Rheindt et al • LEAPFROG PATTERN IN PTILINOPUS FRUIT DOVES 
433 
Number of‘whoop’ 
notes per song 
X 
epia mangoliensis subgularis 
Song 
duration 
T 
epia mangoliensis subgularis 
Duration of ‘whoop’ 
notes in sec 
epia mangoliensis subgularis 
Song pace 
(‘whoops’/sec) 
epia mangoliensis subgularis 
FIG. 3. Boxplots of measurements for all four vocal parameters of all three taxa (outliers not depicted for ease 
of interpretation). 
(b) had to meet the requirement 
x a + t a SD a <,x b -t b SD b 
where fj is the value of the Student’s /-score at 
the 97.5 percentile of the /-distribution for n - I 
degrees of freedom. The Isler criterion is 
substantially more discriminating than the famil¬ 
iar /-test or Mann Whitney U -test because it uses 
the standard deviations of the sample points, not 
the much smaller standard deviations of the taxon 
mean. 
We conducted discriminant analysis with an 
equal prior probability of group assignment and 
with assignment probabilities computed on the 
basis of means that exclude the value of the 
sample under scrutiny. All statistical analysis was 
conducted using Program SPSS (SPSS Institute 
Inc.. Chicago, IL, USA). 
RESULTS 
Measurements, including means and standard 
deviations, of all four vocal parameters varied 
among taxa (Appendix, Fig. 3). A Mann-Whitney 
U -test contrasting all four parameters among the 
three taxa demonstrated significant differences for 
most comparisons, except for song duration 
between subgularis and mangoliensis as well as 
song pace between epia and mangoliensis (Ta¬ 
ble 1). Thus, songs of all three taxa are distinct in 
a range of vocal characters, Judged against the 
Isler criterion, only three of the 12 taxon- 
parameter comparisons were diagnosable (Ta¬ 
ble 1), namely number of notes and song pace 
between epia and subgularis as well as song pace 
between mangoliensis and subgularis. This result 
suggests subgularis is vocally more differentiated 
than the other two taxa are from each other. 
A preliminary inspection of recordings indicat¬ 
ed there may be within-taxon differentiation of 
epia songs. Our epia recordings are from two 
separate regions of Sulawesi (Fig. 1), and we 
analyzed means and standard deviations of all 
parameters for geographical regions separately 
(Table 2). Only song pace was significantly 
different between North and Central Sulawesi as 
