Vargas-Castro et al. • SONG OF THE CLAY-COLORED THRUSH 
449 
syllable or song type sharing (Mundinger 1982, 
Hughes et al. 1998. Molles and Vehrencamp 
1999, Rasmussen and Dabelsteen 2002, Nicholson 
et al. 2007). We printed spectrograms of each 
syllable type in the repertoire of all males and 
mixed them into a single syllable pool. Three 
different observers blindly looked for matches 
between syllable types. A pair of syllables from 
two different males was considered to be a shared 
type only if all observers agreed on that match, 
We re-examined similarity of shared types 
using the spectrogram correlator tool of Raven 
Pro 1.4 software (Cornell Laboratory of Ornithol¬ 
ogy, Ithaca. NY, USA). The correlator compares 
two syllables to each other and provides a 
correlation value based on the similarity of the 
frequency-temporal pattern of the sounds. We 
applied a 900-6,000 Hz band pass filter and 
selected the normalize and linear power options in 
the correlator configuration settings. We random¬ 
ly selected 10 different renditions of that syllable 
type from each male for each pair of shared 
syllables, and ran the correlator using those sets of 
renditions to obtain 100 correlation values. We 
used the largest number of renditions available if 
the number of renditions of a given syllable type 
sung by one male was ■ 10 (2 cases). We also 
selected 10 syllables for each pair of males at 
random from both males and compared those 
random sets to obtain 100 additional correlation 
values. Mean spectrogram correlations of shared 
and random syllables were later analyzed in a 
paired comparison. 
We calculated the proportion of repertoire 
sharing between pairs of males using the formula: 
2N s /(R i + R 2 ), following McGregor and Krebs 
(1982). N„ is the number of shared syllable types 
among two males and R| and R> are the repertoire 
sizes of each male, respectively. Wc substraeted 
the proportion of repertoire sharing from I to 
obtain a dissimilarity measure of the repertoires 
between pairs of males. 
Statistical Analysis .—We used a paired /-test to 
examine w'hether males have as many types of 
shared syllables as indiv idual syllables (exclusive to 
each male) in their repertoires. We used a /-test to 
examine if the relative frequency of individual and 
shared syllables in the songs was not different from 
that expected by the proportion of individual and 
shared types in the repertoires. All tests were two- 
tailed; assumptions of normality and homogeneity 
of variance w ere met (Zar 1996). We conducted a 
Mantel test using repertoire dissimilarity values and 
distance between song perches to test if males that 
were closer to each other shared a higher proportion 
of their repertoires (Sokal and Rohlf 1995). The P- 
value of the Mantel lest correlation was calculated 
from 9.999 permutations. We used R 2.14.0 (R 
Development Core Team, Vienna, Austria) to 
conduct statistical tests. 
RESULTS 
The song of Clay-colored Thrushes consists of 
a sequence of syllables that is highly variable in 
length. Overall variation of number of syllables 
per song (mean = 7 syllables. CV range = 30- 
9407 ), song length (mean = 4.07 sec. CV range - 
31-101%), and silent intervals between songs 
(mean = 3.94 sec, CV range = 58-131%) was 
high within and across individuals, but song 
tempo had low variation (mean = 1.9 syllables/ 
sec. CV range = 7-11%) (;? = 8 males. 44-150 
songs/malc). 
Wc identified 91 different syllable types from 
all birds. Male repertoire size ranged from 10 to 
17 syllable types (Table I). The cumulative plots 
of syllable types all clearly reached asymptotes, 
demostrating that complete repertoires were 
detected for all males during the time of the study 
(Fig. 3). Males sang a given syllable type only 
once during the study period in few cases. 
Only single syllables were shared among males 
rather than complete songs. There were 51 
matches considered as shared syllable types 
between pairs of males by at least one observer. 
Seventeen of these 51 possible matches were 
identified as shared types by all three observers 
(Table I). These final matches consisted of 13 
different syllable types. Comparisons within pairs 
of males showed that mean spectrogram correla¬ 
tions were constantly higher for shared types than 
for random syllables (Table 2). 
Most of the males (6/8) shared syllables with 
three or more males (Table 1). Males shared on 
average (,± SD) 4 ± 2 syllables, which corre¬ 
spond to 25 ± 16% of their repertoires (Table I). 
Repertoires were composed of more individual 
syllable types (syllables exclusive to each male) 
than shared syllable types (t 7 = 4.689, P = 0.002) 
(Table I). The frequency of individual and shared 
syllables in songs did not differ from that 
expected by chance according to the proportion 
ol individual and shared types in the repertoires 
(/7 = 0.047, P = 0.96) (Fig. 4). Males that had 
closer song perches did not have more similar 
repertoires (Z = -0.036. P = 0.48) (Table I ). 
