
(1993) had observed higher growth for sablefish caught 
in pots than in trawls. Preliminary analyses indicated 
that growth of fish recaptured by longline was similar 
to that of fish recaptured in pots; therefore, the groups 
of fish captured by the 2 fixed gears were combined for 
comparison with fish recaptured in trawls. 
We used a “best subsets” approach to evaluate the 
potential role of independent variables in growth, first 
including all factors in the nonlinear regression, then 
estimating subset models with individual factors re¬ 
moved. Because the growth model was not based on 
an explicit likelihood calculation, it was not readily 
adaptable to the commonly used Akaike information 
criterion approach for model comparison. Therefore, 
Mallows’s Cp (Mallows, 1973), an appropriate metric of 
model fit for least squares regression, was calculated to 
compare among different formulations of the model in 
order to evaluate support for the different potential ex¬ 
planatory variables. The full model, with 6 parameters, 
was based on the following equation: 
FL2 = FLl ^ days * exp (Pi * FLl 
P 2 * days -I- P 3 ^ sex - 1 - P 4 * depth 1 
•+ Ps * depth2 -I- Pe * gear), (3) 
where FLl 
FL2 
days 
sex 
depthl 
depth2 
gear 
= initial fork length; 
= recapture fork length; 
= days at large; 
= a dummy variable for sex ( 0 =female, 
l=male); 
= depth at initial capture; 
= depth at recapture; 
= a dummy variable for recapture gear ( 0 = 
fixed, l=trawl); and all P are coefficients. 
We then evaluated all possible subset models with 1-5 
parameters and calculated Mallows’s Cp for each to 
evaluate how well each model balanced parsimony and 
fit to the data. The full model has Cp equal to the num¬ 
ber of parameters by definition, whereas subset models 
that adequately account for variance in the data set 
but with fewer parameters will have a Cp that closely 
matches their reduced number of parameters. We ex¬ 
amined the Cp values to determine if a simpler model 
(i.e., one with fewer parameters) was appropriate for 
describing growth differences. 
We also examined growth of dispersing individu¬ 
als to determine whether their growth rates differed 
from those of fish categorized as residents and whether 
growth varied with the distance moved. We used resid¬ 
uals from the full growth model in ANOVAs, conducted 
separately by sex, to first compare growth of dispers¬ 
ing fish with growth of resident fish. For the dispers¬ 
ers only, linear regressions were then used to compare 
growth with the distance moved. 
All statistical analyses were performed with SYS- 
TAT, vers. 13, software (Systat Software Inc., San Jose, 
CA). 
Results 
Initial size distributions 
Size distributions at tagging differed by depth zone 
(Figs. 1 and 2). For the first set (1996-1998), 5291 fish 
were tagged from depth zone 1 and 2218 fish from 
depth zone 2. For the second set (2003-2004), 4923 fish 
were tagged from zone 2 and 4968 from zone 3. In zone 
1 there were 2 clear size modes at 34 cm and 44 cm 
FL (tagging set 1, Fig. 1) that were not present in the 
deeper zones. In zones 2 and 3, fish sizes followed a 
continuum, likely representing a broad range of fish 
ages. Modal size was larger (57 cm FL) in zone 3 than 
in zone 2 (54 cm FL). Very large fish (>70 cm FL) oc¬ 
curred at all depths but were relatively rare overall, 
with n=56 in zone 1 (1.1% of total), n=130 in zone 2 
(1.8%) and n=47 in zone 3 (0.9%). 
For both tagging sets combined, sex was determined 
for 690 recaptured fish. The sex ratio varied with ini¬ 
tial capture depth, ranging from 58% females in depth 
zone 1 to 62% in depth zone 2 and 81% in depth zone 
3, indicating an increasing bias toward females at in¬ 
creasing depths. 
Probability of recapture 
In total, 2614 tagged fish were recaptured up to De¬ 
cember 2016, with 1254 (16.7%) from tagging set 1 and 
1360 (13.7%) from tagging set 2. The potential roles of 
fish size and initial capture depth on the probability 
of recapture were tested with logistic regressions, con¬ 
ducted separately for each tagging set. For tagging set 
1, recapture rates increased with fish size (Wald sta¬ 
tistic: 10.4, P<0.001) and were greater in depth zone 2, 
with 23.6% fish recaptured, than in zone 1, with 13.8% 
fish recaptured (Wald statistic: 4.4, P<0.001). For tag¬ 
ging set 2 , recaptures again increased with fish size 
(Wald statistic: 13.7, P<0.001), but were greater in zone 
2 (20.9%) than in zone 3 (6.7%, Wald statistic: 22.1, 
P<0.001). 
One of the objectives with tagging set 2 was to deter¬ 
mine whether discard mortality is influenced by depth 
of capture or by the temperature gradient experienced 
by fish captured in cold deep waters and released in 
warm surface waters. Tagging set 1 was not included in 
this comparison because of limited sampling at warm 
surface temperatures and deeper depths. For tagging 
set 2 , recapture rates differed markedly between depth 
