vailable online at www.sciencedirect.com
Rumble vocalizations mediate interpartner distance in African
elephants, Loxodonta africana
KATHERINE A. LEIGHTY, JOSEPH SOLTIS, CHRISTINA M. WESOLEK & ANNE SAVAGE
Education & Science, Disney’s Animal Kingdom
(Received 30 January 2008; initial acceptance 11 March 2008;
ﬁnal acceptance 7 May 2008; published online 29 August 2008; MS. number: A08-00062R)
The ability to utilize contact calls to facilitate reunions with social partners has been documented in a num-
ber of species showing a ﬁssion/fusion social organization. Field observations and playback experiments
suggest that African elephants use low-frequency rumble vocalizations to reunite with their herd members
following periods of ﬁssion. Using a digital audio and GPS recording collar system, we documented the
production of rumbles and subsequent movements of ﬁve adult female African elephants at Disney’s
Animal Kingdom, Bay Lake, Florida, U.S.A. This recording system allowed us to identify the producer of
each rumble and to document the effect of rumbles on the movements of herd members relative to the
caller. Our ﬁndings provide the ﬁrst empirical evidence that spontaneously produced elephant rumble vo-
calizations function in part to mediate the spatial relationships of group members. Overall, the production
of rumbles resulted in a net decrease in distance between the caller and her social partners. This approach
behaviour was enhanced if the partner was highly afﬁliated with the caller, if the partner replied with
a rumble of her own, and if the pair was initially far apart (61 m). Rumble production was likely to result
in avoidance behaviour only when there was no rumble reply by the partner and the dyad was close
together prior to the initial call. These results suggest that a general function of elephant rumbles is to
promote spatial cohesion among separated group members, but they may also mediate a variety of other
close-distance social interactions.
Ó2008 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
Keywords: African elephant; antiphonal exchange; contact call; ﬁssion/fusion; Loxodonta africana; rumble; vocal
The vocalizations of animals can function to mediate
proximity both within and between social groupings.
Speciﬁcally, the production of sounds by socially living
animals can be used to pronounce territorial boundaries,
advertise sexual states, alert others to the presence of
predators, note the location of resources and promote
group cohesion (reviewed in: Boinski & Campbell 1996;
Boinski & Garber 2000). As social groups move within
their environment, individuals can become spatially sepa-
rated and move beyond the range of visual contact. Con-
tact call vocalizations provide a means by which animals
can maintain contact with group members over both
short and long distances (Oda 1996; Ramos-Fernandez
2005). Studies have documented the presence of contact
calling in a wide variety of species. These observations
note an attraction of group members to contact calls, an
increased likelihood of contact calling when animals
become separated from their group and/or as the group
becomes more dispersed, and an increased frequency of
contact calling during reunions as compared to disper-
sions (zebra ﬁnches, Poephilia guttata:Butterﬁeld 1970;
Carolina chickadees, Poecile carolinensis:Smith 1972;
greater spear-nosed bats, Phyllostomus hastatus:Wilkinson
& Boughman 1998; mongooses, Helogale parvula:Estes
1991; golden brown mouse lemurs, Microcebus ravelobensis:
Braune et al. 2005; ringtailed lemurs, Lemur catta:Oda
1996; squirrel monkeys, Saimiri oerstedi:Boinski 1991).
Perhaps the clearest examples of spatial separations of
social groups occur within ‘ﬁssion/fusion’ societies.
These groups show ﬂuid association patterns while
group members maintain permanent social relationships
Correspondence: K. A. Leighty, Animal Programs Administration, P.O.
Box 10000, Lake Buena Vista, FL 32830, U.S.A. (email: katherine.
0003e3472/08/$34.00/0 Ó2008 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
AN IM AL BE HA VI OU R , 2008, 76, 1601e1608
(Ramos-Fernandez 2005). This social system is typiﬁed by
groups breaking into smaller subgroupings to seek out dis-
persed resources and then reuniting after a period of time.
