Content uploaded by Joyce H Poole
Author content
All content in this area was uploaded by Joyce H Poole on Jul 04, 2014
Content may be subject to copyright.
• Proposes several new call types.
• Describes a broad range of behavioral contexts and
examines, qualitatively and quantitatively, the acoustic
signals with which they are associated.
• Makes a fi rst attempt to discriminate between proposed
contextual call sub- types based on a number of acoustic
measurements.
• Uses these and contextual differences to illustrate how
vocal signals form an essential and integral component
of the complex dynamics of elephant society.
I begin by describing what is known about elephant
acoustic communication and vocal repertoire, in general,
and then examine in detail how elephants in Amboseli use
acoustic signals in the context of what we know about their
sociality.
Elephant Sound Production and Detection
African elephants produce a broad range of sounds from
very low frequency rumbles to higher frequency trumpets,
snorts, barks, roars, grunts, and cries (e.g., Berg 1983;
Poole et al. 1988; Poole 1994a; Langbauer 2000; Leong
et al. 2003; Poole and Granli 2004; Stoeger- Horwath et al.
2007) as well as a range of idiosyncratic, novel, and imi-
tated sounds (Poole et al. 2005).
Two features make elephant vocal production unusual:
their very large size and the inclusion of the trunk in the vo-
cal tract (Soltis 2009). Elephants are capable of producing
powerful, very low frequency sounds in part due to their
large body size and correspondingly large vocal organs. The
The fission- fusion society of elephants is built
upon a complex network of social relationships
within and between families, bond groups, and clans
(see chapters 13 through 16) and between individual males
(see chapters 17 and 18). Added to this multi- layered social
network are fl eeting interactions and temporary consortships
that form between reproductively active males and cycling
females. This elaborate system of associations, partnerships,
coalitions, and enduring relationships is in part established,
mediated, and maintained via an intricate suite of acoustic
signals (see also chapter 10). The survival of females and their
offspring depends upon the cohesion and coordination of the
family and upon their ability to compete with other groups
for access to scarce resources (McComb et al. [2001], and see
also Mutinda, Poole, and Moss, chapter 16). Consequently,
most calls produced by female and juvenile elephants give
emphasis to the importance of the social unit. Family mem-
bers call to reinforce bonds between relatives and associates,
to care for calves, to reconcile differences between “friends,”
to defend close associates, to form coalitions against aggres-
sors and predators, to coordinate movements, and to keep
in contact with one another over long distances (Poole et al.
[1988]; Poole [1994a]). Adult male elephants lead relatively
more independent lives (but see chapter 17) than do females,
where reproductive success and survival depend upon an in-
dividual’s ability to detect sounds made by others; males tend
to use calls to advertize their sexual state, identity, and rank
(Poole 1999a; chapter 18).
In this chapter, I build upon an established body of work
on elephant communication to examine the calls emitted by
African elephants, Loxodonta africana, from the standpoint
of social function. This chapter does the following:
Chapter 9 Behavioral Contexts of Elephant
Acoustic Communication
Joyce H. Poole
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
126 Joyce H. Poole
Elephant Acoustic Communication Background
Soltis (2009) provides a detailed review of the literature on
African elephant vocal communication; therefore, I provide
only a summary here.
Berg (1983) made the fi rst descriptions of African ele-
phant call types in captivity and showed that elephants
produced more high frequency calls (e.g., trumpets) during
periods of high excitement. Soon thereafter, Payne, Lang-
bauer, and Thomas (1986), also working in a captive set-
ting, discovered that the low frequency rumbling sounds,
or rumbles, made by Asian elephants (Elephas maximus)
contain frequencies below the level of human hearing, or
infrasound. Payne, Langbauer, and Thomas described fun-
damental frequencies of 14 to 24 Hz, with some calls pro-
duced at very high sound pressure levels. Since very low fre-
quency sound attenuates more slowly than higher frequency
sound (Dneprovskaya, Iofe, and Levitas 1963; Eyring 1946;
Ingard 1953; Wiley and Richards 1978), Payne, Langbauer,
and Thomas proposed that elephants might use very low
frequency sounds at high sound pressure levels to commu-
nicate with one another over long distances. Their discov-
ery prompted a fl urry of studies on infrasound in free- range
conditions.
In Amboseli, Poole et al. (1988) established that the
rumbles of African savannah elephants, Loxodonta afri-
cana, also extend into the infrasonic range and documented
frequencies as low as 14 Hz and sound pressure levels of
up to 103 dB (re 20 µPa extrapolated to 5 m from source),
although later, some of these calls have been shown to be
as low as 8 to 9 Hz (this chapter). Subsequently, Payne,
Thompson, and Kramer (2003) established that African
forest elephants, L. cyclotis, also produce low frequency
rumbles with infrasonic components as low as 5 Hz.
Poole et al. (1988) also described several well- known
and commonly heard rumble sub- types (e.g., greeting-
rumbles, contact- calls and answers, “let’s- go”- rumbles,
musth- rumbles, female-rumbles, chorus-rumbles, post-
copulatory- rumbles, and the mating- pandemonium). Using
specifi c examples in which vocal exchanges between affi li-
ated females or mating pairs resulted in specifi c changes in
behavior, they argued that long- term records on the behav-
ior of elephants and the contexts of these specifi c rumble
sub- types showed that elephants make use of low frequency
rumbles in the spatial coordination of groups and as they
search for mates (see also Poole and Moss [1989]). Much
later, Soltis, Leong, and Savage (2005a) documented that
vocal exchanges or “anti- phonal calling” such as that ob-
served among wild elephants (e.g., Moss 1981; Poole et al.
1988; Poole and Moss 1989; Poole 1999a; McComb et al.
2000) also occurs among females in a captive setting and
that, as we have argued, these constitute true communica-
tion events (Leighty et al. 2008).
majority of elephant sounds are laryngeal in origin, pro-
duced by air passing from their large lungs over the vocal
chords or larynx, a structure reportedly 7.5 cm long, sus-
pended on the hyoid apparatus (Sikes 1971). The elephant’s
vocal tract includes the extent of the trunk, the honeycomb
nasal passages in the skull, and the length from the lips to
the larynx and has been estimated to be 2.5 m (Soltis 2009).
This extended resonator is also unusual in that most of it
(some 1.8 m of trunk) lies outside the cranium, thus pro-
viding great fl exibility in both length and shape and affect-
ing the sounds produced accordingly. Two more anatomical
features may infl uence sound production by affecting the
size and shape of the elephant’s vocal tract. First, the struc-
tures of the hyoid apparatus (a series of bones at the base
of the tongue) and the musculature that support the tongue
and the larynx in elephants are different from other mam-
mals. The hyoid apparatus of elephants has fi ve rather than
nine bones, and these are attached to the skull by muscles,
tendons, and ligaments, rather than by bones, as in most
other mammals. This rather loose arrangement allows for
a greater movement and fl exibility of the larynx (Shoshani
1998). Second, this looser arrangement also houses a pha-
ryngeal pouch, a structure unique to elephants located at
the base of the tongue, which in addition to providing an
emergency source of water, appears to function in the pro-
duction of low frequency calls (Shoshani 1998).
Elephants also produce trumpets, snorts, and some idio-
syncratic sounds by blowing air through the trunk. By alter-
ing the positioning of the trunk and the speed and duration
of air moving through it, elephants can produce an aston-
ishingly versatile mixture of sounds via this “second voice,”
as Soltis (2009) refers to it.
Research by Heffner and Heffner (1980, 1982) demon-
strated that elephants have very good low frequency hear-
ing (frequencies below 17 Hz not tested) and very accu-
rate localization skills. Together with Sirenians, they are
unique among modern mammals in having reverted to a
reptilian- like cochlear structure (Fischer 1990) that may fa-
cilitate greater sensitivity to lower frequencies (O’Connell,
Hart, and Arnason 1998). Furthermore, elephants are able
to detect seismic vibrations, or Rayleigh waves, through
two possible means: (1) Bone conduction and the use of
massive ossicles of their middle ears (Reuter, Nummela,
and Hemila 1998); and (2) mechanoreceptors in the toes or
feet that are sensitive to vibrations (O’Connell, Hart, and
Arnason 1998). The tip of an elephant’s trunk has densely
packed layers of cells called Pacinian corpuscles that are
extremely sensitive to vibrations (Rasmussen and Munger
1996). O’Connell- Rodwell et al. ([2006] and personal com-
munication) and Bouley et al. (2007) suggest a high density
of these cells in the front of the foot and along the edges,
consistent with the notion that elephants are sensing seismic
signals when they suddenly freeze.
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
Behavioral Contexts of Elephant Acoustic Communication 127
As noted by Soltis (2009), call types can be separated
into sub- types from a structural or functional standpoint,
and the relationship between the two can take many forms.
Following on from Poole et al. (1988), Poole (1994a) pro-
posed a long list of rumble and other call sub- types based
on behavioral context and sound quality. Yet, working in
captivity and based on acoustic structure, attempts by Le-
ong et al. (2003) to subdivide rumbles into fi ve sub- types
was largely unsuccessful, and they and later Soltis, Leong,
and Savage (2005b) concluded that rumbles are best viewed
as a single call type with graded variation. Also working in
captivity, Stoeger- Horwath et al. (2007) drew similar con-
clusions for calf rumbles. Using cluster analysis on calls re-
corded in the wild, however, Wood et al. (2005) were able
to divide rumbles into four sub- types. Other researchers
have separated rumbles into a priori categories and then,
using multivariate statistical analysis, have established dif-
ferences in acoustic structure across the previously defi ned
groups. In this manner, researchers have shown differ-
ences in rumble structure between individual callers (wild:
McComb et al. [2003]; captive: Soltis et al. [2005b]) and
between broad social contexts in which calls were produced
(wild: Wood et al. [2005]; captive: Soltis, Leong, and Sav-
age [2005b]; Soltis et al. [2009]). Poole and Granli (2004),
too, were able to document structural differences in trumpet
sub- types produced during specifi c contexts associated with
exuberant play behavior. While these studies demonstrated
statistically different means across call sub- type categories,
each of the studies showed overlap in acoustic structure,
which is consistent with graded rather than discrete struc-
tural sub- types.
While such overlap in acoustic structure exists, play-
back experiments using rumbles and trumpets recorded
under specifi c contexts show that elephants are able to dis-
cern subtle differences between these call sub- types and re-
spond appropriately. For instance, the playback results of
Poole (1999a) show that elephants distinguish between two
rumble sub- types (post- copulatory or estrus- rumbles and
musth- rumbles), and these responses are different, again,
from those of elephants to playbacks of contacts- calls
(Poole unpublished, McComb et al. [2001]). Playbacks
also show that elephants can discriminate between different
trumpets (Poole et al., in preparation).
More recently, researchers have started to look into the
effect that context has on rumble structural forms. Soltis,
Leong, and Savage (2005b); Soltis et al. (2009); and Li
et al. (2007) provided evidence that rumbles produced
by lower- ranking individuals during agonistic interac-
tions with those of higher rank show increased and more
variable fundamental frequencies and amplitudes and in-
creased call durations compared with the rumbles they pro-
duced in periods of social tranquility. Similarly, Wood et al.
(2005) documented that rumbles associated with “social
Since the early work of Payne, Langbauer, and Thomas
(1986) and Poole et al. (1988), theoretical and experimental
studies have confi rmed that elephants are able to broadcast
and detect acoustic signals over several kilometers (Larom
et al. 1997; Langbauer et al. 1991; Garstang et al. 1995),
and playback experiments have shown that they respond
appropriately to specifi c calls (Langbauer et al. 1991; Poole
1999a; McComb et al. 2000, 2001, 2003). Moreover, Mc-
Comb et al. (2003) and chapter 10 (McComb, Reby, and
Moss) showed that elephants are capable of discriminat-
ing between individual voices at distances of up 2 km. The
range of acoustic communication may more than double
at night, due to temperature inversions that typically form
before sunset and decay with sunrise and could travel up to
10 km (Garstang et al. [1995], though even under the best
conditions the distance is likely to be signifi cantly less for
individual recognition). Furthermore, O’Connell- Rodwell,
Arnason, and Hart (1997) discovered that when an ele-
phant emits a low frequency rumble, a corresponding seis-
mic wave with similar characteristics is transmitted in the
ground (O’Connell- Rodwell et al. [2006] were able to show
via playback experiments that elephants are able to detect
this seismic component and to discern subtle social cues
from this information [O’Connell- Rodwell et al. 2007]).
Recently, there has been increasing focus on the vo-
cal repertoire of the African elephant, producing a grow-
ing body of literature on the acoustic structure of different
call types and sub- types, and attempts to describe the social
signals contained therein. Expanding on Berg’s early work,
Leong et al. (2003), Stoeger- Horwath et al. (2007), and
this chapter have described a variety of different call types.
These are described in detail below and may be listened to
online at www.elephantvoices.org / multimedia- resources /
elephantvoices- calls- database- call- types.html.
Adding to the typical call types of African elephants,
Poole et al. (2005) discovered vocal imitation in elephants,
documenting the fi rst case in a non- primate terrestrial mam-
mal, and strengthening the hypothesis that a primary se-
lection pressure for vocal learning involves the communi-
cative demands of maintaining social relationships in fl uid
societies. Poole et al. suggested that the evolution of vocal
imitation in humans, some birds, bats, dolphins, and ele-
phants might result from the need to use acoustic signals
to maintain individual- specifi c bonds when animals sepa-
rate and reunite. This hypothesis predicted vocal learning in
other species where long- lived social bonds are based upon
individual- specifi c relationships, with fl uid group member-
ship, and where vocal communication is used for main-
taining contact and individual or group recognition. The
capacity for vocal imitation is a further indication of the
elephant’s highly fl exible vocal tract and suggests that we
should expect them to exhibit a wide range in sound pro-
duction.
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
128 Joyce H. Poole
formant dispersion and increased amplitude and duration
when they were interacting with one another. Furthermore,
Stoeger- Horwath et al. (2007) and Wesolek et al. (2009)
showed that the rumbles of infants who were denied ac-
cess to the breast were associated with more energy in the
interactions and agitation” were characterized by increased
and more variable fundamental frequencies as well as de-
creased duration. Soltis et al. (2009) also showed that high-
ranking opponents produced rumbles with features that sig-
nal large body size: decreased fundamental frequency and
BOX 9.1 METHODOLOGY
Equipment and Field Techniques
Between 1984 and 1990 and from 1998 through 2006, I re-
corded elephant vocalizations in Amboseli. During 1998, I also
recorded from elephants in Laikipia and Maasai Mara, Ke-
nya, and from semi- captive orphan elephants in Tsavo, Kenya.
I made recordings through 1990 on a Nagra IVSJ (see Poole
1999a); between 1998 and 2003, at 44.1 kHz on an HHB PDR
1000 DAT recorder (frequency response: 8 Hz: –0.43dB; 12Hz:
–0.26 dB; 15 Hz: –0.22 dB; 20 Hz: –0.15 dB; 60–1,000Hz no
roll off); after 2003, on a modifi ed Nagra Ares BB (frequency re-
sponse: 10 Hz: –1 dB; 20 Hz: –0.4 dB, 50 Hz: –0.2 dB, 100 Hz:
–0.1 dB; 200–20,000 Hz no roll off). The majority of recordings
were made with an Earthworks QTC1 omni- directional micro-
phone (frequency response: 4Hz–40kHz ±1dB).
Field data included the general area and specifi c loca-
tion, date, tape and track number, track start and end times,
elapsed time, channel settings, group size and type, and the
individuals present. When a call was heard, the elapsed time
was noted as well as information on the call type, caller, and
distance to the source. I use the term call type to refer to the
broad structurally differentiated categories of sounds and the
term context- type to refer to a priori subtypes initially differen-
tiated from a combination of sound quality and social context.
Call types are in italics (e.g., rumble, roar, trumpet, etc.), while
specifi c call context- types are italicized and hyphenated (e.g.,
begging- rumble, musth- rumble, low- intensity- husky- cry).
Caller, call type, and context- type were recorded with level
of confi dence (A: certain, B: fairly confi dent, C: educated
guess, D: no idea). Also noted was whether the decision re-
garding call context- type was based primarily on the quality of
the sound or on its behavioral context. A call context- type as-
signed confi dence “A” required that both the behavioral con-
text and sound quality matched the context- type designation.
Finally, any contextual or other comments about the situation
or the behavior of the calling animal were noted.
After January 1999, I focused recording sessions on mem-
bers of the EB family (n = ca. 27). Once I located the family I
parked near a sub- group that provided good visibility (5–20
m to the nearest elephant). As I began recording, I noted the
nearest adult female and her nearest adult female neighbor.
Once the elephants moved greater than about 25 m away, I
moved the vehicle again. Though many elephant calls persist
over long distances, the best quality calls and fi eld data are
typically from the closest individuals; accordingly, I varied the
nearest elephant.
