ANIMAL BEHAVIOUR, 2006, 72, 1345e1353doi:10.1016/j.anbehav.2006.03.020

Ontogeny of alarm call responses in meerkats, Suricata suricatta:

the roles of age, sex and nearby conspecifics

LINDA I. HOLLEN & MARTA B. MANSER

Zoologisches Institut, Universitat Zurich

(Received 24 July 2005; initial acceptance 28 September 2005;

final acceptance 26 March 2006; published online 6 October 2006; MS. number: 8630R)

Given the strong selection on prey animals to escape predation, early development of correct avoidancestrategies should be favoured. We studied the development of responses to conspecific alarm calls ina free-ranging population of meerkats in South Africa. Through behavioural observations of naturally oc-curring predator encounters and playback experiments, we monitored responses of young individualsfrom emergence (3 weeks) to 6 months of age and compared them with those of adults (>12 months).Although the total proportion of responses differing from those of adults was low during the observedperiod, the probability of responding like adults increased with age. Female young, who remained in closercontact to adults than did male young, were also more likely to show adultlike responses. The largestproportion of non-adultlike responses was shown before reaching independence at 3 months of age,and during this time young commonly ran immediately to a nearby individual when hearing an alarmcall. After playbacks of alarm calls, young also reacted more slowly, resumed foraging sooner and spentless time vigilant than did adults. We conclude that young may need experience during early developmentto associate an alarm call correctly with the type of threat and appropriate response. Older group membersmay also serve as indirect models, perhaps helping young to form this association.

� 2006 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.

Animals are predisposed to learn about features in theirenvironment that are relevant to their survival, but giventhe strong selective force that predation exerts on animals,in particular young individuals, one might expect avoid-ance strategies to be fully functional upon a first encoun-ter with a predator. However, if predation risk varies inspace and time (Lima & Dill 1990), or if an environmentalchange causes animals to be exposed to previously unfa-miliar predators (Berger et al. 2001), learning would allowresponses to be adjusted to local conditions. Furthermore,the protection of young from predators is an essentialcomponent of parental care (Clutton-Brock 1991), so thepresence and form of parental care may influence the be-haviour and survival of young. In species where parentalcare is present, young may rely on their parents to defendthem against predators or have the opportunity to learnhow to avoid them (e.g. Hodge 2003; Platzen & Magrath2004), whereas young that do not receive parental caremay be under higher pressure to have functional

Correspondence: L. Hollen, Verhaltensbiologie, Zoologisches Institut,Universitat Zurich, Winterthurerstrasse 190, CH-8057 Zurich,Switzerland (email: lindah@zool.unizh.ch).

1340003e3472/06/$30.00/0 � 2006 The Association for the S

antipredator behaviour from birth (e.g. Impekoven 1976;Miller & Blaich 1986; Goth 2001).

Research on the development of alarm call responses inmammals has focused mainly on nonhuman primates(reviewed in Seyfarth & Cheney 1997) and ground squir-rels (Mateo 1996a,b; Hanson & Coss 2001). In both non-human primates and ground squirrels, the appropriateresponses to alarm calls seem to develop gradually withage, suggesting that young individuals need experienceto associate alarm calls correctly with the type of threatand correct response. For example, infant vervet monkeys,Chlorocebus (formerly Cercopithecus) aethiops, of 3e4months of age rarely responded like adults, whereasmost infants older than 6 months did so (Seyfarth & Che-ney 1986). The need for experience was further supportedby Hauser (1988), who found that infant vervet monkeysexposed to superb starling, Spreo superbus, alarm calls ata high rate responded appropriately to these calls at anearlier age than did infants exposed to these calls at a lowerrate.

Given that learning how to avoid predators might becostly to acquire through individual experience, it isperhaps not surprising that there is substantial evidencefor social influences on antipredator behaviour in a widerange of taxa, including fish, birds and mammals

5tudy of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.

ANIMAL BEHAVIOUR, 72, 61346

(reviewed in Griffin 2004). In addition to observationalconditioning (Cook et al. 1985), where individuals acquirealarm responses to previously neutral stimuli, the expo-sure to alarm behaviour of conspecifics can also enhancethe specificity of juvenile responses (Seyfarth & Cheney1986) or cause correct responses to develop more quickly(Mateo & Holmes 1997).

