510 
THE WILSON JOURNAL OF ORNITHOLOGY • Vol. 123. No. 3. September 2011 
TABLE 1. Candidate model set accounting for variation in hourly provisioning rate among nine time periods at two 
study sites. Fort Simpson, NT and Chinchaga, AB. Data include 591 hrs of observations from 37 Swainson’s Thrush nests. 
Model LL 
K 
AIC, 
AAIC, 
L(modellx) b 
AIC. 
Base + study site X time period -1,279.83 
21 
2,603.29 
0 
1 
1 
Base + time period - 1,314.07 
12 
2,652.68 
49.38 
1.89 X 10 " 
0 
Base" -1,568.99 
4 
3,146.05 
542.76 
1.38 X 10"" 8 
0 
Base + study site -1,568.91 
5 
3,147.92 
544.62 
5.45 X 10-" 9 
0 
a Base model includes nestling age + nestling number. 
“ Model likelihood given data x = cxp(— Vi AAIC, I. 
variables as main effects. These analyses were 
restricted to the night period. 
We used logistic regression to examine if the 
probability of fledging one or more young was 
greater for nests provisioned at night. We used a 
zero-truncated Poisson regression to examine if 
more young were fledged from successful nests 
that were provisioned at night. We used linear 
regression to examine if night-provisioned young 
fledged at an earlier age compared to non-night 
provisioned young. 
RESULTS 
We viewed 591 hrs of video from 37 nests. 
Twenty-nine nests were viewed at ~4 and 8 days 
of age whereas eight nests were only available to 
be viewed at one age. We viewed all nine time 
periods for 34 nests. We were unable to view the 
night period on three occasions, once at Fort 
Simpson and twice at Chinchaga, because of 
camera infrared failure. We excluded these days 
from all analyses focused on night provisioning. 
Variation in hourly provisioning rate was best 
explained by addition of an interaction between 
study area and time period to the base model 
(Table 1). The evidence ratio in support of this 
interaction model was 5.29 X I0 10 over the next 
best supported model. The overwhelming support 
for this model occurred because provisioning 
during the night period was not recorded at 
Chinchaga. Removing the night period from the 
analysis did not provide support for a study area 
by time period interaction (Fig. 1) indicating 
similar prey delivery rates at all other times in 
the two study areas. 
We found weak evidence that nests provisioned 
during the night period had lower provisioning 
rates during six of seven remaining periods 
compared to nests not provisioned during the 
night period (Table 2). We restricted these 
analyses to Fort Simpson where we recorded 
provisioning during the night period and excluded 
the single observation day with camera failure, 
The best supported model included night provi- 
sioner (i.e., provisioned during the night period) 
status as an interaction with time period. The 
AIC,- values of all models in the candidate set 
were within 2.25 suggesting addition of night 
provisioncr status added little explanatory power 
to the overall model. The evidence ratio indicates 
the interaction model was 3.1 times more likely 
compared to the base model. Support for the 
interaction reflects an increased provisioning rate 
of - 2/hr by night provisioners over non-night 
provisioners during post-dusk compared to other 
periods (Fig. 2). Non-night provisioners made an 
average (± SE) of 1.55 ± 0.18 additional 
provisioning trips/hr compared to night provision¬ 
ers during the remaining periods (pre-dawn to late 
pm). Night provisioners made an estimated 2.11 
(95% Cl: 1.61-2.77) trips/hr during the night 
period. 
FIG. 1. Predicted hourly provisioning rate by Swain- 
son's Thrush at Fort Simpson. NT (gray bars), and 
Chinchaga, AB (white bars) from the best-fitting model 
(Table I). Nestling age and nestling number are set to 
average values of 6.19 and 3.60, respectively. 