Elephant ages were categorized in two ways: by absolute
age (years) and by age class (0A, 0B, 1A, 1B, 2, 3, 4, and 5; see
General Methods). The terms infant (<6 months), calf (0–4.9
years), juvenile (5–9.9 years), and adult (>10 years) were used.
Sound Acquisition and Measurement
Sounds from 1998 to 2006 were systematically logged from
fi eld notes into a custom- designed MS Access Database with
a Visual Basic Interface (n = 6,592). Using the sound anal-
ysis program SIGNAL RTSD, calls were acquired through a low
pass anti- alias fi lter onto a Gateway 2000 and saved to disk as
.wav fi les (n = 3,934). Low frequency calls were acquired at a
sample rate of 2,000 Hz and displayed at a range of 0–500Hz,
while higher frequency calls were acquired at 22,050 Hz and
displayed at range 0–5,000 Hz. Due to the high frequency
of overlapping calls, measurement was not automated. Time-
frequency measures were taken in Signal 4.0 from the spec-
trograph view using the cursor, and additional measurements
were taken using the selection function in Raven 1.0 (see table
9.1 for a description of measurements).
Statistical Analysis
Of the measured calls, 70 percent (confi dence A, n = 2,014)
were recorded from 59 elephants whose identities were
known. Of known callers, some contribute as few as one call
to the data set, while all except one adult female member
of the EB family each contribute over 30 calls. Nevertheless,
once call context- types and level of confi dence are considered,
sample sizes per individual become considerably smaller. Many
call types (e.g., roars) and some context- types (e.g., calls di-
rected at predators) are only rarely produced. In addition, the
use of particular context- types is unevenly distributed across
individuals by age, sex, reproductive state, and even by person-
ality. For example, confi dent elephants are more likely to emit
“let’s go”- rumbles (see below), male calves are more likely to
protest when denied access to the breast, begging- rumbles
are only produced by calves, musth- rumbles only by males in
musth, estrous- rumbles only by females in estrus, and moth-
ers with small calves contribute more calls to infants to the
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
Behavioral Contexts of Elephant Acoustic Communication 129
picture of how elephants in a natural setting use vocaliza-
tions to communicate with one another and to compare the
form that these calls take under a wide range of situations
(see box 9.1 and table 9.1).
higher frequencies compared to rumbles by infants in other
contexts.
These studies provide fertile ground with which to
examine calls recorded in the wild from known individuals
in known behavioral contexts. Here I attempt to present a
data set than those without. The data set is, therefore, unbal-
anced across individuals and contains calls from individuals at
different ages across many years. The analysis is further com-
plicated because we know that it is possible to discriminate
between callers (McComb etal. 2003; Soltis, Leong, and Sav-
age 2005b) and because we know that emotional state (Soltis,
Leong, and Savage 2005b), as well as age and body size, affect
call duration and fundamental frequency, two key variables
(see fi gure 9.3). Due to the complexity of the data set, my sta-
tistical approach was tailored to the availability of specifi c data
and is explained in further detail within the relevant sections.
In general, however, I used a combination of non- parametric
descriptive statistics, and stepwise discriminant function anal-
ysis (DFA) with cross- validation to examine patterns across ele-
phant calls. In each of a series of stepwise DFAs, I partitioned
the data, using two- thirds of the data set to train the model
and the remaining one- third to test which group the calls be-
longed to. This cross- validation method characterizes the abil-
ity of the model to recognize the group membership of the
calls. In some discriminations, individual callers were largely un-
known (for instance, in the case of trumpets and other high fre-
quency calls). In these cases, the data were partitioned by ran-
dom selection, two- thirds for training and one- third reserved
for cross- validation. In the analysis of the more commonly
heard and recorded rumbles, in which the caller was known
with confi dence A or B, the data were fi rst divided into those
produced by adult females and those by calves. Within each
data set, calls were partitioned such that individuals who con-
tributed to training the model were different from those used
in cross- validation. Adult female calls (cadenced, “let’s go,”
contact, greeting, little- greeting, and coo) included individu-
als across a broad spectrum of ages, from age 10 to 55, while
calls produced by calves (begging, separated, as- touched,
baroo, and grumbling) included between the ages of 0–9.9
years. Calls emitted by both adult females and calves (e.g., coo,
little- greeting, grumbling, etc.) were considered within the
set containing the largest sample size. The inclusion of such a
broad range of ages and the partitioning of the data such that
different individuals contributed to the testing and validation
means that the results presented are very conservative.
Adult Female Rumbles
The adult female rumbles analyzed using DFA included es-
trous, female- chorus, mating- pandemonium, cadenced, “let’s
go,” contact, greeting, little- greeting, and coo. Adult female
reproductive calls were very unevenly distributed among indi-
viduals and due to the nature of the sample, I was not able to
use cross- validation. These three calls were therefore analyzed
separately (see under Mating Signals).
The overlapping nature of greetings and little- greetings
prevented the inclusion of a number of predictor variables;
they were, therefore, also analyzed separately (see below).
Greetings and little greetings were compared using a balanced
data set consisting of four calls of each type contributed by six
females (Echo, Eleanor, Eliot, Elspeth, Enid, Erin, and Eudora)
ranging in age from 10 to 55 years.
The data set for the remaining adult female calls (ca-
denced, “let’s go,” contact, and coo- rumbles) was unbal-
anced with respect to individuals, so I partitioned the data,
using two- thirds of the data set with six individuals (Echo,
Eudora, Enid, Eleanor, Eliot, and Elspeth) to train the model
and the remaining one- third with four different individuals
(Ella, Erin, Edwina, and Emma) to test call membership. The
inclusion of such a broad range of ages and the partitioning
of the data such that different individuals contributed to the
testing and validation mean that the results presented are
very conservative.
Calf Calls
Among calves, call duration and fundamental frequency
change very rapidly with increasing age, thus making dis-
criminations between proposed call types more diffi cult. In
addition, due to the short longitudinal window, the data set
was even more unbalanced than that for adult elephants. I
was, therefore, forced to compare groups of similar context
types, rather than across the entire data set. I fi rst used a step-
wise DFA with cross- validation to test whether discrimination
between rumbles given in the context of begging versus those
given in the context of separation from mother was possible.
Next, I used a stepwise DFA with cross- validation to discrimi-
nate between three proposed call types, two given by calves
in the context of being comforted and another given in con-
text of being thwarted. In each analysis, I used two different
groups of individuals, one set to train the model and another
for cross- validation. Since the rumbles in the latter discrimina-
tion were represented by characteristic contours, I added a
score for contour to the model (see table 9.1).
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
130 Joyce H. Poole
that the sound was produced. Laryngeal calls are those orig-
inating in the larynx and trunk calls are those produced by a
blast of air through the trunk. I further categorized putative
call types by ear and by manner features (those that specify
the manner of articulation: in other words, how the tongue,
lip, and other vocal organs are involved in making a sound)
Range in Frequency and Duration of Elephant Calls
Elephant calls recorded in Kenya (mostly in Amboseli) con-
tain frequencies ranging over more than 10 octaves from
a minimum fundamental frequency of 8 Hz measured in
some rumbles to maximum frequencies of around 10,000
Hz produced in some trumpets and snorts. Indeed, within
a single call the fundamental frequency may range over 4
octaves, starting with a rumble at 27 Hz grading into a roar
at 470 Hz. Sounds produced by African savannah elephants
range in duration from less than a tenth of a second to al-
most 15 seconds (n = 2,299; see table 9.2 and fi gure 9.1).
Call Type Classifi cation
The vocalizations produced by African elephants can be di-
vided into acoustically and structurally distinct call types
within which there are varying degrees of graded variation.
Following from the work of Stoeger- Horwath et al. (2007),
I divided calls into two main categories based on the way
Table 9.1 Acoustic measures
Main acoustic feature Parameter measured
Manner features Feature 1: Tonal —a call or portion of a call showing distinct harmonics (frequency contours of the harmon-
ics in tonal signals: fl at (0); ascending, descending (1); bent, bent left, bent right, slightly arched, slightly
wavy (2); bimodal, multi- modal, jittery, wavy (3); skewed left, skewed right (4); arched (5).
Feature 2: Tonal with noise —A sound displaying distinct harmonics with noisy sections (frequency contours
above apply).
Feature 3: Noisy harmonics—a call or portion of a call showing noisy harmonic bands (frequency contours
above apply).
Feature 4: Noisy—a call or portion of a call with no distinct harmonics or overtones.
Feature 5: Pulsatile
Feature 6: Transient noisy
Feature 7: Transient tonal
a. Duration: Cursor placement—Signal Signal duration. Units ms.
b. Fundamental frequency: Cursor placement—Signal F0 start, F0 end, F0 minimum, F0 maximum. Units: Hz.
c. Location of F0 maximum: Cursor placement—Signal The location of F0 maximum (elapsed time from F0 start to F0 maximum) as a percentage of duration
d. Curve: Cursor placement—Signal Range F0 / Duration
e. Dominant frequency: Selection “Maximum
Frequency”— Raven
The frequency at which maximum power occurs. Units: Hz.
f. Bandwidth: Selection—Raven. Upper frequency bound of the selection / highest visible frequency of the call. Units: Hz. Not a particularly
reliable measure due to variable recording distances.
g. Maximum time: Selection—Raven In a spectrogram view, the fi rst time in the selection at which a spectrogram point with the power equal to
Maximum Power occurs. Units: ms.
h. Peak time: Selection—Raven In a waveform view, the fi rst time in the selection at which a sample with amplitude equal to peak ampli-
tude occurs. Units: ms.
i. Sound pressure levels Measurements are (re 20 μPa) extrapolated to 5 meters and from previously published work (Poole etal.
1988). Units: dB. Available for a few call types.
Table 9.2 Time frequency characteristics across all measured calls
All calls N Median I- Q range Range
Duration (ms) 2219 3,248 1,671–4,818 87–14,372
F0 start (Hz) 2108 18.0 14.3–26.2 8.6–727
F0 max (Hz) 2099 20.5 16.4–29.6 11.7–1,016
F0 min (Hz) 2113 16.6 13.0–24.0 8.4–658
Max Freq. (Hz) 1893 41.0 31.2–107.4 13.7–4,078
Curve 2098 0.001 .001–.002 .000–1.740
Bandwidth 2091 237 168–559 38–32,000
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
Behavioral Contexts of Elephant Acoustic Communication 131
Accordingly, after listening to many thousands of calls
and examining and measuring 2,366 spectrograms from
hundreds of individuals in a free- range setting, I propose
10 primary call types used in communication: the laryngeal
call types (rumble, rev, roar [with sub- types noisy, tonal,
and mixed], cry, bark, grunt, and husky cry); and the trunk
call types (trumpet, nasal trumpet, and snort). In addition
to these principal call types are three idiosyncratic calls:
croak, truck- like, and squelch. Besides these call types, ele-
phants frequently emit composite calls that grade from one
type into another. This rich range of amalgamated calls in-
cludes snort rumbles, roar rumbles, rumble roar rumbles,
cry rumbles, bark rumbles, and trumpet rumbles. Compos-
ite calls are most likely to be produced when elephants are
disturbed or excited.
A list of call types is presented in table 9.3 and illustrated
in fi gure 9.2, while broad structural characteristics of each
are presented in table 9.4.
Laryngeal Call Types
Examples of laryngeal call types may be heard at www
.elephantvoices.org / multimedia- resources / elephantvoices
- calls- database- call- types.html?catid=3.
Rumble. Rumbles (Poole et al. [1988]; table 9.4a; fi gure
9.2a) are the most frequently heard call type in the reper-
toire of all age / sex classes of African savannah elephants
and are easily distinguished from other types by their very
low frequencies and clear harmonic structure. During a sub-
set of 2,055 minutes of recording with the EB family in the
based on visual inspection of the spectrogram, for instance
tonal versus noisy features (Fant 1960; see also table 9.1),
and by structural differences in time and frequency charac-
teristics.
I tried to fi t measured calls into previously established
nomenclature based on available qualitative descriptions,
spectrograms, and acoustic detail as well as exchange of
sound fi les and discussions with Angela Stoeger- Horwath,
Kirsten Leong, and Joseph Soltis. Working on captive ani-
mals and following on from Berg (1983), these authors have
established the mutually exclusive call types: rumble, rev,
croak, trumpet, snort, chuff (Leong et al. 2003) and, ad-
ditionally, bark and grunt (Stoeger- Horwath et al. 2007).
Each of these call types are also produced by free- ranging
elephants (though I exclude chuff, which I believe to be an
elephant cough / sneeze), although, due to the broader and
more complex contextual possibilities offered in a natural
setting, call types are heard in different frequencies and cir-
cumstances than in captivity, often resulting in different in-
terpretations. I will not attempt to discriminate between
the previously described types, but will, where possible,
examine how they are used in a natural social setting. To
these call types, I propose three additional types: cry, husky
cry, and nasal trumpet (see below).
The elephant’s capacity for vocal production learning
or imitation creates the potential for the description of call
types that, while structurally unique, may not be socially
relevant (e.g., croak [Leong et al. 2003; this chapter] and
truck- like call [Poole et al. 2005]). I also describe these call
types (croak, truck- like, and squelch), listing them sepa-
rately as Idiosyncratic.
0
100
200
300
400
500
600
700
0 2000 4000 6000 8000 10000 12000 14000 16000
Duration (ms)
Minimum F0 (Hz)
Figure 9.1 Duration and minimum fundamental frequency of all elephant calls. Note the
learned truck- like calls produced by a single individual stand out as distinct from the other
calls at approximately 50 Hz and multiple durations.
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
132 Joyce H. Poole
and excitement level as well as the call context- type (see
below). Some rumbles are highly modulated, powerful
sounds; others rise or fall in pitch; some rumble contours
are fl at, some undulate, and still others are jittery. Rumbles
range in duration from less than half a second to almost
12 seconds and may be emitted as a soft fl uttering whisper or
an explosive throaty resonance with sound pressure levels
up to 103 dB extrapolated to 5 m from source (Poole et al.
1988). Very powerful rumbles tend to be both highly modu-
lated and rather noisy. Rumble bandwidths range from 38
to 996 Hz, with some of the variance being explained by the
distance of the microphone from the source. Although there
is considerable variation within age classes, the rumbles of
older individuals are signifi cantly longer in duration and
lower in frequency than those of younger individuals (Dura-
tion: Spearman R = 0.636; t (n – 2) = 28.6; n = 1,210, P
< 0.001; minimum F
0
: Spearman R = –0.759; t (n – 2) =
–40.52, P < 0.001; fi gure 9.3).
fi rst half of 2000, I made note of 519 calls in my fi eld notes.
Of those, 474 (91.3 percent) were rumbles. Since many
rumbles would have gone undetected (due to their very low
frequencies), this percentage is likely to be an underesti-
mate. Similarly, in Leong’s study 87 percent of all calls were
rumbles. A large number of specifi c stimuli evoke rumbling
sounds and, unlike other types of calls, it is not possible to
generalize in a few sentences the contexts in which rumbles
are emitted. These are, instead, described in detail in the sec-
tions that follow.
Rumbles originate in the larynx and resonate as they fi l-
ter through the pharyngeal pouch, the nasal passages of the
skull, and through the trunk. Rumbles can be produced with
the mouth open or closed, though louder, more modulated
rumbles tend to be associated with open mouths. Rumbles
are highly variable, graded calls. Rumbles in this data set
contain fundamental frequencies ranging between 8.4 and
34.2 Hz (table 9.4a) depending upon the caller’s age, size,
Table 9.3 A comparison of nomenclature (numbers represent structurally distinct call types)
Poole (this chapter)
Stoeger- Horwath etal.
(2007)
Leong etal.
(2003)c Berg (1983)
Laryngeal calls
1 Rumble Rumble Rumble Growl
Rumble Rumble Rumble (noisy) Rolling growl
2 Rev — Rev —
3 Roar / Roar rumble Roar / Roar rumble —aRoar
Tonal roar Tonal roar — Cry
b
Noisy roar Noisy roar — —
Mixed roar Mixed roar — —
Pulsated roar — —
4 Cry / Cry rumble Tonal- roar — —
5 Bark / Bark rumble Bark / Bark rumble — Bark
6 Grunt Grunt —(No infants) —(No infants)
7 Husky cry Rumble — (No infants) —(No infants)
Trunk calls
8 Trumpet Trumpet Trumpet Trumpet / Trump
Pulsated trumpet — — Pulsated trumpet
9 Nasal trumpet — — —
10 Snort Snort Snort Snort
Imitated and idiosyncratic
11 Truck- like — — —
12 Croak — Croak —
13 Squelch — — —
Note: The dashed lines indicate that the calls were not mentioned by the authors.
a Soltis (2009) reports roars from same elephant group.
b Soltis (2009) interprets Berg’s “Cry” as a Trumpet.
c Based on recordings sent to me by Leong, I believe her class chuff is a wheeze / sneeze.