Contact calls appear to be utilized by many species living
in ﬁssion/fusion societies to facilitate this reunion or ‘fu-
sion’ of their subgroupings. In primates, Guinea baboon,
Papio papio, subgroups have been observed to exchange
bark vocalizations when dispersed on the savanna until
the point in time when they reunite (Byrne 1981). Male
chimpanzees, Pan troglodytes, are more likely to produce
long-distance pant-hoot vocalizations when preferred
social partners are nearby (but out of visual range) rather
than absent or in close proximity (Mitani & Nishida
1993). The ‘whinnies’ of spider monkeys, Ateles geoffroyi,
have been observed to be exchanged by afﬁliated individ-
uals when their subgroups are approaching (Teixidor &
Byrne 1999; Ramos-Fernandez 2005). In the ﬁssion/fusion
societies of social carnivores, contact calls are produced
less frequently, perhaps to avoid alerting prey species to
their presence (Holekamp et al. 2000). Lions, Panthera
leo, and coatis, Nasua nasua, will increase contact call rates
when travelling and foraging in conditions of low visibil-
ity (Kaufmann 1962; Schaller 1972). Spotted hyaenas,
Crocuta crocuta, on the other hand are highly vocal and
their whoop vocalization serves as a long-distance contact
call to which offspring and relatives quickly approach
(Estes 1991). Finally, among marine mammals, dolphins
are thought to use whistles to maintain contact with
pod members (Tursiops truncates:Caldwell & Caldwell
1965;Stenella longirostris:Lammers et al. 2006).
African elephants show a classic ﬁssion/fusion social
system. They live in matriarchical family groups of related
adult females and their dependent offspring. These groups
frequently and consistently associate with other family
groups to form larger ‘bond groups’ (Douglas-Hamilton
1972; Moss & Poole 1983; Wittemyer et al. 2005; Wit-
temyer & Getz 2007). The movements of family groups
of a shared bond group have been documented to be coor-
dinated, even when separated by several kilometres on the
savanna (Martin 1978; Poole et al. 1988). These groups
show parallel movement, synchronous direction changes
and mutual approach over large distances (Martin 1978).
It has been proposed that family groups coordinate their
movements by using low-frequency ‘rumble’ vocalizations
as contact calls (Poole et al. 1988; Langbauer et al. 1991).
Rumbles are the most frequent vocalization of elephants
and their structure affords long-distance propagation
(Langbauer et al. 1991; Langbauer 2000). Elephant rum-
bles are termed ‘infrasonic’ because their fundamental
frequency ranges between 13 and 35 Hz (Poole et al.
1988) and they experience little environmental attenua-
tion because of their low-frequency nature. In fact, ﬁeld
experiments have documented that elephants detect and
respond to playbacks of rumble vocalizations at distances
of over 2 km (Langbauer et al. 1991; McComb et al. 2003).
Furthermore, McComb et al. (2003) found that elephants
are capable of recognizing a rumble as that of a bond or
family group member from distances of 2.5 km, although
discrimination was more accurate at 1e1.5 km.
Detailed study of the functions of elephant rumbles has
been challenging in the wild because the infrasonic nature
of these vocalizations makes it nearly impossible for
human observers to detect and localize the source of
individual rumbles (Langbauer 2000). The use of a collar-
mounted audio-recording system in the captive herds at
Disney’s Animal Kingdom, Bay Lake, Florida, U.S.A., has
allowed for both the detailed description of the acoustic
characteristics of these vocalizations and the determina-
tion of the identity of each rumble’s producer (see Leong
et al. 2003; Soltis et al. 2005a). Early analyses of data cap-
tured via this audio collar system revealed the presence of
clusters of rumble vocalizations (Leong et al. 2003). These
rumble clusters consisted of antiphonal call and reply
sequences between social partners (Leong et al. 2003;
Soltis et al. 2005a). These antiphonal sequences were
most likely to take place between individuals that were
highly afﬁliated (deﬁned by time spent in close proxim-
ity), even when one individual had moved out of the vi-
sual range at the time of the vocal exchange (Soltis et al.
2005a; Leighty et al. 2008). Furthermore, both Clemins
et al. (2005) and Soltis et al. (2005b) found that adult
females produce acoustically distinctive rumble vocaliza-
tions. Thus, these ﬁndings appear to complement observa-
tions of vocal recognition made in the wild, in that
elephants have acoustically distinct rumbles and they
preferentially respond to the rumbles of their afﬁliated
partners, even when they are out of visual contact
(McComb et al. 2003).
In the present study we examined the effect of elephant
rumble vocalizations on the spatial relationships of group
members. Each member of our herd was outﬁtted with
a digital audio-recording collar that documented its
rumble production as well as its subsequent movements
using an on-board Global Positioning System (GPS). We
sought to determine whether the production of rumble
vocalizations resulted in attraction or avoidance, or had
a random effect on the subsequent distance from other
group members. We explored three factors that were likely
to inﬂuence proximity after a rumble was produced: the
partner’s degree of afﬁliation with the caller, the pro-
duction of an antiphonal reply rumble by the partner and
the initial interpartner distance of the dyad prior to the
call. Data presented here represent the ﬁrst systematic
study of spontaneously produced elephant rumble vocal-
izations and their inﬂuence on the subsequent move-
ments of group members.
