432 
THE WILSON JOURNAL OF ORNITHOLOGY . Vol 123. No. 3. September 2011 
X 
-z 
lw rwni "’v.ir (wm -i 
B 
0*5 1 1.5 2 2.5 Time (sec) 
FIG. 2. Sonograms of a typical song boul by (A) epia (North Sulawesi, recording by Mike Catsis). (B) subgularis 
(Peleng, recording by FV), (Q mangnliensis (Taliabu. recording by Pete Davidson). All three recordings are at an equal 
scale. Y-axis denotes frequency in kHz (2 kHz per lick mark). 
Sound Analysis .—Sonograms of vocalizations 
were prepared and analyzed using Program Syrinx 
(Version 2.6 h) from John Burt (available at www. 
syrinxpc.com). Levels of background noise were 
set to an equal level, otherwise default settings 
were used. P. subgularis is occasionally heard 
giving low-pitched alarm-type calls, but its vocal 
repertoire is overwhelmingly restricted to a single 
song type that consists of a series of -5-30 
‘whoop' notes delivered at a uniform pace 
(Fig. 2); this is the only vocalization included in 
our analyses. We measured four vocal parameters 
in our recordings: (1) number of ’whoop' notes 
per song, (2) duration of individual ‘whoop’ notes, 
(3) song duration, and (4) pace of song delivery 
(‘whoop’ notes/sec). Wc refrained from including 
frequency parameters in our analysis of geograph¬ 
ic variation in the vocalizations because we know 
they vary strongly and not on a geographic scale, 
but at an individual level and even within 
individuals based on song context. We routinely 
heard single individuals give the same hooting 
vocalizations at various frequencies in our field 
observations of Wallacean pigeons and doves, 
including also the genera Macropygia and Tur- 
ettoena. Two individuals at times, as in P v 
mangoliensis on Taliabu (FER, pens. obs.). induce 
hit-1 iM ° ulten ' n * «»*» •« increasingly 
tspate qUe " C,ei dUri " S “ P™ 1 ™'* territorial 
Most recordings contained more than one song 
bout (157 song bouts in 35 recordings; Appendix). 
We U-ealed recordings rather than song bouts as 
the sample points of statistical analysis, because 
song bouts within a recording are presumably 
strongly autocorrelated, as they are given by the 
same individual roughly within the same 5-min 
period. Different recordings are less likely to 
exhibit autocorrelation, because in each taxon 
only 33-56% of recordings constituted second or 
third contributions by the same recordist (Appen¬ 
dix), the remainder having been recorded by 
ditterent people on different occasions. Even 
recordings by identical recordists were often 
identified as having been uttered by different 
birds (as in recordings by FER and FV. Appen¬ 
dix). Thus, means of all lour vocal parameters 
w ere computed for each recording, and means and 
standard deviations were evaluated for each taxon 
separately (Appendix). A Mann-Whitney U-test 
was used to assess the significance of inter-taxon 
vocal dillerences (a = 0.0J to account for non¬ 
conservatism of this test). We also used Isler el 
al.’s (1998) vocal diagnosability criterion (hence¬ 
forth the Isler criterion), which is based on two 
conditions: (1) there must be no overlap between 
the ranges of measurements between two laxa; 
and (2) means () and standard deviations (SD) of 
the taxon with the smaller set of measurements (a) 
and the taxon with the larger set of measurements 