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
Behavioral Contexts of Elephant Acoustic Communication 133
me by Leong, which sounded buzzing or revving, I was able
to locate four revs in my collection, although none of these
occurred in a known context. Therefore, very little more
than what Leong described is currently known about these
infrequent call types. Their very rare occurrence in the wild
suggests to me that they may be an artifact of captivity and
possibly another example of a learned call.
Rev. The rev (table 9.4a; fi gure 9.2a) has been described by
Leong et al. (2003) as a short tonal harmonic vocalization,
less than a second in duration, and almost always followed
immediately by a rumble. The rev has a harmonic structure
and appearance similar to a short rumble, but its funda-
mental frequency at between 50 and 90 Hz is signifi cantly
higher than any known rumble. Based on recordings sent to
Figure 9.2a Comparative laryngeal call types.
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
134 Joyce H. Poole
elements. Tonal roars are those in which greater than 50 per-
cent of the call is tonal (that is, containing clear harmon-
ics). Noisy elements, if there are any, occur only at the end
of the call. Mixed roars alternate between noisy and tonal
or contain noisy components, making up more than 50 per-
cent of the call. Noisy roars tend to be unmodulated, while
tonal roars may be highly modulated, some reaching a clear
Roar. Roars (Berg [1983]; Stoeger- Horwath [2007]; table
9.4a; fi gure 9.2a) are powerful, highly variable bellowing,
screaming, shrieking, or squealing sounds. I used the defi ni-
tions of Stoeger- Horwath et al. (2007) in describing roars.
Using manner features they divided roars into three main
types: noisy roars, tonal roars, and mixed roars. Noisy roars
are totally noisy, in other words they contain no harmonic
Figure 9.2b Comparative trunk and learned call types.
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
Behavioral Contexts of Elephant Acoustic Communication 135
or poked has a bellowing quality, as does that of a calf who
is denied access to the nipple. Roars that occur during in-
tense social excitement such as during a greeting ceremony,
the birth of a calf, or in response to the arrival of a musth
male are more likely to occur as a rumble roar rumble. Simi-
larly, the roars of a lost calf may often be in the form rumble
roar rumble, though the quality of these is different.
Roars elicit the support or attention of others. For in-
stance, roars by calves draw the attention of caretakers,
those by an estrous female attract males, and those by a
family member who has been tusked draw the immediate
response and support of close associates.
Cry. Infants and calves under the age of fi ve years may emit
a cry or cry rumble (fi gure 9.2a) in situations in which they
are in some form of distress. Of the 20 cries whose spectro-
grams have been measured, nine were made by calves either
denied access to the breast or a food item, two were pro-
testing unwanted touching by an older elephant, one was
stuck on its side unable to get up, and eight more were in
unknown circumstances. Cries are very short whimpering
sounds lasting less than half a second in duration. A rumble
without an inhalation follows many cries. One composite
rumble cry rumble given by a calf begging for a piece of
palm heart was both preceded and followed by a rumble.
The rumble portion in the composite calls (n = 12) typi-
cally lasts longer than the cry portion (median (ms) = 850;
inter- quartile range = 547–1,628; range = 335–3,824).
Although the cry is structurally most like a short tonal
roar, the two types sound very different. I used a stepwise
crescendo before falling. All roars sub- types may start,
end, or start and end, with a rumble; these composite calls
are termed roar rumbles, rumble roar rumbles, or rarely,
rumble roars. Roars produced by infants are generally
higher pitched than roars produced by adults and the terms
“screaming,” “shrieking,” “squealing,” or “crowing” bet-
ter describe the quality of the calls by younger individuals.
As elephants become larger roars take on a more resonant
“bellowing” or “roaring” quality.
Elephants in some form of distress emit most roars, with
the vast majority being produced by infants, calves, and
juveniles. Of the 120 roars that I measured, 17 were pro-
duced by calves begging for, or being denied access to, the
breast or a food item; 10 were calves protesting unwanted
or rough contact by another elephant; 23 were emitted
by a calf or juvenile who had become separated from its
mother; 30 were the recipients of agonistic behavior (gener-
ally pushed, tusked, poked, or kicked); and 1 was pounced
upon by a lion. Among adults roars were utilized in ad-
ditional contexts. Adults roar when chased or tusked (in-
cluded above), when attacking predators (2), when chased
by males during estrus (11), and during highly exciting
social events (9). An additional 17 roars were recorded in
unknown contexts.
While it is not possible to distinguish statistically between
these roars with such small sample sizes, it is fair to say that
roars produced in these diverse contexts do sound different.
For example, the roar of an estrous female being chased by
males pulsates, the roar of a female threatening a lion is a
deafening scream, and the roar of an elephant being tusked
Figure 9.2c Idiosyncratic calls: (a) Truck- like, Malaika imitates the sound of distant trucks; (b) Squelch; (c) Croak, Gail.
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
136 Joyce H. Poole
sound produced by infant elephants in the fi rst days of life.
Stoeger- Horwath et al. (2007) report that in captivity the
infants cease producing grunts by two months of age. In the
wild situation grunts are barely audible and, therefore, dif-
fi cult to record. The very few grunts that I measured were
associated with attempts to suckle.
Husky Cry. The term husky cry (table 9.4a; fi gure 9.2a) re-
fers to gruff or husky sounds produced only by newborn
and infant elephants up to about four months of age. Husky
cries (n = 32 measured) range from barely audible breathy
exhalations to surprisingly powerful hoarse rasping sounds.
They are usually made during moments of distress and are
heard frequently during the fi rst hours and days of life as
the unstable infant struggles to remain standing and as it
is touched and fondled by family members (13 of the calls
whose spectrograms were measured). Each call brings a vo-
cal response and additional fondling by mother and allo-
mothers, which in turn elicits more vocalizing by the infant.
Infants may also emit a husky cry when separated from
their mothers and frightened by something (n = 5), when
touched and woken by their mothers (n = 6), when engaged
in rough play (n = 1), or when frightened by a loud sound or
discriminant function analysis (DFA) (see box 9.1) to com-
pare cries (n = 22) with tonal roars (n = 12). The two call
types were signifi cantly different (Wilks’ lambda: 0.087; F
(5,17) = 35.860; P < 0.0001). In the training set (see box
9.1), 100 percent of the calls were correctly classifi ed (re-
duced to 91 percent in cross- validation; table 9.5).
Bark. The bark (Stoeger- Horwath et al. [2007]; table 9.4a;
fi gure 9.2a) is a transient and primarily noisy call that dif-
fers from a noisy roar in its very short duration. Similar to
the roars, barks may be combined with a rumble, produc-
ing a composite bark rumble. In the Amboseli population,
barks and bark rumbles were heard and recorded very infre-
quently. Barks occurred in contexts similar to roars: three
occurred in the context of begging (though these were re-
corded from the Tsavo orphans when calves were pushing
and shoving to reach the bottle), seven occurred when ele-
phants were tusked or pushed, one during rough play and
one from a 10- year- old female as she was mated for the fi rst
time. Eight occurred in unknown contexts.
Grunt. The grunt (Stoeger- Horwath et al. [2007]; table
9.4a; fi gure 9.2a) is a soft, short, beeping or honking
Table 9.4a Broad structural characteristics of laryngeal call types
Call
types Callers
Manner
features Duration (ms) F0 max (Hz) F0 min (Hz) Curve
Maximum
frequency (Hz) Bandwidth (Hz)
Rumble Infant–
adult
Tonal / tonal
with noise
3,795 (2,525–5,025)
300–11,857; 1,729
18.6 (15.7–23.2)
10.6–53.7; 1,701
14.7 (12.6–18.8)
8.4–34.2; 1,700
.007(.003–.009)
.000–.060; 1661
35 (29–45)
14–312; 1347
213 (160–285)
38–996; 1,620
RevaCalf–
adult
Tonal with
noise
721 (347–836)
331–1,125; 6
79 (53–98)
35–117; 6
68 (43–88)
35–92; 6
.013 (.012–.018)
0–.053; 6
156 (127–282)
71–439; 6
570 (527–700)
238–4,615; 6
Roar
Noisy
roar
Infant–
adult
Noisy 1,797 (1,147–2,172)
330–3,000; 59
346 (248–435)
108–568; 43
260 (183–314)
39–525; 49
.055 (.018–.085)
0–.359; 42
409 (291–632)
150–2,849; 59
3,524 (2,418–4,299)
1,364–8,791) 49
Tonal
roar
Infant–
juvenile
> 50% Tonal;
noise occurring
only at end
1,253 (1,108–2,137)
467–3,004; 12
518 (448–715)
371–740; 12
280 (241–339)
131–435; 12
.193 (.089–.356)
.038–.689; 12
409 (312–561)
215–687; 12
3,407 (2,118–5,000)
1,560–8,674; 10
Mixed
roar
Infant–
adult
Alternating
tonal and noisy
elements; with
noise more
than 50%
1,554 (1,108–2,062)
532–3,922; 38
376 (277–496)
172–776; 38
278 (191–328)
108–658; 38
.073 (.036- .117)
0–.425; 38
557 (353–732)
183–4,078; 36
3,746 (2,968–5,143)
1,657–8,885; 37
Cry Infant–
calf
Tonal 327 (289–456)
178–528; 20
451 (334–519)
183–800; 20
274 (246–315)
159–547; 20
.314 (.246–.322)
.049–1.497; 20
388 (279–642)
279–1,273; 20
3,596 (1,835–5,460)
973–6,968; 18
Bark Infant–
adult
Transient and
noisy
456 (392–620)
245–787; 12
Not
measurable
Not
measurable
Not
measurable
420 (345–566)
194–796; 12
4,234 (2,242–5,153)
941–6,094; 12
GruntaInfant Tonal 218 (184–237)
139–255; 8
319 (269–429)
250–479; 8
259 (211–350)
205–439; 8
.24 (.19–.32)
.13–.69; 8
448 (345–487)
259–753; 8
2,232 (2,549–2,072)
1,258–3,116; 8
Husky
cry
Infant Tonal with
noise
976 (715–1,244)
27–1,544; 33
37 (34–45)
25–69; 32
31 (25–33)
18–62; 33
.007 (.004–.009)
.000–.019; 35
66 (59–90)
35–414; 33
584 (409–685)
87–1,000; 33
Notes: Figures represent median (inter- quartile range) range and n. All calls that had been assigned a level of sureness “A” are included.
a View with caution due to low sample size and uncertain classifi cation.
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
Behavioral Contexts of Elephant Acoustic Communication 137
Table 9.4b Broad structural description of Trunk Call types
Trunk
calls Callers
Manner
features
Duration
(ms)
F0 max
(Hz)
F0 min
(Hz) Curve
Max frequency
(Hz)
Bandwidth
(Hz)
Trumpet Infant–adult Noisy
harmonics
724 (506–1,044)
198–3,385; 224
426 (372–515)
178–917; 211
318 (266–372)
108–606; 215
.142 (.087–.220)
.012–.630; 211
460 (351–790)
105–2,405; 211
8,654 (1,274–21,931)
5,192–11,100; 223
Nasal
Trumpet
Calf–adult Tonal with
noise
709 (578–864)
403–2,732; 23
118 (68–146)
34–329; 19
62 (56–107)
32–246; 21
.050 (.026–.076)
.003–.420; 19
241 (131–306)
32–563; 23
3,729 (2,501–4,965)
786–10,425; 23
Snort Infant–adult Noisy 783 (488–884)
338–1,374; 18
Not measurable Not measurable Not measurable 86 (65–130)
43–517; 16
9,285 (4,973–11,152)
470–16,874; 18
Note: Figures represent median (interquartile range) range and n.
Table 9.4c Learned calls: Imitated and idiosyncratic calls
Production
learning Callers
Manner
features
Duration
(ms)
F0 max
(Hz)
F0 min
(Hz) Curve
Max frequency
(Hz)
Bandwidth
(Hz)
Croak Two individuals Noisy har-
monic
4,123 (1,651–4,643)
1,128–9,032; 6
Not
measurable
Not
measurable
Not measurable 611 (579–633)
526–633; 6
4,254 (4,000–4,293)
3,862–5,661; 6
Squelch Calf–adult Noisy har-
monic
204 (117–378)
87–554; 13
306 (50–343)
30–377; 7
245 (42–290)
28–352; 7
.232 (.022–.883)
.004–.896; 7
215 (65–428)
59–1920; 12
2,372 (1,078–5,940)
411–9,545; 13
Truck- like One individual Tonal 6,984 (4,763–9,305)
685–14,372; 60
60 (58–62)
45–597; 60
52 (50–54)
45–66; 60
.001 (.001–.002)
.000–.060; 60
Not measured 168 (161–173)
95–252; 59
Note: Figures represent median (interquartile range) range and n.
if receiving rough treatment (n = 3). Most of these involve
some form of being touched. The remaining calls (n = 5)
were in unknown contexts.
Husky cries contain noisy harmonics similar in struc-
ture to some rumbles yet their sound quality is dissimilar.
Comparing the husky cries (n = 32) and rumbles (n = 110)
of calves less than a year of age reveals signifi cant differences
(Wilks’ lambda 0.412, F = 14.466, P < 0.0001; table 9.5). In
the training set 91 percent of calls were correctly classifi ed
(87 percent using cross- validation procedures; table 9.5).
0
5
10
15
20
25
30
35
40
0 2000 4000 6000 8000 10000 12000
Duration (ms)
Minimum F0 (Hz)
0A
0B
1A
1B
2
4
5
Figure 9.3 Call duration and minimum fundamental frequency of rumbles (males and females combined) by age class.
The rumbles of older individuals are signifi cantly longer in duration and lower in frequency than those of younger indi-
viduals. Duration: Spearman R = 0.636; t (n – 2) = 28.6; n = 1,210, P < 0.001; Minimum F0: Spearman R = –0.759;
t (n – 2) = –40.52, P < 0.001.
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
138 Joyce H. Poole
birth, 5 with presence of a musth male, 6 with a mating, 3
with a greeting ceremony, 2 with the rescue of a calf); 16
were made by frightened calves separated from their moth-
ers; 13 by elephants chasing predators; 1 by a male express-
ing irritation at noise in the research vehicle; 1 by a female
in response to a member of her family being pushed; and
another by an elephant responding to a distant trumpet.
Nasal Trumpet. Nasal trumpets (table 9.4b; fi gure 9.2b) are
noisy with underlying tonal structure and sound like a large
man blowing his nose. They are much noisier than trum-
pets, considerably lower in frequency, and almost invari-
ably heard in the context of play. Of the 22 nasal trumpets
whose spectrograms I have measured, 18 were in the con-
text of play while 4 were associated with the intense excite-
ment surrounding an elephant birth.
In discriminating between nasal trumpets and trumpets,
it was not possible to balance data sets by individual since
unidentifi ed callers produced these calls during exuber-
ant play or frenzied social events. Thus, to test the valid-
ity of this newly proposed call type, I performed a stepwise
DFA on all measured nasal trumpets (n = 23) and trumpets
(n = 243; Wilks’ lambda = 0.482, F (8,217) = 19.711, P
< 0.0001; table 9.5) in which 96 percent of the calls were
correctly classifi ed (94 percent in cross- validation). Funda-
mental frequency values mainly separate the two call types.
Snort. Snorts (Leong et al. [2003]; Stoeger- Horwath et al.
[2007]; table 9.4b; fi gure 9.2b) are short, noisy, broadband
sounds produced by blowing air purposefully through the
trunk. Elephants may snort when they are surprised by
Trunk Call Types
Examples of trunk call types may be heard at www
.elephantvoices.org / multimedia- resources / elephantvoices
- calls- database- call- types.html?catid=4.
Trumpet. Trumpets are produced by a forceful expulsion
of air through the trunk and come in several forms (Berg
[1983]; Poole and Granli [2004]; table 9.4b; fi gure 9.2b).
Trumpets are mainly tonal sounds, though harmonics are
overlaid with noise. Most last less than a second (though
extremely long trumpets may last over three seconds), with
fundamental frequencies falling around 300 Hz and peak
energy concentrated at approximately 450 Hz (table 9.4b).
An elephant can produce a wide variety of sounds by
varying the speed of air she forces through her trunk, by the
shape in which she holds her trunk, and by her own body
posture and movement. Elephants tend to trumpet when
they are highly stimulated—in situations in which they
may be fearful, surprised, aggressive, playful, or socially ex-
cited—and the quality of trumpeting varies with the context
(see descriptions below). Trumpeting is often associated with
intensely social events such as a birth, mating, or greeting
ceremony, where group participation is important. In these
situations I postulate that rumbling may defi ne the context
(e.g., greeting, mating, etc.) while trumpeting may function
as a kind of “exclamation mark,” expressing the very high
level of excitement and stressing the importance of the event.