Subjects and Housing
A herd of ﬁve unrelated adult female African elephants
(three with dependent calves), housed at Disney’s Animal
Kingdom, Bay Lake, FL, U.S.A., served as subjects of this
study (estimated ages 20e27 years) (Table 1; also see Soltis
et al. 2005a). Data presented here were collected between
January and November 2006. Animals were maintained in
two connected outdoor yards (1.639 ha and 2.321 ha) as
a single herd during the day and returned to the barn at
night. These naturalistic outdoor yards contain pools, in
which multiple adult elephants can fully submerge, mud
wallows, multiple scratching surfaces, and large rock rings
ANIMAL BEHAVIOUR, 76,5
to allow animals to move out of visual contact of other
Audio and GPS Data Acquisition
Each of the ﬁve adult females was outﬁtted with a collar-
mounted data acquisition system. The system was de-
signed in 2005 by Brian Walter’s of Walt Disney World
Ride & Show Engineering (Bay Lake, FL, U.S.A.) and
constructed by Acumen Instruments Corporation (Ames,
IA, U.S.A.). The collars were constructed of ﬁre hose (2.24e
2.77 m) and weighed between 4.0 and 4.5 kg. Collars were
fastened using a magnetic lock and key system that pro-
vided durability while allowing for easy removal. Mounted
to each of the collars was an 18 12 10 cm metal water-
tight containment box that housed both digital audio and
GPS recording units powered by two serial 6 V lead-acid
batteries. The recording of audio and GPS data was con-
trolled via a microprocessor and a compact ﬂash card
that also served as the data storage device. One hour of
synchronized audio and GPS data was collected between
1200 and 1700 hours, two times per week, with a total
of 40 observation hours used in these analyses.
Audio data was recorded with a Knowles MR-8406
waterproof microphone that was exposed to the air in
an opening in the metal containment box. The acquired
audio signal was run through an ampliﬁer and an 8-pole
Butterworth low-pass ﬁlter with a cut-off frequency of
5 kHz. The signal was then digitized at a sample rate of
10 kHz (Texas Instruments ADS320 audio-digital con-
verter) and saved as an uncompressed wav ﬁle to the
compact ﬂash card. The audio device had a ﬂat frequency
response (3 dB) between 8 Hz and 4.5 kHz. The GPS
device (Xemics no. RGPSM002, Camarillo, CA, U.S.A.)
acquired data via a satellite antenna embedded in the
ﬁre hose with the internal receiver positioned on the
back of the elephant’s neck. The device was programmed
to collect location data every 10 s during the observation
period (accuracy 5 m).
All elephants were trained using operant conditioning
to engage in collaring and decollaring procedures. Collars
were put on the elephants before their release into the
outdoor yards as part of their morning husbandry routine
and removed at night to allow for battery recharging.
Upon removal of the collar, we downloaded the contents
of each compact ﬂash card for analysis. This research was
approved by Disney’s Animal Care and Welfare Commit-
tee/IACUC in July of 1998 with the most recent adden-
dum approved in April of 2005.
Video Documentation of Behaviour
During each 1 h observation period, four camera per-
sons were positioned around the elephant yard to record
behaviour. At the onset of each observation, an airhorn
was sounded, and subsequent GPS time was recorded to
allow for the synchronization of the video, audio and
GPS signals during data analysis. Each camera (Panasonic
MiniDV camcorder no. PV-DV402) was then assigned a fo-
cal animal based on its location relative to each subject in
the enclosure. Cameras remained on the focal animals
until the animals moved out of view, at which point the
cameras were switched to alternate, visible animals. All an-
imals were recorded for approximately equal periods over
the course of the study. At the conclusion of each observa-
tion period, the digital video from each camera was trans-
ferred to computer and rendered as a quad-split digital
video ﬁle using Adobe Premiere Pro (Adobe Systems Incor-
porated, San Jose, CA, U.S.A.; version 2.0) to allow for
improved viewing of the herd during behavioural analysis.
An instantaneous scan sampling technique with a 5 min
interval was used on each of the 40 1-hour quad-split
videos to determine the overall afﬁliation of each dyad.
Afﬁliation was deﬁned as the percentage of time that the
dyad spent in proximity. Individuals were deﬁned to be
in ‘proximity’ when they were within 8 m (approximately
two elephant body lengths) of each other (see Soltis et al.
