Bacheler et al.: Influence of soak time and fish accumulation on the catches of reef fishes 
229 
trap saturation. For instance, if catch reaches an as- 
ymptote because most or all of the individuals in a lo- 
cal area have been caught, then catch is likely a good 
index of abundance. But catch may not reflect actual 
abundance if trap saturation occurs because of space 
limitation in the gear, negative species interactions, 
increasing avoidance of the gear due to individuals al- 
ready being caught, handling time limitations, or bait 
deterioration or consumption of bait (Kennedy, 1951; 
Munro, 1974; Olin et ah, 2004). If traps become satu- 
rated at a level of catch unrelated to actual abundance, 
then statistical models for censored data may be useful 
(Bagdonavicus et ah, 2011). Another factor that may 
affect trap saturation is changes in feeding motivation 
of fishes with time of day or light levels. In our study, 
changes in feeding with time of day were not related 
to catch because all trapping was done during daylight 
hours. Light levels, therefore, were driven by water 
clarity more than by time of day. However, preliminary 
occupancy and W-mixture modeling for a few species 
has shown that water clarity does not appear to influ- 
ence the detection probability of traps. 2 In our study, 
we could not identify the exact mechanisms respon- 
sible for trap saturation; therefore, this topic clearly 
requires more research. 
Catches of Tomtate, Stenotomus spp., and Sand Perch 
declined with increasing soak times, indicating that at 
least some individuals of these 3 species may have es- 
caped from the trap. These results are consistent with 
our own observations and a growing body of literature 
that indicates that some fish, crab, and lobster species 
frequently escape from traps and pots (Jury et al., 2001; 
Cole et al., 2004; Sturdivant and Clark, 2011). Tomtate, 
Stenotomus spp., and Sand Perch were among the 3 
smallest fish species examined in our study, and their 
small size may have allowed them to escape through 
the narrow trap entrance more easily than could spe- 
cies of larger size. Size was the only life-history trait 
(Table 1) that influenced catch. An alternative explana- 
tion for the decreased catch of these 3 species during 
soak times over 100 min is that they had a longer time 
over which they may have been exposed to and eaten 
by predatory fish species caught in the same trap. We 
consider this explanation less likely because the diets 
of predatory fishes caught in traps only occasionally 
contain freshly consumed Tomtate and Stenotomus spp. 
(Goldman 3 ), but the 2 explanations are not mutually 
exclusive. Experimental work should be conducted with 
underwater video to quantify entry and exit rates of 
reef fishes in fish traps — research that can provide 
significant insights into the catch dynamics and spe- 
cies interactions of marine organisms (e.g., Jury et ah, 
2001; Cole et al., 2004; Sturdivant and Clark, 2011). 
2 Bacheler, N., and L. Coggins. 2012. Unpubl. data. Beau- 
fort Laboratory, Southeast Fisheries Science Center, National 
Marine Fisheries Service, NOAA, Beaufort, NC 28516. 
3 Goldman, S. 2012. Personal commun. Marine Resources 
Monitoring, Assessment, and Prediction Program, South Caro- 
lina Department of Natural Resources, Charleston, SC 29422. 
Delta-GAMs provided a convenient analytical ap- 
proach that helped us understand the influence of soak 
time and fish accumulation on the catch of reef fish 
species in a multispecies trap survey. The primary ben- 
efit of a delta-GAM is that the effects of soak time and 
fish accumulation can be understood after accounting 
for variation in total fish catch that might be due to 
all the other predictor variables in a model (Li et al., 
2011). By accounting for soak time, fish accumulation, 
and other predictor variables, we found an improve- 
ment over previous (primarily gill net) studies that 
focused on only those predictor variables that were di- 
rectly related to the gear saturation process itself (e.g., 
Minns and Hurley, 1988; Hansen et ah, 1998; Akiyama 
et ah, 2007). A secondary benefit of delta-GAMs is that 
they can account for zero-inflation. It is important to 
note that delta-GAMs, which are composed of separate 
submodels that must be combined, have been criticized 
as less elegant than the recently developed zero-inflat- 
ed GAMs to account for zero-inflation (Chiogna and 
Gaetan, 2007; Liu and Chan, 2011). In our study, zero- 
inflated models were challenging to work with because 
they rarely converged, and, when they did, model solu- 
tions were often unreasonable. 
There were some potential drawbacks of our experi- 
mental design and analyses. First, the range of soak 
times used in our study (50-150 min) was relatively 
narrow, and broader insights into the catch dynamics 
of species in traps would have been possible if large 
numbers of traps had been soaked for much shorter 
or longer periods of time. Second, the fishacc variable 
was made up partially of the catch of each individual 
species (Olin et ah, 2004; Li et al., 2011), but we do 
not consider this aspect of our study to be a problem 
because we were interested primarily in the shape of 
the relationship between catch and fish accumulation, 
not necessarily in the significance of fishacc in the 
delta-GAMs. Third, the fishacc variable did not distin- 
guish large, predatory species from smaller, nonpreda- 
tory species caught in the trap; future analyses could 
separate the catch of potential predators from smaller 
species. 
Conclusions 
We showed that the rate at which reef fish species 
entered traps in long-term programs of fishery-inde- 
pendent surveys decreased either over a range of soak 
times or over a range of fish accumulation levels. Trap 
saturation occurred for all 8 reef fish species that we 
examined; therefore, we recommend that future stud- 
ies use catch standardization on raw catch or CPUE 
data (Maunder and Punt, 2004). It is also extremely 
important to understand the exact mechanisms that 
cause fish to saturate fishing gears, and our results in- 
dicate that these mechanisms may vary considerably 
among species. Ultimately, whether catch or CPUE can 
be used to index abundance will depend on a clear- 
