Leadership In Elephants: The Adaptive Value Of Age

Downloaded from rspb.royalsocietypublishing.org on October 3, 2011Leadership in elephants K. McComb et al.directly how groups of African elephants (Loxodontaafricana) with leaders of different ages performed at assessingfine-scaled differences in the risk involved. In doing so, webuilt on techniques previously developed to assess the factorsaffecting social knowledge in elephant groups [10].Female African elephants live in matrilineal familyunits led by the oldest female, or matriarch, who plays akey role in coordinating group movements and responsesto threat [10,18 – 20]. Other than humans, lions are themain natural predators of African elephants [21 –23],typically preying on calves of less than 4 years of age[23,24]. Although it is often assumed that lionesses domost of the hunting, male lions are in fact considerablymore effective and successful predators when it comesto targeting the largest prey—namely elephants and buffalo [23,25,26]. These large-bodied species often reactto lion attacks with cooperative and aggressive defencemechanisms, presenting formidable opponents giventheir combined strength and dangerous weaponry[25,27]. In such situations, male lions, being on average50 per cent larger in body mass and considerably morepowerful than females [28], are at a distinct advantage.It is notable that the mean group size for successfulmale lion hunts on elephants is dramatically smallerthan that for successful female hunts (1.8 versus 7 inone key study [23]), a finding also reflected in data onlion hunts of buffalo [25]. Moreover, while hunting success typically increases with group size [22,23], it hasbeen demonstrated that male lions have the capabilityof overpowering a young elephant even when huntingalone [19,23]. Older elephant matriarchs should bemore experienced in facing lion encounters than youngermatriarchs, and hence we might expect them to be betterable to identify the greater risk posed by larger groups oflions and by male lions in particular. Our study, usingindividually known elephants in Amboseli NationalPark, Kenya (see §2), provides an unusual opportunityto examine directly how the age of a leader may influencethis vital ecological knowledge in a wild population.Abilities to discriminate between different levels of predatory threat were tested by giving elephant family unitsplaybacks of three lions versus a single lion roaring and,within these categories, roaring from male versus femalelions (see §2). In a wide range of social species, individualsmay benefit through increased attentiveness, group-defensivebehaviour and, in some cases, even approaches to harass predators (mobbing), which can serve both to directlydiscourage attack and to gain further information about thethreat involved [19,29,30]. While mobbing is not widespreadamong mammals [17], elephants are sufficiently large andwell coordinated to search out and direct group attacks onpotential predators [19,27,31]. In conjunction with ourprevious research on social knowledge in African elephants[10], we predicted that groups led by older matriarchswould show greater attentiveness, defensive bunching behaviour and increased likelihood of mobbing approach whenfaced with lion groups (as opposed to singletons) and withmales (as opposed to females).2. MATERIAL AND METHODS(a) Study populationAll playback experiments were conducted on a natural population of African elephants in Amboseli National Park inProc. R. Soc. B (2011)3271Kenya between May 2007 and July 2009. This study population encompasses approximately 1500 elephants, with 58distinct family groups. The Amboseli Elephant ResearchProject (AERP; http://www.elephanttrust.org) has longterm demographic and behavioural data on the entirepopulation, including detailed ages for all elephants bornafter 1971, while ages for older individuals were estimatedusing criteria that are accepted as standard in studies of African elephants [18,32]. All elephants in the population arehabituated to the presence of AERP research vehicles.