Of the 224 trumpets whose spectrograms have been
measured and contexts were known, 107 were associated
with play; 85 with intense social excitement (69 with a
Table 9.5 Results of stepwise DFAs for the three newly proposed call types: Cry, husky cry, and nasal trumpet
Call types Cries and tonal roars Husky cries and infant rumbles Nasal trumpets and trumpets
n in the training set 23 79 156
Eigenvalue overall 15.9 1.43 1.073
% Variance 100 100 100
Wilks’ lambda 0.059 0.412 0.482
F34.077 14.466 19.711
P<0.001 0.001 <0.001
Expected probabilities .50 .50 .50
Classifi cation % correct 100 91 96
Classifi cation cross- validation
% correct
91 87 94
Variables in the model Duration, max time, location
F0 max, curve, F0 max- min,
bandwidth
Max frequency, duration, band-
width, max time, curve, F0 max- min,
peak time
F0 start, bandwidth, F0 max,
location F0 max, max frequency,
curve, max time
Notes: Each was compared with the call types they most closely resemble. Cries with tonal- roars; husky cries with rumbles produced by calves under a year; nasal
trumpets with trumpets.
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
Behavioral Contexts of Elephant Acoustic Communication 139
odors into the mouth and usually occurred in a series of two
or three croaks. In Amboseli, as far as we know, only two
elephants in the entire population of more than 1,500 indi-
viduals emit this highly unusual sound, and both are adult
females from the same family. For many years only Gail was
heard to croak, but subsequently, I heard Gwen croaking.
Unlike the observations by Leong et al., both Amboseli ele-
phants croak when they are relaxed and feeding, and based
on the lack of reaction and absence of production by other
elephants, it appears to have no communicative function.
Gwen and Gail are sisters and close associates; I suggest that
the croak is another example of vocal learning.
Squelching. Occasionally elephants may be heard to make a
“squelching” sound (table 9.4c; fi gures 9.2b and 9.2c), ap-
parently produced by forcing air through a “scrunched- up”
trunk. Sometimes the individual gives the impression of
having a genuine itch in his or her trunk, but at other times
production of this rather odd bubbling sound seems to be
an end in itself. Squelching is most often heard when ele-
phants are relaxed (for example, standing by the side of a
waterhole or waiting patiently for other family members to
move).
Chorused Calling
There are numerous situations in which elephants call in
chorus. In some cases, these are simply overlapping low to
moderate level rumbles between two or more individuals,
but very often choruses occur as a powerful series of lay-
ered calls that include a variety of different rumbles, trum-
pets, snorts, and roars. These powerful chorused calls typi-
cally occur during excited social events such as a birth, a
greeting ceremony, the arrival of a sexually active musth
male, a mating, a calf rescue, or during offensive or defen-
sive action.
During a sub- set of 2,015 minutes of recordings from
1999, a total of 3,685 vocalizations were counted on the
spectrograms, of which 42 percent occurred as single calls
and 58 percent occurred within a chorus of two or more
overlapping calls. The median number of calls in a chorus
was 2 (inter- quartile interval = 2–4; range 2–29). In the en-
tire data set, the maximum number of overlapping calls in a
chorus was 64, following a mating recorded in the Maasai
Mara population.
Behavioral Contexts and Acoustic Quality of Calls
Since many of the laryngeal and trunk call types described
in the previous section are graded and, based on the mea-
surements taken thus far, it is not possible to discern discrete
something (these may simply be less explosive trumpets)
during intense social excitement or to alert other members
of their group to a new situation. Snorts are usually audibly
distinguishable from the more common “blows” (blowing,
sneezing, wheezing, and coughing sounds) that appear to be
made for the purpose of clearing the nasal passages; snorts
sound more sharp and purposeful than a blow and may also
be distinguished by context.
Vocal Learning: Imitated and Idiosyncratic Sounds
Elephants produce a variety of novel sounds, some of which
are learned through imitation of other elephants or things in
their environment (Poole et al. 2005). Sounds reported thus
far include humming (Randall Moore, personal commu-
nication), squelching (this chapter), croaking (Leong et al.
2003; this chapter), and the imitation of whistling by Asian
elephants (Wemmer and Mishra 1982; Wemmer, Mishra,
and Dinerstein 1985), the imitation of trucks and Asian
elephant chirps by African elephants (Poole et al. 2005),
and the imitation of Korean words by a 16- year- old male
Asian elephant named Kosik at Everland, an amusement
park in South Korea (Kwang- Seog Heo, personal com-
munication; www.wayodd.com / south- korean- elephant
- mimics- human- sounds / v / 4023; www.reuters.com / news /
video?videoId=1231). Many of these unusual or idiosyn-
cratic sounds are produced only rarely, and consequently
I do not have a good sample of recordings. There are two
unusual sounds, croaking and squelching, for which I have
some information and one, the truck- like call, for which I
have numerous recordings from one individual. Examples
of these sounds may be heard at www.elephantvoices
.org / multimedia- resources / elephantvoices- calls- database
- call- types.html?catid=5.
Truck- Like. In 2005, we (Poole et al. 2005) reported the imi-
tation of a truck sound by a semi- captive orphan elephant
in Tsavo (table 9.4c, fi gures 9.2b and 9.2c). With a funda-
mental frequency hovering around 50 Hz and highly vari-
able duration (range: 685 ms to almost 15 seconds), this
sound is unlike any call in the normal repertoire of African
elephants. We were able to show that the elephant was imi-
tating the noise of distant trucks on a highway 3 km away.
Since the publication of the 2005 paper, additional ele-
phants in the small Tsavo group of orphans learned to pro-
duce the same sound.
Croaking. Studying elephants in captivity, Leong et al.
(2003) fi rst reported croaking (table 9.4c; fi gures 9.2b and
9.2c) as pulsatile sounds lasting between 1 and 10 seconds.
These sounds were produced by several different individu-
als and were often associated with the sucking of water or
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
140 Joyce H. Poole
complex, non- parametric nature of the data as elaborated
in box 9.1, in the Statistical Analysis section.
In order to discriminate statistically between specifi c
context- types, I followed the procedure outlined in box
9.1, using a combination of non- parametric descrip-
tive statistics, and stepwise DFA with cross- validation to
examine patterns across elephant calls. Due to limitations
presented by elephant sociality, life histories, and fi eld col-
lection, analyses had to be customized to suit data avail-
ability (see box 9.1) and were thus sub- divided into adult
female reproductive rumbles (estrous, female- chorus,
mating- pandemonium); calf rumbles (begging, separated,
and, separately, as- touched, grumbling, umbrage); adult
female social and caregiving rumbles (cadenced, contact,
“let’s- go,” coo, and separately, greeting, little- greeting);
fi nally six trumpet context- types were compared. Calls
produced by adult male calls were excluded, since only
musth- rumbles were suitably represented. The results of
these analyses are presented in tables 9.6, 9.7, 9.10, and
9.11 and are described further in the text in the following
sections. Further statistical information is presented in the
methodology.
The results presented in the text below and in tables 9.6,
9.7, 9.10, and 9.11 clearly show that elephants produce a
wide variety of rumble forms that are associated with be-
havioral context. While these forms, or context- types, are
certainly highly variable (e.g., refl ecting emotional level and
individual caller in a range of measures), and in some cases
may be only subtly different, the elephants show by their
consistent behavior that they are able to discriminate emo-
tion and meaning in these graded calls. For this reason, I
suspect that more detailed measurements will, in the future,
call sub- types, I ask: Are specifi c behavioral contexts associ-
ated with particular acoustic- structural patterns? And if so,
what clues might this give us for further understanding ele-
phant communication?
Many years before I began examining the structural
characteristics of vocalizations, members of the Amboseli
Elephant Research Project (AERP) collectively gave names
to calls as a way of referring to them when in the fi eld. We
assigned labels based upon a combination of sound qual-
ity and observed patterns of behavior. While the names
of the call types described earlier refl ect sound quality
(e.g., rumble, roar, cry, husky cry, grunt, snort, trumpet,
etc.), the names we gave to what we perceived to be dif-
ferent rumbles tended to refl ect their behavioral context
(e.g., begging- rumble, separated- rumble, contact- rumble,
musth- rumble, estrous- rumble). I refer to these proposed
sub- types as context- types. Although the contextual nam-
ing of calls may have led to a bias in interpretation, the very
subtle differences (at least to the human ear) of many low
frequency calls made it diffi cult to come up with unique
names based purely on sound quality. Where such a dis-
tinction is possible, we have used names that describe the
sound quality (e.g., cadenced- rumble). The remainder of
this chapter gives a broad overview of the different behav-
ioral contexts in which vocalizations occur and a descrip-
tion of the calls with which they are associated. Descriptive
statistics of these call context- types are presented in annex
9.1; context- types include all measured calls included in the
analyses that were designated a level of confi dence “A” (see
box 9.1; n = 2,014).
These results presented below include some 35 context-
types, with a wide range of sample sizes. Some of the pro-
posed context- types, such as those given during group de-
fense, are represented by small sample sizes and cannot
be used in further statistical discriminations. For those
context- types with suffi cient sample sizes, a standard
DFA was fi rst used to indicate whether signifi cant differ-
ences might exist between the many proposed rumble
context- types. For this analysis, I simply selected all
context- types with classifi cation level A and with a sample
size of at least 25 measured calls across all individuals. These
included 12 context- types: anti- predator, begging, cadenced,
grumbling, contact, coo, greeting, “let’s- go,” little- greeting,
musth, separated, and baroo (or “protest”) rumbles (n =
902). I used the following predictor variables: duration, F0
max, F0 max- min, F0 start, curve, high frequency, and maxi-
mum frequency. Although the results suggest that rumbles
do vary with context (Wilks’ lambda = 0.159, F77, 6,473 =
30.148, P < 0.0001; 35 percent classifi ed correctly; expected
0.08), I was not able to balance the groups with respect to
the contribution of calls by different individuals given the
Table 9.6 Results of discriminant function analysis in reproductive rumble types
Adult female rumbles
Estrous- rumble, female- chorus, mating-
pandemonium
N 47
Eigenvalue overall 1.527
Wilks’ lambda 0.291
FF = 2.099
P<0.01
Classifi cation expected
probabilities
0.33
Classifi cation % correct 77
Variables in the model Contour, duration, max time, location F0 max,
frequency modulation, curve, F0 start, peak time,
bandwidth, maximum frequency
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
Behavioral Contexts of Elephant Acoustic Communication 141
toward one another), and then Bunching (see Poole and
Granli, chapter. 8). Once bunched, the elephants vocal-
ize, secrete from the temporal glands, and reach out with
their trunks to touch one another with reassuring gestures.
Through this multi- faceted exchange of information the ele-
phants appear to assess the level of danger present and col-
lectively decide what their response will be. A single adult
female may charge forth, the group may attack en masse,
or members of the group may make a hasty retreat. Their
highly coordinated response is communicated via a com-
plex combination of vocal, olfactory, and tactile signals and
is an area for further research. Unfortunately, I do not have
enough recordings of any of the calls given in these contexts
to include them in statistical analysis.
Alert. When family groups are exposed to an unusual or
disturbing situation, Listening and / or Freezing (Poole and
Granli, chapter 8) behavior may follow a sharp snort or
snort- rumble. Immediately following a snort given in this
context, a number of soft, medium- length rumbles by one
or more individuals may then be heard. The calls of sev-
eral individuals may overlap. Elephants in the calling group
stand alert, listening and looking.
Events that elicit such behavior may include non-
elephant disturbances such as unusual commotion in a ve-
hicle, a helicopter passing overhead, the discovery of Maa-
sai herdsmen in the area, or the roaring sounds of lions. A
disturbing or exciting event among con- specifi cs (such as
a fi ght between two musth males or serious aggression di-
rected at a family member) may also elicit this form of soft
rumbling, as will playbacks of unexpected callers. I refer to
this as comment- rumbling (fi gure 9.4, panel a), so called be-
cause elephants appear to use it to “comment on” or “call
attention to” an unusual or disturbing event.
Defensive and “Mobbing.” An elephant confronting a dan-
gerous predator can produce terrifyingly powerful sounds,
which may include deafening roars, trumpets, and ex-
tremely loud and noisy rumbles. Having bunched together,
older individuals at the fore and calves occupying the center
(see Bunching; chapter 8), one or more individuals may
Advance- Toward or Charge (chapter 8) the predator while
emitting powerful roaring- rumbles, trumpet- blasts, and
noisy roars (fi gure 9.4, panels b, c, and d) that would liter-
ally put fear in the hearts of men! As the elephant charges it
fl ings its trunk toward its adversary and may stop abruptly,
kicking up dust. The primary function of these calls appears
to be to intimidate.
Throughout a confrontation with a dangerous predator
the bunched elephants may continue to vocalize with noisy,
throaty, rolling rumbles, their heads raised, ears extended,
reveal further discriminations than I am able to show at the
present.
Examples of the elephant call context- types that I de-
scribe here may be heard online at www.elephantvoices
.org / multimedia- resources / elephantvoices- calls- database
- contexts.html.
Group Defense
Calls associated with anti- predator behavior include those
used in the context of alerting companions to the presence
of a predator, intimidating, or “mobbing” a predator as
well as those used while taking defensive action. Family
members produce several different call types when they
confront predators or when they fi nd themselves in poten-
tially threatening or frightening situations. These include
rumbles, snorts, trumpets, and roars. Much has been writ-
ten about the complex and highly coordinated defensive
and offensive behavior of elephants in the presence of pred-
ators (Douglas- Hamilton 1972; McComb et al. 2000; see
also chapter 8), but the variety of calls produced, and the
dramatic responses of other elephants to these calls, has re-
ceived little attention.
When exposed to the sound, sight, and smell of lions,
hyenas, humans, or other potentially dangerous predators
or situations, females and calves typically respond by fi rst
Freezing, then rapid Assembly (rapid walking or running
Table 9.7 Discriminant function analysis (DFA) results for common calf rumbles
Calf calls Begging, separated
As- touched,
grumbling, umbrage
N (training) 148 80
Eigenvalue overall 0.647 0.828
Wilks’ lambda 0.607 0.471
F15.208 6.681
P<0.0001 <0.0001
Classifi cation expected
probabilities
.50 .333
Classifi cation % correct 79 70
Classifi cation
cross- validation % correct
60 67
Variables in the model Duration, bandwidth,
maximum frequency,
max time, curve, F0 max
Contour, F0 max- min,
F0 start, duration
Notes: In the begging, separated DFA the training sub- set included calls by Ejac, Eldon,
Elettra, Elmo, Emmet, Ewaso. The validation sub- set included Elaine, Emily Kate, Erica,
Eudora 00, Explorer, and Lewa. In the as-touched, baroo and grumbling DFA, the train-
ing sub- set included calls by Echeri, Ejac, Elettra, Ella 03, Emmet, Erica, Eudora 00,
Europa, and Explorer. The validation sub- set included Charlotte, Elaine, Eldon, Eliot 03,
Elmo, Emily Kate, and Ewaso.
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
142 Joyce H. Poole
Food
Among many species of primates and birds, food calls are a
common type of vocal signal. In Amboseli, resources such
as grass, herbs, shrubs, water, and minerals are fairly evenly
distributed, and, thus, calls to alert members of a group to
an exciting food source may not be particularly relevant.
I have occasionally heard rumbles given in the context of
an individual locating a rare and nourishing food source,
such as a fallen Acacia xanthophloea, yet these instances
are so infrequent that I do not have a large enough sample
of calls to comment on whether Amboseli’s elephants have
food calls. I expect that if such calls exist, they are likely to
temporal glands streaming, and trunks reaching out to
touch one another. These calls have the effect of both in-
timidating the predator and simultaneously calling in sup-
port from family members not in the immediate vicinity. I
refer to the powerful noisy rumbles given in this context
as roaring- rumbles (fi gure 9.4, panel e). Softer calling by
bunched elephants alternatively appear to be involved in
decision making (fi gure 9.4, panel f) and may be followed
by a hasty retreat.
Startled. Startled elephants make sounds associated with
a sudden exhalation of air, which may take the form of a
trumpet or a snort or something in between.
Figure 9.4 Calls produced in the context of predators (note differences in frequency scale). (a) Snort followed by comment- rumble; (b) roaring- rumble as Enid
threatens lions; (c) trumpet- blast at a lion; (d) noisy roar in response to playback of lions roaring; (e) roaring- rumbles as Amy and Alison threaten a lion after it
pounced on Amy’s calf. Note more distant calls of approaching family members. (f)Young female charges my vehicle while trumpeting and rumbling; her behavior
causes her family to bunch together and emit these rumbles while facing toward me.
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
Behavioral Contexts of Elephant Acoustic Communication 143
of four males (Dionysus, RBG, Rashid, Long Left) from
whom I had recorded at least six musth- rumbles. Despite
the small sample size (n = 24), the model was able to clas-
sify 96 percent correctly (cross- validation was not possible
due to the very small sample sizes [Wilks’ lambda 0.056,
F = 2 .723, P < 0.003]). Although musth- rumbles are highly
distinctive, I was not able to statistically compare them with
other rumble sub- types, as males do not produce the other
call types under consideration and females do not produce
musth- rumbles.
Female Choice.
