Vocal production mechanisms in a non-human
primate: morphological data and a model
Tobias Riedea,*, Ellen Bronsonb,1, Haralambos Hatzikiroua,
Klaus Zuberbu ¨ hlerc
aInstitut fu ¨r Theoretische Biologie, Humboldt-Universita ¨t zu Berlin, Invalidenstrasse 43, 10115 Berlin, Germany
bThe Baltimore Zoo, Druid Hill Park, Baltimore, MD, USA
cSchool of Psychology, University of St. Andrews, St Andrews, Scotland, UK
Received 3 July 2004; accepted 5 October 2004
Human beings are thought to be unique amongst the primates in their capacity to produce rapid changes in the
shape of their vocal tracts during speech production. Acoustically, vocal tracts act as resonance chambers, whose
geometry determines the position and bandwidth of the formants. Formants provide the acoustic basis for vowels,
which enable speakers to refer to external events and to produce other kinds of meaningful communication. Formant-
based referential communication is also present in non-human primates, most prominently in Diana monkey alarm
calls. Previous work has suggested that the acoustic structure of these calls is the product of a non-uniform vocal tract
capable of some degree of articulation. In this study we test this hypothesis by providing morphological measurements
of the vocal tract of three adult Diana monkeys, using both radiography and dissection. We use these data to generate
a vocal tract computational model capable of simulating the formant structures produced by wild individuals. The
model performed best when it combined a non-uniform vocal tract consisting of three different tubes with a number of
articulatory manoeuvres. We discuss the implications of these findings for evolutionary theories of human and non-
human vocal production.
? 2004 Elsevier Ltd. All rights reserved.
Keywords: Cercopithecinae; formants; vocal tract; motor pattern; Cercopithecus diana
* Corresponding author. Tel.: C49 30 20938652.
E-mail addresses: email@example.com (T. Riede), firstname.lastname@example.org (K. Zuberbu ¨ hler).
1Present Address: Smithsonian’s National Zoological Park, 3001 Connecticut Ave., NW, Washington D.C. 20008, USA.
0047-2484/$ - see front matter ? 2004 Elsevier Ltd. All rights reserved.
Journal of Human Evolution 48 (2005) 85e96
Adult male Diana monkeys (Cercopithecus
diana) utter predator-specific
response to two of their predators, the crowned
eagle (Stephanoaetus coronatus) and the leopard
(Panthera pardus), hereafter ‘eagle alarm calls’ and
‘leopard alarm calls’. Nearby monkeys respond
predator, suggesting that the calls contain informa-
tion about the external event, that is, the predator
type present (Zuberbu ¨ hler, 2000a, 2003). In pre-
vious work, we have documented that some
prominent acoustic features of Diana monkey
alarm calls are best conceptualised as formants,
the resonance frequencies of the vocal tract (Riede
and Zuberbu ¨ hler, 2003a,b). Formants are the
acoustic product of a series of bandpass filters that
shape the sound primarily produced by actions of
the vocal folds in the larynx and emitted from the
the laryngeal, nasal, and oral cavities. The cross-
sectional diameters and length of the vocal tract
determines the location of the formants’ acoustic
energy. During speech production in humans the
shape of the vocal tract is changing constantly and
rapidly, due to precise movements of the various
articulators, such as the lips, tongue, jaw, or larynx.
For many years, it has been the default
assumption that mammalian vocal tracts, includ-
ing those of non-human primates,
a uniform or flared tube during vocalization
(Lieberman, 1968; Lieberman et al., 1969; Shipley
et al., 1991). In a uniform cylindrical vocal tract
the resonance frequencies are expected to appear
as odd numbered multiples of the first resonance,
and all resonances are evenly spaced. The straight
line in Fig. 1 represents the different combinations
of the first and second formants, which would be
expected under the uniform tube assumption for
different vocal tract lengths and of a given di-
ameter. More recently, various studies challenged
this view, by suggesting that some animal vocal-
isations are the product of non-uniform vocal
tracts (e.g. Owren et al., 1997). For example, we
have previously demonstrated that the location of
the first (F1) and second (F2) formant in Diana
monkey alarm calls cannot be explained by
a uniform vocal tract but must be the result of
a more complex vocal tract geometry (Riede and
Zuberbu ¨ hler, 2003b, see dots in Fig. 1).
Second, in previous work we have shown that
Diana monkey leopard and eagle alarm calls differ
most prominently in the fine structure of the
formants albeit showing very little variability in
the fundamental frequency (Riede and Zuberbu ¨ h-
ler, 2003a,b). In particular, the first and second
formant (higher formants are only rarely detect-
able in the spectrum) of Diana monkey leopard
alarm calls exhibit a threefold stronger decrease in
frequency compared to the spectrographically less
modulated formants in the eagle alarm calls
(Fig. 2), suggesting that these animals are able to
adjust the shape of their vocal tracts during vocal
behaviour independent of the laryngeal source.
