468 
THE WILSON JOURNAL OF ORNITHOLOGY • Vol. 123, No. 3, September 2011 
Arizona Baja 
Song length (sec) 
No. syllables 
No. unique syllables 
]_ ) 
Syllable repeats 
r m 
No. trills 
m 
No. syllables/trill 
Unique notes/trill 
No. phrases 
| 
i 
Notes/phrase 
Peak freq. (kHz) 
Low freq. (kHz) 
High freq. (kHz) 
Bandwidth (kHz) 
-2 -i o 
i 2 
*-»uciwvirn means 
Me F x\co 3 TheTeofXL S r e trn eS ^ Ariz0na * USA ’ ** California Sur, Mexico, and Oaxaca, 
regions) between the means of the iwn^ ° regl °" S reflects ,he number ol pooled standard deviations (pooled from paired 
us^g the correspond!nelsontr'featiire -?h re ?’° nS » U j' om P ar ' son • Bar size reflects how distinguishable the two regions ate 
feature. Baja California songs are more distinguish hT ,nd,cates the region with the larger mean value for the 
Oaxaca songs. 8 s a e from those of Arizona and Oaxaca than Arizona songs are from 
of trills per song and the lowest frequency. Four 
features differed between Oaxaca and Arizona 
songs (all, P < 0.001); Baja California songs were 
different from those in Arizona for 11 features (all, 
P < 0.001) and were different from Oaxaca songs 
for six features (all, P < 0.001: Table I, Fig. 3). 
Values for the number of unique syllables and the 
number of repeated syllables of the Baja California 
population dkl not overlap with values from the 
two mainland populations. Baja California songs 
had more unique and total syllables per song, and 
included more phrases than songs from the^other 
two regions. Songs from Baja California had fewer 
repeats per syllable than those from Arizona or 
Uaxaca, which coincides with fewer trills and 
fewer repeated syllables per (rill than songs from 
the two other regions. Baja California songs also 
were longer and had lower peak and highest 
requencies and a narrower bandwidth than Ar- 
izona songs. Spectral features of Baja California 
and Oaxaca songs differed only in terms of peak 
Ditch Ue H Cy ' Whl e S ° ngS from ° axaca were higher 
hm h !" t ! m,S ° f h 'S hest frequency and had a 
broader bandwidth than those from Arizona 
songfe“e? n f° n V*** Was based «JI 
g features except number of notes per phrase 
because many songs from Oaxaca and Arizona 
lacked phrases. The analysis correctly assigned 
87% of the songs to the correct region: 100% 
were correctly assigned to the Baja California 
region, 90% were correctly assigned to the 
Arizona region, and 50% were correctly assigned 
to the Oaxaca region. Three songs from Arizona 
were misclassified as from Oaxaca and four 
Oaxaca songs were misclassified as from Arizona. 
No Arizona or Oaxaca songs were misclassified as 
Baja California songs. The Baja California songs 
were easily distinguishable from .Arizona and 
Oaxaca songs based on the first discriminant axis 
(97.5% of variance), but Arizona and Oaxaca 
songs were not easily identifiable based on this 
axis, nor on the second discriminant axis (2.5% of 
variance: Fig. 4). Variables with strong loadings 
on the first axis were song length (1.23) and 
number of phrases (1.48); those with strong 
loadings on the second axis were number of trills 
(1.09) and song length (-2.03; values of all other 
loadings are in Table 2). 
DISCUSSION 
We found strong systematic differences be- 
ween .songs of J. p. bairdi and those of two 