A series of highly modulated and typically
overlapping, or “chorused” calls, may be heard when a
musth male joins a family group or when he moves among
the group testing the genitalia and urine of females for
possible receptivity. Adult and juvenile females join in the
rumbling chorus that may also include trumpeting and,
occasionally, roaring. Females may rumble when tested,
in response to hearing a musth- rumble, including a play-
back (Poole 1999a), and may even call upon locating the
scent- trail of a musth male (Poole, unpublished data). As fe-
males call in this context, they characteristically urinate, def-
ecate, and secrete from their temporal glands. When females
are very excited, their chorusing is particularly powerful.
We refer to this type of calling as a female- chorus (Poole
et al. 1988). While female- choruses are relatively common,
it is extremely diffi cult to obtain representative measure-
ments of calls due to their overlapping nature. Figure 9.5,
panel c, shows a section of a female- chorus as Rashid joined
the EB family. What is immediately obvious from exam-
ining spectrograms is that the calls are highly variable in
shape but have in common a high degree of modulation and
energy in the upper harmonics, especially during the peak
of the chorus. The measurements presented in annex 9.1 are
unlikely to refl ect the calls at the peak of a chorus due to the
diffi culty of measurement.
Estrous females typically roar when pursued by a non-
musth male (Moss 1983). The roar may begin as a series
of short, pulsated growling sounds before developing (or
not) into a pulsating roar (fi gure 9.5, panel d), which has
the effect of attracting distant males. We refer to this as
an estrous- roar (or estrous- bellow, Moss [1983]; Poole
[1989b]). Females are more likely to estrous- roar in early
and late estrus—when young, low- ranking non- musth
males chase them—than during peak estrus, when they are
in consort with a high- ranking musth male and, therefore,
have less reason to attract the attention of additional males
(Moss 1983; Poole 1989b).
When a female has been mated, and immediately after
the male dismounts, she begins a series of distinctive rumbles
that are repeated at intervals, as if in song (fi gure 9.5, pan-
els b and f). These long rolling rumbles start at relatively
be heard in forests where food is more unevenly distributed
and elephants, like primates and birds, have an interest in
fruiting trees.
Sexual
Adult male and female elephants live in two essentially
separate social systems and both sexually active musth
males and estrous females are confronted with the problem
of fi nding and attracting suitable mates who may be uncom-
mon in space and time. They locate potential mates through
searching behavior, conspicuous postures (see chapter 8),
the secretion of strong odors, and loud and characteristic
calling.
Male- Male Competition.
During the heightened sexual and
aggressive period of musth, males emit a distinctive rumble
with a characteristically pulsated “put- put- put” or “glug-
glug- glug” quality, like water gurgling through a deep tun-
nel. This musth- rumble (Poole 1987a; Poole et al. 1988;
fi gure 9.5, panel a) is associated with an increased rate of
urine dribbling and a particular ear posture known as an
Ear- Wave (Poole 1987a; chapter 8).
Males emit musth- rumbles in many different, but rather
specifi c contexts, including in aggressive and sexual situ-
ations, while marking, drinking, or wallowing, as well as
in situations where they feel challenged. For example, the
sound of another musth male, an approaching vehicle,
or airplane is often enough to trigger a musth male to
musth- rumble (Poole 1987a). Musth males also frequently
musth- rumble before or after a bout of listening and are
presumably responding to distant elephant calls or using
calling to search for potential mates (Poole and Moss 1989).
Musth- rumbling in the company of females elicits a clus-
ter of overlapping calls that we refer to as a female- chorus
(Poole et al. [1988]; see below). Musth males are, however,
most likely to musth- rumble when they are alone, less often
in the company of females, and least often when guard-
ing an estrous female (Poole 1999a). Since musth males
often listen after musth- rumbling, it is likely that they use
musth- rumbles to advertise their heightened aggressive and
sexual state to rivals and potential mates and to search for
female groups. Once a male has located an estrous female,
he may have neither the need nor the desire to further ad-
vertise his whereabouts.
Some males have very distinctive musth- rumbles. Calls
by different males range from rather short to very long in
duration and from rolling to highly pulsated. The “typi-
cal” musth- rumble is a long, very low pitched, pulsated,
tonal call with considerable overlaying noise. I used a step-
wise DFA to test whether the measurements taken were ca-
pable of discriminating between the randomly selected calls
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
Figure 9.5 Calls produced in a reproductive context (note different frequency scale): (a) Musth- rumble note pulsations; (b) Single estrous-
rumble skewed left; (c) Segment of a female- chorus; (d) Pulsated estrous- roar; (e) Social- roars emitted during mating- pandemonium; (f)Series
of estrous- rumbles by Shirley; (g) Mating- pandemonium; calls by the mated female are those that reach a rapid peak in frequency. Note how the
rate of calling as well as its intensity and frequency modulation decreases with time after the mating.
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
Behavioral Contexts of Elephant Acoustic Communication 145
roars, roar rumbles, barks, bark rumbles, and grunts. An
infant or calf typically initiates begging by approaching or
moving along its mother’s side with its trunk raised, often
touching the mother’s side or leg, and emitting one or a
series of several very soft, short rumbles characterized by an
“rrrrmmmm” sound of descending pitch. This behavior and
associated calling is frequently heard in an elephant family
with infants and calves, and I refer to it as a begging- rumble
(fi gure 9.6, panel a; annex 9.1). Begging- rumbles are typi-
cally fl at or slightly descending in frequency, but about a
third are modulated, rising slightly and then falling. Gen-
erally only elephants under the age of fi ve years emit
begging- rumbles. Although juveniles as old as eight years
of age may be heard to beg in this manner, it is rare and
occurs primarily after the death of a younger sibling, when
an older calf may attempt to access the breast again. Among
free- ranging elephants, this type of rumbling is heard al-
most exclusively when calves are begging for access to the
breast, although I have twice heard a calf emit this sound
when it was begging for food from its mother’s mouth. In
captivity, these calls may be heard with great frequency and
intensity at feeding time, whether food is a bottle of milk or
coconut cakes.
In 1999 and 2000, I recorded a total of 166 begging-
rumbles from the EB family with level of sureness A. Fig-
ure 9.7 illustrates the number of calves in the EB family
during those years and the number of begging- rumbles that
were recorded by age, illustrating that the occurrence of
begging- rumbles declines around three years old, approxi-
mately the age when calves are beginning to be weaned.
A calf’s mother usually responds to a begging- rumble
by stopping and adopting a Suckling- Stance (see Poole and
Granli, chapter 8) whereupon calling ceases. If she denies
the calf access to her breast, begging continues or may es-
calate into a more modulated and noisy call with a whin-
ing tone that I refer to as a grumbling- rumble (see below),
a higher- pitched cry, cry rumble, roar, or roar rumble
(fi gure 9.6, panels b through f). The roars show a wide
range of structural characteristics, ranging from noisy to
tonal. Sound quality of the roaring calls is highly variable
and might be described as squealing like a pig, screeching,
roaring, shouting, yelling, crying, and even crowing like a
rooster.
Eliciting Care or Support.
Young elephants when distressed
also produce a wide range of calls including rumbles, cries,
cry rumbles, roars, roar rumbles, husky cries, and trumpets.
Distress can be defi ned broadly to include instances of being
physically hurt (e.g., pushed, poked, kicked by another ele-
phant); becoming “stuck” somewhere (e.g., as in a mud-
wallow); becoming frightened or alarmed (e.g., separation
from mother or family); or being thwarted (e.g., unable to
access the breast or other resource; see section specifi cally
low frequencies, rise sharply and descend slowly, and may
be repeated for up to 45 minutes (Poole, unpublished data).
Though typically emitted in association with a mating, fe-
males at peak estrus may call in this manner in the absence
of a mating. We refer to this pattern of calling as a post-
copulatory or estrous- rumble (Poole et al. 1988). Inter- call
duration is variable, starting at around 1,500 ms and in-
creasing with time after the mating. The calling female ini-
tiates her song at very high sound pressure levels, gradu-
ally tapering the intensity of her calling. She may reach out
repeatedly to touch her partner’s penis or his semen on the
ground, calling with more vigor at each renewed sniff. Her
calling is associated with loud, rhythmic ear- fl apping and co-
pious secretion from her temporal glands. Estrous- rumbles
have the effect of attracting (often distant) males (Langbauer
et al. [1991]; Poole 1999a; see also chapter 8).
Following a mating, relatives and associates join the call-
ing female in what we have termed a mating- pandemonium
(Moss [1983]; Poole et al. [1988]; fi gure 9.5, panels e and g).
The arrival and participation of the mated female’s family
increases her own level of excitement and calling. Adult
females, juveniles, and calves all participate, overlap-
ping the distinctive estrous- rumbles of the mated female
with powerful rumbles, trumpets, and roars of their own.
True pandemonium exists: it is diffi cult to tell who is call-
ing and whether the component calls can be classifi ed as a
particular context- type (though the estrous- rumbles stand
out as distinctive, both audibly and structurally). For now
I refer to all trumpets and roars that occur during matings,
greetings, births, and other highly social, excitable events
as social- trumpets and social- roars (see Social Integration,
below). While the social- trumpets do appear to be dis-
tinctive (see below), it may be the tempo of the repeated
estrous- rumble that distinguishes the event rather than any
detectable differences in the rumbling, trumpeting, or roar-
ing of the other participants. What is notable is that many
of the rumbles made by females other than the mated female
appear to be bimodal or multi- modal.
I used a standard DFA with 11 variables (see table 9.6)
to compare adult female reproductive calls: estrous- rumble,
female- chorus, and mating- pandemonium. The results were
signifi cant (Wilks’ lambda = 0.291, F = 2.099, P < 0.01; the
model classifi ed 77 percent correctly). The results should be
viewed with caution, however, as the data set was unbal-
anced, and due to the nature of the sample, I was not able
to use cross- validation. Three females contributed to the
estrous- calls (n = 15), 15 females to the female- chorus (n
= 21), and 4 females to the mating- pandemonium (n = 9).
Mother- Offspring
Begging. A broad range of call types may be heard in the
context of begging, including rumbles, cries, cry rumbles,
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
146 Joyce H. Poole
begging are different from those emitted by infants who are
distressed for other reasons. At this stage, I assume that they
are not, although calls that are associated with higher levels
of distress appear to be coupled with more noise.
Infants, calves, and even juveniles of either sex emit a
series of characteristic calls when they are separated from
and searching for their mothers. The body posture of these
youngsters (Head- Raising; Tail- Raising; Trunk- Curved-
Under; see Poole and Granli, chapter 8) indicates that they
feel insecure and frightened. A separated calf begins by call-
ing with a barely audible, often pulsated humming- sounding
rumble, “mm- mm- mm- mm.” The calling calf then Freezes
(see Poole and Granli, chapter 8), standing with his head
raised and ears spread, listening, presumably for an answer
to his call, and then calls again. This pattern is repeated over
and over, sometimes increasing in amplitude, until he lo-
cates his mother or other family member. The calf’s typically
rapid movement, as it searches for its family, is likely to be
the cause of the pulsating quality of the call. I refer to calls
of this quality given in this context as a separated- rumble or
lost- call (fi gure 9.8, panel f ).
about begging). Within these broad call types, there is a
wide range of call patterns. To some extent the variety re-
fl ects the developmental stage of the elephant (e.g., infant,
calf, juvenile, or even adult; see under Confl ict section), but
the level of distress is also refl ected.
When moderately distressed, infants produce husky
cries. These sounds are commonly heard in a group with
a newborn and family members respond rapidly to aid the
calling infant, touching it and rumbling softly. Husky cries
range from barely audible, low- intensity sounds to surpris-
ingly powerful abrupt and gravelly sounds. I refer to the
softer, less intense call as a low- intensity- husky- cry and
the louder type as a high- intensity- husky- cry. In the high-
intensity example illustrated in fi gure 9.8, panel a, an adult
male pushed Eudora’s hours- old infant. Infants may also
roar when highly distressed.
Older calves may cry or roar when distressed and several
such sounds are illustrated in fi gure 9.8, panels b through
e. Each of these calls elicited the support of other elephants.
I do not have large enough sample sizes of these call types
to know whether the cries and roars given in the context of
Figure 9.6 Mother–Offspring: Calls produced by calves in the context of begging. (a) Begging- rumble by Ella’s 1999 calf; (b) Mixed
roar in association with begging; (c) Mixed roar rumble by EA calf attempting to suckle; (d) Rumble mixed roar rumble by Elspeth’s
c00 while begging; (e) Cry rumble as begging; (f) Rumble cry rumble by Elaine while begging for a palm heart from mother Edwina.
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
Behavioral Contexts of Elephant Acoustic Communication 147
trumpets emitted in this context as alarm- trumpets (fi g-
ure 9.8, panel g; see table 9.10 for statistical differences
between trumpet context- types).
Receiving Comfort. Calves respond to reassurance and
comfort given by adults and allomothers by emitting two
rather different- sounding calls. When infants and calves are
approached and / or touched in a caring way by an older
member of their family, they may emit a soft, short, low-
pitched call lasting about two seconds in duration. This
rumble has an “aauurrrrr” quality, and the infant or calf
raises its head and lifts its ears as it is calling. I refer to
calls given in this context as as- touched- rumbles (fi gure 9.8,
panel h). The calling calf holds herself in a posture compa-
rable to that exhibited by older individuals in the context of
a little- greeting- rumble (see Social Integration, and chapter
8) and the quality of the associated calls is also similar. My
guess is that rumbles given in this context are precursors
to little- greeting- rumbles. Similar to little- greeting- rumbles,
calls given by calves in the context of receiving an affi liative
gesture are highly variable, ranging from rather soft unmod-
ulated tonal calls to loud, highly modulated and sometimes
rather noisy calls.
Infants and calves who are comforted by a family mem-
ber after having suffered some “injustice” (e.g., pushed,
kicked, tusked, denied access to the breast) or who ex-
perienced something frightening or distressing typically
emit a loud, noisy, and characteristically highly arched
While the rumbles emitted by separated and begging
calves look structurally similar, they are audibly distinct. I
used a stepwise DFA to test whether discrimination between
these two proposed calls was possible (table 9.7). Although
the results suggest they are signifi cantly different, with dura-
tion and bandwidth contributing most to the discrimina-
tion, the ability of the model to correctly classify calls was
not particularly good (79 percent; reduced to 60 percent in
cross- validation). Several factors play a role in the poor dis-
crimination of the model. The calls are very similar, with the
primary audible difference being the pulsating nature of the
call emitted by separated calves—something the measure-
ments were not able to discern. Second, the frequency and
duration of calls change rapidly with age in calves, thereby
confounding the training / validation data sets, particularly
considering the relatively small sample sizes and the spread
in ages of the calves. I therefore ran one fi nal test using a
balanced data set from only three individuals. Once again
the results are signifi cant, but Wilks’ lambda shows that the
model is not particularly good at discriminating between
the calls given in these two contexts (75 percent correct;
see table 9.7). Additional data are necessary to conclude
whether these are two different calls or not. For now we
can say that since they are usually audibly distinct to my
ears they probably carry specifi c information to the ele-
phants, too.
As a lost calf becomes increasingly distressed, rumbling
escalates to shrill trumpeting or load roaring. I refer to
0
10
20
30
40
50
60
70
80
<1 1<2 2<3 3<4 4<5 5<6
Age (years)
Frequency
Begging-Rumbles Individual calves
Figure 9.7 Number of begging- rumbles recorded and calves in age group in EB family.
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
Figure 9.8 Mother–Offspring: Calls produced by calf in the context of other forms of distress. (a) High- intensity- husky- cry, Eudora’s 2000
newborn, as he is pushed by an adult male; (b) Cry rumble, one- year- old infant, Elaine, stuck on her side in a depression; (c) Noisy roar
rumble, infant Explorer, after Eudora pushed him away from her hours- old infant; (d) Noisy roar rumble, Amy’s one- year- old calf as a lion
pounces upon it; (e) Noisy roar, Eudora’s 2000 newborn as it is tusked and kicked by an adult male; (f) Separated- rumble by Edwina’s 1999
calf; (g) Series of trumpets, separated Ejac as he runs to his mother; (h) As- touched- rumble, Emmet, as he is comforted by Ella and Elettra
(note: his call overlaps with a reassuring call by Elettra); (i) Baroo- rumble, Elspeth’s 2000 calf after being pushed away from an infant by Eu-
dora; (j) Eudora, coo- rumble to her infant.
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
Behavioral Contexts of Elephant Acoustic Communication 149
The distress of younger calves brings forth calling by more
participants.
A comparison of acoustic measurements, however, sug-
gests no signifi cant difference between these calls given in
these two reassuring contexts, suggesting that they should
be viewed as one context type. I therefore refer to these calls
collectively as cooing- rumbles (fi gure 9.8, panel j). A com-
parison of coo- rumbles with other common adult rumbles
is presented in table 9.10.
Some rumbles given by mothers have the immediate
effect of calling a wandering calf back to her side. Unfor-
tunately, I have very few examples of calls emitted in this
context, so they are not included in annex 9.1 or in any
analyses.