These observations have lead to the hypothesis
that Diana monkeys possess some control over the
shape of their vocal tract, and that they employ this
ability to communicate about some important
environmental events. In humans, acts of active
changes in vocal tract shape during vocalization
are termed articulation, the result of movement of
the tongue, the mandible, the lips, the larynx and so
on. For example, narrowing the lip aperture causes
a lowering of the formants (Stevens, 1999, p. 284),
an articulatory manoeuvre that can also be
observed in Diana monkeys, especially when pro-
ducing leopard alarm calls (Riede and Zuberbu ¨ h-
ler, 2003 a,b). Evidence for articulation caused by
mouth opening also exists from studies of the vocal
behaviour of rhesus monkeys (Macaca mulatta,
Hauser et al., 1993; Hauser and Scho ¨ n-Ybarra,
1994) and domestic cats (Felis catis, Shipley et al.,
1991). A second articulatory manoeuvre that could
explain the formant behaviour in Diana monkey
alarm calls is the lowering of the larynx (Story
et al., 1996; Fitch, 2000; Fitch and Reby, 2001).
Acoustically, lowering the larynx causes an equal
lowering of all formants due to vocal tract
elongation. As both articulatory manoeuvres have
previously been described in animal communica-
tion, it is likely that they are promising candidates
to explain the formant acoustics of Diana monkey
alarm calls as well. Yet, the relative amount of
variation explained by them and whether or not the
86T. Riede et al. / Journal of Human Evolution 48 (2005) 85e96
acoustic effects they generate are sufficient remains
unknown and needs to be investigated.
Here, we provide measurements of vocal tract
length and shape from three anaesthetised Diana
present results of the dissection of one male speci-
men. We then use these findings to generate the
most likely computational model of a Diana mon-
key vocal tract to test how much formant change
each of several changes in vocal tract geometry can
account for, and whether or not a single change is
sufficient for producing the formant characteristics
observed in the natural alarm calls.
Two adult males and one adult female Diana
monkey (C. diana diana) from the Baltimore Zoo
F2 Frequency (Hz)
F1 Frequency (Hz)
F1 Frequency (Hz)
F2 Frequency (Hz)
Fig. 1. First (F1) and second (F2) formant chart for human vowels and for animal vocalisations. Fig. 1a. Frequency values of F1 and F2
from 50 eagle alarm calls (black open circles; measured in 10 calls from each of 5 adult male Diana monkeys) and for 50 leopard alarm
calls (gray shaded area; measured in 10 calls from each of 5 adult male Diana monkeys). In each call F1 and F2 were measured at the
beginning of the call, in the middle and at the end. Formant data were taken from Riede and Zuberbu ¨ hler (2003b). The monkey data are
plotted within the American-English F1/F2 vowel space mapping (redrawn from data of Lee et al., 1999). Human formant data are for
adult males (dashed ellipses) and for 10-12 year old children (solid ellipses) (Table 2 and Fig. 4 in Lee et al., 1999). Human vowel data
were recorded from subjects saying the target words bead /IY/, bat /AE/, pot /AA/, ball /AO/, boot /UW/ (Lee et al., 1999). The straight
line represents the expected F1 and F2 values for uniform/cylindrical tubes of certain length, according to the equation Fn Z (2n ? 1)c/
4L ; (n Z n-th formant, c Z speed of sound; L Z length of the vocal tract). Fig. 1b. Same formant (F1, F2) data for Diana monkeys as
in Fig. 1a. The open ellipses represent ranges for formant data for male (B_m) and female (B_f) chacma baboon grunts (redrawn from
Owren et al., 1997; Rendall et al., 2004), and for human laughter (man: H_m, woman: H_f) (redrawn from Bachorowski and Owren,
2001), for rhesus monkey coo calls (R_c) and grunt calls (R_g) (estimated after data from Fitch 1997; Rendall et al., 1998) and for male
Red Deer roars (D) (estimated after data from Fitch and Reby, 2001). The straight line, as in Fig. 1a, represents the expected F1 and F2
values for uniform/cylindrical tubes of certain lengths. 4 lengths (8 cm, 20 cm, 30 cm and 80 cm) are indicated by open circles. Note that
axis scaling is different in the two diagrams. Formants in domestic dog growls (crosses) (data from Riede and Fitch, 1999) show
substantial variability because each data point represents a different breed which show high variability in head shape and thus vocal
tract length (from Yorkshire terrier with 8 cm vocal tract length to Rottweiler with approx. 23 cm vocal tract length).
87 T. Riede et al. / Journal of Human Evolution 48 (2005) 85e96
served as subjects. In November 2002, the animals
were anaesthetised with Ketamine (9.5 mg/kg) or a
combination of Tiletamine-Zolazepam (4.3-5 mg/
kg) via intramuscular injection as part of their
routine annual medical examination. Each animal
was then placed in lateral recumbence on a radio-
graphic table. Lateral images of the head-neck
region were taken, one with a relaxed head
position and another with a straightened head
position. A one-centimetre lead reference square
was positioned at the midsaggital level of the head
for calibration. The vocal tract length (VTL) was
determined from tracings of the X-ray images
using a Microtek MRS-600Z scanner and Scion
Image Beta 4.0 (www.scioncorp.com) for measur-
ing. Image clarity was sufficient to delineate the
outlines of the oral vocal tract. The midpoint of
the thyrohyoid bone, which was always visible,
was used as the point of origin for all vocal tract
length measurements. The thyrohyoid appeared
on radiographs just cranial to the glottis. Thus,
our estimate was slightly smaller than the actual
vocal tract length but in a manner consistent
across individuals. We then drew a curvilinear line
from the midpoint of the thyrohyoid cartilage
along the line of the soft and hard palates to the
front of the incisors (Fig. 3). We used this line as
an estimate of the length of the vocal tract using
the calibration squares as a reference. Skull length
was defined as the distance between the front of
the incisors and the external occipital protuber-
ance of the occipital bone. Studies in dogs have
shown that these measurements are reliably repli-
cable (Riede and Fitch, 1999). We estimated the
shape of the vocal tract by measuring the rostro-
ventral distance along the laryngeal, pharyngeal,
and oral cavities, from the glottis to the lips.
