orifices extending 10–13 cm (4–5 in.) above the nostril openings and the ability
to open or constrict the nostril openings is unknown.
The elephant is a very large mammal with a unique vocal tract in the form
of an elongated proboscis or trunk. The lung weight of an African elephant is
about 22 kg (48.4 lbs), the vocal folds measure around 7.5 cm (3 in.), and the
total length of the vocal tract measures between 3 and 4 m (9 and 13 ft). The
elephant can extend this vocal tract by as much as one-fifth (20%) and can
produce sounds as loud as 117 decibels (dB), where 120 dB is the level beyond
which human hearing might be damaged (Soltis, 2010).
SOUND DETECTION
The hearing mechanism of the elephant is also large. Each outer ear or pinna
measures on the order of 1.8 × 1.1 m (5.76 × 3.52 ft or more than 20 ft2) and is
highly mobile so that it can be extended outward and raised (together with the
60 Elephant Sense and Sensibility
FIGURE 9.4 Raised head with fully spread ears. (Pen and ink drawing by author.)
head), indicative of listening (Figure 9.4) (Moss, 1988; Poole et al., 1988, see Figure 18). The middle ear structures of the African elephant are also extremely
large compared with other mammals (Soltis, 2010). The area of the eardrum or
tympanic membrane is about 9 cm2 (1.4 in.2) compared to that of a human of
about 0.7 cm2 (0.07 in.2). The mass of the malleus, incus, and stapes (278, 237,
and 22.6 mg, respectively) are some 10 times that of the human equivalent. The
large sizes of the elephant’s auditory structures may well enhance their ability
to detect frequencies down to and below 10 Hz.
The eardrum vibrates in response to a sound pressure wave and the vibrations
are transferred by the three small bones known as ossicles. The size, shape, and
attachments of the ossicle vary greatly among mammals. In general, the larger the
mammal, such as the elephant, the larger the ossicles. The size of the ossicle is
inversely related to frequency. The ossicles serve to both amplify the sound and
transmit it to the oval window of the cochlea or inner ear. The precise relationship
between the structural features of the external and middle ears and their audi-
tory functions are not well understood in any mammal, let alone in the elephant.
Sensitivity to low frequencies is, however, thought to be associated both with large
tympanic membrane areas and large, compliant middle ear spaces. In Loxodonta,
the linear dimensions of the malleus in the middle ear are about twice those of the
incus. The long arm of the malleus (the manubrium) in Loxodonta is nearly verti-
cal, placing it perpendicular to the horizontal plane and possibly increasing the
sensitivity to low-frequency sounds (Rosowski, 1994, p. 173).
The cochlea has the characteristic spiral shape and is filled with fluid. The
cochlea in Loxodonta has two spiral turns and a total length of the basilar mem-
brane of 60 mm (von Békésy, 1960, pp. 506–509). von Békésy (1960) measured
the resonance of the cochlea dissected from an elephant, finding it responsive
Communication Chapter | 9 61
down to 30 Hz. von Békésy also concluded that of the animals studied, the
elephant cochlea exhibited the sharpest resonance, indicating that elephants may
be well equipped to distinguish between frequencies (Long, 1994, pp. 26–27).
SOUND LOCATION
The size of the elephant’s head and ears may govern how well it can locate
sounds, particularly using low frequencies at large distances. Heffner and
Heffner (1982, 1984) show that an Asian elephant can localize low-frequency
sounds (<1 kHz) to within an azimuth angle of 1°, which is about as well as
humans can. Langbauer et al. (1991), in a playback experiment, was able to
show that African elephants can locate the source of a low-frequency call at an
estimated range of 2 km (1.25 mile). Sounds are localized by using interaural
time differences (ITDs) and interaural level differences (Palmer, 2004). Low
frequencies work best in generating a time delay between the ears, whereas
highs and lows (peaks/troughs) in the sound record work best at high frequen-
cies. The ITD for humans is on the order of 0.7 s but is unknown and probably
much larger for an elephant based both on the size of the head and pinnae as
traveled by a wave front between the eardrums is larger than the skull perimeter
for low frequencies but equal to the skull perimeter for high frequencies (Kuhn,
1977, 1987). For humans, the effective acoustic circumference for low frequen-
cies is 150% that of high frequencies (Brown, 1994, pp. 64–69). The size and
position of an elephant’s ears and the presence of the trunk and forehead struc-
ture would suggest that this expansion in perimeter distance for low-frequency
signals might be considerably greater for elephants (Kuhn, 1977).
