SHORT COMMUNICATIONS 
837 
TABLE 1. Mean ± SD and potential for information coding (PIC) values for 25 variables measured for 20 pee-ah-wee 
songs for 16 male Eastern Wood-Pewees. T = temporal measures (sec); TP = temporal proportion measures; F = 
frequency measures (kHz): and FM = frequency modulation measures (kHz). CV h = between individual coefficient of 
variation; CV W = within individual coefficient of variation. PIC = CVyC\\ v . 
Variable 
Mean = SD 
F t ,uu.‘ 
CV b 
CV W 
PIC 
T1 (A to G) 
1.18 ± 0.16 
136.6 
13.31 
4.66 
2.9 
T2 (A lo C) 
0.12 ± 0.02 
264.5 
17.68 
5.08 
3.5 
T3 (D to F) 
0.29 ± 0.07 
315.9 
24.24 
6.06 
4.0 
T4 (F to G) 
0.77 ±0.16 
137.3 
21.31 
7.85 
2.7 
T5 (A to B) 
0.07 ± 0.01 
67.4 
20.70 
9.33 
2.2 
T6 (B to C) 
0.04 ± 0.01 
52.9 
21.64 
10.34 
2.1 
T7 (D to E) 
0.10 ± 0.03 
66.5 
35.70 
18.15 
2.0 
T8 (E to F) 
0.19 ± 0.06 
200.8 
30.49 
9.36 
3.3 
T9 (B to E) 
0.14 ± 0.04 
67.9 
25.17 
12.30 
2.0 
TPI (T5/T1) 
NA 
NA 
23.56 
9.93 
2.4 
TP2(T2/TI) 
NA 
NA 
22.27 
6.47 
3.4 
FI (A) 
3.46 ± 0.24 
83.9 
6.88 
2.93 
2.3 
F2 (B) 
4.55 ± 0.24 
360.2 
5.30 
1.15 
4.6 
F3(C) 
3.35 ± 0.20 
184.8 
5.92 
1.80 
3.3 
F4(D) 
3.62 ± 0.27 
306.4 
7.35 
1.76 
4.2 
F5 (E) 
4.81 ± 0.30 
357.4 
6.17 
1.37 
4.5 
F6(F) 
3.37 ± 0.20 
545.7 
6.09 
1.14 
5.3 
F7(G) 
4.05 ± 0.28 
224.2 
7.03 
2.01 
3.5 
FM1 (A to B) 
1.08 ± 0.20 
43.1 
18.82 
10.88 
1.7 
FM2 (B to C) 
1.19 ± 0.21 
153.9 
17.98 
5.94 
3.0 
F.M3 (D to E) 
1.19 ± 0.33 
277.5 
28.06 
7.67 
3.7 
FM4 (E to G) 
0.68 ± 0.22 
245.3 
18.85 
4.98 
3.8 
FM5 (F to G) 
0.68 ± 0.22 
123.5 
32.85 
13.57 
2.4 
FM6 (A to G) 
0.64 ± 0.28 
71.8 
43.04 
23.60 
1.8 
FM7 (A lo E) 
1.34 ± 0.28 
90.4 
20.67 
9.31 
2.2 
' F-values for one-way ANOVAs comparing helwccn and within individual variation lor each song variable: all P < 0.0001. 
and wee -000 songs (Robisson et ul. 1993, Charrier 
ct al. 2004). PIC values compare the within- 
individual variation of a particular variable to the 
between-individual variation and provide a foun¬ 
dation for understanding which vocalization 
variables hold the potential to code information. 
PIC values are increasingly used to investigate 
vocalization variability in a wide variety of avian 
toxa including oscines (Charrier et al. 2004, Xia et 
2010), suboscines (Lein 2008), and non- 
passerines (Robisson et al. 1993). 
We first calculated the coefficient of variation 
ICV) for each song variable to calculate PIC 
values. We applied the formula (SD/X)(I + 1/ 
*to)(100) for within-individual CV (CV W ), cor¬ 
seted for sample size, and applied the formula 
(SD/X)(1Q0) for betwccn-individual CV (CV„), 
where SD is standard deviation, X is the sample 
mean, and n is the sample size for an individual 
•Sokul and Rohlf 1995). PIC then equals CV b / 
CV W . Song variables with a PIC > 1 have the 
Potential to code information as their within 
individual variation is less than their between 
individual variation (Robisson et al. 1993, Char¬ 
rier et al. 2004). 
We conducted a principal components analysis 
(PCA) on all data sets to minimize any effects of 
collinearity resulting from variables potentially 
correlated with each other and to reduce the 
number of variables. We applied Varimax rotation 
with Kaiser normalization to all principal compo¬ 
nents. We conducted a discriminant function 
analysis (DFA) on the factors produced by the 
PCA to ascertain whether individual songs could 
be attributed to individual birds. 
DFA avoids over-weighting outliers and highly 
variable measurements. DFA also produces linear 
functions that maximize the probability of cor- 
rectlv assigning data points to groups, in this case, 
the individual identity of a male Eastern Wood- 
Pewee. Each linear function is derived from all 
songs in the data set except for the one being 
classified in a cross-validation process (jackknif¬ 
ing or leave-one-out validation) (Manly 1994). 