2005a). The percentage of time spent in proximity was de-
ﬁned as the number of scans in which the dyad was in
proximity divided by the total number of scans in which
both members of the dyad were visible on the quad-split
video, multiplied by 100. The total number of samples in
which dyads were visible ranged from 299 to 333.
Rumble Localization and Identity
Rumble vocalizations were identiﬁed via visual inspec-
tion of the audio streams downloaded from the collars in
Adobe Audition (Adobe Systems Incorporated; version
1.5). The identity of the producer of the vocalization
was determined in two ways. First, rumbles that appeared
in the audio stream of a single collar were assigned as
being produced by the animal wearing that collar. Second,
since data collection on all collars was synchronized via
the satellite clock acquired by the GPS system, rumbles
that were recorded by multiple collars were assigned to the
individual on whose collar the vocalization had the
highest relative amplitude.
Antiphonal exchanges of rumble vocalizations were
deﬁned using a 30 s window between the offset of the ini-
tial call and onset of the antiphonal reply (see Soltis et al.
2005a; Leighty et al. 2008). In these analyses, we only
used rumbles that were answered by at least one other
group member so that the effects of rumbles that were
answered could be compared to nonanswered rumbles.
Detailed examination of these antiphonal exchanges was
Table 1. Subject information
Date of arrival
Thandi 1981 Zimbabwe 25 Jun 1997
Moyo 1981 Zimbabwe 25 Jun 1997
Fiki 1979 Zimbabwe 4 Oct 1997
Vasha 1986 Zimbabwe 7 Nov 2000
Donna 1984 Africa 24 Nov 2003
LEIGHTY ET AL.: VOCALIZATIONS MEDIATE INTERPARTNER DISTANCE 1603
carried out by randomly selecting 30 initial calls from each
female (with the exception of Vasha who provided 20
because her rumbles were less frequently answered) and
documenting the females that produced an antiphonal
reply to that call within the deﬁned 30 s window. The
140 initial calls selected for analysis resulted in a total of
187 unique antiphonal replies across group members.
Movements Surrounding Antiphonal
We used GPS technology supplemented with digital
video data to assess the movements of all females in the
herd relative to the producer of the initial call. To do this,
we documented the GPS reading of all herd members 60 s
before the production of the initial call, at the onset of the
call, as well as 60 and 120 s following this call. We then
calculated the Euclidean distance in metres between the
producer of the initial call and all other females at each
of these time markers. In those instances in which a GPS
unit did not have a valid reading at the necessary time
marker (i.e. poor satellite connection due to dense cloud
cover or positioning underwater or beneath deep cover),
we determined the interanimal distance using the quad-
split digital video ﬁles. In these cases, animals were located
on the playback screen and their positions were mapped
to a geo-referenced aerial image of the enclosures from
which distance could be extracted.
To quantify approach and avoidance movements within
each caller/partner dyad, we calculated a movement ratio
for each dyad: Movement Ratio ¼Interpartner distance
60 s after initial call/Interpartner distance at initial call on-
set. The same calculation was made using the interpartner
distance observed 120 s after the initial call. These ratios
give the proportional distance between partners at 60
and 120 s following the initial call relative to the distance
between them at the time the call was produced. For
example, doubling the initial distance yields a score of
2.0 and halving the distance yields a score of 0.5 (no
change ¼1.0). We performed a negative log-tranforma-
tion (log(x)(1)) of these ratios that resulted in the follow-
ing desired properties: distributions became normal,
increases in interpartner distance yielded negative values
(‘avoidance’), decreases in interpartner distance yielded
positive values (‘approach’), and avoidances and ap-
proaches of equal magnitude received identical scores
with opposite signs (e.g. doubling the distance results in
a movement ratio of 0.301 and halving the distance
yields an movement ratio of þ0.301).
