Woltmann and Sherry • SURVIVAL AND TERRITORY DYNAMICS OF ANTBIRDS 
17 
TABLE 1. Models of apparent survival ((p) and detection probability (p ) of Chestnut-backed Antbirds at La Selva, 
Costa Rica. “# Par” is the number of parameters included in each model; “a2” indicates a 2-age class system (juveniles 
and adults). 
Model 
AICc 
AAICc 
AICc weights 
Model likelihood 
# Par 
Deviance 
{^Pa2 Psexl 
275.306 
0.000 
0.554 
1.000 
4 
91.277 
{<P- Psexl 
276.988 
1.682 
0.239 
0.431 
3 
95.053 
{cPsex Psexl 
279.050 
3.744 
0.085 
0.154 
4 
95.021 
{CPa2*sex Psex 1 
279.434 
4.128 
0.070 
0.127 
6 
91.139 
(CP- Pa2*sexl 
280.762 
5.456 
0.036 
0.065 
5 
94.613 
{ Cp a 2*sex Pa2*sexl 
283.020 
7.714 
0.012 
0.021 
8 
90.358 
{CPa2 P-l 
287.201 
11.895 
0.001 
0.003 
3 
105.266 
{ CPa2 Pa2l 
289.238 
13.932 
0.001 
0.001 
4 
105.208 
{CPa2*sex Pi 
290.485 
15.179 
0.000 
0.001 
5 
104.336 
f<p- p-l 
290.688 
15.382 
0.000 
0.001 
2 
110.824 
{CPsex Pi 
290.928 
15.622 
0.000 
0.000 
3 
108.994 
{cp. pa2l 
291.788 
16.482 
0.000 
0.000 
3 
109.853 
{CPsex Pa2l 
292.195 
16.889 
0.000 
0.000 
4 
108.165 
3 days apart. Chestnut-backed Antbirds have 
relatively small territories, respond readily to 
playback, and are most responsive to dry-season 
playback, making it unlikely we incorrectly 
declared a territory unoccupied. 
Data Analyses .—We used Program MARK, 
Version 5.1 (White and Burnham 1999) to 
evaluate Cormack-Jolly-Seber (CJS) models of 
apparent survival (cp) and detection (reobservation 
or recapture) probability (p) of territorial birds, 
and to generate parameter estimates of survival 
and detection probability. Models were construct¬ 
ed to examine effects of age (juvenile vs. adult) 
and males or females on estimates of (p and p, and 
were ranked using AICc values. We had no a 
priori reason to expect year effects, nor did we 
have sufficient data to test for them after 
individuals were grouped into age and male/ 
female classes. We used U-CARE Version 2.2 
(Choquet et al. 2005) to test the most parameter¬ 
ized model for evidence of transience and trap- 
dependence. We used the median c approach in 
MARK to test goodness-of-fit of a general model 
to ascertain whether adjustments to model rank¬ 
ings were necessary. The first capture interval in 
our data set was 0.25 (Dec 2004 to Feb 2005), but 
all subsequent intervals were set to one. We 
considered the best supported models as those 
with AAICc < 2, and used AICc-weighted model 
averaging (over all models) to derive estimates of 
cp and p in MARK. 
There were 111 detection events involving 77 
individuals (54 males and 23 females). The U- 
CARE analysis indicated no significant problems 
of transience (P = 0.290) or trap-shyness (P = 
0.464). Data were too sparse in some age-sex 
classes to evaluate a fully time-dependent model. 
A goodness-of-fit test on the next most parame¬ 
terized model, { cp a ge*sex Page*sexh produced an 
estimated median c of 0.991 (SE = 0.018), not 
meaningfully different from 1, and no adjustments 
to model rankings were made. 
We describe annual territory dynamics of 
Chestnut-backed Antbirds following Greenberg 
and Gradwohl (1997) and Fedy and Stutchbury 
(2004), and include only territories within the 
300-ha focal plot. Annual turnover includes all 
changes in territorial positions (i.e., birds that 
disappeared from the study area plus birds known 
to have switched territories). Both turnover and 
switching are presented in terms of “territory- 
years” (i.e., a marked bird monitored from 1 year 
to the next), calculated separately for males and 
females; these are minimum estimates of turnover 
and switching (Morton et al. 2000). 
RESULTS 
Apparent Survival and Detection Probability .— 
The two best models (AAICc < 2) indicated 
differential detection of males and females, but 
differed in whether or not cp varied by age 
(Table 1). Estimates of cp ± SE of adults based 
on weighted model averaging were relatively high 
(males: 0.794 ± 0.037; females: 0.798 ± 0.050) 
compared to estimates of cp for independent 
juveniles (males: 0.629 ± 0.159; females: 0.629 
± 0.168; Fig. 1). Detection probabilities were 
higher for males (adults: 0.916 ± 0.034; juve¬ 
niles: 0.915 ± 0.049) than for females (adults: 
0.544 ± 0.104; juveniles: 0.540 ± 0.115). 