Confl ict
Aggression. During confl icts between females, especially
between females from different families or between a fe-
male and a young male, the matriarch or another large adult
female will come to the aid of a younger member of the
family. During such coalitions, the older female rapidly ap-
proaches the female in confl ict, and as she reaches her side,
they may raise their heads high and rumble in unison as they
prepare for a joint attack on a young male or on another
female(s). A female seriously threatening another elephant
stretches her head forward and outward (Bow- Neck, see
chapter 8), and as she chases her opponent, her aggressive
rumbling may be associated with audible ear- fl apping (fi g-
ure 9.9, panel a). I do not have enough examples of calls
recorded in either of these contexts to include in annex 9.1
or in any analyses, but the specifi c behavior of the elephants
when calling, particularly during coalitions, suggests an
area for further research.
Protesting. The grumbling- rumble mentioned before (fi g-
ure 9, panels b through e), with its distinct undulating or
wavy contour and whining tone, is also emitted by adults.
Grumbling- rumbles vary between one or several undula-
tions, resulting in a rather bimodal distribution of call dura-
tion, with the longer calls sounding distinctly complaining.
When adult males are on the receiving end of a serious
threat, they frequently emit a sound similar in structure to
the shorter grumbling- rumble but of rather different sound
quality. This very short, groaning rumble has the quality
of an outboard or V- 8 engine and is referred to, thus, as
a V- 8- rumble (fi gure 9.9, panel f; annex 9.1). Adult males
may produce this sound in several specifi c circumstances,
including when seriously threatened by a higher- ranking,
often musth, individual at close range (n = 3); when chased
or lunged at by a higher- ranking individual (n = 5); during
rough sparring matches (n = 5); and during rough play (n =
“wooooaaaarrrrr” or “barooo” sound. I refer to this pro-
testing sound as a baroo- rumble (fi gure 9.8, panel i).
In situations where a calf is in confl ict with another (for
example, over access to food or when one is pestering an-
other), a calf may emit a call with a distinct “whining” tone.
In such calls of longer duration, a very clear undulating or
wavy pattern is distinguishable with frequencies rising and
falling. I refer to this as a grumbling- rumble (see under
confl ict and fi gure 9.9, panels b through e). A number of
grumbling- rumbles in my collection come from the Tsavo
orphans. At night, in their stockade, without the moderat-
ing infl uence of adults, the calves pestered one another (e.g.,
repeatedly touching or sucking on one another’s ears, etc).
I used a stepwise DFA with cross- validation to test
whether discrimination between these three proposed call
types, as produced by calves, was possible (table 9.7). Since
the majority of baroo- rumbles are highly arched and the
grumbling- rumbles display a wavy contour, I included values
representing contour in the model (see table 9.1). Overall,
the model was able to discriminate between the three calls
well above the level of chance (70 percent and 67 percent
with cross- validation; table 9.7), but discriminations were
much higher for the baroo and grumbling- rumbles (73 per-
cent and 78 percent, respectively, in cross- validation) than
for the as- touched- rumbles (only 40 percent). As mentioned
before, the as- touched- rumble is highly variable, perhaps
due to the numerous situations in which a calf may be fon-
dled, some of which may be unwelcome! At least two of
these calls appear to be rather clear signals of behavioral
context with emotional state refl ected in the contour and in
the degree of modulation, noise, and intensity.
Reassuring and Calling Calves.
Calls directed toward calves
are one of the more frequently heard calls in a family group.
These calls are, by and large, rather similar soft, unmod-
ulated, low- pitched rumbles of medium duration. Many
of these calls appear to function to generally reassure and
bond with calves.
I distinguish two broad behavioral contexts in which
adult and juvenile females rumble to calves. One con-
text is when they greet, touch, suckle, or generally “coo
over” a calf in the absence of an expression of distress by
the calf. Calls emitted in this context are most often di-
rected toward infants and are especially frequent when
there is a newborn in the family. Juvenile females may also
call in this manner when they encourage infants to try to
“suckle” from them. The other context is when adult or
juvenile females are reassuring a calf following an event
that has caused the calf to give some form of distress call.
Husky cries, begging- rumbles, separated- rumbles, as-
touched- rumbles, baroo- rumbles, alarm- trumpets, cries,
and roars all elicit this type of calling by family members.
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
a.
b.
c.
d.
e.
h.
f. g.
j.i.
Figure 9.9 Calls produced during confl ict: (a) Geraldine, as she threatens Echo’s family; note loud earfl aps; (b) Grumbling- rumble, Malaika,
as keeper prevents her going where she wants; note characteristic undulating pattern; (c) Grumbling- rumble, Emily, when Imenti sucks on her
ear, note undulation; (d) A short grumbling- rumble, female during scuffl e as a male mounts Ella; (e) Grumbling- rumble by young male denied
access to salt lick; (f) V- 8- rumble, Vronsky, when sparring with Ella; (g) Trumpet, Eleanor, when mother, Erin, pushes her; (h) Mixed roar, young
male as tusked by an adult female; (i) Mixed roar, young male when pushed at salt lick; (j) Mixed roar, Echeri, as access to a palm frond is de-
nied by another elephant.
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
Behavioral Contexts of Elephant Acoustic Communication 151
bonds with individuals who are important to them. I will
describe some of the more common of these greeting and
bonding behaviors.
A relatively frequent event in an elephant family is one
in which an elephant (usually older and female) approaches
another (often in parallel and from behind). If the pair is
closely bonded, typically the approaching elephant emits a
soft and relatively short rumble to which the elephant being
approached responds by Head- Raising, Ear- Lifting, and
sometimes Backing- Toward (see chapter 8) the approaching
elephant and emitting a relatively short rumble of medium
duration and moderate intensity (although these, again,
are variable). These rumbles are associated with the onset
of temporal gland secretion, or Temporin. Examination of
spectrograms indicates that in almost every case both the
approaching and approached elephant rumble (though
the softer call of the former may be drowned out by the
more powerful call of the latter) with their calls overlap-
ping. In some cases, nearby close family may join in. Ele-
phants may also call in this manner when one approaches
another face- to- face, although this pattern is less common.
The Tsavo Orphan Keepers were able to elicit the same re-
sponse from the orphaned calves by individually calling out
their names. Members of an elephant family appear to use
this common vocal exchange as a way of saying something
like, “Hello, its good to be near you again,” or perhaps,
“You are important to me.” I refer to calls given in this con-
text as little- greeting- rumbles (fi gure 9.10, panels a and b;
annex 9.1).
Little- greeting- rumbles may occur between any mem-
bers of an elephant family, including between two males,
but they are most common between females. Of the 45
events in which Echo participated, she was mother to the
other participant in 26 events (of 6 possible pairs), grand-
mother in 9 (of 6 possible pairs), and great- grandmother
in 2 (of 4 possible pairs). Examining just those little-
greeting- rumbles, which occurred between EB females
who were over age 10 in 2000, little- greetings occurred
more often between mother- daughter, sister- sister, and
grandmother- granddaughter pairs than expected by chance
and less often between aunt- niece, great- grandmother–
great- granddaughter, or pairs who were second cousin or
greater (X2 = 107.84, df = 6, P < 0.0001; table 9.8). While
the number of little- greeting exchanges between aunt and
niece is lower than expected by chance some pairs are very
high. In each of these pairs, the niece was an infant prior
to the aunt having her fi rst calf, and the aunt was the most
likely candidate in the family for taking the role of allo-
mother.
Calves and their caretakers (often sisters and aunts)
emit the as- touched- rumble / coo- rumble exchange under
somewhat similar circumstances, and I believe that these
2). Two calls occurred in unknown circumstances. This call
appears to be produced only by males and seems to indicate
submission; the attacker usually stops when his or her (in
one case) opponent has made this call.
In an aggressive and / or competitive situation, an ele-
phant may roar when pushed, poked, tusked, chased, or
thwarted by another individual (fi gure 9.9, panels h through
j). I recorded 32 instances of such calls. Seventeen elephants
had been pushed, 5 poked or tusked, 2 chased, and an-
other 2 chased and then tusked, 2 lunged at, and 1 kicked;
3 occurred in unknown contexts. Such roars attract the at-
tention of family members who may rush to the caller’s aid
either individually or as a coalition. A roar by an individual
from another family elicits listening behavior but no assis-
tance. Roars may be longer and more pulsated if the calling
elephant is being chased.
Following an aggressive interaction, the elephant who
has been chased or thwarted may shake his or her head vig-
orously in “annoyance” and blow air through the trunk, re-
sulting in a trumpet or snort (fi gure 9.9, panel g).
Reconciliation. Vocalizations associated with reconciliatory
behavior often follow protesting calls, particularly within
a family group. Just as distressed calves are reassured by
older individuals, a juvenile or adult who roars, or emits a
protesting call in the context of received aggression, is typi-
cally approached by other members of the family and physi-
cally and vocally consoled. Vocalizing family members may
reach out to touch the aggrieved elephant who may, in turn,
respond with more grumbling- or baroo “woe- is- me”- type
rumbles. Typically, such an event causes the gathering to-
gether of several family members including the victim, the
mother of the victim, or another leading female(s), or the
matriarch as well as the perpetrator of the aggression. As
they stand together, they call with overlapping rumbles and
reach out to touch one another and the victim. This recon-
ciliatory behavior is likely to be a key to the maintenance of
close bonds in an elephant family.
Social Integration
Bonding. Greetings between members of a close social
group, including family, bond group, or, on rare occasions,
clan members, take many forms. Greeting- Ceremonies as
described by Moss (1981, 1988; see also chapter 8) are the
most intense form and typically occur when closely bonded
elephants come together after a period of separation. We
know, however, that many of the visual and tactile behav-
iors that occur during a Greeting- Ceremony may also be
seen in a wide variety of situations where elephants are not
strictly greeting one another but rather giving a show of sol-
idarity or, as with social grooming in primates, reinforcing
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
152 Joyce H. Poole
Head- Raising, Ear- Lifting, and streaming with Temporin.
Audible Rapid- Ear- Flapping (see chapter 8) may be heard
as calling explodes with a burst of powerful, throaty, highly
modulated, and overlapping rumbles that we refer to as
greeting- rumbles (Moss 1988; Poole et al. 1988; fi gure 9.10,
panel c).
The longer the separation and the closer the relation-
ship between the calling individuals, the more intense the
greeting is likely to be. A simple greeting between two indi-
viduals typically involves a few overlapping rumbles, while
the reunion of two closely bonded families may involve
calls probably develop into the little- greeting- rumble ex-
change. I venture to propose that the relationships estab-
lished through early vocal and tactile caregiving by moth-
ers, sisters, and other allomothers form the basis of close
bonds, which are then strengthened and reinforced through
the customary little- greetings between closely bonded pairs
of adults.
When members of a social group come together after
a period of more prolonged separation, they approach
one another face- to- face, and as calling begins, they
turn, “spin,” or “pirouette” to stand in parallel while
Figure 9.10 Calls produced in a bonding context: (a) Little- greeting, Enid then Echo, as Echo approaches Enid; (b) Little- greeting,
Erin then Enid, as Erin approaches Enid; note typical partial overlapping of calls; (c) Greeting- rumbles, Emerald, several more distant
elephants, Elfrida, another, and then Emerald; (d) Rumble noisy roar rumble, Ebony as she greets mother, Echo, after period of sepa-
ration; (e) Social- trumpet, member of Echo’s family during excitement surrounding birth of Ella’s 1999 calf; (f) Greeting- ceremony,
Echo’s family, lasting three minutes. Note the softer rumbles prior to and following the intense calling that occurs between 20 and 80
seconds.
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
Behavioral Contexts of Elephant Acoustic Communication 153
structurally both context- types are highly variable. As
described previously, I used stepwise DFAs with cross-
validation to compare greeting- rumbles and little- greeting-
rumbles. Since duration and fundamental frequency are
key predictor variables, and these change substantially with
age, the model, though signifi cant, was not very accurate
in discriminating between the two call types (classifi ed 73
percent correctly). Nevertheless, cross- validation using 11
randomly selected calls of each type showed that the model
was able to discriminate between the two calls at a level well
above chance (81 percent).
Elephants are highly expressive and demonstrative ani-
mals, and they vocalize loudly and in chorus under a wide
variety of different circumstances. Behavior similar to the
Greeting- Ceremony may occur following the birth of an ele-
phant, a mating, the rescue of a calf, an aggressive interac-
tion with another group, when the group has been threat-
ened in any way, or when close associates are reunited. In
all of these cases, a series of often powerful, overlapping
rumbles—including a wide variety of contours and often in-
terspersed with trumpets and roars—may be heard. It may
be more appropriate to term these collectively as “Bonding-
Ceremonies”: a signal to participants and to more distant
listeners that the callers are members of a supportive unit
and that, together, they form a united front.
Trumpeting in this context typically overlaps with the
lower frequency rumbling produced by other individu-
als. I refer to trumpets produced during social events as
social- trumpets (see table 9.11). Based on the measurements
I used, these trumpets are not signifi cantly different from
harmonic- play- trumpets, though they are signifi cantly dif-
ferent from the other proposed trumpet context- types (see
table 9.11).
Social- trumpets, in effect, function as a form of excla-
mation mark, defi ning the level of signifi cance of an event.
While the context- type of rumbling may indicate the type
of event (e.g., mating, greeting, confl ict), I propose that the
frequency, and perhaps even the placement, of trumpeting
may be an indication of the level of excitement and “impor-
tance” that the elephants collectively confer on an event. In
a sense, the use of social- trumpets can be viewed as a simple
form of syntax, qualifying the sequence of calls.
During moments of peak excitement female adults, juve-
niles, and even calves may emit powerful roars. Of the nine
roars recorded during exciting social events, all except one
were noisy roars, and all except one were combined with
a rumble, with fi ve of these being of the form rumble noisy
roar rumble (fi gure 9.10, panel d). Similar to social- trumpets,
roaring appears to signify the emotional intensity of an event.
I suspect that additional data may well reveal these roars to
be different from roars emitted during distress.
up to 50 or more elephants and last for up to fi ve minutes
(I. Douglas- Hamilton, personal communication). Dur-
ing more intense Greeting- Ceremonies (fi gure 9.10, panel
f), females may exhibit any or all of the following: Head-
Raising, Ear- Lifting, Rapid- Ear- Flapping, Mouth- Opening,
profuse secretion of Temporin, Urination, Defecation, Spin-
ning, Tusk- Clicking, and Reach- Touch (see chapter 8). As-
sociated with intense greetings and these bonding behaviors
are social- trumpets and social- roars (see below).
Prior to an elephant Greeting- Ceremony, calling may
be heard back and forth between two converging groups.
These initial calls sound like, and may be better classifi ed as,
contact- calls (see Logistical). As the groups converge, the
tone of rumbling explodes into a series of powerful, modu-
lated, and overlapping greeting- rumbles, which give way to
softer, less modulated rumbles that may continue for some
minutes (see below). These softer rumbles are structurally
quite different from the initial intense greeting- rumbles. The
more forceful initial rumbles vary enormously, although
they all show some degree of modulation in frequency
and are of medium to long duration. Unfortunately, these
powerful calls are diffi cult to measure due to their overlap-
ping nature. Within a single Greeting- Ceremony, arched,
skewed, arched with a wiggly contour, bimodal, bimodal
and skewed, and multi- modal rumbles occur. During in-
tense greetings, elephants reach out toward other individu-
als or actually touch them with their trunks (Reach- Touch,
chapter 8). While the variability in calling may simply re-
fl ect the intensity of excitement, it may, alternatively, reveal
additional information, such as a caller’s signature or per-
haps even reference to specifi c individuals, and is an area for
further research.
While the behavioral contexts associated with little-
greeting- rumbles and greeting- rumbles are very distinct,
Table 9.8 Little–greeting–rumbles: Frequencies between pairs in the EB family
Pairs
No. of
possible
pairs
Observed
no. of
events
Expected
no. of
events
Mother- daughter 14 61 21.7
Sisters 10 23 15.5
Grandmother- granddaughter 4 9 6.2
Aunt- niece 18 21 27.9
Cousins 5 3 7.8
Second cousins or greater 35 16 54.3
Great- grandmother–
great- granddaughter
1 2 1.6
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
154 Joyce H. Poole
caller is often distant and such rumbling often inaudible to
human listeners though discernible on spectrograms), ex-
plosive rumble that is preceded by an abrupt lifting of the
head as if listening. Her response is seemingly stating, “I am
over here.” The initial caller, upon hearing an answer, may
respond with another call, often associated with a more re-
laxed posture, as if sending confi rmation that an answer has
been received. Nearby family members may also add their
voice to the second or third phase of the sequence, and call-
ing back and forth may continue intermittently over hours
until the callers meet again (fi gure 9.11, panel a).
