438 
THE WILSON JOURNAL OF ORNITHOLOGY • Vol. 124. No. 3. September 2012 
we primarily use data from February as that month 
was sampled more regularly. We use the term 
observation to refer to both visual and auditory 
records of individuals: the great majority of 
observations were auditory. Locations ot all birds 
seen or heard were noted on scale maps ot the plots 
as one of us (JGB) w alked slowly with many stops 
along transects. - 1.0-1.4 km was covered during 
a morning; starting positions were distributed 
throughout the plots and rotated between monthly 
samples to ensure, as much as possible, that all 
parts of the plots were covered early in the morning 
when vocal activity was greatest. Each plot took 
— 12-13 days to cover: transects were not covered 
more than once during a given sample. Total effort 
expended (i.e.. number of hrs and number of km) 
was equivalent between plots and among samples. 
Observations started well before light and contin¬ 
ued for up to 3 hrs of the morning. Periods of rain 
occasionally interrupted or ended counts early. 
Statistical Analyses: Capture Data. —Sample 
effort (number of mist-net hrs. where one mist net 
open 1 hr equals one mist-net-hr or I mn/hr) 
varied slightly among samples. Consequently, 
some comparisons of number of captures were 
based on captures per 100 mn/hr. We used all 
captures or only individuals (recaptures omitted) 
depending on the comparison. Capture rates 
(birds/100 mn/hr) and number of individuals 
captured per year were compared between plots 
using paired /-tests. We used Pearson's correlation 
coefficient to test if annual variation in captures 
was similar between plots. We used coefficients 
of variation and variance-ratio tests to compare 
variation in capture rates between plots. Mean 
recapture distances were calculated for individu¬ 
als with > three captures; individual means were 
used to calculate plot means (individuals as 
replicates). Mean recapture distances per individ¬ 
ual were compared between plots using /-tests. 
Comparisons of mean between-year recapture 
distances were based on the distance between 
the last capture location in I year and the first 
capture location in the subsequent year. 
Spatial distribution of captures (i.e.. clumped, 
random, uniform; also for observations) was 
evaluated using Program PASSaGE Version 
2.0.10.18 (Rosenberg and Anderson 2011). PASS¬ 
aGE computes means and variances from counts 
(e.g., number of coptures/net) and calculates a 
series of indices to describe patterns of variation. 
We used the Index of Dispersion (ID: based on 
the variance-to-mean ratio) as an indication of 
whether distribution of captures was clumped (ID 
> 1.0). random (ID = 1.0). or uniform (ID < 1.0). 
Departure from random distribution was evaluated 
with a Chi-square test (Rosenberg and Anderson 
20! 1). We also provide values of Morisita’s Index, 
the scaled probability that two points chosen at 
random are in the same quadrat (or from the same 
mist net): higher values indicate a more clumped 
distribution. 
We used PASSaGE to calculate spatial auto¬ 
correlation indices (i.e.. correlograms: Moran’s I) 
to examine if number of captures at one location 
was correlated with captures at nets at different 
distances (using 50-m increments, the closest 
distance between nets). Moran's 1 (Moran 1950) 
ranges from 1 to - I with an expected value of -0 
for large sample sizes and no spatial autocorrela¬ 
tion. Significance levels of correlations were 
examined with permutation tests. 
Net locations were classified by habitat and we 
used Chi-square tests to examine if number of 
captures differed among habitats with expected 
numbers based on the number of nets in different 
habitat categories. 
Statistical Analyses: Observation Data. —Num¬ 
ber of observations per year (Feb samples) was 
compared between plots using paired /-tests. 
Variation across years was compared between 
plots with correlation analyses. We used several 
approaches in PASSaGE to examine distribution 
patterns of observations. We created a 50 X 50-m 
(0.25 ha) grid overlay for each plot (i.e.. 400 grid 
cells per plot) and counted the number of 
observations per grid cell; we selected 50-m grid 
si/e as that corresponds to the mean distance 
between nets. We calculated dispersion indices 
(ID) for each plot based on the distribution of 
counts among cells, as was done for number of 
captures per mist net. We used these same grid¬ 
cell data to calculate correlograms and Moran's 1 
to examine spatial autocorrelation among cells. 
Significance of correlations was examined with 
permutation tests. 
We used second-order statistics (Ripley's K- 
function; Rosenberg and Anderson 2011) to 
evaluate patterns of clumping based on all 
observations within the plot. Second-order statis¬ 
tics are based on the co-occurrence of pairs of 
points and answer the question: are there more 
points within a given distance from a specified 
point than one would expect by chance (Rosen¬ 
berg and Anderson 2011). If true, points are 
spatially aggregated. Ripley's /f-function is a 