(b) Playback procedureA total of 72 playbacks were conducted on 39 elephant familygroups (1– 4 playbacks per group), with a mean of five adults( 2.6 s.d.) in each family. All of the playbacks were carriedout between 16.00 and 19.00 h, the time at which lionsbecome active.The lion roars used in the study had been recorded in theSerengeti National Park, Tanzania using SennheiserMKH816T microphones and Panasonic SV-250 digitalaudio tape recorders [33,34]. Eight different playbackexemplars were presented, consisting of two matched pairsof one lion versus three lions roaring for both male andfemale lions [33,34].In each playback, a single bout of lion roars (mean duration: males, 39 7 s; females, 40 5 s) was broadcast tothe subjects from a fieldwork vehicle that was located100 m from the periphery of the family group. The lionroars were played through custom-built loudspeakersmanufactured by Bowers & Wilkins, Steyning, UK andIntersonics Inc., Northbrook, IL, USA. The Bowers &Wilkins loudspeaker was powered by Alpine PDX-1.1000and MRP-T222 amplifiers and the Intersonics loudspeakerby a Kenwood KAC-PS400M amplifier. In all playbacksthe peak sound pressure level at 1 m from the loudspeakerwas standardized at 116 dB, comparable with that of naturallion roars [33]. Sound pressure levels were measured with aCEL-414/3 sound level meter.The first exemplars played to family groups were randomized; repeat playbacks for each sex of lion were thensystematically paired with the first exemplars presented, sothat across playbacks two factors (number of lions and sexof lion) were systematically varied in turn, while holdingothers constant. To avoid habituation, a minimum periodof 7 days was left between playbacks to the same family.Playbacks were not given to groups with calves of less thanone month as our previous work had indicated that thepresence of such very young calves may result in abnormallyhigh sensitivity to perceived threats over this short timeperiod [10].The behavioural responses of the elephants to playbackwere observed through binoculars and recorded on video,alongside live commentary, using a Canon XM2 videocamera. From video analysis, we assessed five key behaviouralmeasures that described the responses of the family group withparticular reference to the matriarch (developed from [10]):— Matriarch prolonged listening. Matriarch continues to exhibit evidence of listening response for more than 3 minafter playback, where ears are held in a stiff extendedposition, often with the head slightly raised.— Matriarch bunching. Defensive response to perceivedthreat by adult females (including the matriarch) andtheir young, which results in the diameter of the family

Downloaded from rspb.royalsocietypublishing.org on October 3, 20113272K. McComb et al. Leadership in elephantsgroup decreasing after the broadcast of the playbackstimulus (calculated in terms of elephant body lengths).— Matriarch bunching intensity. The rate at which a defensivebunch of adult females (including the matriarch) andtheir young occurs. This measure classifies the overalllevel of threat response, scoring bunching intensity on a4-point scale as follows:(i) 0: No bunching occurred.(ii) 1: Subtle reduction in the diameter of the group, elephants remain relaxed and continue with preplayback behaviours (.3 min for bunch formation).(iii) 2: Group forms a coordinated bunch, pre-playbackbehaviours such as feeding interrupted (1 –3 min forbunch formation).(iv) 3: Fast and sudden reduction in the diameter of thegroup, elephants very alert (,1 min for bunchformation).— Matriarch approach. Distinct change in the direction andapproach of the matriarch towards the source of playback.— Group to matriarch. Adult females and/or their youngrespond to playback by moving towards the matriarch.An independent observer who did not have access to thevideo commentary and was blind to the playback sequencesecond-coded 15 per cent of the video records; an overallagreement of 95 per cent was achieved on the binaryresponse variables and the Spearman’s r correlationon the scores for matriarch bunching intensity was 0.940(p , 0.001). A video clip illustrating the highest level ofbunching and mobbing approach is available as electronicsupplementary material S1.(c) Statistical analysesThe data were analysed using generalized linear mixed models(GLMMs) in R (www.r-project.org). To test the responses ofelephants to lion playbacks as a function of predatory threat(number and sex of the lions) and the age of the matriarch,five key behaviours were used as dependent variables (seeabove), while the identity of the elephant group was enteredas a random factor. Model selection was performed usingAkaike’s information criterion (AIC) [35], with lower AICscores indicating better models; however, a more complexmodel with more degrees of freedom was only selected over asimpler model when the AIC differed by 2 or more [35]. Inour main analysis, matriarch age was entered as a continuousvariable in each of the GLMMs. A second analysis comparedthe behavioural responses of the oldest age class of matriarchs( 60 years) with those 59 years. In all analyses, potentiallyconfounding variables—the number of adult female elephantsin the group and their average age [10]—were