Contact- calls are used primarily between family and
bond group members, but on occasion, individuals may
be heard to answer the contact- call of a non- family or
Logistical Calls
Maintaining Contact. Elephants use powerful, modulated
rumbles at sound pressure levels of up to 115 dB (extrap-
olated to 1 m) to keep in audible contact with one an-
other over distances of up to 1 to 2 km (Poole et al. 1988;
McComb et al. 2000; chapter 10). We refer to these as
contact- rumbles or contact- calls. A sequence may include
several calls: a caller’s initial contact- rumble is associated
with rhythmic ear- fl apping and is followed by listening be-
havior; the caller’s head is held in an attentive lifted position
with ears cocked as if waiting for a response, as if query-
ing, “I am here, where are you?” An answering elephant
typically responds with an unexpected (because the initial
Figure 9.11 Calls produced in logistical context: (a) Contact- call; Eudora, Edwina, with distant elephants, note nearby and distant
exchange; (b) Contact- call, Echo; (c) “Let’s- go”- rumble, Enid; (d) “Let’s- go”- rumble, Enid; (e) Cadenced- rumble, Enid and Eleanor, part
of hour- long vocal exchange through which they work together to turn entire group around; (f) Cadenced- rumble, Eliot and Elettra, as
they move toward swamp; (g) Cadenced- rumble, Ella, during exchange with daughter Emma after she abruptly moves off.
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
Behavioral Contexts of Elephant Acoustic Communication 155
Departure Signals. An elephant indicates that she wishes
to depart by calling, by exhibiting intention movements,
and by pointing in the direction she wishes to travel using
the axis of her body. Standing intently at the edge of her
group, facing in the direction she wishes to travel, an ele-
phant emits, with slow, rhythmic ear- fl apping, a moderately
loud, low- pitched, long, relatively fl at rumble (fi gure 9.11,
panels c and d; annex 9.1). The calling elephant repeats her
“let’s- go” appeal once every minute or so, sometimes call-
ing for periods of up to half an hour, as she tries to persuade
others to depart. She may gain the support of other indi-
viduals who join in her calling, but typically this is a solo
call that we refer to as the “let’s- go”- rumble (Poole et al.
1988). The “let’s- go” is, in essence, a proposal: “I want to
go this direction, let’s go together.” It is audibly distinguish-
able from most other elephant rumbles due to its charac-
teristically long, drawn- out quality and the posture of the
calling elephant, but it can be confused in the fi eld with the
cadenced- rumble.
The “let’s- go”- rumble is one of the more commonly
heard sounds among a family of elephants. There is a ten-
dency for older females to produce this call more frequently
than younger females (table 9.9). Less commonly, juvenile
and adult males may be heard to call. Although one might
expect the matriarch to be the primary producer of this call,
as she gathers together her charges for a change in direc-
tion, this is not necessarily so. In the EB family, for example,
Echo’s second- eldest daughter, Enid, a young female with
strong leadership qualities, gave the “let’s- go”- rumble no-
tably more than anyone else (table 9.9). Females with ex-
trovert personalities (see box 13.2) are more likely to emit
“let’s- go” calls than those with introvert personalities (R
= 0.71, t (n – 2) = 2.83, n = 10, P = 0.022; Ebony was ex-
cluded since she was so much younger than the others with
personality scores (see box 13.2). While a matriarch may
propose a plan of action using the “let’s- go”- rumble, very
often she simply moves off without making any obvious
audible suggestion, except perhaps a Flap- Slide (see Poole
and Granli, chapter 8) motion with her ears (indicating a
change of activity), presumably expecting (or at least hop-
ing) others will follow without discussion! This is certainly
not always the case, and many a “discussion” and even
“disagreement” takes place regarding plans of action in an
elephant family (J. Poole, personal observation; C. Moss,
personal communication).
Decision Making. Not uncommonly, vocal exchanges
between related adult females may be heard that have the
cadence of a conversation, rising and falling, as fi rst one
individual and then another contributes her voice. Other
females may join the initiating individuals in a sequence
of low- pitched, moderate intensity, relatively fl at, long
bond- group elephant. In these cases, the calling elephants
have been members of the same clan. Contact- calls are rela-
tively frequently heard rumbles, especially during drier peri-
ods when members of a family are more likely to be spread
out over a larger area or split into sub- groups (see Moss and
Lee, chapter 13).
Contact- calls are among the best studied of all elephant
calls (see Poole et al. 1988; McComb, Moss, and Sayialel
2000; McComb, Moss, and Durant 2001; McComb et al.
2003) and are among the most powerful solo (as opposed
to chorused or clustered calls) low frequency calls. Sepa-
rated by long distances, elephants contact- call at high sound
pressure levels, but as might be expected, elephants also
call back and forth to one another over shorter distances
at lower sound pressure levels (J. Poole, personal observa-
tion; captive: Soltis, Leong, and Savage [2005a]). Though
elephants may be able to hear calls over distances of up to
10 km, information regarding individual identity probably
does not travel more than around 2.5 km and usually less
(McComb et al. 2002).
Adult females, juveniles, and calves all use contact- calls,
though, as far as we are aware, adult males in Amboseli
do not use the powerful long- distance contact- call.
Contact- calls are typically powerful, throaty rumbles that
are relatively long in duration. Most contact- calls last
between 4 and 6.5 seconds and are modulated in frequency
contour, typically rising sharply and falling more gradu-
ally (skewed left). There exists a wide range of variation
in the contours of contact- calls, however, that may be re-
lated to whether the individual is the initial caller or the one
answering. McComb et al. (2003; chapter 10) has shown
that the contact- calls of individuals are structurally distinct
and audibly identifi able to other elephants. In other words,
contact- calls contain an acoustic signature. It is possible
that considerable additional information is contained in the
different variations of the calls, perhaps related to excite-
ment level of the callers, their inter- individual distance, the
sequential arrangement (i.e. call, answer, confi rmation), or
logistical or locational information. Future research may
answer some of these questions.
Contact- calls are signifi cantly different in acoustic struc-
ture from other common calls produced by adult females
(table 9.10). They are distinctive both in quality and in
behavioral context from other calls, though they may be
confused with the onset of a greeting- rumble sequence.
This confusion is partly due to the fact that as separated
elephants approach one another, they may continue to
call back and forth; once they come into close proximity,
their calling shifts to greeting- rumbles. In these cases, it
can be diffi cult to determine when contact- calling ends and
greeting- rumbles begin, particularly as both types are loud
modulated calls.
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
156 Joyce H. Poole
Cadenced- rumbles are often heard following a series
of “let’s- go” calls (see “let’s- go”- rumble) and a series of
cadenced- rumbles may contain calls that are diffi cult to dis-
tinguish from “let’s- go”- rumbles. This may be because there
are “let’s- go”- rumbles interspersed among the other calls
or because they are one and the same call. Much additional
research is needed to clarify the temporal pattern and usage
of these calls, but based on the behavior of the elephants
the cadenced- rumble appears to represent a complex level
of consensus building between family members. It appears
as if one individual proposes a course of action (using the
“let’s- go”- rumble, for example) and then a period of ne-
gotiation and consensus building follows, ending in either
concordance or disagreement. Agreement is sometimes evi-
dent when individuals gather together rumbling with heads
raised and touching or when individuals who were sug-
gesting different directions of travel (as indicated by their
Let’s- Go- Stance, chapter 8) compromise on one direction.
Alternatively, the elephants may continue to disagree and
go separate ways.
While cadenced- rumbles sometimes appear to start
“spontaneously” (fi ve bouts calling), they are more often
associated with another event such as: group departure
(four bouts); a group member proposing departure (“let’s-
go”- rumbles; seven bouts); a group member making contact
rumbles. My 1999 and 2000 fi eld notes include 38 refer-
ences to such bouts of calling, including a total of 239 vo-
calizations (including sureness level A and B). Although this
pattern of rumbling may be heard as a solitary call (n = 16),
it is more often a series of overlapping or closely adjacent
calls interchanged over the course of several minutes and up
to an hour, with the longest uninterrupted bout measured in-
cluding 49 calls. The patterning of vocal exchange has such
a cadence of conversation (in particular, what sounds like a
higher pitch / lower pitch exchange) that I refer to this as
a cadenced- rumble (fi gure 9.11, panels e through g). These
calls may overlap, although typically the degree of overlap
is small. Adult females appear to use cadenced- rumbles to
“lend their voice to” a proposed plan of action, usually,
it seems, regarding where to go and when to depart (e.g.,
related to “let’s- go”- rumbling). As mentioned before, this
call is similar in structure to the “let’s- go,” but unlike the
“let’s- go,” I have been unable to detect any particular axis
of the body that would indicate directionality. In fact, the
call is unusual in that the vocalizing females continue to
feed (or whatever activity they were engaged in) without
looking up or showing any particularly attentive behavior. I
have heard elephants from age four and upward (including
juvenile though not adult males) participate in such bouts
of calling.
Table 9.9 Use of “let’s- go”- rumble by members >4 yrs old in the EB family 1999–2003
Elephant Sex Birth A B Total Personality
Echo—Matriarch F 1945 34 13 47 +0.7
Ella—Echo’s sister F 1965 11 10 21 +0.5
Erin—Echo’s daughter F 1969 4 2 6 –2.0
Eudora F 1972 2 4 6 –0.4
Enid—Echo’s daughter F 1982 72 21 93 +1.2
Edwina F 1982 0 –0.6
Eleanor F 1985 2 1 3 –1.0
Eliot—Echo’s daughter F 1985 11 7 18 +0.8
Emma—Ella’s daughter F 1987 2 2 +0.25
Elspeth—Eudora’s daughter F 1988 0 –0.5
Ely—Echo’s son M 1990 0 —
Esau—Ella’s son M 1990 2 2 —
Erwin—Erin’s son M 1991 0 —
Ebony—Echo’s daughter F 1994 1 1 +1.0
Elettra—Ella’s daughter F 1995 8 8 —
Echeri—Erin’s daughter F 1995 1 1 —
Europa—Edwina’s daughter F 1995 3 3 —
Eudora c96—Eudora’s son M 1996 0 —
Ejac—Enid’s son M 1997 1 1 —
Notes: Columns A and B refer to the number of calls with level of sureness A or B. Personality Z scores are from box
13.2 Component 1 “Sensitive to Insecure” and represent a measure of leadership and social integration
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
Behavioral Contexts of Elephant Acoustic Communication 157
Similar to the harmonic- play- trumpet but exhibiting a
more prolonged and resolute quality is a trumpet that is
associated with Mock- Charging behavior (see chapter 8).
Cavorting elephants often chase other species in their envi-
ronment (such as hares, hyenas, wildebeests, monkeys) dur-
ing which they trumpet loudly. We term trumpets associated
with this form of play mock- charge- play- trumpets (fi gure
9.12, panel c; table 9.11). These trumpets are signifi cantly
longer in duration than harmonic- play- trumpets.
Highly spirited lone locomotor play is characterized
by Floppy- Running (Moss 1988) behavior and trum-
peting that is expelled in a sequence of breathy pulses as
the elephant moves at a fast gait. We refer to this trum-
pet as a pulsated- play- trumpet (fi gure 9.12, panel d; table
9.11). Pulsated- play- trumpets are signifi cantly longer in
duration and more modulated in frequency contour than
harmonic- play- trumpets.
As play escalates, a fourth form of trumpet may be
heard that, as air is forced slowly through the upper part
of the nasal passages and reverberates down the length
of the trunk, sounds like a large man blowing his nose.
These trumpets are noisier and signifi cantly lower in fre-
quency than all other trumpets and are referred to as
nasal- play- trumpets (Poole and Granli [2004]; fi gure 9.12,
panel b; table 9.11).
Both nasal- play- trumpets and pulsated- play- trumpets
are associated with exuberant play. It has been my impres-
sion that elephants may be imitating the form of trum-
pet made by nearby playmates, as there is a tendency for
nasal- play- trumpets to be temporally associated with other
nasal- play- trumpets and pulsated- play- trumpets to be as-
sociated with pulsated- play- trumpets. Additional data will
be required, however, to determine whether elephants are
using vocal imitation in play.
Trumpeting by elephants is associated with high levels of
excitement ranging from exuberant play, stimulating social
events, and threatening, frightening, or startling situations.
Just by hearing a trumpet, I can make a fairly accurate as-
sessment of its behavioral context. I performed a stepwise
DFA on six proposed harmonic trumpet sub- types (table
9.11; nasal trumpets have already been classifi ed as a sepa-
rate type), including the three play trumpets previously de-
scribed and the so- called social- trumpet, alarm- trumpet, and
anti- predator- blast. The model was able to classify 60 per-
cent correctly (down to 52 in cross- validation). Though not
particularly high, the discriminations are well above the
level of chance (table 9.11).
Conclusions
Elephants produce discrete call types, but within these pri-
mary types, calls are highly graded (Leong et al. [2003];
from a distance (contact- calling, fi ve bouts); a group mem-
ber rejoining the group (greeting- rumbles; three bouts);
or the reinforcing of bonds (little- greeting- rumbles; seven
bouts). The association between cadenced- rumbles and
little- greeting rumbles may be related to a reinforcement of
bonds in an attempt to gain support for a proposed plan of
action.
As described earlier, I used two stepwise DFAs with
cross- validation to examine the commonly heard calls pro-
duced by adult females (table 9.10). In the model involv-
ing coo- rumbles, “let’s- go”- rumbles, cadenced- rumbles,
and contact- rumbles, 57 percent were correctly classifi ed
(54 percent in cross- validation). The results are well above
what would be expected by chance. Due to the conservative
nature of the model and the spread in the ages of the callers,
I think it is fair to accept the differences in the calls as real.
Play
In the context of play, calves, juveniles, and adults of both
sexes produce a variety of trumpets (Poole and Granli
2004). The majority of these are shrill harmonic sounds;
others are distinctly nasal, while still others are pulsated.
Harmonic- play- trumpets are typically short high- pitched,
high- intensity sounds associated with lone locomotor and
social play (Poole and Granli [2004]; fi gure 9.12, panel a;
table 9.11). Some harmonic- play- trumpets have a fl atter re-
verberating sound, almost like a loud nose- blow, and spec-
trographically these show more noise. Elephants of both
sexes and all ages also produce a shortened form of trumpet
during play referred to by Berg (1983) as a trump.
Table 9.10 Discriminant function analysis (DFA) results for common adult
female rumbles
Adult female rumbles
Cadenced, contact,
“let’s- go,” coo
Greeting, little-
greeting
N (training) 206 56
Eigenvalue overall 1.087 0.474 (100)
Wilks’ Lamda 0.419 0.679
F11.333 6.041
P<0.0001 <0.0001
Classifi cation expected
probabilities
0.25 0.50
Classifi cation % correct 57 73
Classifi cation cross- validation
% correct
54 81
Variables in the model F0 max- min, duration,
bandwidth, max time,
location F0 max
Duration, max time,
F0 max, maximum
frequency
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
158 Joyce H. Poole
on a better understanding of which features are most salient
with regard to signal recognition.
Clearly, some elephant context- types are very distinct,
while others are much more diffi cult for humans to discrim-
inate. The musth- rumble and the estrous- rumble, for ex-
ample, stand out from other rumbles even to the layman,
and DFA shows the more powerful contact- calls, too, to
be highly distinctive. We might expect calls, such as the
estrous- rumble and the more powerful of the contact- calls,
which function in long- distance advertisement and contact
and, thus, rely purely on acoustics, to be discrete. We might
also expect calls that defi ne reproductive state, such as the
musth- rumble and, again, the estrous- rumble, to be highly
distinctive.
“Let’s- go”- rumbles, cadenced- rumbles, and the softer
greeting- rumbles, on the other hand, are more diffi cult to
discriminate between. For calls that are given within the
close context of the family, in which animals are in vi-
sual contact and interacting with known personalities,
more subtle factors may come into play. For instance, the
“let’s- go” signal is extremely distinctive to an observer
(and to the elephants) due not purely to the audible char-
acteristics of the call but to the posture of the calling ani-
mal, the pattern of repetition, and often the emphasis and
sense of purpose with which the call is given. One can lit-
erally hear and see the impatience of the calling animal.
The cadenced- rumble, though similar in acoustic structure,
does not carry the same resoluteness in its tenor. Moreover,
the calling animal’s behavior is signifi cantly different. She
does not stand facing away from her family, waiting in a
Soltis, Leong, and Savage [2005b]; this chapter). Many spe-
cies have repertoires that are acoustically graded yet are
perceptually discrete (Hauser 1997a). Elephant calls, like
human speech, exhibit gradedness on the production end
but discreteness on the perceptual end. The degree to which
within- call type distinctions are made by elephants will ul-
timately depend upon their signifi cance to reproduction and
to survival. Our ability to make the sort of acoustic discrim-
inations that elephants are clearly able to make will depend
Table 9.11 Forward stepwise discriminant function analysis of six
trumpetsubtypes
Harmonic trumpet subtypes
Harmonic- play, pulsated- play,
mock- charge- play, alarm,
anti- predator- blast, social
N (training) 136
Eigenvalue overall 0.794
Wilks’ Lamda 0.262
F4.839
P<0.0001
Classifi cation expected probabilities 0.167
Classifi cation % correct 60
Classifi cation cross- validation % correct 52
Variables in the model Duration, F0 max, curve, F0 max- min,
bandwidth, location F0 max, max
time, and peak time
Note: The model did not consider the pulsated nature of the pulsated- play- trumpet.