A three-way univariate analysis of variance (ANOVA)
was used to determine the effect of social afﬁliation of the
dyad, antiphonal reply production by the partner and the
dyad’s interpartner distance before the initial call on their
log-transformed movement ratios at both 60 and 120 s fol-
lowing the call. Social afﬁliation was deﬁned according to
each dyad’s percentage of time spent in proximity (as de-
ﬁned above). Dyads were categorized as ‘High afﬁliation’
if they were observed to spend 10% or more of their time
in proximity, ‘Moderate afﬁliation’ if they spent between
3 and 10% of their time in proximity, and ‘Low afﬁliation’
if they spent less than 3% if their time in proximity (see
Table 2). For each initial call, the antiphonal reply produc-
tion of each caller/partner dyad was categorized as ‘Reply
produced’ if the partner produced a rumble within 30 s
of the offset of the initial call, and ‘No reply’ if they did
not rumble during the 30 s window. Finally, the interpart-
ner distance of caller/partner dyads 60 s before the initial
call was classiﬁed as ‘Close’ or ‘Far’ based on a median-split
of these scores (median interpartner distance ¼61 m;
range 1e337 m). Inclusion of this factor allowed us to dis-
tinguish between the interactions of relatively close and
distant dyads. ANOVAs were calculated using SPSS (version
15.0, Chicago, Illinois, U.S.A.) and two-tailed alpha was set
to 0.05. The assumptions of ANOVA (normality and
homogeneity of variance) were tested and met according
to the guidelines set forth by Quinn & Keough (2002).
The social afﬁliation of each adult female dyad was
calculated according to the percentage of time they spent
in proximity (see Methods). Dyads ranged in this measure
of afﬁliation from 0.0 to 26.4% with a mean SD score of
7.7 9.04% of time spent in proximity. Of the 10 poten-
tial dyads in this herd, three were categorized as High
afﬁliation, three as Moderate afﬁliation and four as Low
afﬁliation (see Table 2).
The social afﬁliation of caller/partner dyads had a signif-
icant main effect on movement ratios at both 60 and 120 s
following production of the initial call (F
P¼0.026 and F
¼4.381, P¼0.013, respectively).
Caller/partner dyads of High afﬁliation showed approach
behaviour 60 and 120 s after the initial call, while those
of Moderate and Low afﬁliation showed lesser degrees of
approach (Fig. 1).
The production of an antiphonal reply by the social
partner also had a signiﬁcant main effect on the dyad’s
movement ratio at 60 s following the initial call (F
Table 2. Social afﬁliation categorization of each dyad (percentage of
time observed in proximity)
High afﬁliation Moderate afﬁliation Low afﬁliation
ANIMAL BEHAVIOUR, 76,5
12.557, P<0.001), but not at 120 s (F
P¼0.070). Sixty seconds after the initial call, dyads in
which partners produced an antiphonal reply showed ap-
proach behaviour while those that did not produce a reply
showed little average change in their interpartner dis-
tance. This statistical effect was not present at 120 s as
dyads in both response categories moved closer together
by this time (Fig. 2).
Interpartner distance of the caller/partner dyad 1 min
before the initial call had a signiﬁcant main effect on
the dyad’s movement ratio at both 60 and 120 s following
initial call production (F
¼6.358, P¼0.012 and
¼12.691, P<0.001, respectively). Caller/partner
dyads classiﬁed as being ‘far’ apart (61 m) prior to the
production of the initial call showed overall approach be-
haviour at 60 and 120 s after the initial call. Caller/partner
dyads that were classiﬁed as ‘close’ together (<61 m) be-
fore the initial call moved relatively little on average dur-
ing the ﬁrst 60 s and showed some signs of avoidance at
120 s following initial call production (Fig. 3).
There were virtually no interactions between the three
factors (afﬁliation, initial distance, antiphonal reply). Of
the eight possible interactions arising from the analysis of
three factors in two ANOVAs, only one signiﬁcant in-
teraction was documented (afﬁliation by interpartner
distance before the initial call in the post-60 s analysis:
¼3.876, P¼0.021). Thus, we focus here on the
main effects, which were largely additive (Fig. 4).
Results of this study reveal that rumble vocalizations of
African elephants can function to mediate spatial relation-
ships within the herd. While this function has been
suggested based on evidence from ﬁeld observations and
playback studies, data presented here, collected from our
captive herd using a unique digital audio and GPS re-
cording collar system, provide the ﬁrst evidence that
spontaneously produced rumbles inﬂuence the move-
ments of group members over a wide variety of contexts.
Both anecdotal ﬁeld observations and results from play-
back experiments conducted with wild populations have
shown a tendency of female elephants to approach the
calls of their herd members (Martin 1978; Poole et al.
1988; Langbauer et al. 1989, 1991; McComb et al. 2000,
2003). Our ﬁndings expand on these observations and
show that vocalization-induced approach behaviour is
mediated by a dyad’s social afﬁliation. That is, while all
dyads of the herd tended towards approach behaviour
following the production of the initial call, those that
were highly afﬁliated did so to the greatest degree (Fig. 1).
