456 
THE WILSON JOURNAL OF ORNITHOLOGY • Vol. 124. No. 3. September 2012 
habitat to ensure recording of different birds, 
sometimes skipping the neighbor of the bird just 
recorded if there was noticeable movement. 
Territory measurements were not conducted, but 
the distance from one singing male to the next was 
on the order of 10-15 m, and studies elsewhere 
found territory sizes of —900-1,500 in 2 (Catterall 
et al. 1982). The small territories made it 
relatively easy to keep track of the location of a 
bird previously recorded as attention was placed 
on the next subject. However, the high density 
created some problems as singing was both loud 
and frequent during the dawn chorus and often 
songs of an individual being recorded had parts 
overlapped by another. It had been previously 
noted in a different location that Silvereyes utter 
songs independently of one another (i.e., no 
alteration, etc.; Slater 1991, 1993). The birds 
appeared to be well habituated to people, perhaps 
from the frequent tourists near these habitats and, 
together with the low light conditions. I was able 
to get close to singing birds. These conditions 
were optimal for near-field recording, but it was 
difficult to obtain songs free of overlapped 
syllables. 
Recordings were made with a Marantz cassette 
deck (PMD222, Marantz America Inc., Itasca, IL. 
USA) connected to a Sennhciser ME62 micro¬ 
phone (Sennheiser Corp.. Old Lyme, CT, USA) 
mounted in a 45-cm parabolic reflector. A 
constant frequency marker and known click rate 
(Seiko SQ-44. Seiko Instruments. Maidenhead, 
UK) were recorded before and at random times 
during a recording session so that tape speed 
could later be checked for accuracy. No changes 
in tape speed were found at the time of digitizing 
the songs for acoustical analyses. 
Acoustical Analyses. —Vocal recordings were 
digitized with Sound Blaster Audigy 2NX (Crea¬ 
tive Technology Ltd., Creative Labs Inc., Milpitis, 
CA, USA) with 16 bit accuracy at a rate of 44.1 kHz 
to a personal computer. The favorable recording 
conditions provided good quality sounds, but low 
frequency noise (below -300 Hz) was filtered 
out. Sound spectrograms were produced with Real 
Time Spectrogram (Model 5129 Version 2.3: 
Kay Elemetrics Corp (now KayPentax). Lincoln 
Park, NJ, USA) with settings of 256 points, and 
Hamming Window weighting. I used Sound 
Analysis Pro software (Version 1.04: Tcherni- 
chovski et al. 2000, Baker and Loguc 2003, 
Tchemichovski and Mitra 2004) following digiti¬ 
zation to quantify the spectral features of the 
syllables constituting songs. Sound Analysis Pro 
creates spectral derivatives of a sound, such as a 
song syllable, and automatically extracts measure¬ 
ments of a set of acoustic variables (Fee et al. 1998. 
Ho et al. 1998). The values of each acoustic 
variable were averaged over a succession of narrow 
and overlapping windows of time (data window 
9.27 ms, advance window 1.36 ms) over the 
duration of the given syllable. Six variables were 
quantified for eaeh syllable examined by Sound 
Analysis Pro: sy llable duration, pitch, fm (frequen¬ 
cy modulation is the change in frequency with 
time), am (amplitude modulation is the change in 
sound power with time), goodness of pitch (how 
much of the sound energy is concentrated in the 
pitch), and Wiener entropy (a measure of the 
breadth and evenness of a sound: pure tones have 
low entropy, white noise high entropy). Mathe¬ 
matical and verbal descriptions of these features 
are provided in Tchemichovski et al. (2000) and 
Tchemichovski and Mitra (2004). 
Analysis Strategy. —The only published research 
providing a detailed description of song of any 
Zosterops is that of Slater (1993) (Capricorn 
Silvcreye, Z. lateralis chlorocephalus ), and initially 
I thought I would be able to follow the analysis 
approach used in that study. Slater’s study, 
however, turned out not to be an appropriate model: 
it did not present actual sound spectrograms of 
songs hut instead made line drawings of syllables to 
create a catalog of syllable ‘types', a method that 
can be an impediment to comparative studies. I 
tried to follow this methodology by examining 
sound spectrograms of songs, and systematically 
comparing syllables looking for matches, but found 
that syllables w ere difficult to sort into categories of 
'types': there was a great deal of variation and 
grouping and splitting often had a highly arbitrary 
aspect to it, The population (and subspecies) studied 
by Slater (1991. 1993) may have exhibited more 
stereotypy in song elements than the populations I 
recorded, hut without a published presentation of 
actual spectrograms of syllable ‘types', and the 
variants subsumed in these categories, a compari¬ 
son of Slater's results to my data was impossible. It 
became necessary to formulate a different method 
for comparing population samples of Zosterops 
songs for the purposes of my research. 
I present a typological classification to demon¬ 
strate the problem, based upon sound spectro¬ 
grams. to show the great diversity of song 
elements and, evidently, an essentially continuous 
generation of new syllables constituting Silver- 
