Perez Mena and Mora • GEOGRAPHIC SONG VARIATION IN THE CUBAN TODY 77 
Current vegetation cover 
| Forest 
Original vegetation cover 
jUjij Grassland 
fl Forest 
Cayo Sabinal (3) 
Cayo Coco (9) 
100 
Kilometers 
Cupeyal (19) 
Ojitode Agua (11) 
(3) 
Pico Turquino (8) Baitiquirf (17) 
FIG. 1. Original vegetation (Del Risco 1989) and current forest cover on the island of Cuba. The seven provinces and 
the recording sites involved in this study are identified. The number of birds recorded in each site is in parentheses. 
distance between individuals. There are no 
acoustic studies of the Cuban Tody. Thus, to test 
this prediction, we recorded the song of the 
species across Cuba, characterized it quantitative¬ 
ly, and examined geographic variation in songs by 
using a discriminant function analysis. 
METHODS 
Field Recordings.— Acoustic recordings of the 
Cuban Tody were obtained between 2001 and 
2006 at 15 sites in seven provinces of Cuba: Isla 
de la Juventud (1 site), Pinar del Rio (3), 
Matanzas (4), Ciego de Avila (1), Camagiiey 
(1), Santiago de Cuba (1), and Guantanamo (4) 
(Fig- 1). Recordings were made from February to 
August to include most of the reproductive 
period of the species (Garrido and Kirkconnell 
2000 ). 
We recorded songs from 116 solitary adults 
perching in the forest up to 10 m from the 
microphone. Adults are easily recognized because 
they combine the ventral pale gray color with a 
brilliant red throat patch that is absent in young, 
which are entirely pale gray below (Garrido and 
Kirkconnell 2000). We could not test for vocal 
differences between males and females as they 
can not be identified by plumage (Raffaele et al. 
1998, Garrido and Kirkconnell 2000). Recordings 
were made using a Marantz PMD 222 tape 
recorder and Sennheiser ME66/K6 microphone. 
Care was taken to avoid signal saturations and to 
maximize signal-to-noise ratio. 
Acoustic Analysis. —A data base of 1,371 songs 
(11.67 ± 4.52 songs/bird) including 8,885 notes 
was analyzed. Songs were digitized to examine 
the quality of the recordings with 16 bit accuracy 
and a sampling rate of 44,100 Hz using BatSound, 
Version 2.1 (Petterson Elektronic AB, Uppsala, 
Sweden). A note was defined as a continuous 
tracing on a spectrogram following Nelson et al. 
(1996), while a song was defined following 
Baptista (1974) and Staicer (1989) as an arrange¬ 
ment of notes forming a coherent unit. 
Songs were analyzed with Avisoft-SAS Lab 
Pro 4.3 (Avisoft Bioacoustics, Berlin, Germany). 
Spectrograms were made using consecutive Fast 
Fourier Transforms (FFT’s) and Hamming win¬ 
dows with a 92% overlap. A 512 points FFT was 
chosen to attain a frequency resolution of 86 Hz 
and a time resolution of 0.18 msec. 
We used an automatic two-threshold algorithm 
for note separation with the additional start/end 
threshold set at — 20 dB. This algorithm minimiz¬ 
es the effect that different note amplitudes may 
have on the measurements. The following param¬ 
eters were automatically measured: (1) note 
duration (time between start and end of a note 
measured in msec in the spectrogram), (2) peak 
frequency (frequency in kHz corresponding with 
the maximal intensity in the power spectrum), (3) 
initial and (4) final frequency (values of frequency 
measured, respectively, at the beginning and at 
end of the note), (5) bandwidth (calculated as the 
difference between the lower and higher values of 
frequency measured at 20 dB below peak intensity 
in the power spectrum), and (6) entropy (used as 
an estimate of the tonal-noisy structure of the 
note: zero for pure-tone signals and 1 for random 
noise). We counted the number of notes in each 
song and calculated the interval between notes 
and songs. All values of the acoustic parameters 
are given as means ± SD. 