Our results also indicate a signiﬁcant increase in ap-
proach behaviour among dyads in which partners pro-
duced an antiphonal reply compared to those that did not
High Moderate Low
Figure 1. Mean movement ratio (1 SE) of caller/partner dyads at
60 and 120 s following rumble vocalizations across afﬁliation cate-
gories (high, moderate, low). Values show the proportional change
in interpartner distance relative to the distance at the time of the ini-
Reply produced No reply
Figure 2. Mean movement ratio (1 SE) of caller/partner dyads at
60 and 120 s following rumble vocalizations across reply categories
(reply produced, no reply). Values show the proportional change
in interpartner distance relative to the distance at the time of the ini-
rior to initial call
Figure 3. Mean movement ratio (1 SE) of call/partner dyads at 60
and 120 s following rumble vocalizations across distance categories
(close, far). Values show the proportional change in interpartner dis-
tance relative to the distance at the time of the initial call.
LEIGHTY ET AL.: VOCALIZATIONS MEDIATE INTERPARTNER DISTANCE 1605
answer the initial call (Fig. 2). The production of reply
rumbles in conjunction with approach behaviour has
also been documented in ﬁeld-based playback experiments
(Langbauer et al. 1989, 1991; McComb et al. 2000, 2003).
Reply rumbles may serve to inform the caller of the loca-
tion or impending approach of the social partner, further
supporting the claim that rumbles may function to facili-
tate social reunions.
Movement within dyads following the production of
the initial call was also signiﬁcantly inﬂuenced by inter-
partner distance before to call production. Members of
dyads that were far apart (61 m) before production of the
initial call showed approach behaviour in the ﬁrst 2 min
following call production (Fig. 3). That approach behav-
iour increased in magnitude with increasing distance
from the caller further supports the proposal that rumbles
maintain group cohesion within ﬁssion/fusion societies of
elephant herds. In contrast, when dyad members were
close together before the initial call they showed relatively
little movement over the ﬁrst minute and showed some
degree of avoidance behaviour in the second minute fol-
lowing call production (see below).
Figure 4 shows the additive effects of all three factors
(afﬁliation, reply production, initial interpartner distance)
on the movements of dyad members in the ﬁrst minute
following the initial call. Overall, dyads within the herd
tended to decrease their interpartner distance after the
production of a rumble, suggesting that a general func-
tion of rumbles in African elephants is to maintain group
cohesion. Examination of the particular results in Fig. 4
may help generate more speciﬁc hypotheses for future
testing. For example, the top panels of Fig. 4 show inter-
partner movement when dyads were initially far apart
(61 m). There was a general decrease in interpartner dis-
tance, particularly when an antiphonal reply was pro-
duced and the members of the dyad were highly
afﬁliated (Fig. 4, top left panel). In this latter context,
dyads reduced their interpartner distance by 56.2% rela-
tive to their initial distance (i.e. on average these dyads
reduced their distance by nearly one-half). This ﬁnding
shows the potential for rumbles to function speciﬁcally
as long-distance ‘contact calls’ that allow for animals to
decrease interpartner distance or reunite when out of
the visual range, as argued above. In addition, some of
these cases could also be instances of ‘distress calls’ that
elicited approaches and aid from closely bonded social
partners (e.g. Gouzoules et al. 1984; Gouzoules & Gou-
zoules 1995; Slocombe & Zuberbu
When pairs were relatively close together (<61 m), the
cohesion effect was weaker than when pairs were initially
far apart (see Fig. 4, bottom panels). Nevertheless, when
there was a reply to the initial rumble among dyads
with strong social bonds, rumble production on average
still resulted in approach behaviour within the dyad.
Rumbles produced under these circumstances may func-
tion to mediate close-distance social interactions of afﬁli-
ated females and therefore serve to reinforce existing
social bonds (Berg 1983; Poole et al. 1988). Some of these
cases may also be instances of ‘greeting rumbles’, which
are given during reunions of bond group members follow-
ing periods of separation (Moss 1981; Berg 1983; Poole
et al. 1988).
Movement ratio Movement ratio
Reply produced No reply
Interpartner distance before antiphonal call
Figure 4. Mean movement ratio (1 SE) at 60 s following rumble production across the three factors (afﬁliation, reply production, initial in-
terpartner distance). Values show the proportional change in interpartner distance relative to the distance at the time of the initial call. Values
above the dotted line indicate approach and values below the line indicate avoidance.
