Content uploaded by Svetlana S Gogoleva
Author content
All content in this area was uploaded by Svetlana S Gogoleva on Aug 29, 2016
Content may be subject to copyright.
Bioacoustics
The International Journal of Animal Sound and its Recording, 2008, Vol. 18, pp. 99–132
© 2008 AB Academic Publishers
TO BARK OR NOT TO BARK: VOCALIZATIONS
BY RED FOXES SELECTED FOR TAMENESS OR
AGGRESSIVENESS TOWARD HUMANS
S.S. GOGOLEVA1, J.A. VOLODIN1, 2*, E.V. VOLODINA2 AND L.N. TRUT3
*Email: volodinsvoc@yahoo.com
1Department of Vertebrate Zoology, Faculty of Biology, Lomonosov Moscow State
University, Vorobievy Gory, Moscow, 119992, Russia
2Scientic Research Department, Moscow Zoo, B. Gruzinskaya, 1, Moscow, 123242,
Russia
3Laboratory of Evolutionary Genetic, Institute of Cytology and Genetics, Siberian
Branch, Russian Academy of Sciences, Pr. Lavrentjeva, 10, Novosibirsk, 630090,
Russia
ABSTRACT
In this study we classify call structures and compare vocalizations toward humans by
captive red foxes Vulpes vulpes, articially selected for behaviour: 25 domesticated,
or “Tame” animals, selected for tameness toward people, 25 “Aggressive” animals,
selected for aggression toward people, and 25 “Unselected” control foxes, representing
the “wild” model of vocal behaviour. In total, 12,964 calls were classied visually from
spectrograms into ve voiced (tonal) (whine, moo, cackle, growl and bark), and three
unvoiced, or noisy (pant, snort and cough) call types. The classication results were
veried with discriminant function analysis (DFA) and randomization. We found that
the Aggressive and Unselected foxes produced the same call type sets toward humans,
whereas the Tame foxes used distinctive vocalizations toward humans. The Tame
and Aggressive foxes had signicantly higher percentages of time spent vocalizing
than the Unselected, in support of Cohen & Fox (1976) hypothesis that domestication
relaxes the selection pressure for silence, still acting in wild canids. Unlike in dogs,
the “domesticated” Tame foxes did not show hypertrophied barking toward humans,
using instead the cackle and pant. We conclude that the use of a certain call type for
communication between humans and canids is species-specic, and not is the direct
effect of domestication per se.
Keywords: vocalization, domestication, vocal communication, nonlinear phenomena,
articulation, red fox, Vulpes vulpes, Canidae
INTRODUCTION
Vocal behaviour of fox-like canids has been the subject of a long-
standing research tradition, especially in relation to questions of
systematics, structural variability, contexts of production for different
100
call types, their functions and, most recently, species conservation.
Studies reviewing vocal catalogues in Canidae, as a rule, demarcate
between vocal repertoires of fox-like and wolf-like canids (Cohen & Fox
1976; Schassburger 1987). The structural variability and contexts of
vocalizations were reported in wild red fox Vulpes vulpes by Newton-
Fisher et al. (1993) and in captive swift fox V. velox by Darden &
Dabelsteen (2006). Individual variability of serial barking in wild
Arctic fox Alopex lagopus (Frommolt et al. 2003) and in captive swift
foxes (Darden et al. 2003) were studied to determine call functions
and the possible applicability of vocal identity to acoustic-based
monitoring in the wild (Terry et al. 2005).
This study considers another problem, still little studied – the
effects of captivity and domestication on vocalization. The few reports
(Cohen & Fox 1976; Budde 1998; Monticelli & Ades 2001) provide
little insight into the general processes that affect call structure
during domestication. The domestic dog Canis familiaris, whose
vocal behaviour is considered to be a result of domestication of its
wild ancestor, the timber wolf Canis lupus, is not a good model for
estimating the effects of domestication on vocalization. Because the
domestic dog and the timber wolf have a long evolution as independent
species (Tchernov & Valla 1997; Sablin & Khlopachev 2002); their
vocal repertoires may have differed signicantly already at an early
stage of domestication. Therefore, the modern timber wolf may not
represent an undomesticated “default” stage for vocal repertoire of
the domestic dog. Ideally, to estimate the effects of domestication on
vocalizations, domesticated and undomesticated individuals within a
species should be compared.
A good model for such a within-species analysis comes from
farm-bred red foxes, selected either for tame behaviour or for enhanced
aggressiveness toward humans, in comparison with unselected
controls (Belyaev 1979; Trut 1999; Gulevich et al. 2004; Trut et
al. 2006). According to Belyaev (1979), directional selection simply
for tame behaviour toward humans provoked the domestication,
which has resulted in the evolution of the dog from the wolf.
Testing of this hypothesis started in 1960, with the beginning of
the directional articial selection of farm foxes for positive attitudes
toward people in Novosibirsk (Russia). Further, in 1970, directional
selection of previously unselected foxes for enhanced aggressiveness
toward people was started. Additionally, throughout these times, a
population of foxes unselected for behaviour has been living on this
farm (Trut 1999). The “Unselected” foxes show escape reaction toward
people and keep a maximum possible distance apart from a human.
Distinctive from the Unselected foxes, both “Tame” and “Aggressive”
foxes are not afraid of humans; but the Tame foxes are positive to
people, whereas the Aggressive foxes are negative to people. These
behavioural differences between the Tame and Aggressive foxes are
101
genetically determined, as has been proven by cross-foster parenting
and by embryo transplantation experiments (Trut 1980). The study
of captive foxes differing in tameness may help to estimate both the
quantitative and the qualitative shifts in vocal behaviour that have
occurred under domestication. The control population of the Unselected
foxes represents a “wild” model of vocal behaviour.
One particular question – of the hypertrophied barking of
the dog in comparison with the wolf, may also be examined with
the captive red fox population. Cohen & Fox (1976) proposed that
hypertrophied barking in dog resulted from the relaxation of selection
pressure for silence that still acts in the wolf. If so, we should expect
that Tame foxes will have an enhanced barking rate compared with
those unselected for tameness. Furthermore, with the captive fox
population it is possible to determine whether Tame foxes will prefer
barks to all other vocalizations or whether the relaxation of selection
pressure for silence affects all vocal types equally. Another enigmatic
question is whether hypertrophied barking in domestic dogs has
evolved as a vocal response toward humans (Feddersen-Petersen
2000; Yin 2002). Do Tame foxes use any specic vocalizations toward
humans compared to foxes not selected for tameness?
To answer these questions, we must determine the range of vocal
diversity produced by red foxes in captivity, i.e., we need to describe
the repertoire of their vocal structures. The existing descriptions of
the vocal repertoire of the red fox (e.g., Cohen & Fox 1976; Tembrock
1976; Newton-Fisher et al. 1993) must be revised, because previous
attempts based on limited sample sizes do not provide a detailed
evaluation of vocal variability in captivity and do not consider fox
vocalizations directed toward humans.
Furthermore, the revised catalogue of red fox vocalizations
should include the analysis of nonlinear phenomena. Many canids
produce nonlinear phenomena (e.g., Wilden et al. 1998; Riede et al.
2000; Volodin & Volodina 2002; Volodina et al. 2006a) that greatly
enhance variability in call structure. Such nonlinear phenomena as
subharmonics, deterministic chaos and frequency jumps emerge from
slight shifts in the operation of mammalian vocal apparatus – the
paired vocal folds that create a system of coupled oscillators (Wilden
et al. 1998; Fitch et al. 2002). Thus, the nonlinear phenomena merely
represent different working modes of the same voice source (Berry
et al. 1996). Previously, the appearance of nonlinear phenomena
was used to argue for subdividing calls into separate vocal types, or
attributing these variants to “mixed”, or “intermediate” vocalizations
(Cohen & Fox 1976; Tembrock 1976; Schassburger 1987; Newton-
Fisher et al. 1993).
Also, all earlier studies in the red fox ignored articulation effects
– another factor promoting vocal variability. The articulators (soft
palate, mandible, tongue, lips etc.) are extrinsic to the vocal source
102
(the vocal folds in the larynx) and can modify the vocal signal (Fant
1960). Given that these articulation effects greatly inuence the nal
structure of the vocal output, they deserve as much consideration
as nonlinear phenomena. Unfortunately, means to recognize and
analyze these articulation effects are not as advanced as those that
exist to recognize and analyze nonlinear phenomena (Wilden et al.