excludedduring model selection as they generated higher AIC scores,with no significant main effects for the variables nor forinteractions between them and the sex and number of lions.3. RESULTSIn all cases, the best model (see §2) to describe theresponses of elephants to playbacks included the numberof lions, lion sex, matriarch age, and the interaction betweenlion sex and matriarch age (electronic supplementarymaterial S2). The number of lions had significant effectson four out of five of our behavioural measures, with playback of three lions generating elevated responses aspredicted (table 1 and figure 1). The ability to make thisProc. R. Soc. B (2011)basic distinction did not depend on the age of the matriarch,with no significant interactions occurring between matriarchage and the number of lions for any of our response variables. However, matriarch age (entered as a continuousvariable) had marked effects on behaviour in response toplaybacks of male versus female lions, with significant interactions between age of matriarch and sex of lions(electronic supplementary material S2 and table 1). Morespecifically, sensitivity to male (versus female) lionsincreased with matriarch age, matriarchs being morelikely to engage in prolonged listening and defensivebunching, and showing a higher bunching intensitywhen faced with male lion roars as they get older(table 1 and figure 2a; see electronic supplementarymaterial S3 for scatter plots of raw data). Groups witholder matriarchs were also more likely to approach thesource of playback when male roars were presented(table 1 and figure 2b). In addition, as reproductiveactivity shows a marked decline in female elephantsbetween the ages of 55 and 60 years [18,36,37], we investigated the sensitivity of the oldest age class (60 years andabove) to the particular threat posed by male lions in asecondary analysis. Comparing these with matriarchs of59 years and younger (matriarch age was categorical inthis analysis: age 59, n ¼ 53; age 60þ, n ¼ 19), therewere significant interactions between matriarch age classand lion sex in our measures of attentiveness and defensive bunching (table 1). The oldest matriarchs weremore likely to engage in prolonged periods of listeningand exhibited greater defensive bunching (higher probability of bunching and greater bunching intensity) inresponse to male versus female roars, demonstratingtheir key role in identifying and responding to this mostserious threat (table 1 and figure 2c,d ). Other potentiallyconfounding variables—number of females in the groupand mean age of females other than the matriarch—were excluded from the final models because theirinclusion produced poorer results when model selectionwas carried out using AIC scores (see §2).4. DISCUSSIONThese results demonstrate that while elephant familygroups react more strongly to three lions than a singlelion roaring irrespective of matriarch age, those led byolder matriarchs show a greater sensitivity to the moreserious threat posed by male lions across all our behavioural responses, including measures of attentiveness,bunching and mobbing approach. The superior abilityto detect the presence of male lions at an early stage—evident in the greater probability of older matriarchsengaging in prolonged periods of listening and defensivelybunching with their groups—is likely to have significantsurvival benefits affording better protection for vulnerablecalves in particular. Previous researchers have speculatedthat elephant groups can derive fitness benefits from theimprovement in ecological knowledge conferred by anolder matriarch, suggesting that this enables better utilization of scarce resources and predator avoidance duringperiods of drought [15]. Our work provides the firstdirect experimental evidence that older matriarchs arein fact able to make better decisions when faced with ecological challenges—in this case, the presence of dangerouspredators. It thus bridges an important gap between

