Underwood et at: Behavior-dependent selectivity of Limanda ferruginea in the mouth of a bottom trawl 
437 
Figure 3 
Percentages of different swimming directions ofyellowtail flounder (Limanda 
ferruginea) after they left the substrate for each initial response of (A) run 
and (B) slope observed in analysis of video footage from 5 tows of a bottom 
trawl in June 2010 on the southern Grand Bank off eastern Newfoundland. 
Orientations were related to the trawling direction, represented in the figure 
by the arrow. There were 112 observations of fish with a run response and 29 
observations of fish with a slope response. Observations from all 3 locations, 
port, starboard, or in the middle of the trawl, are combined together. 
in the mouth of the trawl than the nonherded fish (run, 
4.3 s, and slope, 3.0 s, versus rise, 1.1 s; Table 4) and 
maximum time spent in the mouth of the trawl was ob- 
served for individuals that had a run response (maxi- 
mum time: run, 31.9 s, versus slope, 6.5 s). 
Change in response 
The majority of yellowtail continued their initial be- 
havioral response to the footgear, but 40% of individ- 
ual yellowtail showed a behavioral shift (Fig. 4). The 
change in response always resulted in the fish moving 
farther away from the substrate. Location, swimming 
direction, fish length, gait, residence time, and start 
density alone did not have a significant influence on 
change in herding response, but initial response sig- 
nificantly influenced the observed change in herding 
response (Table 3). Almost all changes occurred in in- 
dividuals that initially exhibited run behavior (58 of 60 
individuals); in contrast, all individuals that initially 
moved upward continued to do so. 
Capture outcome 
Of the fish that left the substrate, 37% escaped through 
or under the footgear. Most fish escaped by swimming 
across the mouth of the trawl toward the outer foot- 
gear and by finding gaps in the footgear (40 of 55 in- 
dividuals; Table 4). The remainder collided with the 
gear or were passively overtaken by the footgear. Fish 
length, gait, residence time, and 
start density alone did not have a 
significant influence on capture, but 
initial orientation and response sig- 
nificantly influenced the probability 
of capture (Table 3). Individuals in 
the peripheral locations that were 
facing inward and were assumed 
to have been previously herded (in- 
dicated by the striped area in Fig. 
2) were twice as likely as all other 
individuals to be caught (caught-to- 
escape ratio, 3.0:1 and 1.5:1, respec- 
tively). Nearly all yellowtail (> 97%) 
that had an initial response of slope 
or rise were caught; whereas, only 
half of the fish that exhibited a run 
response (52%) to the footgear were 
captured. 
Discussion 
This detailed analysis of the be- 
havior of yellowtail in the central 
part of the mouth of an approach- 
ing bottom trawl revealed that in- 
dividual fish responded in different 
ways and that the response of a fish 
had consequences for its probability 
of being caught. Some fish swam along the bottom in 
front of the moving trawl (run behavior), whereas other 
fish gradually left the bottom (slope behavior) and oth- 
ers swam directly upward (rise behavior). The behav- 
ioral decision of a fish was linked to its initial orienta- 
tion. Fish that stayed in the middle location along the 
footgear tended to be oriented in the trawling direction 
and swam along the bottom, whereas fish at peripheral 
locations were usually oriented inward and swam up- 
ward. Leaving the bottom as a rule resulted in a fish 
being caught. 
There are some limitations to our study. For in- 
stance, we assumed that only individuals that were in 
peripheral locations and facing inward were herded by 
the gear before our observation, but we recognize that 
some of the fish in the middle location may also have 
encountered the gear but were not categorized as pre- 
viously herded. Likewise, some of the fish in peripheral 
locations could have maintained an inward orientation 
without having reacted to the gear. One solution to this 
conundrum may be the use of electronic tagging of in- 
dividual fishes (Engas et al., 1998; Winger, 2004), an 
approach that could provide information about gear en- 
counters by recording the position of an individual fish 
in relation to the sweeps before it enters the mouth of 
a trawl. Another potential bias in our analysis is the 
undercounting of fish that did not react to the footgear 
and were passed over by the trawl. These fish could 
not be distinguished from the substrate and, therefore, 
were not included in the database. 