1998). To date, there are very few studies on articulation effects in
nonhumans. Shipley et al. (1991) demonstrated the effects of mouth
opening on the vocal output in the domestic cat Felis catus. Further,
there are anatomical and bioacoustical data on the role of articulation
in modifying the leopard alarm call in Diana monkeys Cercopithecus
diana (Riede et al. 2005). Finally, in the Monk parakeet Myiopsitta
monachus, the tongue articulatory effects on vocal output were
conrmed experimentally using air forced through vocal tract post
mortem (Beckers et al. 2004).
In the present study, we classify the call structural diversity and
compare vocal behaviour of captive red foxes, selected for tameness,
selected for aggressiveness and unselected for any behaviour.
MATERIALS AND METHODS
Subjects and study site
Our subjects were 75 captive adult female red foxes, aged from 1 to 2
years, kept at the experimental fur farm of the Institute of Cytology
and Genetics, Novosibirsk, Russia. Since 1960, foxes at this farm have
been strictly selected for tame behaviour toward humans. In addition,
since 1970 previously unselected farm foxes were strictly selected
for enhanced aggressiveness toward humans (Trut 1999; Gulevich
et al. 2004; Trut et al. 2006). During this long-term selection work,
following an outbreeding scheme, more than 50,000 foxes were reared
and tested for behaviour toward humans (Trut 2001; Kukekova et al.
2004).
We recorded calls between 6 July and 18 August 2005 from
foxes derived from three selection groups: 25 Tame foxes (selected for
tame behaviour toward humans, 44–45 generations since the start
of selection), 25 Aggressive foxes (selected for aggressive behaviour
toward humans; 34–35 generations since the start of selection) and 25
Unselected foxes, representing a control group of animals not selected
for behaviour. We recorded only female foxes, because we would not
have the necessary sample of males (sex ratios in industrial fur
populations is usually 1:4 in favour of females).
The focal foxes were recorded in their home individual outdoor
wire mesh cages measuring 70 × 85 × 90 cm with shelters 70 × 85
× 85 cm. The cages were arranged in batteries of 50 cages per row,
103
with two rows opposite each other and a 1.7 m wide pass between
them and covered with a slated roof. The Tame and Aggressive foxes
were not kept in separated batteries, whereas the Unselected foxes
were kept separately from the other selection groups. The criterion
for selection of subject foxes among approximately 400 farm foxes was
their readiness to produce calls toward humans (according to reports
of the farm staff). On the farm, it is prohibited to become “familiar” to
any particular fox, because it may inuence the behaviour of animals
– only routine procedures or scientic or biomedical tests may be
done.
Data collecting
We used a Marantz PMD-222 cassette recorder with an AKG-
C1000S cardioid electret condenser microphone, and Type II chrome
audiocassettes EMTEC-CS II. The system had a frequency response
of 0.04–14 kHz at a tape speed of 4.75 mm/s.
All the sound recordings were done by the same individual
researcher, who was unfamiliar to the foxes, during a single record
session per animal. The researcher approached to the focal fox’s cage
and started a recording, which lasted from 4 to 6 minutes. While
recording, the researcher stood 0.5–1 m from a focal fox cage, affecting
the animal by her presence. The threshold for calling varied between
individuals. If an animal did not vocalize spontaneously or stopped
vocalizing, the researcher additionally provoked it by moving a hand
toward the cage, opening a cage door, or caressing an animal. The
distance between microphone and focal fox varied from 25 to 100 cm;
the orientation of animal to microphone also varied, but mostly was
frontal or lateral. If a neighbouring animal called during a recording,
the calls of the focal fox were identied by the researcher's voice.
Of the 75 examined foxes, 53 were separated from other examined
foxes by a minimum of one cage (i.e., they were not neighbours). The
remaining 22 examined foxes were neighbours (11 pairs of adjacent
cages), but these ‘neighbouring’ foxes have never been tested during
one day. During one day, 8–12 tests could be conducted. The total
duration of recordings was 450 minutes.
Call analysis
Calls were analysed using Avisoft SASLab Pro software v. 4.33
(Avisoft Bioacoustics, Berlin, Germany). Call recordings were
digitized with a 22.05 kHz sampling frequency and 16-bit precision,
with each recording a separate le. Spectrograms for analysis were
created using Hamming window, FFT-length 1024 points, frame 50%
104
and overlap 87.5%. These settings provided a bandwidth of 56 Hz, a
frequency resolution of 21 Hz and time resolution of 5.8 ms. Calls for
spectrogram illustrations were digitized with a 11.025 kHz sampling
frequency and created using Hamming window, FFT-length 512
points, frame 50% and overlap 87.5%.
We classied each call visually to one of eight structural types
and checked it for the presence of nonlinear phenomena and/or
articulation effects. We considered sound utterances as separate calls
if they were separated with a silent interval longer than 20 ms. In
total, we analyzed 12,964 calls from 75 foxes.
We measured the duration of each recording and the duration
of each call with the standard marker cursor in the main window
of Avisoft. Other measurements were made only for selected call
samples. To make these samples, we took 1 to 3 calls per call type
per animal, who produced calls of the given type, and which were
of good quality for the given type i.e. not superimposed with calls of
other animals and with background noise.
For the voiced calls (see Results), we measured four fundamental
frequency parameters using the reticule cursor in the spectrogram
window of Avisoft: the initial (f beg), nal (f end), maximum (f max)
and minimum (f min) fundamental frequencies. For each call, we
measured the frequency of maximum amplitude (f peak). For the
unvoiced calls (see Results) we also measured the lower (quart 25),
medium (quart 50) and upper (quart 75) quartiles of mean power
spectrum.
For the growls and whines with a rhythm articulation effect
(see Results), we also measured the pulse period, from the beginning
of a previous sound pulse to the beginning of the following sound
pulse. Further, we calculated the pulse rate as an inverse value to
the pulse period. For the cackles and pants (see Results), we measured
the period between the consecutive calls within a series, from the
beginning of a preceding call to the beginning of the following one.
All measurements were exported automatically to Microsoft Excel
spreadsheets.
We then calculated the average values for all measured
parameters for each individual, and further calculated the averages
of these averages for each call type. Since the numbers of calls of
each type differed substantially between individuals, this approach
avoided pseudoreplication (Leger & Didrichsons 1994).
For calls assigned to the type whine (see Results), we registered
the presence or absence of nonlinear phenomena (Wilden et al. 1998;
Fitch et al. 2002; Volodina et al. 2006a). We registered all prominent
frequency jumps within calls. Also, we registered the appearance of
subharmonics and/or deterministic chaos in cases where the total
duration of the call portions bearing these nonlinear phenomena was
at least 30 ms for calls shorter than 300 ms, not less than 10% of
105
duration for calls of duration from 300 to 1000 ms, and at least 100
ms for calls longer than 1000 ms.
For the whines, we also noted the presence or absence of
articulation effects (see Results). Flutter was registered if two and
more inversed U-modulations of the fundamental frequency, one
after another, occurred in a call. Rhythm was registered if two or
more successive sound pulses, breaking a tonal vocalization, occurred
within a call. Babble was noted if at least one part with U-shaped
modulation of the fundamental frequency was presented in a call.
Statistical Analyses
All statistical analyses were made in STATISTICA, v. 6.0 (StatSoft,
Inc., Tulsa, USA). All means are given as mean ± SD.
We used the discriminant function analysis (DFA) forward
stepwise procedure to conrm our visual classication of call types
separately within voiced and within unvoiced calls. For each call
type, the values were normally distributed for most parameters
(Kolmogorov-Smirnov test). Since the DFA is relatively robust to
departures from normality (Dillon & Goldstein 1984), this was not
an obstacle to the application of this test.
To validate the DFA results, we performed a randomization
analysis (Solow 1990). For that, we did 500 permutation procedures
with software macros specially created for STATISTICA software.
For example, for the voiced calls, each permutation procedure
included a random permutation of all measured voiced calls among
5 randomization groups (the number of voiced call types), followed
by DFA standard procedure. Then we created a distribution of 500
received classication percentages to randomization groups and
estimated a position of the observed value of assignment to type
within this distribution. If the observed value exceeded 95% or 99% of
the values within this distribution, we established that the observed
value differed signicantly from the random one with probability p <
0.05 or p < 0.01 respectively (Solow 1990). For the unvoiced calls, the
randomization analysis was done similarly, but with 3 randomization
groups, (the number of the unvoiced call types). The randomization
procedure allowed us to compensate for the unequal samples for the
different call types, included in the DFA, since such inequality may
otherwise affect the correctness of a classication (Titus et al. 1984).