ANIMAL BEHAVIOUR, 76,5
Only certain combinations of factors resulted in an
average increase in interpartner distance (Fig. 4, bottom
right panels). Such avoidance behaviour occurred when
dyads were close together before the initial call and an
antiphonal reply was not produced by the partner, and
the strongest case occurred when dyads were also weakly
afﬁliated (bottom right panel). These cases may involve
competition over resources and/or close-range dominance
interactions. In these situations, subordinate animals may
simply disperse upon detecting the vocalization of a nearby
dominant animal, or some of these rumbles could be spe-
ciﬁc warning calls by dominants that function to displace
subordinates from valued resources.
While our ﬁndings do face the limitations of being
collected from a herd formed of unrelated females in
a captive environment with plentiful food resources, we
found that a general function of rumbles across a wide
variety of contexts is to promote spatial cohesion among
group members. We would expect to see an ampliﬁcation
of this cohesion effect in situ where herds may have
higher degrees of afﬁliation and undergo greater degrees of
ﬁssion due to larger herd size and wider distribution of
resources. Furthermore, our ﬁnding that rumbles may also
mediate avoidance behaviours at closer distances also has
the potential to be ampliﬁed in situ, particularly during
periods when the bond group is fused in the dry season,
because of the potential for greater within-group resource
competition (Wittemyer et al. 2005). To continue to disen-
tangle the complex functioning of elephant rumbles, our
future investigations will look for variation in acoustic
structure and differential movements and behaviours of
social partners following their production across a variety
of speciﬁc social contexts.
The project would not have been possible without the
support and assistance of the keepers and managers of the
Elephant team. We thank the Disney’s Animal Kingdom
Education and Science and Animal Operations teams for
support of the project and assistance in data collection.
Acumen Instruments Corporation designed and main-
tained the GPS recording collars. The work was supported
in part by the National Science Foundation (NSF-IIS-
Berg, J. K. 1983. Vocalizations and associated behaviors of the
African elephant (Loxodonta africana) in captivity. Zeitschrift fu
Boinski, S. 1991. The coordination of spatial position: a ﬁeld study of
the vocal behaviour of adult female squirrel monkeys. Animal
Boinski, S. & Campbell, A. F. 1996. The huh vocalization of white-
faced capuchins: a spacing call disguised as a food call. Ethology,
Boinski, S. & Garber, P. (Eds) 2000. On the Move: Group Travel in
Primates and Other Animals. Chicago: University of Chicago Press.
Braune, P., Schmidt, S. & Zimmermann, E. 2005. Spacing and
group coordination in a nocturnal primate, the golden brown
mouse lemur (Microcebus ravelobensis): the role of olfactory and
acoustic signals. Behavioral Ecology and Sociobiology,58, 587e596.
Butterﬁeld, P. A. 1970. The pair bond in the zebra ﬁnch. In: Social
Behavior in Birds and Mammals (Ed. by H. Crook), pp. 249e278.
New York: Academic Press.
Byrne, R. W. 1981. Distance vocalizations of Guinea baboons (Pa-
pio papio) in Senegal: an analysis of function. Behaviour,78,
Caldwell, M. C. & Caldwell, D. K. 1965. Individualized whistle
contours in bottlenose dolphins (Tursiops truncatus). Nature,
Clemins, P. J., Johnson, M. T., Leong, K. M. & Savage, A. 2005.
Automatic classiﬁcation and speaker identiﬁcation of African ele-
phant (Loxodonta africana) vocalizations. Journal of the Acoustical
Society of America,117, 956e963.
Douglas-Hamilton, I. 1972. On the ecology and behaviour of the
African elephant. Ph.D. thesis, University of Oxford.
Estes, R. D. 1991. The Behavior Guide to African Mammals: Including
Hoofed Mammals, Carnivores, Primates. Berkeley: University of
Gouzoules, S. & Gouzoules, H. 1995. Recruitment screams of
pigtail monkeys (Macaca nemestrina): ontogenetic perspectives.
Gouzoules, S., Gouzoules, H. & Marler, P. 1984. Rhesus monkey
(Macaca mulatta) screams: representational signaling in the
recruitment of agonistic aid. Animal Behaviour,32, 182e193.
Holekamp, K. E., Boydston, E. E. & Smale, L. 2000. Group travel in
social carnivores. In: On the Move: Group Travel in Primates and
Other Animals (Ed. by S. Boinski & P. Garber), pp. 587e627.
Chicago: University of Chicago Press.
Kaufmann, J. H. 1962. Ecology and social behavior of the coati,
Nasua narica, on Barro Colorado Island, Panama. University of
California Publications in Zoology,60,95e222.
Lammers, M. O., Schotten, M. & Au, W. W. L. 2006. The spatial
context of free-ranging Hawaiian spinner dolphins (Stenella longir-
ostris) producing acoustic signals. Journal of the Acoustical Society
of America,119, 1244e1250.