Portions of a carcass from an adult male Diana
monkey were obtained from the Baltimore Zoo.
Time (100 ms / unit)
Fig. 2. Spectrogram and time series of a leopard and an eagle alarm call uttered by a male Diana monkey. Note that the first (F1) and
the second (F2) formant behave differently in the two calls. There is a downward modulation at the beginning of the leopard call but
not in the eagle alarm call.
88T. Riede et al. / Journal of Human Evolution 48 (2005) 85e96
The specimen was collected in August 2003 and was
deep-frozen within 36 hours post-mortem. Un-
fortunately, the thorax had been severely damaged
during pathological examination, thus the dimen-
sion of the airsac cavity could not be completely
investigated. The specimen was then cut in the
median plane and each half was photographed.
Modelling the vocal tract
A popular vocal tract model used for simulation
of vowel sounds is the ‘‘Wave-Reflection Ana-
logue’’ (or ‘‘Wave Digital Filter’’) (Smith, 1992,
1998). The model divides the three-dimensional
vocal tract into a finite number of cylinders of cer-
tain length. Reflection coefficients are calculated at
each cylinder junction, based on the relative areas
of adjoining sections. In this model, waves pro-
pagate through the system as a function of the
reflection coefficients, which determine the incident
and reflected components of the pressure waves at
each junction at each step in time. This has the
effect of transporting a wavefront from section to
section. We chose the commonly used practice of
setting the reflection coefficients at the sound
source and at the mouth opening at 1 (e.g. Olesen,
1995, p. 29). The Wave-Reflection-Analogue is an
attractive method for acoustic modelling of the
vocal tract because computations are performed
serially in time-synchrony with the acoustic wave
propagation (Story et al., 1996). Computations are
efficient because the equations describing the wave
propagation take on a digital filter structure in
their final form.
Radiography and dissection
Vocal tract length
The oral vocal tract lengths in the three
radiographed individuals were 9.7 and 10.5 cm
(males A and B) and 8.9 cm (female). The oral
vocal tract length in the dissected individual (male
B) was 10.1 cm (Table 1).
Vocal tract shape
During vocalization, the animals’ head is lifted
so the mouth points forward. We observed that in
this head position, the epiglottis was touching the
soft palate. The vocal tract had a non-uniform
shape caused by a major constriction separating
the front oral cavity from the laryngo-pharyngeal
Fig. 3. Schematic drawing of the head-neck region of a Diana monkey with details as seen in the dissection as well as in the lateral
x-ray. T - tongue, Tr - trachea, uL - upper lip, lL - lower lip, L - larynx, P - palate, dashed line 1 - oral vocal tract length, dashed
line 2 - nasal vocal tract length, arrows indicate the dorso-ventral distances of the oral vocal tract.
89T. Riede et al. / Journal of Human Evolution 48 (2005) 85e96
cavity. The constriction which is built by the
tongue and the palate, extends over 2 to 4 cm.
Fig. 4 summarises radiograph measurements of
rostroventral distances in the vocal tract as
a function of the distance from the glottis.
The length of the vocal folds in the freshly
dissected specimen was 11 mm on both sides.
Cercopithecines possess airsacs branching off from
the larynx, located subcutaneously in the throat
and ventral thorax area, and these structures are
very likely to affect the acoustic output (Gautier,
1971). In our specimen the air sac projected
medially between the thyroid cartilage and the epi-
glottis toward the subcutaneous area of the neck.
However, because of severe damage to the thorax
of the specimen we were not able to reconstruct the
dimensions of the air sac.
Modelling the vocal tract
Our goal was to simulate the formant behaviour
as observed in natural alarm calls, particularly to
account for the formant transitions that character-
ise the leopard alarm calls. Is a single vocal tract
adjustment, such as the movement of larynx,
mouth, or jaw, sufficient to achieve the naturally
observed formant behaviour?
Vocal tract non-uniformity
Formants in Diana monkey alarm calls are the
likely product of a non-uniform vocal tract (Riede
and Zuberbu ¨ hler, 2003a,b), in the simplest case
a two-tube model. A two-tube vocal tract can
produce vowels with a low second formant, like
the /a/ vowel, as demonstrated for instance by
Stevens (1972). The lowest F2 values are achieved
if both segments are of the same length. However,
in our case a two-tube model, created by a narrow
laryngo-pharyngeal tube attached to a wider oral
tube, was unable to generate the formant patterns
in the natural Diana monkey alarm calls, mainly
due to the insufficient lowering of the second
formant (see calculations in the next paragraph).
Interspersing a third narrow connecting tube
between the laryngo-pharyngeal back tube and
the frontal oral tube is known to lower the second
formant dramatically while the first formant is
mildly increased (e.g. Stevens, 1972). Finally,
adding a small fourth tube will allow for separate
lip aperture and variable mouth opening, hereby
accommodating natural observations.