Downloaded from rspb.royalsocietypublishing.org on October 3, 2011Leadership in elephants K. McComb et al.3273Table 1. Results of GLMMs investigating matriarch behavioural responses to lion roar playbacks (a) with matriarch ageentered as a continuous variable and (b) with two age classes of matriarchs—the oldest age class (60 years and over) and allyounger matriarchs ( 59 years).dependent variable(a)matriarch prolonged listeningmatriarch bunchingmatriarch bunching intensitymatriarch approachgroup to matriarch(b)matriarch prolonged listeningmatriarch bunchingmatriarch bunching intensitymatriarch approachgroup to matriarchparametersestimates.e.Z-valuep-valuelion numberlion sexmatriarch agelion sex matriarch 22.03120.1571.8900.00080.040.880.06lion numberlion sexmatriarch agelion sex matriarch agelion numberlion sexmatriarch agelion sex matriarch agelion numberlion sexmatriarch agelion sex matriarch agelion numberlion sexmatriarch agelion sex matriarch 30.830.060.510.030.00070.040.830.04lion numberlion sexmatriarch age classlion sex matriarch age classlion numberlion sexmatriarch age classlion sex matriarch age classlion numberlion sexmatriarch age classlion sex matriarch age classlion numberlion sexmatriarch age classlion sex matriarch age classlion numberlion sexmatriarch age classlion sex matriarch age 010.840.660.98theoretical predictions about how knowledge might beexpected to affect leadership [1,9] and empirical studies,which to date have been largely confined to observationalaccounts [6,7,15].It has recently been suggested that prey species withlarge brains relative to their body size may be better atevading predators because they can more effectivelyadjust their behavioural responses to specific encounters[38]. Our results indicate that elephants can indeed makeimportant distinctions between threat levels even within apredator class, in parallel with research showing their ability to distinguish human ethnic groups that pose differentlevels of threat on the basis of olfactory and visual cues[39]. Moreover, the results demonstrate how the accumulated knowledge of the oldest individuals may have an overProc. R. Soc. B (2011)riding influence on the effectiveness of anti-predatordecisions made by the social group as a whole, and theyhighlight the vital role of such individuals in natural populations [10]. This is a particularly important considerationwhere relatively large-brained social species have theopportunity to accumulate ecological information overlong lifespans and considerable disparities in knowledgebetween different individuals are therefore likely to arise.While we cannot be sure that matriarch age will correlate with experience of lions in every instance, it isreasonable to assume that as matriarchs get older theywill accumulate exposure to the particular risks posedby lions. Although elephants are relatively impregnableto most predators by virtue of their large body size andaggressive group defence, lions remain a very significant

Downloaded from rspb.royalsocietypublishing.org on October 3, 20113274K. McComb et al. Leadership in elephants(b) 1.6(38)(38)1.40.6mean bunching intensityprobability of elephants bunching(a) 0.70.50.40.3(34)0.20.11.21.00.8(34)0.60.40.2001 lion3 lions1 lion3 lionsFigure 1. Elephant response to the scale of predation threat as simulated by one versus three roaring lions: (a) probability ofmatriarch bunching and (b) matriarch bunching intensity (mean s.e.m.). Number of playbacks given in parentheses.mean bunching intensity(7)(8)(8)1.00.50(c)(12)(6)(11) 4041–5051–60age of matriarch (years)0.31.5(8)1.0(25)0.5 5960 matriarch age class (years)(8)(6)0.2(12)(8)(11)0.1(11) 4061 (9)(30)(9)0.402.00probability of approach1.5(b) 0.5(9)(11)2.0(d)0.6probability of prolongedlisteningmean bunching intensity(a)0.5(7)41–5051–60age of matriarch (years)(9)(30)0.4(8)0.30.261 (25)0.10 5960 matriarch age class (years)Figure 2. Behavioural responses of matriarchs and their family groups to male (grey bars) versus female (white bars) lionsdepicted as a function of increasing matriarch age for illustration (see table 1 for detailed analysis and electronic supplementarymaterial S3 for raw data). (a,c) Mean bunching intensity ( s.e.m.); (b) probability of approach; and (d) probability ofprolonged listening. Number of playbacks given in parentheses.threat and have the potential to shape the evolution ofspecific anti-predator behaviour [23,27]. Despite this,the occurrence of lion predation on elephants is typicallya rare event [24]. Moreover, the sex ratio of adult lions inmost populations is heavily female-biased [40,41], with a1 M : 1.8 F ratio in the Amboseli population [42], whichmay lower the exposure to male-specific cues further still.Our experiments test elephant abilities to make a rathersubtle distinction—between characteristics defining maleand female roars—and while sex-specific acoustic differences have been documented in lion roars [43], theseare not overtly obvious to human listeners. It is perhapsnot surprising therefore that considerable experience,Proc. R. Soc. B (2011)built up over a long lifespan, should enable olderindividuals to perform better in this task.In common with humans and certain other socialanimals, in particular pilot and killer whales, femaleelephants can have a significant post-reproductive lifespan—although it is currently unclear whether theyexhibit menopause [18,37,44,45]. While reproductiveactivity declines among this oldest age class, they showno signs of increased vulnerability to predators, and infact are often the largest individuals within familygroups [31]. Indeed, our results suggest that this oldestage class of females can provide significant benefits fortheir groups in the context of decision-making about