Proportions of time spent vocalizing (total duration of calls
within a recording divided by the duration of a recording) between
the selection groups were compared with White T-criterion and one-
way ANOVA, since distribution of values for proportions of time spent
vocalizing did not differ from normal for all the selection groups (p >
0.05, Kolmogorov-Smirnov test).
106
RESULTS
Vocal structures
We subdivided all recorded vocalizations into two structural classes:
the voiced and unvoiced calls. The voiced calls showed signs of
production from a vocal source (i.e., larynx with vocal folds): a tonal
spectrum with a fundamental frequency and its harmonics, sometimes
bearing nonlinear phenomena and/or articulation effects. The unvoiced
calls lacked a fundamental frequency and had an explosive wideband
spectrum, revealing their non-vocal nature, i.e., their production not
with vocal folds but with another source, most probably turbulence
(vortices), occurring during passage of air through a narrowest vocal
tract.
Further we subdivided calls into eight types: ve voiced, or
tonal calls (whine, moo, cackle, growl and bark), and three unvoiced
or wideband calls (pant, snort and cough). We observed also many
transitional calls from one type to another. Tables 1 and 2 present
the mean values for the measured parameters of these call types.
Voiced calls
Whine. These are tonal calls, variable in amplitude, pure or complicated
with either nonlinear phenomena or articulation effects or both (Figure
1). The duration varied from 51 to 2607 ms (mean 711 ± 502 ms)
among animals and could reach a maximum of 7100 ms (Table 1).
The maximum fundamental frequency varied from 0.32 to 1.21 kHz
(mean 0.66 ± 0.21 kHz) among animals. The frequency modulation
varied both in depth and shape (Table 1). The maximum amplitude
frequency (mean 0.85 ± 0.53 kHz) coincided with the fundamental
only in 107 of 161 (66.5%) whines, and shifted to higher harmonics
in the rest 33.5% whines.
The whines could contain any of the three nonlinear phenomena
(subharmonics, deterministic chaos and frequency jumps) and any
of the three articulation effects (utter, rhythm and babble). More
Figure 1. Spectrogram of four whines from three female red foxes. The short
call at 0.5 s is considered to be a separate whine, since it is separated from
the following call with a silent interval longer than 20 ms.
107
Figure 2. Spectrogram (below) and waveform (above) of whines with the
articulation effect utter, recorded from two female red foxes. The waveform
shows the rhythmic amplitude modulation.
than one nonlinear phenomenon and/or articulation effect could occur
within a whine call.
Articulation effects occurring in whines. The utter was a
repeatedly produced inverse-U modulation of the fundamental
frequency contour (Figure 2). The period from the beginning of a
preceding inverse-U modulation to the beginning of the following one
could vary both within and between calls. The rhythm was recognizable
from the short breaks of a call spectrum (the fundamental frequency
and its harmonics), resulting in brief, broadband, rapid-onset pulses,
sounding as a vibration or chirr (Figures 3, 4, 7). The pulse rate
varied from 32 to 88 pulses/s (Hz) among animals (mean 57 ± 15
Hz, N animals = 24, n calls = 58) and could vary even within a call.
Sometimes the rhythm occurred over an entire call (usually over a
short call), and in this case, the call structure looked like a sequence
of pulses, without any tracks of the fundamental frequency (Figure
4). It seems that the rhythm in whines arises when a caller for a
moment perfectly blocks the air ow through the vocal tract. The
babble is a U-shaped modulation of the fundamental frequency, with
an abrupt amplitude shift to higher frequencies at the beginning of
each U-shaped fragment (Figures 4, 5). Usually, before this shift, an
abrupt weakening of the sound amplitude is visible, sometimes even
leading to a small silent break in a call spectrum. It is proposed
that the babble results from the work of the articulators (primarily
– tongue and mandible) during the production of the whines.
Nonlinear phenomena occurring in whines. Subharmonics
represent the appearance of additional frequency bands of ½, ⅓ and
¼ of the fundamental frequency (Figure 6). Deterministic chaos shows
108
TABLE 1
Mean ± SD values for parameters of ve voiced call types of captive red foxes: duration - whole call duration, f beg – initial
fundamental frequency, f end – nal fundamental frequency, f max – maximum fundamental frequency, f min – minimum
fundamental frequency, f peak – maximum amplitude frequency. N – number of individual foxes, n – number of calls.
Call type Call parameter
duration, ms f beg, kHz f end, kHz f max, kHz f min, kHz f peak, kHz
Whine 711 ± 502 0.53 ± 0.21 0.49 ± 0.16 0.66 ± 0.21 0.41 ± 0.15 0.85 ± 0.53
N = 59; n = 4810 N = 57; n = 161
Moo 347 ± 324 0.28 ± 0.09 0.28 ± 0.09 0.36 ± 0.10 0.24 ± 0.08 0.28 ± 0.09
N = 49; n = 1189 N = 48; n = 139
Cackle 61 ± 10 0.38 ± 0.06 0.39 ± 0.08 0.49 ± 0.05 0.33 ± 0.07 0.76 ± 0.27
N = 19; n = 1006 N = 18; n = 146
Growl 668 ± 428 0.20 ± 0.05 0.18 ± 0.05 0.23 ± 0.07 0.16 ± 0.04 0.19 ± 0.06
N = 19; n = 367 N = 19; n = 46
Bark 106 ± 16 0.86 ± 0.06 0.53 ± 0.08 1.12 ± 0.12 0.52 ± 0.10 1.31 ± 0.40
N = 2; n = 11
109
TABLE 2
Mean ± SD values for parameters of three unvoiced call types of captive red foxes: duration – whole call duration, f peak – maximum
amplitude frequency, quart 25 – lower quartile, quart 50 – medium quartile, quart 75 – upper quartile. N – number of individual
foxes, n – number of calls.
Call type Call parameter
duration, ms f peak, kHz quart 25, kHz quart 50, kHz quart 75, kHz
Pant 42 ± 9 1.00 ± 0.55 1.17 ± 0.31 2.62 ± 0.60 4.57 ± 0.72
N = 16; n = 2387 N = 16; n = 60
Snort 77 ± 30 0.22 ± 0.05 0.39 ± 0.23 1.42 ± 0.70 4.08 ± 0.92
N = 45; n = 1547 N = 45; n = 129
Cough 72 ± 19 0.85 ± 0.61 1.09 ± 0.31 2.39 ± 0.54 4.64 ± 0.86
N = 45; n = 1647 N = 45; n = 127
110
Figure 3. Spectrogram (below) and waveform (above) of a female red fox
whine with articulation effect rhythm at 0.2–1.6 s and multiple nonlinear
phenomena: a segment with deterministic chaos at 1.7–2.2 s, a tonal segment
at 2.2–2.6 s, a second segment with deterministic chaos at 2.6–2.9 s and a
frequency jump at 2.9 s.
Figure 4. Spectrogram of four whines recorded from the same female red fox,
with the articulation effect rhythm. The rst, third and fourth calls contain
rhythm throughout the entire call whereas the second call – only has it in a
few call segments. Also, in the second call, a segment with the articulation
effect babble is visible at 0.5–1.0 s.
Figure 5. Spectrogram of a whine with the articulation effect babble.
111
Figure 6. Spectrogram of three whines with nonlinear phenomena from
three female red foxes. The rst whine bears subharmonics, the second one
– subharmonics and deterministic chaos, and the third one deterministic
chaos.
Figure 7. Spectrogram of two whines with nonlinear phenomena and
articulation effects from two female foxes. The rst whine begins with
deterministic chaos, further superimposed with rhythm. The second whine
begins and ends with segments of deterministic chaos, and also contains four
frequency jumps and a subharmonic segment at 1.7–1.8 s.
Figure 8. Spectrogram of a female red fox moo.
112
Figure 9. Spectrogram of a female fox call, transitional from moo to whine
at 1.5 s.
Figure 10. Spectrogram of a natural series of female red fox cackles.
Figure 11. Spectrogram (below) and waveform (above) of a female red fox
growl.
113
Figure 12. Spectrogram (below) and waveform (above) of a female red fox
call, transitional from whine to growl. The second part of the tonal segment
at 0.4–1.0 s contains a peculiar frequency modulation – the articulation
effect utter.
Figure 13. Spectrogram of female red fox barks. The intervals between calls
are not natural; the rst four calls were recorded from one fox, the fth
and the sixth calls from another. The sixth call is transitional from bark to
whine.
