Ethology
Biphonation May Function to Enhance Individual Recognition in
the Dhole, Cuon alpinus
Elena V. Volodina*, Ilya A. Volodin* , Irina V. Isaeva & Carolyn Unckà
* Scientific Research Department, Moscow Zoo, Moscow, Russia
Department of Biology, Lomonosov Moscow State University, Moscow, Russia
à University of Western Ontario, London, Canada
Correspondence
Elena V. Volodina, Scientific Research
Department, Moscow Zoo, B. Gruzinskaya, 1,
Moscow, 123242, Russia.
E-mail: volodinsvoc@yahoo.com
Received: November 21, 2005
Initial acceptance: December 24, 2005
Final acceptance: December 24, 2005
(J. Lazarus)
doi: 10.1111/j.1439-0310.2006.01231.x
Abstract
Biphonation (two independent fundamental frequencies in a call spectrum) represents one of the most widespread nonlinear phenomena in
mammalian vocalizations. Recently, the structure of biphonations was
described in detail; however, their functions are poorly understood. For
the dhole (Cuon alpinus), biphonic calls represent a prominent feature of
vocal activity. In this species, the biphonic call is composed of two frequency components – the high-frequency squeak and the low-frequency
yap, which also occur alone as separate calls. In this study, we test the
hypothesis that the complication of call structure, resulting from the joining of these calls into the biphonic yap–squeak may enhance the potential
for individual recognition in the dhole. We randomly selected for analysis
30 high-frequency squeaks, 30 low-frequency yaps and 30 biphonic
yap–squeaks per animal from five subadult captive dholes (450 calls in
total). Discriminant analysis, based on 10 squeak parameter values,
showed 80.7% correct assignment to a predicted individual. For 10 yap
parameters, the correct assignment was only 44.7%. However, the analysis based on 10 parameters of the biphonic yap–squeak, selected as best
contributing to discrimination, showed 96.7% correct assignment to a
predicted individual. The results provide strong support for the hypothesis
tested showing that the joining of two independent calls into a common
vocalization may function to enhance individual recognition in the dhole.
Introduction
Biphonation is one of the nonlinear phenomena in
mammalian vocalizations that evoked a burst of
research interest in recent years (Wilden et al. 1998;
Riede et al. 2000; Fitch et al. 2002). Biphonation
may be recognized: (1) by appearance of two independent fundamental frequencies in a sound spectrum, (2) by different contours of frequency
modulation in some frequency bands on a spectrogram, (3) by the appearance of additional frequency
bands, representing linear combinations of two independent frequencies, which may be calculated using
the formula n Æ f + m Æ g (f and g are two independent
Ethology 112 (2006) 815–825 ª 2006 The Authors
Journal compilation ª 2006 Blackwell Verlag, Berlin
frequencies; n and m are integers) (Wilden et al.
1998; Volodin & Volodina 2002). At least four probable mechanisms for production of biphonic vocalizations have been proposed: (1) asynchronous
vibration pattern of the left and right vocal fold
(Berry et al. 1996; Tigges et al. 1997); (2) involvement of vocal fold extensions (vocal membranes) in
the production of a second fundamental frequency
(Brown & Cannito 1995; Mergell et al. 1999; Riede
et al. 2000); (3) vortex-shedding at the glottal constriction inducing a whistle-like sound (Solomon
et al. 1995; Herzel & Reuter 1997; Wilden et al.
1998); and (4) source-tract coupling (Herzel & Reuter 1996; Mergell & Herzel 1997). These proposals
815
Biphonation for Individual Recognition in the Dhole
come from three sources: from analysis of actual animal and human calls (Brown & Cannito 1995; Herzel & Reuter 1996, 1997; Tigges et al. 1997; Wilden
et al. 1998; Riede et al. 2000), from experiments on
excised larynges (Solomon et al. 1995; Berry et al.
1996; Brown et al. 2003), and from computer simulation models (Mergell & Herzel 1997; Mergell et al.
1999). However, research of the role of various
mechanisms in production of biphonic calls in particular species is limited, and sound production mechanisms in nonhuman mammals remain poorly
understood (Peters et al. 2002).
Functional interpretations of nonlinear phenomena
in vocalizations are also scarce, and nearly entirely
absent with concern to biphonation. These phenomena are not under central nervous system control, but
arise from the physics of a mammalian sound production apparatus and thus may not have an adaptive
meaning for a caller (Wilden et al. 1998; Fitch et al.
2002). For example, subharmonics, chaos and biphonations occur in unhealthy voices both in humans
(Herzel 1993; Herzel et al. 1995; Herzel & Reuter
1996) and in nonhuman mammals, such as Japanese
macaque Macaca fuscata (Riede et al. 1997), domestic
dog Canis familiaris (Riede et al. 2001) and domestic
cat Felis catus (Riede & Stolle-Malorny 1999).
Although some nonlinear phenomena in voices
may not have evolved as a structural feature with a
communicative meaning a priori, they may have
been adopted subsequently (Fitch et al. 2002). Moreover, there are some data confirming the communicative functions of biphonation. In two penguin
species – the king penguin Aptenodytes patagonicus
and the emperor penguin A. forsteri – beating generated by interaction of two close frequencies enhances
the ability of calls to propagate through obstacles of
numerous penguin bodies in a colony and provides
additional cues for both parent–chick and mate–mate
recognition (Aubin et al. 2000; Lengagne et al. 2001;
Aubin & Jouventin 2002). In the killer whale Orcinus
orca, differences in degradation and directionality of
the lower- and higher-frequency components in
biphonic calls provide information about direction of
movement of a caller (Miller 2002). In addition,
among proposed functions of nonlinear phenomena
are accenting call formant structure for delivering
information about size of a caller and breaking of
monotony in vocal sequences in order to escape
habituation in listeners and to attract their attention
to a caller (Fitch & Hauser 2002; Fitch et al. 2002).
Biphonic calls have been recorded in a few terrestrial mammalian taxa, and besides such specialized
group as bats (Kanwal et al. 1994), they have been
816
E. V. Volodina, I. A. Volodin, I. V. Isaeva & C. Unck
recorded primarily in primates (Brown & Cannito
1995; Brown et al. 2003) and canids (Wilden et al.