The formant values for (a) natural alarm calls
and (b) different tube dimensions of the model are
given in Table 2. Our anatomical results indicated
a VTL in adult Diana monkeys of about 10 cm.
However, in a uniform cylindrical vocal tract the
first formant would be expected to be much higher
than measured in Diana monkey alarm calls, while
the second formant would be much lower (Fig. 1).
A 10 cm uniform 1-tube model (1 cm diameter)
Distance from the glottis
(% of total vocal tract length)
Fig. 4. Vocal tract shape measured as dorso-ventral distance of
the oral vocal tract (as indicated in Fig. 3) in a distance from the
glottis, measured in 3 monkeys, 2 males and 1 female.
Morphological data of study animals
name/sex body mass
oral vocal tract
nasal vocal tract
oral vocal tract
length (cm) - dissected
90T. Riede et al. / Journal of Human Evolution 48 (2005) 85e96
resulted in the following formants: F1Z754 Hz,
F2Z2264 Hz (Table 2, #1), that is, a 23% lower
F1, and a 61% higher F2 compared to the mea-
surements at the beginning of natural leopard
alarm calls (Table 2). A 10 cm non-uniform 2-tube
model generates the following formants: F1Z
1391 Hz, F2Z2110 Hz (Table 2, #2), that is now
a 48% higher F1 and still a 50% higher F2
compared to the measurements at the beginning of
natural leopard alarm calls. A 10 cm non-uniform
3-tube model generates substantially smaller devia-
tions from the expected values (F1Z949 Hz,
F2Z1489 Hz), although F2 still deviates by
approximately 6% (Table 2, #3). A 10 cm non-
uniform 3-tube model with a very short 4thtube,
achieved by cutting off a short segment from tube
C, to simulate the lip aperture, lead to formant
values that resembled the natural conditions most
closely at the beginning of the leopard call
(F1Z945 Hz, F2Z1403 Hz, Table 2, #4).
Our anatomical and observational data com-
bined with the physics of tube resonance acoustics
therefore suggest a 3-tube model with a short
fourth tube attached to the end as the best model
to explain Diana monkey vocal behaviour (Fig. 5).
Leopard alarm calls are the likely product of
articulation, as evidenced by the two formants
undergoing a dramatic downward modulation
during the first half of a call (Fig. 2) (Riede and
Zuberbu ¨ hler, 2003b). In our model, closing the
mouth corresponded with a lowering of the tube D
diameter, leading to a decrease in the first two
formant frequencies. However, with this manoeu-
vre the second formant decreased more strongly
than the first formant, unlike in the natural call
(Table 2, #5). This suggested that Diana monkeys
engaged in additional articulatory manoeuvres
while producing leopard alarm calls. To accom-
modate for these findings we added the following
two articulatory manoeuvres to the model which
brought its formant values to a very good match
with those of the natural calls: (a) elongation of
tube A by about 4% (corresponding to a laryngeal
lowering of about 4 mm; see Table 2, #6), (b)
narrowing of tube C by about 17% (corresponding
to a rising of the mandible by about 4 mm, leading
to narrowing of the frontal cavity accompanying
mouth closing; see Table 2, #7). Finally, in leopard
alarm calls the first and second formants increase
slightly from the middle of the call to the end of
the call (Table 2). Our model could simulate this
effect by slight adjustments in the mouth opening
and the larynx position (Table 2, #8).
Formant values in natural calls and in the computational modal
(Leopard alarm calls)
beginning of call
middle of call
end of call
# tube dimensions
(diameter/length in cm)
1 A 1.0/10.0 - B 1.0/0.0 - C 1.0/0.0 - D 1.0/0.0
2 A 1.0/5.0 - B 1.0/0.0 - C 3.0/5.0 - D 1.0/0.0
3 A 1.0/0.88 - B 0.3/3.0 - C 2.4/6.0 - D 2.4/0.0
4 A 1.0/0.88 - B 0.3/3.0 - C 2.4/5.75 - D 1.5/0.25
5 A 1.0/0.88 - B 0.3/3.0 - C 2.4/5.75 - D 0.71/0.25 950 1390
6 A 1.0/1.27 - B 0.3/3.0 - C 2.4/5.75 - D 0.71/0.25 805 1370
7 A 1.0/1.27 - B 0.3/3.0 - C 2.0/5.75 - D 0.71/0.25 782 1194
8 A 1.0/1.23 - B 0.3/3.0 - C 2.0/5.75 - D 0.76/0.25 795 1227
Formant values for natural leopard alarm calls were determined
from monkey vocalisations recorded in the Taı¨ forest, Ivory
Coast (see Riede and Zuberbu ¨ hler, 2003b). Formant calcula-
tions are based on different tube model dimensions as well as
mouth openings. tube dimensions: #1 - uniform tube 1 cm
diameter and 10 cm length; #2 to #5 - multi-tube-approxima-
tions (tubes A to D see Fig. 5) with respective dimensions
(diameter/length in cm); #6 to #8 - articulatory manoeuvres.
Fig. 5. Schematic drawing of the 3-tube-approximation of the
monkey’s vocal tract indicating the lengths (1A, 1B, 1C, 1D) and
diameter dimensions (dA, dB, dC, dD) used for the calculation.
Tube D represents the mouth opening, and is either as wide as
tube C in diameter (dCZdD) or narrows down (dC!dD) in
order to simulate the closing of the lip aperture.