Downloaded from rspb.royalsocietypublishing.org on October 3, 2011Leadership in elephants K. McComb et al.predators. As a result of typically no longer having youngoffspring, they may be able to take a more active role ingroup defence and could be more attentive to ecologicalsignals. However, if families with older matriarchs weresimply more reactive, we would predict an overall increasein sensitivity to lion roars rather than the greaterdifferential response to male versus female roars reportedhere. Instead, older matriarchs appear more adept atassociating male lion roars with a greater level of threat.Observational data also indicate that it is individuals of60 years and older who have the greatest success inleading large-scale foraging movements [16]—againsuggesting that in decisions relating to ecological knowledge, the level of experience that these oldestindividuals possess may prove critical. Recent researchon humans has demonstrated the key role that grandmothers can play in enhancing the reproductive successof their offspring by assisting in child-rearing after theirown reproduction has ceased [46,47]. Our findings indicate that through their greater ecological knowledge,older females can also represent a crucial resource inother long-lived cognitively advanced mammals that livein kin-based social groups.Leadership is known to be crucial in the coordination ofhuman groups, but we are only now beginning to appreciate its role in animal societies [3,4]. In order to fullyunderstand how animal groups are organized and achievecollective action, it is vital to determine the advantagesthat a leader may bring to group decision-making. Ourstudy provides the first empirical evidence that individualswithin a social group may benefit directly from the influence of an older leader because of their enhanced abilityto make crucial decisions about predatory threat. It thusgenerates important insights into selection for longevityand the evolution of a sophisticated social structure thatsupports the coordinating role of a single older leader.Such abilities are likely to be highly relevant in otherlarge-brained, long-lived social species where older individuals play a key role in coordinating group activities.This work complies with the Association for the Study ofAnimal Behaviour/Animal Behaviour Society guidelines forthe use of animals in research, and received approval fromthe Ethical Review Committee at the University of Sussex.We thank the National Geographic Society (grant no. 797706), the Leverhulme Trust (grant no. F/00230/AC) and theAmarula Elephant Research Programme for funding thiswork. We are grateful to the Kenyan Office of the Presidentand to Kenya Wildlife Services for permission to conductthe research in Amboseli National Park, to Kenya Airways,P. B. Allen and Bowers & Wilkins, for logistical supportand provision of equipment, and to Jon Grinnell for accessto lion recordings. We also thank the Amboseli Trust forElephants for facilitating this study, Larissa Conradt,Leanne Proops, David Reby and Paul Thompson forproviding comments on the manuscript, Leanne Proopsfor second coding of video records, Nathalie Pettorelli forstatistical advice, and Andy Radford, Tim Caro and twoanonymous referees for helpful suggestions for revisions.REFERENCES1 Conradt, L. & Roper, T. J. 2003 Group decision-making inanimals. Nature 421, 155–158. (doi:10.1038/nature01294)2 Conradt, L., Krause, J., Couzin, I. D. & Roper, T. J. 2009‘Leading according to need’ in self-organizing groups.Am. Nat. 173, 305 –312. (doi:10.1086/596532)Proc. R. Soc. B (2011)32753 King, A. J., Johnson, D. D. P. & Van Vugt, M. 2009The origins and evolution of leadership. Curr. Biol. 19,911 –916. (doi:10.1016/j.cub.2009.07.027)4 King, A. J. 2010 Follow me! I’m a leader if you do; I’m afailed initiator if you don’t? Behav. Proc. 84, 671 –674.(doi:10.1016/j.beproc.2010.03.006)5 King, A. J., Douglas, C. M. S., Huchard,

The Bowers & Wilkins loudspeaker was powered by Alpine PDX-1.1000 and MRP-T222 amplifiers and the Intersonics loudspeaker by a Kenwood KAC-PS400M amplifier. In all playbacks the peak sound pressure level at 1 m from the loudspeaker was standardized at 116 dB, comparable with that of natural lion roars [33].