1998; Volodin & Volodina 2002). Biphonation has
been recorded in timber wolf Canis lupus (Nikol’skii &
Frommolt 1989; Frommolt 1999), domestic dog (Solomon et al. 1995; Volodin et al. 2005b; Volodina et al.
2005), dog–wolf hybrid (Riede et al. 2000) and jackal
Canis aureus (A. Pojarkov, pers. comm.), but are especially prominent both in the African wild dog Lycaon
pictus (Wilden 1997; Wilden et al. 1998; Robbins
2000) and in the dhole (Volodin & Volodina 2002). In
canids, biphonic calls may occur as irregular events
and not in all individuals, or attend specific states,
such as frustration in domestic dogs (Volodin et al.
2005b; Volodina et al. 2005). But, in African wild
dogs and dholes, biphonic calls occur regularly among
calls attending short-distant affiliative interactions in
a pack and make up 60% and 44% of vocal emissions
in this context, respectively. Moreover, they occurred
in all individuals in these species (Wilden et al. 1998;
Volodin & Volodina 2002).
The dhole is a pack-living canid, communally
hunting on large prey and inhabiting areas with
complex relief in mountains and in locations with
dense vegetation, with primary breeding by a dominant pair and other group members functioning as
helpers (Cohen 1977; Johnsingh 1982; Karanth &
Sunquist 1995; Venkataraman et al. 1995; Venkataraman 1998). Dholes typically show very high vocal
activity, that attend all, even small movements of
pack members and all contacts among the animals
(Sosnovskii 1967; Cohen 1977, 1985; Johnsingh
1982). In captivity, the dhole vocal repertoire
includes 11 call types, based on three vocal components: the low-frequency tonal (with fundamental
frequency varying from 0.5 to 1.4 kHz), the highfrequency tonal (with fundamental frequency
varying from 5.5 to 10.8 kHz), and the pulsed component. Only one call type has a biphonic structure,
resulting from simultaneous production of the highand low-frequency components (Volodin et al.
2001). Biphonic calls (call type yap–squeak), alongside with ‘clear’ yaps and squeaks and frequency
jumps from squeak to yap, occurred primarily during
peaceful interactions among group members and in
the context of spontaneous movements in an enclosure (Volodin et al. 2001; Volodin & Volodina 2002).
In these situations, the occurrence of biphonic calls
(yap–squeaks) among contact calls (yaps, squeaks
and yap–squeaks) varied in 14 individual dholes
from 20% to 92%, and was not related significantly
to age, sex or litter membership (Volodin & Volodina
2002).
Ethology 112 (2006) 815–825 ª 2006 The Authors
Journal compilation ª 2006 Blackwell Verlag, Berlin
E. V. Volodina, I. A. Volodin, I. V. Isaeva & C. Unck
It is not clear if yap–squeak is the same call type
known as ‘mixed yip-yack cackle’ after Johnsingh
(1982). Moreover, it is difficult to relate it to other
call types reported onomatopoetically in earlier literature, because of the absence of spectrograms in
these papers. Moreover, biphonic calls are difficult to
discern from some other call types by ear. Therefore,
based on the context of usage of these calls in captivity, we propose their function to be as peaceful
vocalizations and that these calls may promote individual recognition in dholes. Furthermore, we propose that the complexity of the call type, composed
from two independent frequencies, may play a special role, enhancing reliability of the recognition.
Here we test using discriminant analysis, if the biphonic calls, composed of the high- and low-frequency components, provide better potential for
individual identification, than the monophonic calls
consisting of one of these components.
Animals and Methods
Calls were tape-recorded from five subadult dholes
(aged 7.5–11 months) from two litters born in captivity. The first litter of one male and two females
(no. 10, 11, 12) was born in March 19, 1999 in
Moscow Zoo (Russia). The second litter of three
males was born in April 24, 1999 in Volokolamsk
Moscow Zoo Brooder, but only two of them (no. 14,
15) provided the necessary number of calls to be
included in the analysis. All the recordings were
made during November 2, 1999 to February 21,
2000. All the animals were housed with their littermates. Parents were housed together with pups (first
litter) or separated from them by wire mesh (second
litter).
The sound recordings were made with a SONY
WM-D6C recorder (Sony Corp., Tokyo, Japan) and
MCE-100 unidirectional microphone (LOMO, St.
Petersburg, Russia). Frequency responses of both
systems were 40–15 000 Hz. Distance to animals
during the recordings varied from 2 to 8 m. The
sounds were produced spontaneously without stimulation from observers. All individuals from the same
litters could be reliably identified by their coloration
pattern. Simultaneous video recordings to identify
calling individuals were made with a SONY TRV-65E
video camera.
For analysis, we randomly selected 90 high-quality
calls per individual (30 yaps, 30 squeaks and 30
biphonic yap–squeaks, 450 calls for five dholes in
total). Ownership of each call was confirmed independently by two observers during recordings which
Ethology 112 (2006) 815–825 ª 2006 The Authors
Journal compilation ª 2006 Blackwell Verlag, Berlin
Biphonation for Individual Recognition in the Dhole
was confirmed additionally on the basis of video
recordings.
All spectrograms of these calls were analyzed with
Avisoft SASLab software (ª R. Specht). Digital processing used fast Fourier transform (FFT) with
22.05 kHz sampling frequency, Hamming window,
FFT length 512 points, frame 50%, overlap 93.8%,
that provided 1.45 ms time resolution and 43 Hz frequency resolution.
Biphonic calls of dholes comprise two independent
vocal components that also occur as separate vocalizations. The fundamental of the low-frequency component f0 is about 1 kHz, and the fundamental of
the high-frequency component g0 is higher than
5 kHz (Fig. 1). We measured seven frequency and
three temporal parameters for each component
occurring alone as yap and squeak calls and 20
parameters for each biphonic yap–squeak (Table 1)
For the biphonic calls, the high- and low-frequency
component parameters were measured after highand low-pass filtration of 5 kHz, applied alternately.
Temporal parameters were measured from the
spectrogram window using a standard marker cursor.