91T. Riede et al. / Journal of Human Evolution 48 (2005) 85e96
In this study we were interested in the vocal
tract structure and motor patterns underlying
alarm call production in wild Diana monkeys.
We have collected a set of anatomical data, which
allowed us to estimate the length and shape of the
vocal tract in this species with satisfactory
accuracy. We used these measurements in combi-
nation with a number of theoretical considerations
to construct a simple computational model of the
Diana monkey vocal tract that simulated the
natural calls remarkably well. The model consisted
of a three-tube non-uniform vocal tract, with an
additional short forth tube, which could engage in
three types of simultaneous articulatory manoeu-
vres to change the overall shape of the vocal tract:
mouth closing, jaw movement, larynx movement.
Vocal tract shapes change considerably during
all sorts of activities, such as food intake (see for
instance Lieberman and McCarthy, 1999) and still
X-ray images from dead or anaesthetized animals
may not necessarily capture the entire range of
vocal tract mobility during vocalization. To
address questions of motor patterns underlying
vocal behaviour, the ideal approach is to observe
the movement of the articulators in the animal’s
vocal tract during vocalisation, using cine-radio-
graphic or related techniques (e.g. Fitch, 2000;
Riede et al., 2004). For a number of reasons this
approach was unsuitable here. First, there are no
other species known which produce these acoustic
features and could serve as a comparison. Second,
Diana monkeys are a highly endangered species,
which excludes any kind of risky or invasive
research. Computational modelling thus may be
the best approach to identify likely mechanisms of
vocal production and articulation.
Vocal tract length
Earlier interpretations of the formant patterns
in Diana monkeys relied on simple skull measure-
ments (Riede and Zuberbu ¨ hler, 2003b). In this
study, we were able to provide more accurate data
using radiography and dissection. The vocal tract
length in 3 anaesthetised and 1 dead Diana
monkeys ranged between 9 and 11 cm. The true
vocal tract length during vocalisation might be
slightly different due to protrusion of the lips or
temporary up-and-down-movements of the larynx.
Laryngeal lowering during vocalisation has been
reported in piglets (Sus scrofa), domestic dogs
(Canis familiaris), goats (Capra hircus), cotton-top
tamarins (Saguinus oedipus) (Fitch, 2000), and red
deer (Cervus elaphus; Fitch and Reby, 2001). We
therefore concluded that changes in vocal tract
length, caused by mild larynx lowering, are
a mechanism likely to take place during Diana
monkey alarm calling.
Vocal tract shape
Non-uniformity is an important characteristic
of the human vocal tract (e.g. Story et al., 1996). It
has been a tacit assumption that non-human
primate vocalisations are the result of a uniform
tube-like or a flared tube-like vocal tract, resulting
in F1 and F2 to be situated along the straight line
in Fig. 1. Our data from Diana monkeys show that
this is not true for all non-human primates. Our
anatomical data indicate that the larger frontal
oral cavity and the smaller laryngo-pharyngeal
cavity are separated by a constriction, the Isthmus
faucium, which is built by the tongue and the
palate. This kind of non-uniformity has been
reported in a baboon (Zhinkin, 1963, p. 162), in
cotton-top tamarins (Fig. 2 in Fitch, 2000), and it
is further supported by formant data in domestic
dogs (Riede and Fitch, 1999; Fig. 1b this study).
Non-uniformity of this kind is not reported for the
vocal tracts of cats and pigs (cats: Shipley et al.,
1991; pigs: Fitch, 2000).
In 3-month old infants, the larynx and the root
of the tongue begin to descend into the pharyngeal
cavity (Laitman et al., 1977; George, 1978). As
a consequence, the tongue becomes increasingly
mobile and serves as the anterior wall of the
pharynx, an arrangement that permits prominent
alternations in the cross-sectional area and shape
of the vocal tract. In older infants and adult
humans, temporary deviations from uniformity
are normal during vowel production (Fig. 1b). The
92T. Riede et al. / Journal of Human Evolution 48 (2005) 85e96
formant characteristics of the vowels are the result
of unique shapes of the vocal tract (Titze, 1994;
Story et al., 1996). In non-human primates, the
larynx is normally positioned higher in the
pharynx (Keleman, 1948, 1969; Laitman et al.,
1977) although a laryngeal descence during early
ontogeny has been described in chimpanzees
(Nishimura et al., 2003). This position can be
altered by specific muscles (Negus, 1949), but non-
human primates are thought to make only re-
stricted use to change the shape of their vocal
tracts during vocalization
McCarthy, 1999), similar to humans when pro-
ducing a schwa vowel (Lieberman, 1968). Here we
demonstrated that the vocal tract in Diana
monkeys during leopard alarm call production is
likely to undergo a complex motor pattern not
described for a non-human primate before.
Modelling the Diana monkey vocal tract
As such, the model was able to explain two
acoustic ’anomalies’ in the Diana monkey alarm
calls, that is, the close proximity of the first and
second formant in both alarm call types (explained
by a 3-tube approach) and the prominent formant
downward modulation in the leopard alarm calls
(explained by the simultaneous movement of at
least 3 articulators).
It was assumed earlier that the non-human
mammalian vocal tract cannot create vowels as /u/
or /o/ because the production of these vowels
requires an extreme constriction in the vocal tract.