Figure 14. Spectrogram of the serial bark of a female red fox, derived from
the Unselected group. The overlap of a neighbour fox moo is visible at
approximately 0.4 kHz. The call was recorded out of human-related context,
outside this study.
114
Figure 15. Spectrogram of a natural series of female red fox cackles (rst
and three last calls) and pants (the remaining calls).
Figure 16. Spectrogram of ve snorts recorded from three female red foxes
(the rst from one, the second and third from the second, and the fourth
and fth from the third). Notice that each snort includes the short explosive
onset, passing into the more prolonged exhalation, looking like a “cloud” of
wideband noise on the spectrogram.
Figure 17. Spectrogram of ve coughs, the rst three from one female red fox
and the last two from another.
115
Figure 18. Spectrogram illustrating the structural distinctiveness between
the short whines containing the articulation effect rhythm (the rst two
calls), snorts (the third and forth calls) and coughs (the two last calls). The
fth call is transitional from snort to cough.
Figure 20. Spectrogram of two whistles of a female red fox from the Unselected
group. The call was recorded outside the design of this study.
Figure 19. Spectrogram (below) and waveform (above) of two whoops of a
female red fox derived from the Unselected group.
116
Figure 21. Percentages of calls of each type, recorded from the Unselected,
Aggressive and Tame foxes.
Figure 22. Mean percentages of time spent vocalizing for the Unselected,
Aggressive and Tame foxes, whiskers represent SD.
117
the sound amplitude uniformly distributed over the call spectrum,
with the residual frequency bands at the ranges of the fundamental
frequency and its harmonics (Figure 6, 7). Frequency jump represents
a momentary shift of fundamental frequency of a call (Figure 3, 7).
Moo. A low-amplitude tonal call, with an accented fundamental
frequency band and poorly visible harmonics (Figure 8). The
fundamental frequency was usually restricted to the 0.2–0.4 kHz
range. The mean duration 347 ± 324 ms varied from 55 to 1808 ms
among animals (Table 1). Unlike whines, the moo showed very weak
frequency modulation (Table 1). The maximum amplitude frequency
(mean 0.28 ± 0.09 kHz) shifted to higher harmonics only in 2 of 139
(1.4%) moo calls, coinciding with the fundamental frequency band in
the remaining 98.6%. Evident in transitional calls from moo to whine,
just at the point of transition from the one call type to another, is
an abrupt enforcement of the amplitude, accompanied by the equal
distribution of the sound amplitude among harmonics (Figure 9). It
seems that the mouth was closed during moo and the moment of
transition from moo to whine coincided with the opening of the fox's
mouth.
Cackle. Cackles were tonal calls shorter than 100 ms, produced
in semi-regular series with an intercall period from 170 to 400 ms
(mean 210 ± 50 ms, N animals = 18, n periods = 146), that varied
both within and between series (Figure 10). The mean duration of 61
± 10 ms ranged from 39 to 79 ms between animals. The fundamental
frequency was usually restricted within the 0.3–0.5 kHz range and
did not exceed 0.8 kHz (Table 1). The frequency modulation was
upward, downward or bell-shaped. The cackle was often interspersed
with pants or whines. Unlike whines, cackles were shorter and had
a nal noisy segment, probably representing a peculiar exhalation.
Whines, when alternating with cackles in series, also often had the
similar end noisy segment. In such cases, we formally classied calls
longer than 100 ms as whines, and calls shorter than 100 ms as
cackles.
Growl. A low-amplitude and low-frequency call, with obligate
pulsation, varying from 22 to 35 pulses/s (Hz) between animals
(mean 27 ± 3 Hz, N animals = 19, n calls = 46) (Figure 11). The
fundamental frequency was usually restricted within the 0.25–0.3
kHz range and did not exceed 0.4 kHz (Table 1). The fundamental
frequency was poorly traceable, often broken into separate pulses.
The maximum amplitude frequency coincided with the fundamental
frequency band (Table 1). The mean duration 668 ± 428 ms ranged
from 202 to 1806 ms between animals (Table 1). Often, a tonal onset
(whine or moo) preceded the growl (Figure 12).
Bark. This short, explosive high-amplitude tonal call had
a clear inverse-U frequency modulation (Figure 13). Barks were
recorded only from two foxes. The maximum fundamental frequency
118
was 1.12 ± 0.12 kHz; the mean duration was 106 ± 16 ms (Table
1). The frequency modulation of the bark (mean 0.60 ± 0.22 kHz)
was the deepest among the voiced calls (Table 1). The barks could
alternate with whines. The transitional calls from bark to whine were
also registered (Figure 13).
Throughout our study, the focal foxes never emitted the serial
barks – a sequence of serial calls (Figure 14) – toward the researcher.
However, serial barks were heard regularly from foxes of all selection
groups (Tame, Aggressive and Unselected) in contexts without
humans. Foxes probably produced serial barks toward conspecics or
in response to stimuli not related to the appearance of the researcher
nearby.
Unvoiced calls
Pant. These were low-amplitude short exhalations, produced repeatedly
in a semi-regular series or interspersed with cackles and whines
(Figure 15). Pants were structurally similar to cackles, but did not
contain the voice (tonal) component. The intercall period ranged from
130 to 260 ms (mean 180 ± 40 ms, N animals = 15, n periods = 59),
and varied both within a series and from series to series. The mean
duration of the pants 42 ± 9 ms ranged from 30 to 63 ms among
animals.
Snort. Low-amplitude harsh explosive exhalations, probable
produced through the nose. The tonal component was missing, so the
voice folds apparently were not involved (Figure 16). The maximum
amplitude frequency was 0.22 ± 0.05 kHz and did not exceed 0.37
kHz (Table 2). Usually, the explosive onset passed into the prolonged
exhalation, looking like a “cloud” of wideband noise on a spectrogram.
The mean duration of 77 ± 30 ms ranged from 38 to 157 ms among
animals (Table 2).
Cough. This short harsh explosive call was higher than the
snort in amplitude and, unlike the snort, was produced through the
open mouth (Figure 17). The maximum amplitude frequency was
0.85 ± 0.61 kHz, noticeable higher than that of the snort (Table 2).
A comparison of quartiles for the cough and snort showed also that
the cough’s energy was shifted to higher frequencies (Table 2). The
amplitude spacing over a call spectrum was crucial to distinguishing
between these two types of the unvoiced calls. The mean cough
duration 72 ± 19 ms varied from 32 to 123 ms among animals (Table
2). Short whines with the articulation effect rhythm were distinguished
from the coughs by the presence of residual fundamental frequency
(Figure 18). Transitional calls from snort to cough, with the snort-like
beginning passing into one or two coughs, intermediate in intensity
between snort and cough, were also registered (Figure 18).
119
Throughout our research, focal foxes never emitted toward the
researcher the whoop – the low-amplitude noisy call with unclear
pulsation (Figure 19). Unselected and Tame foxes produced these calls
probably toward neighbouring conspecics. We recorded a single case,
when an Unselected female fox produced a whoop toward a human,
but it was not a subject of this study.
Also, we registered a single case when an Unselected female
fox produced whistles toward a human, but this fox also was not
a subject of this study. Whistles were rhythmically organized high-
frequency calls (Figure 20).
Transitional calls
We recorded many transitional calls, both between the types and
between the classes. The most widespread were transitional calls
from moo to whine (Figure 9), from whine to growl (Figure 12), from
moo to growl and from snort to cough (Figure 18). The transitional
calls involving growl occurred even more often than “the clear” growl.
Rarely, we registered the transitional calls from bark to whine (Figure
13), from whine to cough, from whine to snort, from moo to cough and
from moo to snort.
In the following computations of calls belonging to different
call types within each recording all transitional calls with the growl
were considered as growls, transitional with the whine but without
the growl were considered as whines, transitional with the moo but
without growl or whine – as moos, and transitional from the snort
and cough – as coughs.
Classication of call types with DFA
The DFA conrmed the visual classication of call types. For the
DFA, we took from 1 to 3 calls per animal, which provided calls of
good quality for the given type. All whines, included in this analysis,
were free of any nonlinear phenomena or articulation effects.
For the voiced calls, the DFA forward stepwise procedure
included all 6 measured parameters. The mean value of correct
assignment to call type was 66.4%, that was signicantly higher (p
< 0.01) than the random value (mean 34.4 ± 1.6) calculated with the
randomization. The correct assignment value varied from 90.9% for
the bark to 19.6% for the growl (Table 3). Only for the growl the
correct assignment was lower than random, however, the growl was
distinctive from all other voiced calls by the presence of the pulsation.