Fundamental frequency parameters for each component were measured using a free reticule cursor. All
measurements were exported automatically into an
Excel database. The number of fundamental frequency extrema were counted visually from spectrograms according to Tooze et al. (1990). Frequency
parameters of the high-frequency components were
measured by g0 only, whereas those of the low-frequency components by the most well-expressed frequency band (f0, f1 or f2). Values of peak frequency
and bandwidth of peak frequency (at distance
)10 dB from peak) of the corresponding component
Fig. 1: Spectrograms illustrating three call types of the dhole: left –
the high-frequency call or squeak; middle – the low-frequency call or
yap; right – the biphonic call or yap–squeak. Designations: g0 – fundamental frequency of the high-frequency component; f0 – fundamental
frequency of the low-frequency component; f1 and f2 – harmonics of
f0; g0–f0 – the linear combination of f0 and g0
817
Biphonation for Individual Recognition in the Dhole
E. V. Volodina, I. A. Volodin, I. V. Isaeva & C. Unck
Table 1: Call parameters used in the statistical analyses
Call parameters
High-frequency
component
Low-frequency
component
Start fundamental frequency (kHz)
End fundamental frequency (kHz)
Maximum fundamental frequency (kHz)
Minimum fundamental frequency (kHz)
Peak frequency (frequency with the maximum amplitude in the power spectrum) (kHz)
Bandwidth of peak frequency (Hz)
Number of fundamental frequency extrema (total number of peaks and depressions)
Duration from start to maximum frequency point of a component (s)
Duration from maximum frequency point to end of a component (s)
Proportion: duration from start to maximum frequency point of a component/duration of a component
g0_ini
g0_end
g0_max
g0_min
g_peak
g_bandw
g_extrem
g_dur_inc
g_dur_dec
g_k_max
f0_ini
f0_end
f0_max
f0_min
f_peak
f_bandw
f_extrem
f_dur_inc
f_dur_dec
f_k_max
were taken automatically from the mean power
spectrum. For the high-frequency component, the
peak frequency band coincided with fundamental
frequency g0, otherwise, for the low-frequency component, it could fall on one of the harmonic bands
from f0 to f2, predominately on f1. We also used
one calculated parameter: the duration from the
start to the maximum frequency point of a component divided by the duration of the component
(k_max) (Table 1).
For each of the three call types, values were normally distributed for most parameters (Kolmogorov–
Smirnov test). As parametrical anova and discriminant analysis are relatively robust to departures from
normality (Dillon & Goldstein 1984), this was not an
obstacle to the application of these tests.
We performed one-factor anova, with ‘individual’
as the grouping variable, to compare variability of
the parameters within and between individuals for
each call type. Then we used the standard discriminant analysis procedure based on 10 parameters for
each call to determine whether calls could be
assigned to the correct caller. Because the biphonic
calls contained a double set of parameters, 10 from
the low frequency, and 10 from the high frequency,
we had to reduce this number to 10 to escape
increasing discriminability of biphonic calls simply
because of an increase in the number of parameters
entered into the analysis. To select 10 of the 20
parameters available for biphonic calls, we conducted a stepwise discriminant analysis and took the 10
parameters that best contributed to the discrimination. For statistical comparison of correct assignment values resulting from discriminant analyses for
the biphonic and non-biphonic calls, we used a
2 · 2 chi-squared test.
To validate results of discriminant analysis, we
performed cross-validation analysis and randomization. For cross-validation analysis, call samples for
818
each dhole were randomly split half-and-half, providing a training set (75 calls) and a test set (75 calls)
for each call type. Then the classification of one half
of the dataset was made, with the discriminant function derived from the other half.
Randomization was applied in order to calculate the
expected level of correct assignment by discriminant
analysis if the calls we analyzed were randomly distributed between individuals. For this procedure, we
created five randomization groups for each call type.
Each group of 30 calls consisted of six randomly selected calls taken from each of five dholes. After that, we
conducted a standard discriminant analysis and calculated the probabilities of correct assignment of calls to
the randomization groups. These probabilities were
taken as random values for each call type. Differences
between the random and actual values of correct
assignment were tested with a 2 · 2 chi-squared test.
All the analyses were performed in STATISTICA,
version 6.0 (StatSoft, Inc, Tulsa, OK, USA).
Results
The one-way anova revealed the highly significant
individual differences in all call parameters for individuals both for squeaks (10 parameters) and yap–
squeaks (20 parameters), but only for four call
parameters of yaps (f0_end, f0_min, f_dur_dec,
f_k_max). For the other six call parameters of yaps,
interindividual variability did not exceed intra-individual variability. Table 2 shows mean values and
standard deviations for some call parameters (for
four parameters both for the high-frequency squeak
and the low-frequency yap as well as for eight
parameters of biphonic yap–squeak calls) and results
of anova-based interindividual comparison for each
call type for five individuals. All other parameters of
squeaks and yap–squeaks, not included in Table 2,
showed significant differences at the p < 0.001 level.
Ethology 112 (2006) 815–825 ª 2006 The Authors
Journal compilation ª 2006 Blackwell Verlag, Berlin
E. V. Volodina, I. A. Volodin, I. V. Isaeva & C. Unck
Biphonation for Individual Recognition in the Dhole
Table 2: Values (mean SD) of high-frequency call (squeak), low-frequency call (yap) and biphonic call (yap–squeak) parameters for five dholes
and results of ANOVA-based interindividual comparisons for each call type
Call parameters
Female 10
Squeak (n ¼ 150, 30 per animal)
g0_max, kHz
8.76 0.91
g_dom, kHz
8.43 1.01
g_dur_inc, ms
52 36
g_dur_dec, ms
50 44
Yap (n ¼ 150, 30 per animal)
f0_max, kHz
0.94 0.11
f_dom, kHz
1.62 0.77
f_dur_inc, ms
38 18
f_dur_dec, ms
33 13
Yap–squeak (n ¼ 150, 30 per animal)
g0_max, kHz
9.27 0.49
g_dom, kHz
8.62 0.88
g_dur_inc, ms
29 27
g_dur_dec, ms
83 38
f0_max, kHz
1.02 0.09
f_dom, kHz
1.57 0.62
f_dur_inc, ms
29 21
f_dur_dec, ms
52 18
Male 11
Female 12
Male 14
Male 15
F4,145
8.08
7.87
80
35
0.26
0.26
32
40
7.51
7.13
99
21
0.30
0.28
40
30
5.92
5.72
78
44
0.14
0.13
38
34
8.13
7.88
38
69
0.16
0.13
34
31
169**
139**
13**
7**
1.00
1.54
40
29
0.15
0.80
21
19
1.00
1.67
34
32
0.18
0.94
16
19
1.01
1.72
42
21
0.12
0.76
18
13
0.95
2.07
37
20
0.10
0.84
14
13
1.6; p ¼ 0.16
1.8; p ¼ 0.13
1.0; p ¼ 0.42
4.1*
7.91
7.74
41
75
1.22
1.90
37
49
0.14
0.14
39
44
0.10
0.61
24
12
7.27
6.84
71
47
1.13
1.67
8
42
0.16
1.20
40
36
0.14
0.66
14
16
5.98
5.76
55
63
1.16
2.08
17
39
0.29
0.19
35
41
0.09
0.54
14
14
8.25
7.76
26
88
1.27
1.50
48
29
0.25
0.10
37
47
0.09
0.52
14
11
511**
78**
8**
5*
27**
5**
24**
13**
*p < 0.01; **p < 0.001.