Figure 9.12 Calls produced in the context of play: (a) Harmonic- play- trumpet; (b) Nasal- play- trumpet; (c) Mock- charge- play- trumpet,
juveniles while chasing a hare; (d) Pulsated- play- trumpet.
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
Behavioral Contexts of Elephant Acoustic Communication 159
and less noisy with time. Comparing individual calls from
the beginning of a greeting or immediately following a mat-
ing with those at the end reveal enormous differences. Are
elephants able to acoustically recognize each individual call
as being related to a greeting, a mating, or the arrival of a
musth male? Or are they, like us, only able to recognize the
greeting in the entirety of the sequence by its overall pattern
of calling?
Irritation and anger are also clearly encoded in elephant
calls, as described by Soltis et al. (2009). The calls of com-
plaining or protesting elephants, such as grumbling- rumbles,
have an audibly whining tone. The spectrograms of these
calls show a rising and falling frequency contour and
patches of noise. It is this undulating frequency and alter-
nating noisy and tonal sound that gives these calls their dis-
tinctive complaining tone. The noisy component is also evi-
dent on other complaining calls such as the baroo- rumble
and the V- 8- rumble.
If the level of excitement and anger can be heard in the
voices of elephants and seen in the spectrographic struc-
ture of calls, what else does the infi nite variability in the
frequency contours of calls encode? Greeting- rumbles or
bonding- rumbles, in particular, show an extreme range in
the frequency contour of calls. They may be fl at, slightly
arched, highly arched, bimodal, multi- modal, skewed left or
right. What does all this variability signify to the elephants?
Considering that elephants are capable of vocal produc-
tion learning and have unusual vocal tract fl exibility and
a large resonance chamber, we should not be surprised
that elephants have the capacity for producing such a wide
variety of sounds. Elephants are intelligent and complex
social animals; the increasing understanding of elephant
communication generated by this chapter, together with the
very signifi cant research of my many colleagues refl ects this
intricacy. As noted by Soltis (2009) and seen clearly in this
chapter, the separation of elephant call types into sub- types
can take many forms. If individual identity, age and body
size, relative dominance, level of arousal or affect, and spe-
cifi c behavioral context all contribute to the shape a call
takes, it is easy to understand why a rumble given by one
individual in the context of a Little- Greeting, for example,
might take many forms.
In this chapter, I have presented the broad range of con-
texts in which African elephants are heard to vocalize and
endeavored to make a fi rst attempt to discriminate between
some of the more common calls that elephants make in dif-
ferent behavioral contexts. The discriminations made thus
far are based on rather limited and fundamental structural
measurements. In the future, more detailed observations,
measurements, and analyses are likely to reveal a highly ver-
satile repertoire of calls matching the multi- faceted interac-
tions and relationships observed among wild elephants.
determined and impatient manner, but is, in activity and
posture, integrated with the others. The call is not repeated
every minute or so but is used in a vocal exchange with an-
other or several other elephants. Accordingly, we can defi ne
these as two discrete signals since the message is perceptively
distinct. And while the structural differences between the
calls may not always be obvious to us at this stage, the ele-
phants appear able to distinguish between them. It is likely
that the difference is one of tenor, a subtle shift in emphasis.
Two other calls that may be confused based on acous-
tic structure (both being calls short in duration) are the
softer versions of little- greeting- rumbles and coo- rumbles.
Little- greeting- rumbles are calls given in exchange between
two closely bonded individuals, most commonly between
mother and daughter or between sisters, and are typically
composed of a softer fl atter call and a louder, more mod-
ulated response. Mothers or allomothers (who are often
older sisters) give the soft, fl at coo- rumble as a small greet-
ing or reassurance to calves who respond, similarly, with
a louder, modulated call. The adult- calf / juvenile- calf pairs
would be the same pairs likely to engage in little- greetings
later in life. Thus, it seems likely that these calls, a pair of
little- greeting- rumbles, and the coo- rumbles and the calf
responses, as- touched- rumbles, are related. The bonds
formed as calves through this tactile and vocal exchange
with older individuals create an enduring relationship sus-
tained through acoustic exchange and mutual cooperation.
Arousal level is clearly encoded in elephant calls, as
was fi rst noted by Berg (1983). Since then, detailed anal-
yses have characterized the expression of affect in the
structural patterns of rumbles (Soltis, Leong, and Savage
2005b; Soltis et al. 2009; Li et al. 2007; Wood et al. 2005;
Stoeger- Horwath et al. 2007; Wesolek et al. 2009). Soltis
et al. 2009 focused on dominance interactions, showing that
the calls of females tended to contain higher, more variable
fundamental frequencies when they were the lower ranking
of a pair interacting aggressively. The fact that elephants’
voices refl ect affective state helps to explain the broad vari-
ability in the calls of individual females recorded in similar
behavioral contexts. For example, although coo- rumbles
and calf responses and little- greeting- rumbles are affi liative
gestures, they are typically given between pairs of unevenly
ranked individuals, and usually the younger individual’s call
is of higher and more variable fundamental frequency and
of higher amplitude.
Estrous- rumbles, greeting- rumbles / bonding- rumbles,
mating- pandemonium, and roaring- rumbles (heard when
elephants are mobbing predators) are all marked during the
peak of excitement by increased amplitude, increased noise,
and increased modulation, with energy distributed in the
upper harmonics (rather than in the second harmonic as in
most rumbles), with calls becoming softer, less modulated,
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
ANNEX 9.1 BEHAVIORAL CONTEXTS AND ASSOCIATED CALL AND CONTEXT TYPES
Behavioral
context
Call
type
Call
sub- type Manner features Duration (s) F0 max (Hz) F0 min (Hz)
Max frequency
(Hz) Bandwidth (Hz)
Anti-
predator
Rumble
Comment Tonal; fl at 3,438 (2,636–3,808)
2,501–4,129; 5
15.2 (15.1–15.4)
13.8–19.0; 5
13.0 (13.0–13.8)
12.2–16.8; 5
29.3 (29.3–29.3)
29.3–29.3; 2
166 (143–183)
160–180; 2
Anti-
predator
Tonal, upper har-
monics noisy;
arched
4,081 (3,273–6,233)
2,156–11,781; 32
21.8 (18.7–26.6)
14.2–33.7.4; 32
15.2 (12.4–19.0)
10.3–24.6; 32
35.2 (29.3–
128.0)
15.6–154.3; 24
253 (211–560)
162–701; 30
Roar Anti-
predator
2,538
2
290
1
235
1
215
2
4,435
1
Trumpet Blast Noisy harmonics 1,269 (952–1,398)
583–2,188; 18
533 (476–658)
368–757; 18
345 (322–393)
312–498; 18
668 (457–949)
293–2,028; 17
5,004 (4,295–5,595)
2,940–6,449; 8
Mating
Rumble
Musth Tonal with noise 4,723 (4,054–5,160)
1,739–8,744; 76
17.1 (16.4–17.9)
12.5–28.2; 77
11.7 (10.5–12.3)
8.4–14.0; 77
35.2 (33.2–35.2)
25.4–169.9; 70
124 (103.0–155)
55.0–417.0; 77
Female-
Chorus
Tonal; arched 3,797 (3,210–5,304)
1,958–7,107; 25
19.1 (17.7–21.4)
14.7–39.2; 25
14.7 (13.8–17.4)
10.8–25.8; 25
29.3 (26.4–34.2)
13.7–58.6; 20
221 (168–297)
107–562; 20
Estrous
Tonal, upper har-
monics noisy;
highly arched
skewed left
5,667 (4,777–6,060)
4,367–8,036; 14
24.8 (22.5–26.8)
18.3–34.9; 18
17.3 (15.1–17.4)
12.3–17.7; 18
129 (127–131)
35–203; 5
376 (337–400)
180–929; 18
Mating
Pandemo-
nium
Tonal, upper har-
monics noisy; vari-
able
4,152 (3,367–4,933)
1,957–8,657; 22
18.0 (16.1–18.8)
14.8–22.8; 21
14.0 (13.4–16.1)
10.5–17.9; 22
32.2 (27.3–36.2)
25.4–121.1; 16
294 (235–419)
177–818; 20
Roar Estrous Variable pulsated
mixed noisy
2,377 (1,852–3,418)
566–3,506; 11
197 (160–362)
153–602; 10
131 (103–174)
96–458; 11
172 (140–430)
107–585; 11
1,483 (1,184–3,746)
352–4,200; 11
Parent–
Offspring
Rumble
Begging Tonal; primarily
fl at–descending
1,659 (1,218–2,256)
511–4,151; 139
24.4 (22.5–26.3)
16–33.5; 137
22.4 (20.7–23.5)
12.7–28.1; 139
46.9 (43.0–49.7)
29.3–242.2; 134
188 (139–283)
63–645; 139
Separated Tonal; fl at–de-
scending
2,480 (2,070–3,250)
621–5,163; 98
24.8 (21.4–26.2)
16.1–36.0; 97
21.8 (18.5–23.4)
13.4–27.8; 98
44.9 (39.1–49.7)
29.3–288.1; 98
162 (123–228)
63–670;98
As- touched Tonal and with
noise; variable
1,544 (1,138–2,185)
415–3,404; 26
28.5 (25.2–33.1)
18.9–48.5; 26
22.8 (20.5–24.5)
17.2–34.1; 26
54.7 (48.8–66.4)
43–238.7; 26
337 (294–405)
145–872; 25
Baroo Tonal with noise;
arched
2,037 (1,514–2,727)
630–6,752; 37
32.2 (27.3–36.3)
16.1–53.7; 37
24.8 (21.5–28.0)
11.4–32.8; 37
63.9 (55.6–81.1)
27.3–303.1; 36
418 (296–700)
95–1,000; 37
Coo Tonal; contour pri-
marily fl at
3,388 (2,695–4,404)
1,113–6,811; 167
16.4 (14.7–18.0)
12.0–24.8; 163
14.5 (12.8–16.0)
8.7–21.8; 166
31.2 (27.3–34.7)
15.6–91.8; 164
186 (143–226)
38–684;166
Husky cry
Low-
intensity
Tonal with noise 735 (637–940)
328–1,282; 18
35.5 (32.0–40.4)
25.1–54.1; 17
31.4 (27.5–33.2)
19.5–50.4; 18
62.5 (50.8–74.2)
35.2–101.6; 18
584 (408–693)
276–946; 16
High-
intensity
Tonal with noise 1,245 (891–1,458)
770–1,544; 13
41.6 (37.6–44.4)
33.9–51.0; 13
32.1 (28.0–33.9)
23.4–41.3; 12
88.0 (74.2–93.8)
66.4–414.1; 13
600 (571–800)
391–1,000; 13
Roar
Suckle
protest
Mixed, noisy, tonal 1,410 (994–1,845)
606–3,004; 16
459 (414–651)
172–727; 16
337 (205–443)
109–658; 16
345 (366–858)
301–2,093; 16
4,582 (3,182–7,102)
2,070–8,834;12
Separated Mixed, noisy, tonal 2,055 (1,852–2,329)
467–3,564; 23
370 (279–447)
236–568; 23
278 (205–304)
163–526; 23
391 (311–632)
214–2,062; 23
2,968 (2,418–4,117)
1,816–5,143; 23
Cry Protest Tonal; descending 314 (218–453)
107–528; 21
454 (337–522)
183–800; 21
276 (250–315)
159–547; 21
388 (279–642)
236–1,273; 20
3,198 (1,745–5,460)
724–6,968;19
Trumpet Alarmed Noisy harmonics,
arched
620 (491–701)
435–830; 8
687 (656–737)
625–791; 8
458 (437–458)
390–469; 8
761 (691–1,044)
624–1,317; 8
9,425 (5,538–11,133)
4,811–11,235; 8
Confl ict
Rumble
Grumbling Tonal with noise;
wavy
2,142 (1,520–3,390)
816–7,434; 60
24.6 (19.7–28.5)
14.8–39.2; 60
19.8 (17.1–23.0)
12.4–27.8; 60
46.9 (35.2–62.5)
15.6–246; 60
302 (208–423)
92–1,000; 60
V- 8Tonal with noise 1,271 (1,103–1,738)
994–1,806; 6
17.8 (15.3–24.0)
15.1–28.1; 4
13.2 (19.2–14.5)
11.2–23.4; 4
32.2 (15.1–86.4)
14.6–124; 4
424 (316–450)
308–661; 6
Roar
Agonistic
and rough
contact
Noisy, mixed, pul-
sated, tonal
1,422 (1,105–2,056)
330–5,419; 40
362 (251–448)
108–776; 31
260 (168–314)
39–507; 35
477 (332–679)
172–2,849; 39
3,688 (2,733–5,000)
1,583–8,886; 35
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
Behavioral
context
Call
type
Call
sub- type Manner features Duration (s) F0 max (Hz) F0 min (Hz)
Max frequency
(Hz) Bandwidth (Hz)
Social
integration
Rumble
Little-
greeting
Tonal with noise;
variable
3,370 (2,571–3,997)
1,206–7,265; 167
18.2 (15.9–22.0)
12.1–30.9; 167
15.1 (13.1–18.6)
10.4–22.9; 166
35.2 (29.3–43.0)
19.5–209.0; 157
207 (173–250)
55–712; 160
Greeting Tonal with noise;
variable
4,801 (4,079–5,953)
1,743–10,294; 123
16.9 (15.0–20.0)
11.7–40.2; 122
12.9 (12.0–15.4)
9.9–26.8; 123
31.2 (27.3–37.1)
15.6–231; 85
250 (215–389)
135–991; 102
Bonding Tonal with noise;
variable
2,842 (2,451–3,759)
1,479–4,446; 11
24.4 (19.7–27.1)
18.7–35.3; 12
18.3 (15.2–19.7)
13.2–21.1; 12
168 (160–178)
29.3–207; 7
301 (261–564)
156–628;9
Contact call
Variable tonal and
with noise; variable
arched
5,008 (4,035–6,329)
1,844–9,478; 180
19.4 (17.1–21.7)
13.7–31.0; 174
13.4 (12.3–15.0)
10.3–25.5; 180
35.1 (29.2–39.5)
15.6–218.8; 174
253 (171–470)
57–1,015; 175
“Let’s- go”Tonal; fl at, slightly
modulated
5,234 (4,308–6,521)
1,164–9,229; 123
15.6 (14.7–16.8)
12.8–21.3; 123
12.5 (12.0–13.4)
9.6–19.6; 123
27.8 (26.3–29.3)
14.6–40.9; 122
228 (178–246)
41–615; 113
Cadenced Tonal; fl at, slightly
modulated
5,106 (4,243–5,624)
3,036–7,675; 84
15.3 (14.4–16.4)
12.6–22.1; 84
12.8 (12.2–13.6)
11.0–18.0; 84
27.8 (26.3–29.3)
13.7–113.0; 70
190 (159–237)
42–645; 76
Harmonic
Trumpet Social Noisy harmonics 670 (469–936)
273–1,951; 70
394 (349–450)
179–841; 68
276 (255–321)
138–606; 68
453 (323–818)
132–2,019; 69
9,700 (6,450–11,000)
1,274–19,000; 70
Roar Social
Noisy, mixed 1,755
(1,280–1,891)
1,055–2,112
9
341
(239–419)
220–624
8
233
(148–260)
139–452
9
378
(253–483)
172–619
8
3,185
(2,152–4,640)
1,556–5,160
8
Play
Trumpet
Mock–
charge
Noisy harmonics 1,116 (717–1,461)
536–2,559; 8
521 (458–581)
375–600; 7
386 (317–417)
90–464; 8
450 (424–632)
329–1,009; 8
6,794 (3,733–10,094)
3,036–11,128; 8
Pulsated Noisy harmonics;
pulsated
1,494 (956–1,712)
215–3,385; 29
460 (407–554)
336–732; 29
333 (301–387)
124–519; 29
460 (373–878)
105–1,324; 29
6,636 (4,802–9,086)
1,326–11,256; 29
Harmonic Noisy harmonics;
arched
546 (405–765)
198–2,486; 65
407 (371–478)
178–790; 61
342 (271–380)
108–496; 65
415 (345–604)
108–2,405; 63
7,200 (4,600–10,900)
1,600–32,500; 65
Nasal
trumpet Nasal- play
Noisy harmonics;
fl at slightly modu-
lated
676 (576–809)
445–2,426; 19
101 (67–144)
34–329; 16
61 (46.6–107)
22.7–122.9; 19
248 (132–306)
32–563; 19
3,729 (2,256–4,927)
786–10,425; 19
Notes: Manner features list sound quality and primarily observed contours. Figures represent median (interquartile range) range; n.
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal
and injures the author and publisher.