The maximum fundamental frequency and call duration were the
main factors contributing to discrimination.
For the unvoiced calls, the DFA forward stepwise procedure
also included all 5 measured parameters. The mean value of correct
120
assignment to call type was 72.0%, signicantly higher (p < 0.01) than
the random value (mean 42.1 ± 1.6) calculated with randomization.
The correct assignment varied from 91.0% for the snort to 47.7% for
the pant, and exceeded the random value for all call types (Table 4).
The lower quartile of the power spectrum and call duration were the
main contributors to discrimination.
Comparison of vocal behaviour in Tame, Aggressive and
Unselected foxes
Both the numbers of callers producing certain call types and overall
occurrence of each call type within a selection group differed greatly
between the examined fox groups (Table 5, Figure 21).
Foxes from the different selection groups used distinctive call
sets. Only whine, moo and growl occurred in all the three groups. The
Unselected and Aggressive foxes never produced cackle or pant calls,
while the Tame foxes never produced cough or snort. The bark was
TABLE 3
Assignment of red fox voiced calls to a predicted call type, based on
discriminant analysis.
Actual group Predicted group membership Total Correctly
(Call type) classied (%)
Whine Moo Cackle Growl Bark
Whine 90 30 29 1 11 161 55.9
Moo 11 101 26 1 0 139 72.7
Cackle 4 23 134 0 0 161 83.2
Growl 0 37 0 9 0 46 19.6
Bark 1 0 0 0 10 11 90.9
Total 106 191 189 11 21 518 66.4
TABLE 4
Assignment of red fox unvoiced calls to a predicted call type, based on
discriminant analysis.
Actual group Predicted group membership Total Correctly
(Call type) classied (%)
Pant Snort Cough
Pant 42 10 36 88 47.7
Snort 2 121 10 133 100
Cough 18 23 92 133 69.2
Total 62 154 138 354 72.0
121
the rarest vocalization, recorded only from two Aggressive individuals
(Table 5). The Unselected and Aggressive foxes used the same call
type sets, whereas the Tame foxes used the perfectly distinctive call
type set (Figure 21).
Proportions of time spent vocalizing differed signicantly
between the selection groups (Figure 22). Unselected foxes showed the
lowest values, Tame foxes intermediate values, and Aggressive foxes
the highest values (all differences are signicant, White T-criterion,
p < 0.001). A one-way ANOVA (factor – selection group) also showed
that the proportions of time spent vocalizing differed signicantly
between the examined fox groups (F2,72 = 12.2, p < 0.001).
DISCUSSION
Vocal structures of fox-like canids: a comparative analysis
All the calls examined in this study were produced by captive adult
female red foxes toward an unfamiliar human. For this reason, we
cannot compare the production contexts for particular call types
recorded in our study with those reported in nature. In the present
study we classied the calls by their structures. The structural
classications for red fox calls were provided by three earlier studies
(Cohen & Fox 1976; Tembrock 1976; Newton-Fisher et al. 1993).
Some data were also provided by Movchan & Orlova (1990). Also, the
structural classications are available for related species, the swift
fox (Darden & Dabelsteen 2006) and the Arctic fox, both in nature
(Safronov et al. 1979), and in captivity (Ovsjanikov et al. 1988).
TABLE 5
Numbers of Unselected, Aggressive and Tame fox callers, provided calls of a given
type. % – percent of callers within a group.
Call type Selection group
Unselected Aggressive Tame
n % n % n %
Whine 12 48 23 92 24 96
Moo 24 96 22 88 3 12
Cackle 0 0 0 0 19 76
Growl 9 36 8 32 2 8
Bark 0 0 2 8 0 0
Pant 0 0 0 0 16 64
Snort 24 96 21 84 0 0
Cough 22 88 23 92 0 0
122
TABLE 6
Published data for vocalizations of red fox, swift fox and Arctic fox. Names are the terms given to a call type by the respective
author(s); the left column shows call type terms given for red fox in this study. Sources are given below. “–” means “no data”.
Call type Red fox Swift fox Arctic fox
Type Structure Type Structure Type Structure
Duration, ms F0, kHz Duration, ms F0, kHz Duration, ms F0, kHz
Whine Scream1 3000–4000 1.2–2 Whine5 150±140 0.87±0.18 Whine6 20–1000 0.4–0.6
Mew1 – – Whimper-whine5 410±110 1.03±0.11
Coo1 – – Full-whine5 240±80 0.96±0.14
Whimper2 – – Squabble5 300±160 –
Whine3 550 1 Yip-whine5 60±30 0.97±0.28
Whine4 650±350 1.06±0.15 Yip5 40±20 1.23±0.23
Meow5 670±200 1.03±0.04
Whine Whimpering– – – Scream5 420±90 – Scream6 730±280 1.45±0.33
with chaos crying1
Moo Hum5 310±140 0.14±0.06 Rumble6,7 – Low-frequency
Moan5 80±30 0.11
Cackle Chuckle5 50±10 0.41±0.10 Calm cackle6 57±2 0.5–1
Whine-chuckle5 50±20 0.54±0.09 Whining cackle6 76±4 1.6–1.8
Cackle7 62±15 –
Pant Panting1 – –
Panting2 – –
Growl Growl4 1170±210 – Growl5 410±220 – Growl6 240±10 0.4–0.6
Growl &
growl/scream1 – –
Growling2 – –
Snort Snorting2 – –
Cough Yelping2 – – Chitter5 70±30 –
Cough4 – –
Bark Bark1 <500 0–2 Bark5 310±70 – Single bark7 401±46 –
Bark2 – Not visible Yaps6 436±7 –
Bark4 830±360 –
Yell bark4 760±240 –
1Cohen & Fox (1976); 2Tembrock (1976); 3Movchan & Orlova (1990); 4Newton-Fisher et al.(1993); 5Darden & Dabelsteen (2006); 6Ovsjanikov
et al. (1988); 7Safronov et al. (1979).
123
TABLE 6
Published data for vocalizations of red fox, swift fox and Arctic fox. Names are the terms given to a call type by the respective
author(s); the left column shows call type terms given for red fox in this study. Sources are given below. “–” means “no data”.
Call type Red fox Swift fox Arctic fox
Type Structure Type Structure Type Structure
Duration, ms F0, kHz Duration, ms F0, kHz Duration, ms F0, kHz
Whine Scream1 3000–4000 1.2–2 Whine5 150±140 0.87±0.18 Whine6 20–1000 0.4–0.6
Mew1 – – Whimper-whine5 410±110 1.03±0.11
Coo1 – – Full-whine5 240±80 0.96±0.14
Whimper2 – – Squabble5 300±160 –
Whine3 550 1 Yip-whine5 60±30 0.97±0.28
Whine4 650±350 1.06±0.15 Yip5 40±20 1.23±0.23
Meow5 670±200 1.03±0.04
Whine Whimpering– – – Scream5 420±90 – Scream6 730±280 1.45±0.33
with chaos crying1
Moo Hum5 310±140 0.14±0.06 Rumble6,7 – Low-frequency
Moan5 80±30 0.11
Cackle Chuckle5 50±10 0.41±0.10 Calm cackle6 57±2 0.5–1
Whine-chuckle5 50±20 0.54±0.09 Whining cackle6 76±4 1.6–1.8
Cackle7 62±15 –
Pant Panting1 – –
Panting2 – –
Growl Growl4 1170±210 – Growl5 410±220 – Growl6 240±10 0.4–0.6
Growl &
growl/scream1 – –
Growling2 – –
Snort Snorting2 – –
Cough Yelping2 – – Chitter5 70±30 –
Cough4 – –
Bark Bark1 <500 0–2 Bark5 310±70 – Single bark7 401±46 –
Bark2 – Not visible Yaps6 436±7 –
Bark4 830±360 –
Yell bark4 760±240 –
1Cohen & Fox (1976); 2Tembrock (1976); 3Movchan & Orlova (1990); 4Newton-Fisher et al.(1993); 5Darden & Dabelsteen (2006); 6Ovsjanikov
et al. (1988); 7Safronov et al. (1979).
124
Specifying the boundaries to distinguish between different call
types represents the main problem for data comparison between
studies. The primary peculiarity of the present classication is the
involvement of new approaches for the description of calls, based on
concepts of nonlinear dynamics and sound production mechanisms
(Wilden et al. 1998; Fitch et al. 2002; Volodina et al. 2006a). In this
study, the boundaries for whine are especially wide. The whine is
a tonal vocalization that often involves nonlinear phenomena and
articulation effects that change the call sound drastically. Whereas
other authors usually consider each vocal variant as a separate call
type, in this study we united the extremely variable whine calls into a
single call type, and registered only the presence or absence of certain
nonlinear phenomena and/or articulation effects in them. Below, we
compare our call classication with those of previous authors (Table
6).