For yaps, significant individual differences did occur
also in f0_end (p < 0.001), f0_min (p < 0.05) and in
f_k_max (p < 0.05).
Three discriminant analyses were performed,
based on: (1) 10 parameters of the squeak; (2) 10
parameters of the yap; and (3) 10 parameters of the
yap–squeak, selected using a stepwise discriminant
analysis.
For the squeak, 80.7% correct assignment was
achieved (Table 3, Fig. 2a), significantly more than
the random value of 30.7% (v2 ¼ 73.96, d.f. ¼ 1,
p < 0.001), being calculated by using the randomization procedure. The first discriminant function correlated basically with fundamental frequency
parameters, primarily with g0_max and g0_end, as
well as with g_peak, and explained 85.72% of the
variance. Contribution of other parameters was
small. The second discriminant function described
8.88% of the variance and correlated mainly with
g_dur_inc, g_k_max and g0_ini (Table 4). Cross-validation analysis showed 84.0% correct assignment
for the training call set (n ¼ 75, 15 calls per individual), with 66.7–100% for particular individuals. Correct assignment for the test call set (n ¼ 75, other 15
calls per individual) did not differ significantly from
the training percentage of assignment, and showed
72.0%, varying from 33.3% to 100% among individuals (v2 ¼ 2.49, d.f. ¼ 1, p ¼ 0.11).
For the yap, the discriminant analysis showed only
44.7% correct assignment to individual (Table 3,
Ethology 112 (2006) 815–825 ª 2006 The Authors
Journal compilation ª 2006 Blackwell Verlag, Berlin
Table 3: Assignment of dhole calls to a predicted individual, based
on discriminant analysis of the squeak, yap and yap–squeak call
parameter values
Predicted group
membership
Actual group
Squeak
Female 10
Male 11
Female 12
Male 14
Male 15
Total
Yap
Female 10
Male 11
Female 12
Male 14
Male 15
Total
Yap–squeak
Female 10
Male 11
Female 12
Male 14
Male 15
Total
10
11
12
14
15
Total
Correctly
classified (%)
17
0
0
0
0
17
4
22
1
0
1
28
5
3
25
0
2
35
0
0
1
30
0
31
4
5
3
0
27
39
30
30
30
30
30
150
56.7
73.3
83.3
100
90.0
80.7
13
12
4
1
3
33
8
5
4
5
2
24
2
6
16
3
1
28
4
6
2
11
2
25
3
1
4
10
22
40
30
30
30
30
30
150
43.3
16.7
53.3
36.7
73.3
44.7
30
0
0
0
0
30
0
28
2
0
1
31
0
1
28
0
0
29
0
0
0
30
0
30
0
1
0
0
29
30
30
30
30
30
30
150
100
93.3
93.3
100
96.7
96.7
Fig. 2b). This value of correct assignment did not differ significantly from the random value of 33.3%
(v2 ¼ 3.59, d.f. ¼ 1, p ¼ 0.06). Both frequency and
819
Biphonation for Individual Recognition in the Dhole
E. V. Volodina, I. A. Volodin, I. V. Isaeva & C. Unck
Table 4: Values of correlation between squeak call parameters and
the two first discriminant functions; eigenvalues and percent variance,
described by each function
Parameters
Root 1
Root 2
g0_ini
g0_end
g0_max
g0_min
g_peak
g_bandw
g_extrem
g_dur_inc
g_dur_dec
g_k_max
0.503
0.730
0.915
0.511
0.827
0.142
0.062
)0.107
0.036
)0.059
)0.593
0.074
)0.158
)0.528
)0.169
0.215
)0.246
0.680
)0.537
0.672
Eigenvalue
Percent variance
5.56
85.72%
0.58
8.88%
Table 5: Values of correlation between yap call parameters and the
two first discriminant functions; eigenvalues and percent variance,
described by each function
Fig. 2: Scatterplots showing separation produced by the first two discriminant functions of three call types for five dholes: (a) based on
parameters of the high-frequency call or squeak; (b) based on parameters of the low-frequency call or yap; (c) based on parameters of the
biphonic call or yap–squeak
temporal parameters contributed to discrimination.
The first discriminant function was related primarily
to parameters f0_end, f_dur_dec and f_k_max, that
820
Parameters
Root 1
Root 2
f0_ini
f0_end
f0_max
f0_min
f_peak
f_bandw
f_extrem
f_dur_inc
f_dur_dec
f_k_max
)0.112
0.559
)0.100
0.318
0.301
)0.131
0.229
0.065
)0.484
0.423
)0.007
)0.235
0.205
)0.294
0.141
0.444
)0.202
)0.261
)0.037
)0.043
Eigenvalue
Percent variance
0.42
61.23%
0.17
24.24%
described 61.23% of the variance only. The second
discriminant function was founded on f_bandw,
f0_min and f_dur_inc and described 24.24% of the
variance (Table 5). Cross-validation analysis showed
53.3% correct assignment for the training call set
(n ¼ 75, 15 calls per individual), with 40.0–60.0%
for particular individuals. Correct assignment for the
test call set (n ¼ 75, other 15 calls per individual)
was only 34.7% (26.7–46.7% between individuals),
significantly lower than the results for the training
set (v2 ¼ 4.57, d.f. ¼ 1, p < 0.05). So, the yap parameters showed low ability to discriminate between
individuals.
For the yap–squeak, the stepwise discriminant
analysis selected three parameters of the high-frequency component and seven parameters of the
low-frequency component that contributed mostly
to discrimination of calls to individuals (Table 6).
