184 
Fishery Bulletin 113(2) 
outliers by using a standardized sequential outlier re- 
jection routine (Kotwicki et al., 2011). Arithmetic means 
were then calculated by treatment for each trawl per- 
formance measure and for trawl speed. Because of the 
symmetrical placement (port and starboard) of all but 
one bottom contact sensor, the number of data sets 
produced that could be used to calculate means were 
double for the distances from the seabed to the foot- 
rope corners and wing ends and to 2 positions along 
the lower bridles than for the distance from the seabed 
to the center of the footrope (which was collected with 
a single sensor), or for acoustic measurements of door 
and wing spreads. 
Generalized linear models were used to compare the 
effects of different depths on our net performance mea- 
sures between the 3 towing treatments. Because we 
could not control for effects of current on trawl speeds 
but we suspected speed has an effect on wing spread 
(Weinberg et al., 2002), we included trawl speed in the 
model to account for any confounding effects it might 
have on the analyses in this study. Also, because wing 
spread directly correlates with headrope height with 
this low-opening style of net, we elected not to include 
the vertical opening of the trawl in the modeling pro- 
cess. Mean door and wing spread and mean distance off 
bottom at each position were treated as independent 
variables (Vj) in the following model: 
Vi -depth + trawl speed + treatment 
+ depth:treatment + trawl speed:treatment, (4) 
where the “treatment” term accounted for possible dif- 
ferences in the intercepts between treatments, and 
interaction terms accounted for possible differences 
in the slopes between treatments. Significance of the 
treatment term indicates difference in magnitude of 
the Vj. Significance of the interaction term indicates 
differences between treatments on the effects of depth 
or trawl speed on Vj. 
Terms in both models were selected with the help of 
the glmulti package, vers. 1.0.6 (Calcagno, 2012), in R, 
vers. 2.15.1 (R Core Team, 2012) by running each model 
with all possible combinations of terms and choosing 
the one with the smallest Akaike’s information crite- 
rion corrected for finite sample size (AICc; Burnham 
and Anderson, 2002). 
Results 
Of the 38 experimental hauls that were completed, 
14 occurred at the shallow site, 10 were conducted at 
the middle site, and 14 occurred at the deep site. The 
number of successful treatments used in the analyses 
of this study varied for each performance measure as 
a result of instrument failures and gear conflicts. Ta- 
ble 1 summarizes the successful efforts of the experi- 
ment. Although the target towing speed was 3.0 kn, 
tidal currents and weather affected the speed of the 
trawl at it moved through water such that the average 
trawl speed per treatment varied between 2.1 and 3.3 
kn. 
Door spread 
For our control treatment, mean observed door 
spreads increased 52% across depths from 46 m at 
the shallow site to 70 m at the deep site (Table 1). 
The GLM showed that each treatment affected the re- 
lationship between door spread and depth differently 
(Table 2, Fig. 3). Treatments with the restrictor line 
were quite effective at mitigating the effect of depth 
on door spread; the slope of the standard survey scope 
ratio was slightly positive and the slope of the modi- 
fied scope ratio was slightly negative across depths. 
On average, observed door spreads for the treatment 
with the restrictor line and standard scope ratio 
ranged between 45 and 51 m, and the treatment with 
the restrictor line and a modified scope ratio held door 
spread nearly constant, between 44 and 45 m (Table 
1). The effect of trawl speed was small and positive 
for all 3 treatments. The significance of the interac- 
tion term between trawl speed and treatment indicat- 
ed differences between the trawl speed effects among 
treatments, but the magnitude of these differences 
were small, positive, and ineffective in mitigating the 
effect of trawl speed (Table 2). 
Wing spread 
For our control treatment, mean observed wing spread 
increased 20% across depths from 15.1 m at the shal- 
low site to 18.1 m at the deep site (Table 1). The GLM 
showed that each treatment affected the relationship 
between wing spread and depth differently (Table 2, 
Fig. 4). As with door spread, treatments with the re- 
strictor line were very effective at mitigating the effect 
of depth on wing spread; the modified scope ratio per- 
formed slightly better than the standard survey scope 
ratio. On average, observed wing spreads for the treat- 
ment with the restrictor line and standard scope ratio 
ranged between 14 and 16 m, and the treatment that 
used the restrictor and modified scope ratio held wing 
spread nearly constant, between 14 and 15 m (Table 
1). The effect of trawl speed was small and positive for 
all treatments, but no differences between treatments 
were detected (Table 2). 
Distance of the footrope off bottom 
Changes in mean distances of the footrope off bot- 
tom were relatively small, 3.5 cm maximum across all 
depth sites. For our control treatment, mean observed 
distances off bottom across depth sites ranged from 1.8 
to 5.3 cm at the footrope center to 2.7 to 5.1 cm at the 
footrope corner and 3.4 to 4.1 cm at the wings (Table 
1) . The GLM indicated that both depth and trawl speed 
positively affected distances of the footrope off bottom 
at the center position, albeit to a small extent, but 
increased with increasing depth and speed; but only 
depth, and not speed, positively affected distances of 
the footrope corners and wings from the bottom (Table 
2) . The use of the restrictor line and modified scope 