Langbauer, W. R. 2000. Elephant communication. Zoo Biology,19,
Langbauer, W. R., Payne, K. B., Charif, R. A. & Thomas, E. M.
1989. Responses of captive African elephants to playback of low-
frequency calls. Canadian Journal of Zoology,67, 2604e2607.
Langbauer, W. R., Payne, K. B., Charif, R. A., Rapaport, L. &
Osborn, F. 1991. African elephants respond to distant playbacks
of low-frequency conspeciﬁc calls. Journal of Experimental Biology,
Leighty, K. A., Soltis, J., Leong, K. & Savage, A. 2008. Antiphonal
exchanges in African elephants (Loxodonta africana): collective
response to a shared stimulus, social facilitation, or true communi-
cative event? Behaviour,145, 297e312.
Leong, K. M., Ortolani, A., Burks, K. D., Mellen, J. D. & Savage, A.
2003. Quantifying acoustic and temporal characteristics of vocali-
zations for a group of captive African elephants Loxodonta africana.
McComb, K., Moss, C., Sayialele, S. & Baker, L. 2000. Unusually
extensive networks of vocal recognition in African elephants.
Animal Behaviour,59, 1103e1109.
McComb, K., Reby, D., Baker, L., Moss, C. & Sayialel, S. 2003.
Long-distance communication of acoustic cues to social identity
in African elephants. Animal Behaviour,65, 317e329.
Martin, R. B. 1978. Aspects of elephant social organisation. Rhodesia
Science News,12, 184e187.
Mitani, J. & Nishida, T. 1993. Contexts and social correlates of
long-distance calling by male chimpanzees. Animal Behaviour,
LEIGHTY ET AL.: VOCALIZATIONS MEDIATE INTERPARTNER DISTANCE 1607
Moss, C. J. 1981. Social circles. Wildlife News,16,1e7.
Moss, C. J. & Poole, J. H. 1983. Relationships and social structure
in African elephants. In: Primate Social Relationships: an Integrated
Approach (Ed. by R. A. Hinde), pp. 315e325. Oxford: Blackwell.
Oda, R. 1996. Effects of contextual and social variables on contact
call production in free-ranging ringtailed lemurs (Lemur catta).
International Journal of Primatology,17, 191e205.
Poole, J. H., Payne, K., Langbauer, W. R. & Moss, C. J. 1988. The
social contexts of some very low frequency calls in African ele-
phants. Behavioral Ecology and Sociobiology,22, 385e392.
Quinn, G. P. & Keough, M. J. 2002. Experimental Design and Data
Analysis for Biologists. Cambridge: Cambridge University Press.
Ramos-Fernandez, G. 2005. Vocal communication in a ﬁssione
fusion society: do spider monkeys stay in touch with close associ-
ates? International Journal of Primatology,26, 1077e1091.
Schaller, G. 1972. The Serengeti Lion: a Study of PredatorePrey
Relations. Chicago: University of Chicago Press.
Slocombe, K. E. & Zuberbu
¨hler, K. 2007. Chimpanzees modify
recruitment screams as a function of audience composition.
Proceedings of the National Academy of Sciences, U.S.A.,104,
Smith, S. T. 1972. Communication and other social behavior in
Parus carolinensis.Publications of the Nuttall Ornithological Club,
Soltis, J., Leong, K. & Savage, A. 2005a. African elephant vocal
communication I: antiphonal calling behaviour among afﬁliated
females. Animal Behaviour,70, 579e587.
Soltis, J., Leong, K. & Savage, A. 2005b. African elephant vocal
communication II: rumble variation reﬂects the individual identity
and emotional state of callers. Animal Behaviour,70, 589e599.
Teixidor, P. & Byrne, R. W. 1999. The ‘‘whinny’’ of spider monkeys:
individual recognition before situational meaning. Behaviour,136,
Wilkinson, G. S. & Boughman, J. W. 1998. Social calls coordinate
foraging in greater spear-nosed bats. Animal Behaviour,55,
Wittemyer, G. & Getz, W. M. 2007. Hierarchical dominance
structure and social organization in African elephants, Loxodonta
africana.Animal Behaviour,73, 671e681.
Wittemyer, G., Douglas-Hamilton, I. & Getz, W. M. 2005. The
socioecology of elephants: analysis of the processes creating mul-
titiered social structures. Animal Behaviour,69, 1357e1371.
ANIMAL BEHAVIOUR, 76,5