Ethology 112 (2006) 815–825 ª 2006 The Authors
Journal compilation ª 2006 Blackwell Verlag, Berlin
E. V. Volodina, I. A. Volodin, I. V. Isaeva & C. Unck
Biphonation for Individual Recognition in the Dhole
Table 6: Values of correlation between 10 yap–squeak call parameters, selected with the stepwise discriminant procedure, and the two
first discriminant functions; eigenvalues and percent variance, described by each function
Parameters
Root 1
Root 2
g0_end
g0_max
g_bandw
f0_ini
f0_max
f_bandw
f_extrem
f_dur_inc
f_dur_dec
f_k_max
0.850
0.792
0.075
)0.086
)0.057
0.032
0.092
0.087
0.051
0.058
)0.077
0.170
)0.117
)0.040
0.509
)0.031
0.059
0.437
)0.321
0.542
Eigenvalue
Percent variance
22.27
87.27%
2.12
8.30%
With these 10 parameters, standard discriminant
analysis provided 96.7% correct assignment (Table 3,
Fig. 2c). This value is significantly higher than the
random value of 32.0% (v2 ¼ 133.88, d.f. ¼ 1,
p < 0.001). In this case, the first discriminant function described 87.27% of the variance and was correlated only with fundamental frequency parameters
of high-frequency component (g0_end and g0_max).
The second discriminant function was based on the
low-frequency component parameters, both temporal (f_k_max and f_dur_inc), and frequency
(f0_max), and described as little as 8.3% of the variance (Table 6). Thus, the high-frequency component parameters contributed more to individual
discrimination of biphonic calls, and the same
parameters (g0_end and g0_max) that contributed
mainly to discrimination of yap–squeaks were
among the three that contributed mainly to discrimination of squeaks (Tables 4 and 6). Therefore, the
first discriminant function of the biphonic call relied
on the high-frequency parameters, whereas the second one on the low-frequency parameters (Table 6,
Fig. 2), that resulted in a small increase in the percentage of correct assignment to individuals. Crossvalidation analysis showed 94.7% correct assignment
for the training call set (n ¼ 75, 15 calls per individual), with 86.7–100% for particular individuals. Correct assignment for the test call set (n ¼ 75, other 15
calls per individual) did not differ from the training
percentage of assignment, and also showed 94.7%,
varying from 86.7% to 100% among individuals.
Finally, a comparison of discrimination percentages to individuals between the yap, squeak and
yap–squeak showed that the discriminability was
Ethology 112 (2006) 815–825 ª 2006 The Authors
Journal compilation ª 2006 Blackwell Verlag, Berlin
significantly higher for the squeak than for the yap
(v2 ¼ 40.02, d.f. ¼ 1, p < 0.001) and for the yap–
squeak than for the yap (v2 ¼ 95.84, d.f. ¼ 1,
p < 0.001) and for the squeak (v2 ¼ 17.55, d.f. ¼ 1,
p < 0.001).
Discussion
The presented data support the hypothesis that the
biphonic calls, representing a combination of the
high- and the low-frequency components, enhances
the potential for individual discrimination in the
dhole. However, the high-frequency squeak had
substantially higher potential for individual discrimination than the low-frequency yap. The yap did not
provide cues to individuality at all, showing a discrimination ability that did not differ significantly
from the random values.
For many canids, discriminant analysis-based
research has suggested a potential for individual
recognition by long-distance calls. Such data were
reported for howling of timber wolves (Tooze et al.
1990), for bark series of arctic foxes Alopex lagopus
(Frommolt et al. 1997, 2003) and swift foxes Vulpes
velox (Darden et al. 2003), for hoo-calls of African
wild dogs (Hartwig 2005) and corresponding to hoocall vocalization of dholes (Durbin 1998). Probably,
the cues to individuality in distant calls of canids
compensate for the absence of visual and olfactory
stimuli that provide cues to individuality in close
proximity.
Our study showed the presence of individual cues
in short-distance, low-intensity calls in the dhole.
For short-distance calls, the necessity of individual
cues is questionable, because the roles of visual and
olfactory channels are considered as much more
meaningful for short-distance communication. However, Owren & Rendall (1997, 2001) showed, that
for group-living primates with a complex system of
subordination, individual cues may also be important for short-distance calls, because calls of particular individuals adopt a role of conditioned stimuli,
evoking pleasant or unpleasant effects.
Consistent with this model, dholes might use individually distinctive short-distance calls to support
stable social relationships within a pack. Biphonic
calls are emitted in peaceful short-distance interactions (Volodin et al. 2001), and thus their production itself may evoke positive affiliative effects in
pack fellows and result in very low intrapack aggression, a characteristic for this species (Johnsingh
1982; Ludwig & Ludwig 2000). Such an effect of
vocalizations has been shown for complex primate
821
Biphonation for Individual Recognition in the Dhole
societies: if high-ranking animals emit groomingassociated affiliative calls when approaching subordinates, positive interactions occur more often, than if
they are silent (Bauers & de Waal 1991; Cheney
et al. 1995).
Although the effect-conditioning model outlines
the role of vocal tract formants as cues to individual
identity in short-distance primate calls, such as lowfrequency non-biphonic grunt and ‘coo’ calls (Owren
& Rendall 1997, 2001; Rendall et al. 1998), in the
dhole the individual cues are based on a very complex call structure, resulting from appearance of a
second fundamental frequency. With the absence of
formant cues in dhole calls (Volodin et al. 2001), just
the use of a second fundamental frequency allows
dholes to enhance strongly the potential for individual recognition of short-distance calls. The key role
of two frequencies in a call spectrum for both parent–chick and mate–mate recognition was also demonstrated for two penguin species (Aubin et al. 2000;
Lengagne et al. 2001; Aubin & Jouventin 2002).
Furthermore, exploitation of calls with two frequencies, lying far apart from each other, may provide additional advantages: cues to orientation and
direction of a movement of pack members emitting
these calls. These proposals come from physical frameworks, suggesting that high frequencies, propagated in the environment, attenuate much more
strongly, than low frequencies (Wiley & Richards
1978; Roberts et al. 1980; Owings & Morton 1998;
Naguib & Wiley 2001). For canids, this effect was
experimentally confirmed for the domestic dog
(Frommolt & Gebler 2004).