As mentioned above, studies in other non-human
primates and the present study suggest that this
constriction is not uniquely human. Moreover,
a modelling approach of the Neandertal vocal tract
suggests that this earlyhominidwasequallycapable
constriction (Boe et al., 2002), despite the fact that
their vocal tract were characterised by a higher
laryngeal position than in the modern humans.
Our model showed that one single change of
vocal tract shape alone could not account for the
formant lowering in leopard alarm calls. It was the
combination of an initially non-uniform vocal
tract and the several simultaneous articulatory
manoeuvres that provided the best match between
simulated and real data. Direct observations of
vocalising Diana monkeys suggested that males
not only narrow the lip aperture but also raise the
mandible. Our model incorporated this fact by
simulating changing the lip aperture and mandible
movement (i.e. narrowing diameter of tubes C and
D; Fig. 5). It also addressed the effect of larynx
descending and its effect on lowering the first and
Modelling approaches rely heavily on anatom-
ical information in order to produce meaningful
output. This is because the same or similar
formant patterns can be obtained with a range of
different vocal tract configurations (see for exam-
ple discussion in Espy-Wilson et al., 2000). In this
study, we only incorporated assumptions that were
based on empirical evidence provided by our
anatomical findings or on facts provided by earlier
Three important questions were not addressed
in this study but await further investigation: (a) the
possibility of nasalization and (b) the acoustic role
of air sacs. The potential acoustic effects have been
discussed elsewhere (Riede and Zuberbu ¨ hler,
2003b). (C) The possibility of tongue movement
has been demonstrated during swallowing (Hiie-
mae et al., 1995) but is unknown to take place
during vocalization. Before incorporating these
structures into models further anatomical inves-
tigations will be necessary.
At least three implications for theories of speech
evolution emerge from this study. First, a recent
study on Neanderthal vocal tracts suggests that
a low larynx alone is not a sufficient anatomical
prerequisite for producing the full range of vowels
(Boe et al., 2002). Our findings on Diana monkey
vocal behaviour corroborate this conclusion.
Second, articulatory effects within a single utter-
ance are very important in human speech,
generated by the rhythmic mandible-generated
open-close alternation of the mouth (MacNeilage
and Davis, 2000). Modern languages share certain
patterns of consonant-vowel co-occurrences, the
result of biomechanical constraints associated
with mandibular movements (MacNeilage, 1998;
93 T. Riede et al. / Journal of Human Evolution 48 (2005) 85e96
MacNeilage and Davis, 2000). Our data show that
these patterns are also present in non-human pri-
mate vocal production. Diana monkeys appear to
engage in the same rhythmic mandible-generated
a leopard alarm call bout. Our modelling data
suggest that articulation in the leopard alarm calls
takes place by narrowing the lip aperture accom-
panied with raising the mandible. Since leopard
alarm calls are uttered in bouts, the basic pattern
would be raising (during vocalisation) and lower-
ing (between calls) of the mandible. Third, two
utterances in humans have been considered most
similar to non-human vocalisation, because of
their non-verbal characteristics, laughing and
infant crying. Laughing is acoustically a highly
variable utterance (Bachorowski and Owren,
2001) without consonants, which demonstrated
a persistent lack of articulation effects in supra-
laryngeal filtering (Bachorowski and Owren, 2001;
Owren and Bachorowski, 2003). Infant crying is
acoustically very variable, but the acoustic vari-
ability can be ascribed to source characteristics
(Zeskind and Collins, 1987; Mende et al., 1990),
suggesting that articulation plays no major role
(Lieberman et al., 1971). Diana monkey leopard
alarm calls do not resemble either of these human
utterances but overlap substantially with the /a/
vowel and the /o/ vowel F1/F2-range of a 10 to 12
years old child with a similar vocal tract length
In most other non-human species, however,
formant frequencies do not vary appreciably within
single calls (e.g. Chacma baboons Papio hama-
drayas ursinus: Owren et al., 1997). However,
articulatory effects within a single utterance have
been described for domestic cats and for red deer.
In domestic cats, jaw movement appear to be the
most powerful articulator in ‘meow’ vocalisations,
responsible for the formant change within a single
call (Shipley et al., 1991). In red deer, formant
variability has been ascribed to extreme lowering of
the larynx during a single call (Fitch and Reby,
2001). In this respect Diana monkeys are a first
example of animal vocalisations where acoustic
and anatomical data suggest non-uniformity of the
vocal tract. Some investigators have argued that
ofthe mouth during
manipulate the position of the lips (Hauser, 1992;
Hauser et al., 1993; Hauser and Scho ¨ n-Ybarra,
1994), larynx (Fitch, 2000; Fitch and Reby, 2001)
and tongue (Hiiemae et al., 1995), thus changing
the length, shape, and thus presumably resonances
of the vocal tract. However, available formant data
in other species suggest more or less uniform vocal
tract conditions (Fig. 1b). In sum, our study shows
that a number of vocal tract adaptations important
in human speech production are also present in
a non-human primate. In Diana monkeys, these
are a non-uniform vocal tract and three different
articulation mechanisms caused by movement
of the mandible, the lips, and the larynx. More-
over, Diana monkeys utilise these mechanisms
to produce two types of vocalisations, which
function to communicate important events in the
One remaining question about the Diana
monkey alarm call system is whether formant
variability is perceptually relevant for conspecific
recipients. So far, playback experiments have
shown that leopard and eagle alarm calls are dis-
criminated by conspecifics (Zuberbu ¨ hler, 2000a),
by other monkeys such as Campbell’s monkeys
(Cercopithecus campbelli) (Zuberbu ¨ hler, 2000b),
and even by birds (Rainey et al., 2004a,b). It
remains unknown which acoustic parameters
recipients attend to when making these discrim-
inations, although formants are very likely candi-
dates (Hienz and Brady, 1988; Owren, 1990a,b;
Sommers et al., 1992).