Whine. This call type has been described for the red fox (Cohen
& Fox 1976; Tembrock 1976; Newton-Fisher et al. 1993), Arctic fox
(Safronov et al. 1979; Ovsjanikov et al. 1988) and swift fox (Darden
& Dabelsteen 2006). The whine, in our terms, corresponds to the
whimper (“pure” whine) and whimpering-crying (whine with a segment
of deterministic chaos) described by Tembrock (1976, Figure 1 and 2
respectively).
According to spectrograms presented by Cohen & Fox (1976), the
whine in our terms corresponds to their whine (Figure 1a in Cohen &
Fox 1976), scream (Figure 1b), mew (Figure 1g) and coo (Figure 1e).
For the screams, Cohen & Fox (1976) report a duration of 3000–4000
ms and longer, which substantially exceeds the mean values, recorded
in the present study (Table 1), but still lies within the range limits
presented here for this call type. For the fundamental frequency,
Cohen & Fox (1976) report a range between 1.2–2 kHz, that is twice
as much as the values measured in this study (Table 1). The mixed
calls, reported by Cohen & Fox (1976), correspond to whines with
nonlinear phenomena and articulation effects in our terms. Among
them, the coo and coo-scream (Figure 1e, 2b in Cohen & Fox 1976),
characterized by “short vertical frequency changes”, correspond to the
whine with utter in our terms; yelp-growl-bark-growl (Figure 2c in
Cohen & Fox 1976) – to whine with rhythm; bark/howl (Figure 2d
in Cohen & Fox 1976) – to whine with bubble; growl/scream (Figure
2e,f in Cohen & Fox 1976) – to whine with deterministic chaos and
rhythm; complex long grunt (Figure 3a in Cohen & Fox 1976) – to
whine with subharmonics and deterministic chaos; bark-yelp (Figure
3c in Cohen & Fox 1976) – to the transitional call from bark to
whine.
Concerning short whines (mew), Cohen & Fox (1976, Figure 1g)
noticed that captive 5-week-old red foxes tend to produce them more
often in comparison with same-age domestic dogs. Also, they noticed
125
that the red foxes retain this call as part of their vocal repertoire
throughout the life, whereas in other canids it occurs only in pups.
In addition, Cohen & Fox (1976) report that the red foxes are the
only canids that produce pure screams in the context of greeting
conspecics. In this study, the whine was the most widespread call
type.
In the swift fox study (Darden & Dabelsteen 2006), such types
as whine, whimper-whine, full-whine, squabble, yip-whine, yip and
meow envelope a wide range of contexts and are a near match in
duration and fundamental frequency to the whine established in our
study (Table 1, 6). The scream, described by Darden & Dabelsteen
(2006) as a noisy vocalization, is emitted by swift foxes under extreme
anxiety and, judging by the presented measurements and spectrogram,
represents a very intensive, piercing vocalization. This call type is
consistent with calls of our whine type, in which the tonal structure
is masked with well-expressed deterministic chaos.
For whines of captive red foxes, Movchan & Orlova (1990) report
a duration of 550 ms and fundamental frequency of 1 kHz, consistent
with our data (Table 1). Whine in our study corresponds to a few
call types described by Newton-Fisher et al. (1993) for wild red foxes
(Table 6). The “pure” whines in our terminology correspond to whines
and whimpers of these authors, while our whines with deterministic
chaos relate to their screams.
Among all descriptions of fox-like canid whines, the values
reported by Ovsjanikov et al. (1988) for the Arctic fox whine are
closest to our data (Table 1, 6). Also, judging by the reported call
characteristics, the various high-frequency tonal calls of Arctic foxes
(chirp, scream, whine and yap, or single bark) are close to whine in
our terminology (Safronov et al. 1979; Ovsjanikov et al. 1988).
Moo. Newton-Fisher et al. (1993) do not describe any calm
close-range call types for the red fox. However, the low amplitude
hum and moan produced by swift fox in agonistic contexts (Darden &
Dabelsteen 2006), correspond well to the moo in our study.
The rumble of the Arctic fox, a low-frequency close-range tonal
call, is the closest in structure to the fox moo (Safronov et al. 1979;
Ovsjanikov et al. 1988). Arctic foxes produce this call when vigilant
under immediate danger and nearly exclusively from or near the
den.
Cackle. A single published report of the cackle in the red fox
is provided by Cohen & Fox (1976) when describing panting with a
tonal component. We did not nd however a published spectrogram
of this call. Newton-Fisher et al. (1993) and Tembrock (1976) did
not mention similar calls. The cackle in the present study coincides
with the chuckle and the whine-chuckle of the swift fox (Darden &
Dabelsteen 2006). Both the cackle and chuckle occur in non-agonistic
interactions between adults and pups or between pups. The cackle
126
was reported to be the most characteristic call for the Arctic fox;
the species produces cackle series during friendly contacts between
family group members (Safronov et al. 1979; Ovsjanikov et al. 1988).
Parameters for two cackle types of the Arctic fox are very similar to
our data for the red fox (Table 1, 6).
Pant. Pant is mentioned without spectrograms by Tembrock
(1976) and by Cohen & Fox (1976). The latter authors supposed that
pant represents an invitation to play in domestic dogs and red foxes
and proposed an interesting analogy between pant in the domestic
dog and laughing in humans, since a similar special facial expression
(semi-open mouth and concomitant panting) occurs during invitation
to play in both species. Concerning the red fox, Cohen & Fox (1976)
noticed that pant may be accompanied with mufed screams, mews
and purrs during greeting. However, Tembrock (1976) placed panting
into the same structural class as yelp and snort, and related these
calls to disturbance contexts. Our data agree much better with those
of Cohen & Fox (1976). Newton-Fisher et al. (1993) did not mention
panting in the acoustic repertoire of the red fox, probably because of
its low amplitude. Darden & Dabelsteen (2006) also did not report
this call in the swift fox. Consistently, the pant was not described for
the Arctic fox, in spite of its obvious relation to the cackle (Safronov
et al. 1979; Ovsjanikov et al. 1988).
Growl. This call type could not easily be related to other call
types reported for the red fox, since, in our observations, its structure
shows overlap with the low-frequency whines and moo. Moreover,
judging by published spectrograms, some of the earlier reported
“growls” are indeed whines with deterministic chaos or with rhythm
in our terms, since they are wideband and some of them contain
the retained fundamental frequency: for example, the growl and
growl/scream (Cohen & Fox 1976, Figure 1d, 2e, 2f); and growling
(Tembrock 1976, Figure 6). The growl parameter values reported by
Newton-Fisher et al. (1993) for wild red foxes are consistent with
our data. The growl parameters, reported by Ovsjanikov et al. (1988)
for the Arctic fox are shorter in duration and higher in fundamental
frequency (Table 1, 6). Darden & Dabelsteen (2006) placed the
growl of the swift fox among noisy vocalizations without the visible
fundamental frequency and harmonics. Spectrograms of the growl do
show visible pulsation (Darden & Dabelsteen 2006, Figure 1n, 2b).
Snort. Tembrock (1976) put snorting together with panting and
yelping into the same structural class of noisy calls with explosive
beginning and supposed that these calls are related to a disturbance
context. Newton-Fisher et al. (1993) and Darden & Dabelsteen (2006)
did not mention snort within their classications.
Cough. Unlike snort, cough is mentioned in all studies of red
fox vocalizations (Cohen & Fox 1976; Tembrock 1976; Newton-Fisher
et al. 1993). This call type is produced in short series in the context
127
of warning pups or other conspecics and in agonistic contexts.
Spectrograms provided by Tembrock (1976: yelping, Figure 7a) and
by Newton-Fisher et al. (1993: cough, Figure 12), are similar to the
cough spectrograms received in our study; however, the cited authors
did not give measurements for this vocalization. For the swift fox,
Darden & Dabelsteen (2006) described the noisy vocal type chitter,
whose duration and medium quartile (2.8 ± 0.47 kHz) coincides well
with the measurements of the red fox cough in the present study
(Table 2, 6). As with the red fox cough, the chitter of swift foxes
occurs in agonistic contexts. Neither snort nor coughs were described
for the Arctic fox.