Our recent data showed that both the biphonic
yap–squeaks and non-biphonic yaps provide information about orientation of a caller to a listener:
when dholes called toward a microphone, the proportion of energy in the higher part of the call spectrum (above 5 kHz) was significantly higher than
when calls were emitted in an orientation away
from a microphone (Volodin et al. 2005a). The data,
available for two dolphin species, also showed equivocal relations between directionality and presence
of two frequencies in call spectra. For the killer
whale, only using biphonic calls provides reliable
information about orientation of a caller to a listener, whereas calls consisting exclusively of the
low-frequency component, do not provide such
information (Miller 2002). On the other hand, for
the Hawaiian spinner dolphin Stenella longirostris,
non-biphonic calls provided information about direction of a caller’s movement just as a consequence of
the difference in directionality and propagation
822
E. V. Volodina, I. A. Volodin, I. V. Isaeva & C. Unck
ability of fundamental frequency and higher harmonics (Lammers & Au 2003).
Taken together, the available data show that a
wide frequency spectrum with widely spaced frequency bands alone is sufficient for coding orientation of a caller to a listener. However, as the
amplitude of higher-ordered harmonics decreases
about 6–12 dB per octave (Titze 1994; Owren & Bernacki 1998), an addition of a second higher fundamental frequency, lying apart from the first one,
makes the biphonic call structure especially suitable
for coding orientation. Indeed, both Miller’s data on
the killer whale and our data on the dhole showed a
tendency for better performance of biphonic calls in
coding orientation of a caller to a listener in comparison with monophonic calls (Miller 2002; Volodin
et al. 2005a).
Therefore, in the dhole, the high-frequency
squeak, occurring singly, possesses the ability to discriminate between individuals, although less well
than the biphonic yap–squeak, but it could not provide cues to the orientation of a caller, because the
high-frequency narrow-band calls are the most difficult to locate (Marler 1955; Klump & Shalter 1984).
On the other hand, the low-frequency yap, occurring singly, possesses the ability to encode the orientation of a caller to a listener that is comparable
with, but not as good as, the biphonic yap–squeak,
and at the same time shows poor discrimination ability. Joined together into a biphonic call, they perform better in both respects.
The combination of enhanced potential to code
individuality with enhanced potential to code orientation of a caller to a listener makes biphonic calls
especially appropriate for delicate communication in
a pack with complex subordination between animals, living in close vicinity. This conclusion is in
accordance with our previous data, suggesting a very
high occurrence of biphonic calls in the dhole (Volodin & Volodina 2002). Moreover, for the African
wild dog, a second extremely social canid species, a
very high level of occurrence of biphonic contact
calls was reported (Wilden 1997; Wilden et al.
1998). Only two canid species – the dhole and the
African wild dog – show such high percentages of
production of biphonic calls, although many other
canids are able to produce two fundamentals simultaneously (Nikol’skii & Frommolt 1989; Solomon
et al. 1995; Riede et al. 2000; Volodin et al. 2005b).
It is probable, that in conditions of dense vegetation
and large social groups, under which these species
live, the acoustical channel, providing information
about individuality and spacing of animals, becomes
Ethology 112 (2006) 815–825 ª 2006 The Authors
Journal compilation ª 2006 Blackwell Verlag, Berlin
E. V. Volodina, I. A. Volodin, I. V. Isaeva & C. Unck
preferable for prompt communication even over
short distances, releasing the visual channel from
communicative load (e.g. Morton & Shalter 1977;
Lamprecht et al. 1985; Manser 1999). However, further research is necessary to reveal the more clear
communicative meaning of biphonic calls, using
playbacks and studying auditory perception.
Unfortunately, to date, there are no data concerning the possible mechanism of production of the
high-frequency squeak-like calls in canids. Our
observations of dholes and domestic dogs (E. Volodina, I. Volodin, unpubl. data) showed that during
emission of the squeak the mouth is closed, with
sound passing exclusively through the nose. As soon
as the animal begins adding a low-frequency yap, it
opens its mouth. Thus the high frequency is emitted
into the environment through the nose, whereas the
lower is emitted through the mouth. These observations are in accordance with X-ray video data
on vocalizing domestic dogs, showing that highfrequency whines were produced nasally, whereas
low-frequency barks were produced with open
mouth (Fitch 2000).
Acknowledgements
We are grateful to Leon Durbin, Robert Robbins, KarlHeinz Frommolt, Andrew Poyarkov and Olga Filatova
for valuable discussion, to Igor Pavlinov and Andrew
Babitsky for consulting in statistics, to Egor Bazykin
for providing the necessary literature, and to John
Lazarus, Gustav Peters and an anonymous referee for
comments and improvements of the manuscript. This
work was supported by grant no. 03-04-48919 from
the Russian Foundation for Basic Research.
Literature Cited
Aubin, T. & Jouventin, P. 2002: How to vocally identify
kin in a crowd: the penguin model. Adv. Study Behav.
31, 243—277.
Aubin, T., Jouventin, P. & Hildebrand, C. 2000: Penguins
use the two-voice system to recognize each other. Proc.
R. Soc. Lond. B. 267, 1081—1087.
Bauers, K. A. & de Waal, F. B. M. 1991: ‘‘Coo’’ vocalizations in stumptailed macaques: a controlled functional
analysis. Behaviour 119, 143—160.
Berry, D. A., Herzel, H., Titze, I. R. & Story, B. H. 1996:
Bifurcations in excised larynx experiments. J. Voice 10,
129—138.
Brown, C. H. & Cannito, M. P. 1995: Modes of vocal variation in Sykes’s monkey (Cercopithecus albogularis)
squeals. J. Comp. Psychol. 109, 398—415.
Ethology 112 (2006) 815–825 ª 2006 The Authors
Journal compilation ª 2006 Blackwell Verlag, Berlin
Biphonation for Individual Recognition in the Dhole
Brown, C. H., Alipour, F., Berry, D. A. & Montequin, D.
2003: Laryngeal biomechanics and vocal communication in the squirrel monkey (Saimiri boliviensis). J.
Acoust. Soc. Am. 113, 2114—2126.
Cheney, D. L., Seyfarth, R. M. & Silk, J. B. 1995: The
role of grunts in reconciling opponents and facilitation
interactions among adult female baboons. Anim.
Behav. 50, 249—257.