We thank the Baltimore Zoo Chimp Forest and
hospital staff for their support in collecting data
and performing examinations on the three captive
animals. The investigation of three Diana mon-
keys was approved by the Baltimore Zoo In-
stitutional Animal Care and Use Committee. We
also thank Jack Bradbury, Cornell University, for
technical support and Sue Anne Zollinger, Indiana
University, for help in the preparation of the
figures. T. Riede was supported by a fellowship
German Academic Exchange Service (DAAD).
94T. Riede et al. / Journal of Human Evolution 48 (2005) 85e96
We are most grateful for the comments of Michael
Owren, Brad Story and two reviewers.
Bachorowski, J.A., Owren, M.J., 2001. Not all laughs are alike:
voiced but not unvoiced laughter elicits positive affect in
listeners. Psychol. Sci. 12, 252e257.
Boe, J.-L., Heim, J.-L., Honda, K., Maeda, S., 2002. The
potential Neandertal vowel space was as large as that of
modern humans. J. Phonetics 30, 465e484.
Espy-Wilson, C.Y., Boyce, S.E., Jackson, M., Alwan, A., 2000.
Acoustic modeling of American English /r/. J. Acoust. Soc.
Am. 108, 343e356.
Fitch, W.T., 1997. Vocal tract length and formant frequency
dispersion correlate with body size in rhesus macaques.
J. Acoust. Soc. Am. 102, 1213e1222.
Fitch, W.T., 2000. The phonetic potential of nonhuman vocal
tracts: comparative cineradiographic observations of vocal-
izing animals. Phonetica 57, 205e218.
Fitch, T.W., Reby, D., 2001. The descended larynx is not
uniquely human. Proc. R. Soc. Lond. B 268, 1669e1675.
Gautier, J.P., 1971. Etude morphologique et fonctionnelle des
annexes extra-larynge ´ es des cercopithecinae; liason avec les
cris d’espacement. Biol. Gabon. 7, 229e267.
George, S.L., 1978. A longitudinal and cross-sectional analysis
of the growth of the postnatal cranial base angle. Am. J.
Phys. Anthropol. 49, 171e178.
Hauser, M., 1992. Articulatory and social factors influence
the acoustic structure of rhesus monkey vocalizations:
a learned mode of production? J. Acoust. Soc. Am. 91,
Hauser, M.D., Evans, C.S., Marler, P., 1993. The role of
articulation in the production of rhesus monkey, Macaca
mulatta, vocalizations. Anim. Behav. 45, 423e433.
Hauser, M.D., Scho ¨ n-Ybarra, M., 1994. The role of lip
configuration in monkey vocalization: experiments using
xylocaine as a nerve block. Brain Lang. 46, 232e244.
Hienz, R.D., Brady, J.V., 1988. The acquisition of vowel
discrimination by nonhuman primates. J. Acoust. Soc. Am.
Hiiemae, K.M., Hayenga, S.M., Reese, A., 1995. Patterns
of tongue and jaw movement in a cinefluorographic
study of feeding in the macaque. Arch. Oral. Biol. 40,
Keleman, G., 1948. The anatomical basis of phonation in the
chimpanzee. J. Morphol. 82, 229e257.
Keleman, G., 1969. Anatomy of the larynx and the anatomical
basis of vocal performance. Chimpanzee 1, 165e186.
Laitman, J.T., Crelin, E.S., Conlogue, J., 1977. The function of
the epiglottis in monkey and man. Yale J. Biol. Med. 50,
Lee, S., Potamianos, A., Narayanan, S., 1999. Acoustics of
children’s speech. Developmental changes of temporal and
spectral parameters. J. Acoust. Soc. Am. 105, 1455e1468.
Lieberman, P., 1968. Primate vocalization and human linguistic
ability. J. Acoust. Soc. Am. 44, 1574e1584.
Lieberman, P., Klatt, D.H., Wilson, W.A., 1969. Vocal tract
limitations on the vocal repertoires of rhesus monkey and
other non-human primates. Science 164, 1185e1187.
Lieberman, P., Harris, K.S., Wolff, P., Russell, L.H., 1971.
Newborn infant cry and non-human primate vocalizations.
J. Speech Hear. Res. 14, 718e727.
Lieberman, D.E., McCarthy, R.C., 1999. The ontogeny of
cranial base angulation in humans and chimpanzees and its
implications for reconstructing pharyngeal dimensions.
J. Hum. Evol. 36, 487e517.
MacNeilage, P.F., 1998. The frame/content theory of evolution
of speech production. Behav. Brain Sci. 21, 499e511.
MacNeilage, P.F., Davis, B.L., 2000. On the origin of internal
structure of word forms. Science 288, 527e531.
Mende, C., Herzel, H., Wermke, K., 1990. Bifurcations and
chaos in newborn infant cries. Phys. Lett. 145 (A), 418e424.
