354 
Fishery Bulletin 11 7(4) 
Table 2 
The top-5 combinations of fixed effects in generalized additive mixed models used to pre¬ 
dict 1) the percentage of time smalltooth sawfish (Pristis pectinata) were active (Activity), 
2) burst probability, and 3) burst frequency. Degrees of freedom (df), log likelihood (logLik), 
and the corrected Aikake’s information criterion (AICc) were used to determine the best-fit 
models, which are indicated with asterisks. The burst frequency model described by only 
hour of day (HOD) was selected as the most parsimonious model, with relative differences 
between AICc values (AAICc) <2 for the top model. All models included individual as a ran¬ 
dom effect. Data used in the models were collected during deployments of acceleration data 
loggers in the Peace River, Florida, between May 2014 and November 2015. 
Model 
df 
logLik 
AICc 
AAICc 
Activity 
-HOD + Temp + Depth + Tide* 
13 
-4068.5 
8137.0 
-HOD + Temp + Depth + Tide + Age class 
14 
-4068.1 
8137.9 
1.2 
-HOD + Temp + Tide 
11 
-4072.8 
8140.9 
4.5 
-HOD + Temp 
6 
-4082.2 
8150.3 
13.4 
-HOD + Temp + Age class 
7 
-4082.1 
8150.9 
17.5 
Burst probability 
-Age class* 
3 
-1914.6 
3835.3 
-Depth 
4 
-1915.9 
3839.7 
4.4 
-Depth + Age class 
5 
-1915.2 
3840.4 
5.1 
-Tide 
7 
-1916.0 
3845.9 
10.6 
-Tide + Age class 
8 
-1915.2 
3846.4 
11.1 
Burst frequency 
-HOD + Age class 
5 
-1021.6 
2053.2 
-HOD* 
4 
-1022.8 
2053.6 
0.4 
-HOD + Tide 
9 
-1018.0 
2053.9 
0.7 
-HOD + Depth + Age class 
7 
-1020.3 
2054.6 
1.4 
-HOD + Depth + Tide + Age class 
12 
-1015.5 
2054.9 
1.7 
likely to burst in the main stem of 
Purported foraging behavior 
the river than in mangrove creek habitats (Tukey’s HSD: 
P<0.05; Fig. 7), and a greater burst frequency was observed 
in the main channel; however, this difference was not sig¬ 
nificant (Tukey’s HSD: P>0.05; Fig. 7). 
Smalltooth sawfish that were fitted with ADLs remained 
in a small portion of the Peace River (Figs. 1 and 7). Both 
size classes remained in enclosed creek habitats during 
the day and in more open habitats in the main stem of the 
river at night. 
Discussion 
This study, the first to deploy ADLs on smalltooth saw¬ 
fish, provides behavioral data that describes their fine- 
scale activity and behavior patterns, including potential 
foraging strategies. Acceleration data indicate that a 
number of factors, including time of day, tidal period, 
depth, age class, and habitat, affect activity and forag¬ 
ing rates in this species. These data will increase our 
understanding of how this critically endangered species 
behaves in its nursery habitats throughout juvenile life 
stages and will assist in management and conservation 
of crucial nursery areas. 
Foraging behavior of sawfish species is seldom observed 
in natural habitats and is generally not well documented, 
although the toothed rostrum is known to aid in stunning 
and pinning prey as well as in electroreception (Wueringer 
et al., 2011, 2012). Because validation trials could not 
be performed to match acceleration traces with specific 
behaviors in our study, it was impossible to identify for¬ 
aging activity with certainty from the acceleration data 
we collected (Whitney et al., 2018). However, by examining 
video footage of feeding sawfish in captivity (Wueringer 
et al., 2011, 2012), and by considering the results of 
ground-truthing trials for data collected on acceleration 
behavior of the largetooth sawfish (senior author and 
A. Gleiss, unpubl. data), we made some deductions about 
the relation of the burst events observed in our study to 
putative foraging activity. Conservatively, it is likely that 
any foraging activity displayed by ADL-tagged smalltooth 
sawfish was included in the 98th percentile burst event 
mask determined for each fish, although bursts tied to 
predator escape or startle responses may also be included 
in this subset of data. 
Many of the burst events we recorded were accompanied 
by acute ascents (see Fig. 3B), indicating that smalltooth 