Cohen, J. A. 1977: A review of biology of the dhole or
Asiatic wild dog (Cuon alpinus Pallas). Anim. Regulat.
Stud. 1, 141—158.
Cohen, J. A. 1985: A note on the behaviour of captive
dholes (Cuon alpinus). J. Bombay Nat. Hist. Soc. 82,
183—187.
Darden, S. K., Dabelsteen, T. & Pedersen, S. B. 2003: A
potential tool for swift fox Vulpes velox conservation:
individuality of long-range barking sequences. J. Mammal. 84, 1417—1427.
Dillon, W. R. & Goldstein, M. 1984: Multivariate Analysis: Methods and Applications. Wiley, New York.
Durbin, L. S. 1998: Individuality in the whistle call of the
Asiatic wild dog Cuon alpinus. Bioacoustics 9,
197—206.
Fitch, W. T. 2000: The phonetic potential of nonhuman
vocal tracts: comparative cineradiographic observations
of vocalizing animals. Phonetica 57, 205—218.
Fitch, W. T. & Hauser, M. D. 2002: Unpacking ‘‘honesty’’: vertebrate vocal production and the evolution of
acoustic signals. In: Acoustic Communication (Simmons, A., Fay, R. R. & Popper, A. N., eds). Springer,
New York, pp. 65—137.
Fitch, W. T., Neubauer, J. & Herzel, H. 2002: Calls out of
chaos: the adaptive significance of nonlinear phenomena in mammalian vocal production. Anim. Behav. 63,
407—418.
Frommolt, K.-H. 1999: Sidebands – facts and artefacts.
Bioacoustics 10, 219—224.
Frommolt, K.-H. & Gebler, A. 2004: Directionality of
dog vocalizations. J. Acoust. Soc. Am. 116,
561—565.
Frommolt, K.-H., Kruchenkova, E. P. & Russig, H. 1997:
Individuality of territorial barking in Arctic foxes, Alopex lagopus (L., 1758). Z. fur Säugetierkunde 62 (Suppl.
2), 66—70.
Frommolt, K.-H., Goltsman, M. E. & Macdonald, D. W.
2003: Barking foxes, Alopex lagopus: field experiments
in individual recognition in territorial mammal. Anim.
Behav. 65, 509—518.
Hartwig, S. 2005: Individual acoustic identification as a
non-invasive conservation tool: an approach to the
conservation of the African wild dog Lycaon pictus
(Temminck, 1820). Bioacoustics 15, 35—50.
Herzel, H. 1993: Bifurcation and chaos in voice signals.
Appl. Mech. Rev. 46, 399—413.
823
Biphonation for Individual Recognition in the Dhole
Herzel, H. & Reuter, R. 1996: Biphonation in voice signals. In: Nonlinear, Chaotic, and Advanced Signal Processing Methods for Engineers and Scientists (Katz, R.
A., Frison, T. W., Kadke, J. B. & Bulsara, A. R. eds).
American Institute of Physics, Woodbury, pp.
644—657.
Herzel, H. & Reuter, R. 1997: Whistle register and biphonation in a child’s voice. Folia Phoniatr. Logop. 49,
216—224.
Herzel, H., Berry, D., Titze, I. R. & Steinecke, I. 1995:
Nonlinear dynamics of the voice: signal analysis and
biomechanical modelling. Chaos 5, 30—34.
Johnsingh, A. J. T. 1982: Reproductive and social behaviour of the dhole, Cuon alpinus (Canidae). J. Zool.
Lond 198, 443—463.
Kanwal, J. S., Matsumura, S., Ohlemiller, K. & Suga, N.
1994: Analysis of acoustic elements and syntax in
communication sounds emitted by mustached bats.
J. Acoust. Soc. Am. 96, 1229—1254.
Karanth, K. U. & Sunquist, M. E. 1995: Prey selection by
tiger, leopard and dhole in tropical forests. J. Anim.
Ecol. 64, 439—450.
Klump, G. M. & Shalter, M. D. 1984: Acoustic behaviour
of birds and mammals in the predator context. Z.
Tierpsychol. 66, 189—226.
Lammers, M. O. & Au, W. W. L. 2003: Directionality in
the whistles of Hawaiian spinner dolphins (Stenella
longirostris): a signal feature to cue direction of movement? Marine Mam. Sci. 19, 249—264.
Lamprecht, J., Kaiser, A., Peters, A. & Kirchgessner, C.
1985: Distance call duets in bar-headed geese (Anser
indicus): co-operation through visual relief of the partner? Z. Tierpsychol. 70, 211—218.
Lengagne, T., Lauga, J. & Aubin, T. 2001: Intra-syllabic
acoustic signatures used by the king penguin in parent-chick recognition: an experimental approach.
J. Exp. Biol. 204, 663—672.
Ludwig, W. & Ludwig, C. 2000: Beobachtungen zur
sozialen Organisation eines Rudels Rothunde (Cuon
alpinus) im Zoo Dresden. Zool. Garten N.F. 70,
39—59.
Manser, M. B. 1999: Response of foraging group members to sentinel calls in suricates, Suricata suricata. Proc.
R. Soc. Lond. B. 266, 1013—1019.
Marler, P. 1955: Characteristics of some animal calls. Nature 176, 6—8.
Mergell, P. & Herzel, H. 1997: Modelling biphonation –
the role of the vocal tract. Speech Communication 22,
141—154.
Mergell, P., Fitch, W. T. & Herzel, H. 1999: Modelling the
role of non-human vocal membranes in phonation.
J. Acoust. Soc. Am. 105, 2020—2028.
Miller, P. J. O. 2002: Mixed-directionality of killer whale
stereotyped calls: a direction of movement cue? Behav.
Ecol. Sociobiol. 52, 262—270.
824
E. V. Volodina, I. A. Volodin, I. V. Isaeva & C. Unck
Morton, E. S. & Shalter, M. D. 1977: Vocal response to
predators in pair-bonded Carolina wrens. Condor 79,
222—227.
Naguib, M. & Wiley, H. 2001: Estimating the distance to a
source of sound: mechanisms and adaptations for longrange communication. Anim. Behav. 62, 825—837.
Nikol’skii, A. A. & Frommolt, K.-H. 1989: Vocal Activity
in the Timber Wolf. Moscow State Univ. Press,
Moscow. [In Russian].
Owings, D. H. & Morton, E. S. 1998: Animal Vocal Communication: A New Approach. Cambridge Univ. Press,
Cambridge.