Bark. This call type is reported by all authors for the red
fox, swift fox and Arctic fox (Tembrock 1976; Cohen & Fox 1976;
Newton-Fisher et al. 1993; Darden & Dabelsteen 2006). Judging by
the presented spectrograms and descriptions, however, the authors
attribute a variety of vocalizations to this type, including both tonal
sounds and those bearing nonlinear phenomena (subharmonics and
chaos). For example, the barks on the spectrograms presented by
Tembrock (1976: Figure 5) and by Cohen & Fox (1976: Figure 3d)
contain chaos, so the fundamental frequency is not visible (Table 6).
For the swift fox, the bark is described as a noisy call produced under
anxiety (Darden & Dabelsteen 2006).
Surprisingly, the mean duration of two single bark types (bark
and yell bark) reported for the red fox by Newton-Fisher et al. (1993)
seems unusually long compared to our study (Table 1, 6), even longer
than the whine duration. The bark spectrograms presented by Newton-
Fisher et al. (1993) also look like whines. Arctic fox single barks
(Safronov et al. 1979) and structurally related yaps (Ovsjanikov et al.
1988) look also closer to whines than to barks. The durations of single
barks in domestic dog are also much shorter than the single barks
of the red fox reported by Newton-Fisher et al. (1993). For example,
the mean bark durations in 10 domestic dogs (n = 4672 barks) varied
from 248 ± 27 to 346 ± 76 ms between situations (Yin & McCowan
2004). In another study, the mean bark duration in 24 domestic dogs
(n = 1268 barks) was 176 ± 31 ms (Chulkina et al. 2006), which is
closer to our results for the red fox.
Other calls reported for fox-like canids but not found in this
study. As we have mentioned above, focal foxes in our study never
produced serial barks toward the researcher. Serial bark, however,
could be regularly heard from foxes of all selection groups in other
contexts. Foxes likely produce serial bark toward conspecics or in
response to stimuli not related to the appearance of the researcher
near a cage. Serial bark is the prominent vocalization of red foxes,
swift foxes and Arctic foxes described by many authors (Safronov et
al. 1979; Ovsjanikov et al. 1988; Newton-Fisher et al. 1993; Darden
et al. 2003; Frommolt et al. 2003; Kruchenkova et al. 2003). For the
128
red fox, Newton-Fisher et al. (1993) described three different types of
serial bark (staccato barks, wow-wow barks, yodel barks), which are
slightly different in structure.
The noisy call type whoop, described for the swift fox (Darden
& Dabelsteen 2006), was repeatedly noticed in farm foxes of different
selection groups outside of our study; but only once toward the
researcher. Darden & Dabelsteen (2006) suggest that swift foxes
emit this vocalization under weak threat. The red foxes in our
study probably addressed whoops to neighbour conspecics, but not
immediately to the researcher.
Newton-Fisher et al. (1993) described two other red fox call
types not found in our study: yell whine and ratchet calls. The rst
one is a tonal call with very high maximum fundamental frequency,
up to 3 kHz. The second one represents the irregular sequence of
wideband calls, variable in duration. In addition, two call types, noisy
pulsed snarl and high-frequency noisy whistle, were described for
swift foxes (Darden & Dabelsteen 2006), but we registered only the
whistle, produced once by a female red fox in human-related context,
but outside our study.
Call types reported for other canids but not found in this study.
Similarly to Newton-Fisher et al. (1993), we did not nd the howl
in the red foxes. Probably, the occurrence of this vocalization is
restricted within the genus Canis, where it functions to maintain the
relations within and among packs (e.g., Lehner 1978; Schassburger
1987; Nikol'skii & Frommolt 1989).
Also, we did not nd in the red fox high-frequency squeaks
(higher than 4–5 kHz), occurring either singly as separate vocalizations
or together with low-frequency tonal components, resulting in calls
with two fundamental frequencies: biphonations or frequency jumps
between the higher and lower frequencies (Wilden et al. 1998). Calls
of this kind were described for the timber wolf (Schassburger 1987;
Nikol'skii & Frommolt 1989), domestic dog (Tembrock 1976; Volodina
et al. 2006a), dingo Canis dingo (Tembrock 1976), African wild dog
Lycaon pictus (Tembrock 1976; Wilden 1997; Wilden et al. 1998) and
dhole Cuon alpinus (Volodin & Volodina 2002; Volodina et al. 2006b).
In the domestic dog, both the high-frequency squeaks and biphonic
calls (whines) are very common toward humans (Volodina et al.
2006a), so we expected to nd them in the same context in the red
fox, especially in the Tame foxes. However, we did not nd any high
frequency squeaks in any of 12,964 calls from 75 foxes.
Effects of selection for tameness or aggressiveness toward
humans on fox vocalization
Surprisingly, within the study population we found vocalizations
specic for foxes selected for tameness, but did not nd any
129
vocalizations specic for foxes selected for aggressiveness. Supposing
the Unselected foxes to be the default state for selection in both
directions, we expected to nd in this control group all the range
of vocal structures. Instead, we found that the Unselected and
Aggressive foxes used just the same call type sets. By contrast, the
Tame foxes used a distinctive call type set, overlapping only in whine,
moo and growl with the other two groups. Therefore, the selection for
aggressive behaviour did not affect the fox vocal behaviour, whereas
the selection for tame behaviour resulted in drastic changing of the
call set, produced toward people.
Based on these data, we can speculate about vocal indicators
of tameness and aggressiveness. Since only Tame foxes produced the
cackle and pant, we can consider these types to be vocal indicators
of tameness. Similarly, common to Aggressive and Unselected
foxes, snort and cough may be considered as vocal indicators of
aggressiveness. The whine, and probably the moo, occurring in all
the selection groups, are not related to any selection for behaviour
and may express another attitude, probable frustration. We suppose
that more precise conclusions concerning emotional content of these
vocalizations could be made with further research with different
kinds of hybrids between the selection groups, differing in degree of
tameness and aggressiveness.
Both the Tame and the Aggressive foxes showed signicantly
higher rates of vocal activity in comparison with the Unselected control
group. These data support the Cohen and Fox (1976) hypothesis
that the lack of fear of humans relaxes the selection pressure for
silence. In wild canids, silence prevents the attraction of predators
and frightening the potential prey.
Domesticated foxes do not show hypertrophied barking, although
they have this call type in their vocal repertoires. Unlike dogs, foxes
contact with humans with the cackle and pant. The closely related red
fox, swift fox and Arctic fox use the cackle for communication with
their pair mates and pups. At the same time, domestic dogs use the
bark and whine for contact with humans (Yin 2002; Yin & McCowan
2004; Chulkina et al. 2006; Volodina et al. 2006a). Why these call
types were hypertrophied in dogs is not perfectly clear. Probably the
using of a certain call type for communication with humans depends
not only on domestication, but is species-specic.
ACKNOWLEDGEMENTS
We thank the staff of experimental fur farm of the Institute of
Cytology and Genetics RAS, Novosibirsk, Russia, for help and support
during work, and personally A.V. Kharlamova. We thank E.S.
Vorobjeva and R. Frey for help with literature, A.A Lisovsky for help
130
with statistic analysis and M.E. Goltsman for valuable discussion.
We are very grateful to the two anonymous referees for valuable
comments and correction of the text. During our work we adhered to
the “Guidelines for the treatment of animals in behavioural research
and teaching” (Anim. Behav., 2006, 71: 245–253) and to the laws of
Russian Federation, the country where the research was conducted.
This study was supported by the Russian Foundation for Basic
Research (grant 06-04-48400) the NIH (grants IR03 TW 007056-0IAI
and ROI MH 07781), and by the Programs of Basic Research of the
RAS Presidium “Biodiversity and gene pool dynamics” and “Molecular
and Cell Biology”.
REFERENCES
Beckers, G. J. L., Nelson, B. S. & Suthers, R. A. (2004). Vocal-tract ltering by lingual
articulation in a parrot. Current Biology, 14, 1592–1597.
Belyaev, D. K. (1979). Destabilizing selection as a factor in domestication. Journal of
Heredity, 70, 301–308.
Berry, D. A., Herzel, H., Titze, I. R. & Story, B. H. (1996). Bifurcations in excised
larynx experiments. Journal of Voice, 10, 129–138.
Budde, C. (1998). The vocal repertoire of the grey crowned crane Balearica regulorum
gibbericeps – a comparison between captive and free-living birds. Advances in
Ethology, 33, 122.