Whether and how animals determine range using infrasound is poorly un-
derstood. Elephant behavior with respect to low-frequency calls by familiar and
unfamiliar elephants indicates an acoustic perception of range (Moss, 1988;
Payne, 1998; Poole, 1996). McComb et al. (2003) used low-frequency playback
calls, which provide evidence of a sense of distance between the listener and
the sender of 1–2.5 km (0.6–1.5 mile). From personal observations in the field
(Etosha National Park, in 1999), the time from initial reaction of elephants at a
waterhole to the arrival (10 min) of a new group of elephants and the pace of the
incoming group (8.0–9.6 kmph) suggest that the range at which their presence is
first detected is between 1.5 and 2.5 km (0.9 and 1.5 mile). Distances in excess
of about 100 m (328 ft) eliminate sight (but not smell) from the potential cues
being used in determining these ranges.
However, it is important to distinguish between analog and digital characteris-
tics of calls. Information contained within the call in analog form depends on the
structure of the call. This structure, even in long-range, low-frequency calls, is at-
tenuated more rapidly than the call itself. When information is transmitted in digital
form, best visualized as Morse code, then the loss of signal (and meaning) is greatly
reduced compared to the analog form and one animal can hear and interpret the call
of another at much greater distances (see discussion below on estrous calls).
62 Elephant Sense and Sensibility
Only candidate theories exist that might explain how and how well elephants
determine the distance of the sound source from their location: there is clearly a
loss of higher frequencies with distance such that ultimately only low frequen-
cies remain. This change in the frequency spectrum may be used by elephants
to estimate range.
For hearing processes to be interpreted, we also need to know the sensitivity
of the elephant’s hearing at different frequencies. This quantity too is poorly
known. The only existing study (Heffner and Heffner, 1982, 1984) suggests that
the threshold of hearing at 17 Hz is 50 dB SPL for an Asian elephant.
The distance or maximum range over which one elephant can hear the call of
another is important for a number o
f reasons that influence the survival of the spe-
cies. In the matrilineal society of an elephant, where mature males are not in the
herd and may be many kilometers away, a means to reach these males is essential.
This need is heightened by the fact that the female may only come into heat for a
short period every 4–5 years. Although at this point the estrous cycle may last for
4 months, the estrous periods are short (2–6 days) (Leong et al., 2003). Only when
offspring are not successfully reared do females cycle over shorter intervals. Not
only must the female be able to let the male know that she is ready to mate, but
she must pass this message on to a number of males such that genetic selectivity
can be exercised (increasing the probability that the fittest and strongest male is
selected for mating). The need then for calls of females in estrous to be heard by
males over a long distance and thus over a large area may well be an added reason
why elephants have retained the ability to generate and hear low-frequency calls.
THE ROLE OF THE ATMOSPHERE
For this to happen, the atmospheric conditions that were pervasive in the forests
in which elephants evolved must be replicated over the dry subtropical savannas
of current-day elephants. Carbon 13 data show that forests (C-3 vegetation) con-
tracted progressively around 10 million years ago (MYA) being replaced by grass-
lands (C-4 vegetation). Early elephants changed fairly rapidly from feeding on
C-3 vegetation to feeding on C-4 vegetation about 8 MYA. This progression from
forest to grassland, with some woodland still present, resulted in mixed feeding
by elephants, which changed little over their subsequent history (Lister, 2013).
As outlined in the Introduction, the dry and often cloudless atmosphere of
the habitat of Loxodonta africana and Elphas maximus, the African and Indian subtropical savannas, go through a pronounced cycle of daytime heating and
nighttime cooling at the surface. When translated to heat being transferred from
the earth’s surface to the air immediately above the surface, the heat is lost from
the surface to the atmosphere before sunset until more than an hour after sunrise.
This results in a strong capping nocturnal inversion forming a channel or duct in
which low- (as well as other) frequency sound can be propagated. This layering
of the atmosphere (cold air at the surface about 100 m (328 ft) thick and warm
air above it), decouples the stronger winds higher up in the atmosphere from
the surface like a layer of oil sliding over water. Calm or low winds occur at the
Communication Chapter | 9 63
surface during the early part of the night, gradually increasing later in the night.
This increase in wind speeds as the night progresses is due to cold air drainage
over sloping terrain. By dawn the cold air has pooled and the winds at the surface
once again subside (Garstang and Fitzjarrald, 1999; Garstang et al., 2005).
Langbauer et al. (1991), in his playback experiment in Namibia, used a
sound pressure level (SPL) at 1 m (3 ft) of 112 dB and a threshold of hearing
of 46 dB, yielding a range of 1 km (0.6 mile). He then argued that because the
sound production of the loudspeaker was limited to half power, the actual range
of hearing was double the calculated value (i.e., 2 km (1.2 mile)).
If, however, the atmosphere is stratified and the sound signal ducted under
an inversion, spherical spreading does not occur and the sound can travel much
further before dropping below the threshold of hearing. Propagation of sound
in a stratified atmosphere can be determined by approximating a solution to a
complicated equation called the Helmholtz acoustic wave equation.