Owren, M. J. & Bernacki, R. H. 1998: Applying linear
predictive coding (LPC) to frequency-spectrum analysis
of animal acoustic signals. In: Animal Acoustic Communication (Hopp, S. L., Owren, M. J. & Evans, C. S.
eds). Springer-Verlag, Berlin, pp. 129—162.
Owren, M. J. & Rendall, D. 1997: An affect-conditioning
model of nonhuman primate vocal signalling. In: Perspectives in Ethology, Vol. 12 (Owings, D. H., Beecher,
M. D. & Thompson, N. S., eds). Plenum Press, New
York, pp. 299—346.
Owren, M. J. & Rendall, D. 2001: Sound on the rebound:
bringing form and function back to the forefront in
understanding non-human primate vocal signalling.
Evol. Anthropol. 10, 58—71.
Peters, G., Koenig, A., Herzel, H. P., Riede, T., East, M. &
Hofer, H. 2002: Powerful sound analysis tools and rudimentary knowledge of sound production mechanisms
in mammals – does this discrepancy matter? Adv.
Ethol. 37, 63.
Rendall, D., Owren, M. J. & Rodman, P. S. 1998: The
role of vocal tract filtering in identity cueing in rhesus
monkey (Macaca mulatta) vocalizations. J. Acoust. Soc.
Am. 103, 602—614.
Riede, T. & Stolle-Malorny, A. 1999: The vocal change of
a kitten with craniocerebellar trauma – a case study.
Bioacoustics 10, 131—141.
Riede, T., Wilden, I. & Tembrock, G. 1997: Subharmonics, biphonations, and frequency jumps – common
components of mammalian vocalization or indicators
for disorders. Z. fur Säugetierkunde 62 (Suppl. 2),
198—203.
Riede, T., Herzel, H., Mehwald, D., Seidner, W., Trumler,
E., Tembrock, G. & Böhme, G. 2000: Nonlinear phenomena and their anatomical basis in the natural
howling of a female dog-wolf breed. J. Acoust. Soc.
Am. 108, 1435—1442.
Riede, T., Herzel, H., Hammerschmidt, K., Brunnberg, L.
& Tembrock, G. 2001: The harmonic-to-noise ratio
applied to dog barks. J. Acoust. Soc. Am. 110,
2191—2197.
Robbins, R. L. 2000: Vocal communication in free-ranging African wild dogs (Lycaon pictus). Behaviour 137,
1271—1298.
Ethology 112 (2006) 815–825 ª 2006 The Authors
Journal compilation ª 2006 Blackwell Verlag, Berlin
E. V. Volodina, I. A. Volodin, I. V. Isaeva & C. Unck
Roberts, J. P., Kacelnik, A. & Hunter, M. J. 1980: Some
consequences of sound interference patterns for birdsong. Acoustic Lett. 3, 141—146.
Solomon, N. P., Luschei, E. & Kang, L. 1995: Fundamental frequency and tracheal pressure during three types
of vocalizations elicited from anaesthetized dogs. J.
Voice 9, 403—412.
Sosnovskii, I. P. 1967: Breeding the red dog or dhole Cuon
alpinus at Moscow Zoo. Int. Zoo Yearbook 7, 120—122.
Tigges, M., Mergell, P., Herzel, H., Wittenberg, Th. &
Eysholdt, U. 1997: Observation and modelling of glottal biphonation. Acoustica 83, 707—714.
Titze, I. R. 1994: Principles of Voice Production. Prentice
Hall, Englewood Cliffs, NJ.
Tooze, Z. J., Harrington, F. H. & Fentress, J. C. 1990:
Individually distinct vocalizations in timber wolves,
Canis lupus. Anim. Behav. 40, 723—730.
Venkataraman, A. B. 1998: Male-biased adult sex-ratios
and their significance for cooperative breeding in
dhole, Cuon alpinus, packs. Ethology 104, 671—684.
Venkataraman, A. B., Arumugam, R. & Sukumar, R.
1995: The foraging ecology of dhole (Cuon alpinus) in
Mudumalai-Sanctuary, southern India. J. Zool. 237,
543—561.
Volodin, I. A. & Volodina, E. V. 2002: Biphonation as a
prominent feature of the dhole Cuon alpinus sounds.
Bioacoustics 13, 105—120.
Volodin, I. A., Volodina, E. V. & Isaeva, I. V. 2001: Vocal
repertoire in the dhole Cuon alpinus (Carnivora, Cani-
Ethology 112 (2006) 815–825 ª 2006 The Authors
Journal compilation ª 2006 Blackwell Verlag, Berlin
Biphonation for Individual Recognition in the Dhole
dae) in captivity. Russian J. Zool. 80, 1252—1267. [in
Russian, translated into English in Entomological
Review, 2001, 81, (Suppl. 2), S346–S361].
Volodin, I., Nagailik, M. & Volodina, E. 2005a: Cues to a
direction of movement in biphonic calls of the dhole
(Cuon alpinus). In: XX Congress of International BioAcoustic Council, Book of Abstracts (Trilar, T., ed.).
Portoroz, Slovenia, p. 36.
Volodin, I. A., Volodina, E. V. & Filatova, O. A. 2005b:
Structural peculiarities, occurrence and functional significance of nonlinear phenomena in calls of terrestrial
mammals. J. General Biol. 66, 346—362. [In Russian].
Volodina, E., Volodin, I. & Filatova, O. 2005: Quantitative analysis of nonlinear phenomena in whines of the
domestic dog (Canis familiaris). In: XX. Congress of
International BioAcoustic Council, Book of Abstracts
(Trilar, T., ed.). Portoroz, Slovenia, p. 37.
Wilden, I. 1997: Phonetische Variabilität in der Lautgebung Afrikanischer Wildhunde (Lycaon pictus) und deren frühe Ontogenese. Aachen, Shaker Verlag.
Wilden, I., Herzel, H., Peters, G. & Tembrock, G. 1998:
Subharmonics, biphonation, and deterministic chaos in
mammal vocalization. Bioacoustics 9, 171—196.
Wiley, R. H. & Richards, D. B. 1978: Physical constraints
on acoustic communication in the atmosphere: implication for the evolution of animal vocalizations. Behav.
Ecol. Sociobiol. 3, 69—94.
825