Negus, C., 1949. The comparative anatomy and physiology of
the larynx. Grune and Stratton, New York.
Nishimura, T., Mikami, N., Suzuki, J., Matsuzawa, T., 2003.
Descent of the larynx in chimpanzee infants. Proc. Natl
Acad. Sci. 100, 6930e6933.
Olesen, M., 1995. A speech production model including the nasal
cavity. Aalborg (DK).
Owren, M.J., 1990a. Acoustic classification of alarm calls by
vervet monkeys (Cercopithecus aethiops) and humans (Homo
sapiens). I. Natural calls. J. Comp. Psychol. 104, 20e28.
Owren, M.J., 1990b. Acoustic classification of alarm calls by
vervet monkeys (Cercopithecus aethiops) and humans
(Homo sapiens). II. Synthetic calls. J. Comp. Psychol. 104,
Owren, M.J., Seyfarth, R.M., Cheney, D.L., 1997. The acoustic
features of vowel-like grunt vocalization in chacma baboons
(Papio cyncephalus ursinus). Implications for production
processes and function.
Owren, M.J., Bachorowski, J.A., 2003. Reconsidering the
evolution of nonlinguistic communication: the case of
laughter. J. Nonverbal Behav. 27, 183e200.
Rainey, H.J., Zuberbu ¨ hler, K., Slater, P.J.B., 2004a. Hornbills
can distinguish between primate alarm calls. Proc. R. Soc.
Lond. B 271, 755e759.
Rainey, H.J., Zuberbu ¨ hler, K., Slater, P.J.B., 2004b. The
responses of black-casqued hornbills to predator vocal-
isationsand primate alarm
tract filtering in identity cueing in rhesus monkey (Macaca
mulatta) vocalizations. J. Acoust. Soc. Am. 103, 602e614.
Rendall, D., Owren, M.J., Weerts, E., Hienz, R.D., 2004. Sex
differences in the acoustic structure of vowel-like grunt
vocalization of baboons and their perceptual discrimination
by baboon listeners. J. Acoust. Soc. Am. 115, 411e421.
Riede, T., Fitch, T., 1999. Vocal tract length and acoustics of
vocalization in the domestic dog (Canis familiaris). J. Exp.
Biol. 202, 2859e2867.
95T. Riede et al. / Journal of Human Evolution 48 (2005) 85e96
Riede, T., Zuberbu ¨ hler, K., 2003a. Pulse register phonation in Download full-text
Diana monkey alarm calls. J. Acoust. Soc. Am. 113,
Riede, T., Zuberbu ¨ hler, K., 2003b. The relationship between
acoustic structure and semantic information in Diana
monkey alarm vocalization. J. Acoust. Soc. Am. 114,
Riede, T., Beckers, G., Blevins, W., Suthers, R., 2004. Inflation
of the esophagus and vocal tract filtering in Ring doves.
J. Exp. Biol 207, 4025e4036.
Shipley, C., Carterette, E.C., Buchwald, J.S., 1991. The effect of
articulation on the acoustical structure of feline vocaliza-
tion. J. Acoust. Soc. Am. 89, 902e909.
Smith, J.O., 1992. Physical modeling using digital waveguides.
Comput. Music J. 16, 73e87.
Smith, J.O., 1998. Principles of digital waveguide models. In:
Kahrs, M., Brandenburg, K. (Eds.), Applications of Digital
Publisher, Boston, Massachusetts, USA, pp. 417e466.
Sommers, M.S., Moody, D.B., Prosen, C.A., Stebbins, W.C.,
1992. Formant frequency discrimination by Japanese mac-
aques (Macaca fuscata). J. Acoust. Soc. Am. 91, 3499e3510.
Stevens, K.N., 1972. Quantal nature of speech. In: David Jr.,
E.E., Denes, P.B. (Eds.), Human Communication: A
Unified View. McGraw-Hill, New York.
Stevens, K.N., 1999. Acoustic Phonetics. MIT Press, Cam-
Story, B., Titze, I., Wong, D., 1996. A Simplified Model for
Simulation and Transformation of Speech. In: Proceedings
of IEEE International Joint Symposia on Intelligence and
Systems, Rockville, MD, pp. 320e327.
Titze, I.R., 1994. Principles of Voice Production. Prentice-Hall,
Englewood Cliffs, New Jersey.
Zeskind, P.S., Collins, V., 1987. Pitch of infant crying and
caregiver responses in a natural setting. Infant Behav. Dev.
Zhinkin, N.I., 1963. An application of the theory of
algorithms to the study of animal-speech-methods of
vocal intercommunication between monkeys. In: Busnel,
R.G. (Ed.), Acoustic Behavior of Animals. Elsevier,
Zuberbu ¨ hler, K., 2000a. Referential labeling in Diana monkeys.
Animal Behaviour 59, 917e927.
Zuberbu ¨ hler, K., 2000b. Interspecific semantic communica-
tion in two forest primates. Proc. R. Soc. Lond. B 267,
Zuberbu ¨ hler, K., 2003. The effects of natural and sexual
selection on the evolution of guenon loud calls. In: Glenn,
M., Cords, M. (Eds.), The Guenons. Plenum Press,
96 T. Riede et al. / Journal of Human Evolution 48 (2005) 85e96