The Propagation of High- and Low-Frequency Sound
The pervasively cold floor of the forest compared to the warmer canopy results in
an inversion of temperature with denser air at and near the floor (lower tempera-
tures) and less dense air at and near the canopy (higher temperatures). Sound in a
fluid (air) travels faster at high temperatures and slower at low temperatures. The
result in the forest is to channel or duct the sound within the canopy.
High-frequency sounds (typically above 1000 Hz or cycles per second) behave
like a beam or ray of light. While such a beam of sound may propagate over long
distances in an unobstructed environment, the short wavelengths (a few centime-
ters or less) of high-frequency sound in a forest will be intercepted by the vegeta-
tion and will quickly dissipate.
Low-frequency sound (below 20 Hz) emanates from its source not as a beam
but as a uniformly propagating surface of a sphere. This too diminishes as a func-
tion of the distance from the source. However, it now decreases as the square of
the distance from the source such that it loses 6 dB for each doubling of the radius
of the sphere. If, however, the low-frequency sound is contained within a duct
(the inversion of the forest canopy), it will propagate down the duct, decreasing
as a function of the linear distance from the source and not as the square of the
distance from the source, traveling now like the high-frequency sound as a ray of
light. But with a huge difference: The low-frequency infrasound has a wavelength
of greater than 17.5 m or 40 ft compared to high-frequency sound with a wave-
length of a few centimeters or less than an inch. The long wavelengths of infra-
sound simply go “around” the trees and vegetation whereas short wavelengths of
high-frequency sound crash into the trees and vegetation and are dissipated.
As elephants evolved in the forest they took advantage of the “inversion or
ducting” conditions, developing the capability of generating and detecting low
frequency or infrasound. They subsequently emerged from the forests onto the
savannas with this infrasonic capability. The amazing fact is that these hot, dry
savannas, which during the day destroy almost all sound propagation, flip sides
before sunset until after dawn to create a pronounced duct or inversion at the
surface ideally suited to the propagation of infrasound.
64 Elephant Sense and Sensibility
This equation has no solution and must be approximated using computers to
successively arrive at an acceptable estimate of a solution. Because helicopter
rotors generate infrasound, they can be detected when flying below radar
(terrain) surveillance height, which simply means flying below the height of
the surrounding topography. The Department of Defense and the National
Acoustics Laboratory in Mississippi spent a large amount of time and money
developing the Fast Field Program (FFP) (Franke and Swanson, 1989; Lee
et al., 1986; Raspet et al., 1985) that would yield a timely and acceptable result.
These researchers, in collaboration with scientists at the University of Virginia
(Garstang et al., 1995; Larom et al., 1997), then applied the FFP to elephant
communication in the presence of these nocturnal inversions.
RANGE OF ELEPHANT CALLS
In the presence of a low-level (surface) inversion, one elephant can hear a loud,
low-frequency call of another elephant 10 km (6 mile) away. In the middle of the
day, however, this distance may be reduced to less than 1 km (0.6 mile). These
calc
ulations were made for calm wind conditions. Calm or low-surface wind
speeds typically occur under strong inversion conditions. Strong winds occur
around the middle of the day when heat-driven turbulent mixing is at a maxi-
mum (Figure 9.5). The effect of wind on the propagation and hearing of sound is complex. Turbulent motions in the atmosphere typically form vortices and rolls
in the atmosphere that are highly destructive to sound of almost all frequencies.
Standing on a bridge over a fast-flowing river, the eddies and whirlpools visible
50
40
30
% min
20
10
12
18
0
06
12
Hour of day (LST)
FIGURE 9.5 Average frequency (%) of wind speeds higher than 4 m s−1 (2.1 mph) (thick black lines) in each hour for the 17 days in eastern Etosha National Park, Namibia. The thin vertical lines delineate net daytime heating (08:00–18:00) from net nighttime cooling (18:00–08:00) at the surface. After Garstang et al. (2005).
Communication Chapter | 9 65
at the water’s surface, gives one a visual picture of what occurs in the atmo-
sphere. Turbulence is certainly the most destructive atmospheric effect on the
range of low-frequency sounds used by elephants. Turbulence near the surface is
closely related to wind speed, with higher turbulence occurring in higher winds.
Since friction constrains wind to be zero, at some point very close the surface,
wind speeds increase with height producing layers of faster-moving air over lay-
ers of slower-moving air. This change in wind speed with height produces a
shearing motion of one layer of air over another, enhancing mixing and turbu-
lence. Wind-driven refraction of sound shortens upwind and lengthens down-
wind communication. Finally, wind creates noise and in particular flow noise
across the elephant’s ears. Hunters and game-watchers know that animals tend to
seek shelter and lie low on windy days. This may be mainly due to the fact that
they have difficulty hearing and thus detecting danger (predators) on windy days.
The net result of the relationship between thermal stratification, wind, and
the propagation and detection of sound is that one can conclude that optimum
communication conditions occur when both thermal stratification and low