Chulkina, M. M., Volodin, I. A. & Volodina, E.V. (2006). Individual, intersexual
and interbreed variability of barks in dog Canis familiaris (Carnivora, Canidae).
Zoologicheskii Zhurnal, 85, 544–555. [In Russian].
Cohen, J. A. & Fox, M. W. (1976). Vocalizations in wild canids and possible effects of
domestication. Behavioural Processes, 1, 77–92.
Darden, S. K. & Dabelsteen, T. (2006). Ontogeny of swift fox Vulpes velox vocalizations:
production, usage and response. Behaviour, 143, 659–681
Darden, S. K., Dabelsteen, T. & Pedersen, S. B. (2003). A potential tool for swift fox
Vulpes velox conservation: individuality of long-range barking sequences. Journal
of Mammology, 84, 1417–1427.
Dillon, W. R. & Goldstein, M. (1984). Multivariate analysis: methods and applications.
Wiley: New York.
Fant, G. (1960). Theory of Speech Production. The Hague: Mouton.
Feddersen-Petersen, D. (2000). Vocalization of European wolves (Canis lupus lupus L.)
and various dog breeds (Canis lupus f. fam.). Archiv fur Tierzucht, 43, 387–397.
Fitch, W. T., Neubauer, J. & Herzel, H. (2002). Calls out of chaos: the adaptive
signicance of nonlinear phenomena in mammalian vocal production. Animal
Behaviour, 63, 407–418.
Frommolt, K-H., Goltsman, M. E. & Macdonald, D. W. (2003). Barking foxes, Alopex
lagopus: eld experiments in individual recognition in territorial mammal. Animal
Behaviour, 65, 509–518.
Gulevich, R. G., Oskina, I. N., Shikhevich, S. G., Fedorova, E. V. & Trut, L. N. (2004).
Effect of selection for behavior on pituitary-adrenal axis and proopiomelanocortin
gene expression in silver foxes (Vulpes vulpes). Physiology & Behavior, 82, 513–
518.
Kruchenkova, E. P., Goltsman, M. E. & Frommolt, K-H. (2003). Rhythmic organization
of Arctic foxes Alopex lagopus beringensis and A. l. semenovi: serial barking: sex,
age and context determinants. Zoologicheskii Zhurnal, 82, 525–533. [In Russian].
131
Kukekova, A. V., Trut, L. N., Oskina, I. N., Kharlamova, A. V., Shikhevich, S.
G., Kirkness, E. F., Aguirre, G. D. & Acland, G. M. (2004). A marker set for
construction of a genetic map of the silver fox (Vulpes vulpes). Journal of Heredity,
95, 185–194.
Leger, D. W. & Didrichsons, I. A. (1994). An assessment of data pooling and some
alternatives. Animal Behaviour, 48, 823–832.
Lehner, P. N. (1978). Coyote vocalizations: a lexicon and comparisons with other
canids. Animal Behaviour, 26, 712–722.
Monticelli, P. F. & Ades, C. (2001). Acoustic aspects of domestication: vocal signals
of alarm and courtship in wild and domestic cavies. Advances in Ethology, 36,
153.
Movchan, V. N. & Orlova, I. N. (1990). The reveal of informative features in rhythmically
organized sound signals of red fox (Vulpes vulpes). Zhurnal Evolutcionnoi Biohimii
i Phiziologii, 26, 398–404. [In Russian].
Newton-Fisher, N., Harris, S., White, P. & Jones, G. (1993). Structure and function of
red fox (Vulpes vulpes) vocalizations. Bioacoustics, 5, 1–31.
Nikol’skii, A. A. & Frommolt, K.-H. (1989). Vocal activity in the timber wolf. Moscow:
Moscow State Univ. Press. [In Russian].
Ovsjanikov, N. G., Rytovskaya, M. V., Menushina, I. E. & Neprintseva, E. S. (1988).
Social behaviour of Arctic foxes (Alopex lagopus): the vocal repertoire. Zoologicheskii
Zhurnal, 67, 1371–1380. [In Russian].
Riede, T., Bronson, E., Hatzikirou, H. & Ziberbuhler, K. (2005). Vocal production
mechanisms in a non-human primate: morphological data and a model. Journal of
Human Evolution, 48, 85–96.
Riede, T., Herzel, H., Mehwald, D., Seidner, W., Trumler, E., Tembrock, G. & Boehme,
G. (2000). Nonlinear phenomena in the natural howling of a dog-wolf mix. Journal
of the Acoustical Society of America, 108, 1435–1442.
Sablin, M. V. & Khlopachev, G. A. (2002). The earliest ice age dogs: evidence from
Eliseevichi I. Current Anthropology, 43, 795–799.
Safronov, O. V., Goltsman, M. E., Kruchenkova, E. P., Ovsjanikov, N. G. & Smirin
V. M. (1979). Acoustic repertoire of Arctic fox. In Ecological frameworks for
conservation, protection and rationale use of predatory mammals, pp. 160–161.
Moscow: MAIK Nauka. [In Russian].
Schassburger, R. M. (1987). Wolf vocalization. An integrated model of structure,
motivation and ontogeny. In Man and Wolf (Ed. by H. Frank), pp. 313–346.
Dordrecht: Dr W. Junk Publishers.
Shipley, C., Carterette, E. C. & Buchwald, J. S. (1991). The effect of articulation on
the acoustical structure of feline vocalization. Journal of the Acoustical Society of
America, 89, 902–909.
Solow, A. R. (1990). A randomization test for misclassication probability in
discriminant analysis. Ecology, 71, 2379–2382.
Tchernov, E. & Valla, F. F. (1997). Two new dogs, and other Natuan dogs, from the
southern Levant. Journal of Archaeological Science, 24, 65–95.
Tembrock, G. (1976). Canid vocalizations. Behavioural Processes, 1, 57–75.
Terry, A. M. R., Peake, T. M. & McGregor, P. K. (2005). The role of vocal individuality
in conservation. Frontiers in Zoology, 2, 1–16.
Titus, K., Mosher, J. A. & Williams, B. K. (1984) Chance-corrected classication for
use in discriminant analysis: ecological applications. American Midland Naturalist,
111, 1–7.
Trut, L. N. (1980). The genetics and phenogenetics of domestic behavior. Problems
in general genetics. In Problems of General Genetic, Proceedings of the XIV
International Congress of Genetics, vol. 2, book 2, pp. 123–136. Wallingford, UK;
NY.
Trut, L. N. (1999). Early canid domestication: the farm-fox experiment. American
Scientist, 87, 160–169.
132
Trut, L. N. (2001). Experimental studies of early canid domestication. In The Genetics
of the Dog, pp. 15–43. Wallingford, UK; NY: CABI Publishing.
Trut, L. N., Kharlamova, A. V., Kukekova, A. V., Acland, G. M., Carrier, D. R., Chase,
K. & Lark, K. G. (2006). Morphology and behavior: Are they coupled at the genome
level? In The Dog and its Genome (Eds. by E. A. Ostrander, U. Giger, K. Lindblad-
Toh), pp. 81–93. Woodbury NY: Cold Spring Harbor Laboratory Press.
Volodin, I. A. & Volodina, E. V. (2002). Biphonation as a prominent feature of the
dhole Cuon alpinus sounds. Bioacoustics, 13, 105–120.
Volodina, E. V., Volodin I. A. & Filatova, O. A. (2006a). The occurrence of nonlinear
vocal phenomena in frustration whines of the domestic dog (Canis familiaris).
Advances in bioacoustics II, Dissertationes SASA, Classis IV: Historia Naturalis,
47, 257–270.
Volodina, E. V., Volodin, I. A., Isaeva, I. V. & Unck, C. (2006b). Biphonation may
function to enhance individual recognition in the dhole, Cuon alpinus. Ethology,
112, 815–825.
Wilden, I. (1997). Phonetische Variabilität in der Lautgebung Afrikanischer Wildhunde
(Lycaon pictus) und deren frühe Ontogenese. Aachen: Snaker Verlag.
Wilden, I., Herzel, H., Peters, G. & Tembrock, G. (1998). Subharmonics, biphonation,
and deterministic chaos in mammal vocalization. Bioacoustics, 9, 171–196.
Yin, S. (2002). A new perspective on barking in dogs (Canis familiaris). Journal of
Comparative Psychology, 116, 189–193.
Yin, S. & McCowan, B. (2004). Barking in domestic dogs: context specicity and
individual identication. Animal Behaviour, 68, 343–355.
Received 11 May 2007, revised 15 November 2007 and accepted 4 December 2007
