236 
Fishery Bulletin 115(2) 
Table 2 
Releases of tagged sablefish (Anoplopoma fimbria) off Newport, Oregon, in September 2003 and May 2004 
(tagging set 2), with mean depth of sampling, surface temperature, fish sizes (measured in fork length [FL]), 
number of fish tagged, and number of fish recaptured for each sampling trip. All fish were captured with pots 
from the same vessel. 
Trip (month/year and depth zone) 
9/03 zone 2 
9/03 zone 3 
5/04 zone 2 
5/04 zone 3 
Mean depth (m) 
355 
1145 
353 
1163 
Surface temperature 
category and mean 
Warm, 15.9°C 
Warm, 16.9°C 
Cool, 14.6°C 
Cool, 14.5“C 
Number of fish tagged 
2460 
2486 
2463 
2482 
Size range (cm FL) 
42.5-85.5 
44.0-90.0 
44.5-78.0 
46.5-87.5 
Mean size (cm FL) 
54.3 
57.7 
57.4 
60.1 
Number of recaptured fish 
and percentage of total tagged 
435 (17.7%) 
140 (5.6%) 
594 (24.1%) 
191 (7.7%) 
capture data were used depending on the availability 
and precision of information reported. Sample sizes are 
noted for each analysis in the results section. Where 
practical, recaptures from both tagging sets were com¬ 
bined for analysis. 
Data analysis 
The potential roles of fish size and initial capture depth 
on the probability of recapture were tested with logistic 
regressions conducted separately for each tagging set 
and by using the following equation; 
Logit (P(x)) = a -F PiZi + ^ 2 X 2 , (1) 
where Pix) 
Zi 
^2 
a 
P 
probability of recapture; 
initial fish length (continuous variable); 
depth zone of initial capture (categorical 
variable); 
a constant; and 
a coefficient. 
For tagging set 2, additional logistic regressions tested 
recapture rates between warm and cool surface tem¬ 
peratures (categorical variable) at the time of tagging. 
Initial inspection of these data suggested that surface 
temperature effects were more evident in smaller fish; 
therefore, logistic regressions were conducted sepa¬ 
rately for small (<55 cm FL), medium (55-65 cm FL) 
and large (>65 cm FL) size classes within each depth 
zone. These categories correspond with the size groups 
tested by Davis et al. (2001) for the rate of body core 
temperature increase after transfer to warmer water. 
Regressions for comparisons of recapture probabil¬ 
ity by surface temperature were calculated with this 
equation: 
Logit iP(x)) = a + PiXj, (2) 
where Xi= a temperature category. 
For all logistic regressions, the Wald statistic was used 
to determine significance of each coefficient. 
Analysis of small-scale movement among depths was 
possible for 1762 fish. We used analyses of covariance 
(ANCOVAs) calculated separately by initial depth zone 
to test for effects of recapture season (spawning: from 
November through April and nonspawning: from May 
through October) on recapture depth, with fish size and 
time at large included as covariates. 
For recaptured fish with location coordinates, great 
circle distances from tagging to recapture location were 
calculated. We divided recaptured fish into residents, 
defined as fish recaptured within 200 km of their tag¬ 
ging locations, and “dispersers,” fish recaptured >200 
km from their tagging locations, according to Beamish 
and McFarlane (1988). Fish without precise location 
coordinates could be categorized as dispersers or resi¬ 
dents on the basis of the general area of recapture; for 
example, fish caught in Alaska waters were dispers¬ 
ers regardless of their exact capture location. Using 
chi-square association tests and analysis of variance 
(ANOVA), we examined characteristics of these fish to 
determine whether the tendency to make long migra¬ 
tions from the tagging location was influenced by fish 
size, sex, or the depth of initial capture. We used linear 
regression to determine whether the proportion of fish 
dispersing was affected by the time at large. 
We compared the growth of individuals whose re¬ 
capture length and sex had been recorded, as well as 
recapture depth and gear used. Growth was expected 
to be influenced by fish size, with smaller fish growing 
at a faster rate than larger fish, and by time at large, 
with growth rate decreasing as the time at large in¬ 
creased. To select an appropriate growth model we first 
considered several age-based models that have been 
reparameterized for size-based application to tag-re- 
capture data, including von Bertalanffy, Gompertz, and 
Schnute growth models, as described by Francis (1995), 
and a nonlinear regression model described for sable- 
