Bacheler et al.: Influence of soak time and fish accumulation on the catches of reef fishes 
225 
95% confidence interval was estimated as the 0.025 
and 0.975 quantiles of the 10,000 point estimates. 
Results 
The assumption of independence of predictor variables 
was met in our study. There were no statistically sig- 
nificant relationships between any of the predictor 
variables included in our analyses (VIF<4), and, in 
particular, soak and fshacc variables were unrelated 
(r 2 <0.001). Fishacc was the total number of individuals 
of all species caught in a trap, whereas the response 
variable (i.e., catch) was the total number of individu- 
als of each species caught in a trap. Moreover, no single 
species composed more than 20% of total trap catch 
(Table 2). 
Overall, 8530 samples collected from chevron traps 
in the 22-year span of MARMAP/SEFIS surveys were 
included in our analyses (annual mean: 388 [SD 179]). 
Sampling occurred between March and October, and 
there were no obvious changes in the seasonality of 
sampling across years. The spatial coverage of the 
surveys, in contrast, expanded southward in the mid- 
1990s to include sampling sites in central Florida. 
Of the 8 reef fish species included in our analyses, 
Black Sea Bass had the highest frequency of occurrence 
(0.408) and Sand Perch had the lowest (0.184; Table 
2) . Unstandardized catch per trap ranged from 0.6 for 
Sand Perch to 10.2 for Black Sea Bass. Mean lengths 
ranged from 15.6 cm fork length for Stenotomus spp. 
to 31.3 cm fork length for Gray Triggerfish (Table 2). 
The binomial GAMs explained 20.6-63.6% of the 
deviance in presence-absence patterns of the 8 reef 
fish species (Table 3). Models that explained the least 
deviance were the ones for Gray Triggerfish (20.6%) 
and Red Porgy (28.8%), and models for Black Sea Bass 
(63.6%) and Stenotomus spp. (61.4%) explained the 
most deviance. All 8 predictor variables were selected 
in the binomial model for each species on the basis of 
UBRE scores, with the exceptions of soak for Black Sea 
Bass and Bank Sea Bass and doy for Stenotomus spp. 
and Sand Perch (Table 3). The fishacc variable was re- 
tained in the binomial models for all species. 
The Gaussian GAMs explained 18.9-64.4% of the 
deviance in the positive-catch values of the 8 reef fish 
species (Table 3). The most deviance was explained by 
the models for Black Sea Bass (64.4%) and Tomtate 
(51.2%), and models for Gray Triggerfish (18.9%) and 
Red Porgy (21.4%) explained the least deviance (Table 
3) . On the basis of GCV scores, soak was excluded from 
the models for Black Sea Bass, Bank Sea Bass, Tom- 
tate, Vermilion Snapper, and Sand Perch. Moreover, cloy 
was excluded from the Vermilion Snapper model, and 
tod was excluded from the models for Vermilion Snap- 
per and Sand Perch. The fishacc variable was included 
in all 8 models. 
Over the range of soak values examined, predicted 
overall catch was invariant to soak for 3 species (Black 
Sea Bass, Bank Sea Bass, and Vermilion Snapper), posi- 
tively related to soak for 2 species (Red Porgy and Gray 
Triggerfish), and negatively related to soak for the re- 
maining 3 species (Tomtate, Stenotomus spp., and Sand 
Perch; Fig. 4). For most species, the influence of soak 
on the probability of obtaining nonzero catch was very 
similar to its influence on the estimated overall catch 
when present (Fig. 4). With a doubling of soak from 60 
to 120 minutes, estimated catch approximately doubled 
for Red Porgy (106% increase) and Gray Triggerfish 
(95%) but increased little for Black Sea Bass (20%), 
Bank Sea Bass (9%), and Vermilion Snapper (8%; Table 
4). The estimated overall catch of Sand Perch (—71%), 
Tomtate (-32%), and Sand Perch (-26%) declined when 
soak doubled (Table 4). Note that confidence intervals 
were larger for Gray Triggerfish and Stenotomus spp. 
than for the other 6 species. 
The relationship between the overall catch of the 8 
reef fish species and fishacc displayed one of 3 patterns 
(Fig. 5). The overall catch of Black Sea Bass, Tomtate, 
Vermilion Snapper, and Sand Perch generally was re- 
lated linearly to fishacc, indicating that the rate of 
catch of these species was not strongly influenced by 
the catch of individuals of all species in the trap. Alter- 
natively, the overall catch of Red Porgy, Bank Sea Bass, 
and Gray Triggerfish reached an asymptote at fishacc 
values between 50 and 100, indicating that individu- 
als of these 3 species were much less likely to enter a 
trap once 50 to 100 total individuals of all species were 
caught in it. The last pattern was displayed by Stenoto- 
mus spp., the overall catch of which was exponentially 
related to fishacc, indicating that Stenotomus spp. were 
more likely to enter a trap once large numbers of indi- 
vidual of all species were caught (Fig. 5). All 8 species 
had relatively narrow 95% confidence intervals sur- 
rounding overall mean estimates. 
Discussion 
Fishery-independent survey data form a critical com- 
ponent of modern stock assessments because they pro- 
duce indices of abundance that are assumed to vary 
in proportion to the actual abundance of a population 
(Pennington and Stromme, 1998). Whether or not indi- 
ces of abundance track actual abundance is a complex 
topic that has been the subject of much research (Har- 
ley et al., 2001; Kimura and Somerton, 2006). The basic 
assumption of this approach is that q does not change 
over space, time, or environmental conditions (Hilborn 
and Walters, 1992). In our study, we tested whether q 
was influenced by 2 additional variables, the length of 
time the trap was soaked and fish accumulation. For 
the species in our study, there was no clear relation- 
ship between life history traits and mechanisms of trap 
saturation (Table 1). We found evidence of trap satura- 
tion for most of the reef fish species examined, but the 
responses were species-specific. Trap saturation was 
observed for some reef fishes across the range of soak 