In contrast to studies on human language development(e.g. Galsworthy et al. 2000; Berglund et al. 2005), differ-ences in communicative skills between the sexes have re-ceived comparatively little attention in studies of animalvocal development (but see Gouzoules & Gouzoules1989; Yamaguchi 2001). To our knowledge, studies onthe development of alarm call responses have not lookedspecifically at sex differences between young. However,this question may be important in species where survivalof young differs between the sexes, such as described formeerkats, where female pups were more likely to survivethan male pups (Russell et al. 2002).

Meerkats provide an excellent opportunity to investi-gate how young individuals develop their antipredatorskills. Pups, which are cooperatively reared by the groupand remain below ground for approximately 3 weeks afterbirth (Clutton-Brock et al. 1999a), face extreme challengesduring early development as they move from safety un-derground to a life above ground. First, meerkats live un-der high predation pressure and are preyed upon byseveral aerial and terrestrial predators, including snakes(Clutton-Brock et al. 1999b), so pups are suddenly ex-posed to a wide variety of predators and alarm calls givenby older group members. Adults utter different alarm calls,eliciting different behavioural responses depending onpredator type and the level of response urgency, allowinggroup members to respond appropriately to calls utteredin a specific context (Manser 2001; Manser et al. 2001).Young individuals, especially before reaching indepen-dence at 3 months of age, are particularly vulnerableand suffer from a high mortality rate (approximately30%) from predation (Doolan & Macdonald 1997; Clut-ton-Brock et al. 1999b). Consequently, young should bestrongly selected to respond appropriately at an early age.

Second, during the first 2 months of foraging with thegroup (age 4e12 weeks), pups depend mainly on othergroup members for food (Doolan & Macdonald 1999), andtherefore stay very close to older individuals. Older pups,however, probably have to spend more time away fromother group members because they must rely on theirown foraging skills. Therefore, if pups use such informa-tion, this change in foraging behaviour during early devel-opment may also influence the acquisition and use ofsocial information.

We investigated the development of alarm call re-sponses based on observations of naturally occurringpredator encounters and playback experiments. We con-centrated on three issues. First, we examined whether theresponses to alarm calls changed with age. Second, weinvestigated whether the responses of male and femaleyoung differed. Third, we examined the influence ofproximity to adult individuals, the number of helpersrelative to the number of pups and the number of nearbyindividuals.

METHODS

We studied the development of alarm call responses in 11groups of free-ranging meerkats along the dry bed ofKuruman River in the southern part of the Kalahari Desertin South Africa (26�580S, 21�490E) from January to July2003 and from October 2003 to June 2004 (details onstudy site provided in Clutton-Brock et al. 1999b). All an-imals were habituated to close observation (<1 m) andmarked for individual identification with hair dye orhair cuts applied to their fur. These marks were smalland were applied noninvasively during sunning at themorning sleeping burrow. Ages of all individuals wereknown because they had been monitored since birth.Pups were defined as animals younger than 3 months,juveniles as 3e6 months, subadults as 6e12 months, andadults as older than 12 months. Pups and juveniles arehereafter referred to as young. The study was conductedwith the permission of Northern Cape Conservation Ser-vice and the ethical committee of Pretoria University,South Africa.

Behavioural Observations

We collected longitudinal data on responses to naturallyoccurring alarm calls using an ad libitum sampling pro-cedure (Martin & Bateson 1993). Whenever an alarm callwas uttered, we noted the following observations on aVisor Pro handheld computer (palmOne, Inc., Milpitas,California, U.S.A.): (1) first response (within 2 s of the ini-tial alarm) shown by the nearest young (one or several) insight; (2) first response shown by at least half of all groupmembers older than 6 months observed during the timeof alarm (taken as a typical adult response). If no specificresponse was shown by more than half of the individuals,the most frequently occurring response was noted; (3)most extreme response (following the first response ifresponse escalated) shown by young; (4) most extremeresponse shown by at least half of the older group mem-bers. If the response did not escalate, the most extremeresponse equalled the first response.

We classified responses as ‘look briefly’, ‘watch contin-uously’, ‘move’, ‘move to bolthole’, ‘move below ground’or ‘mob’ (Table 1). If an individual’s behaviour did notchange following an alarm call, it was scored as ‘not re-sponding’. First responses where young looked up briefly

Table 1. Response codes recorded during natural observations ofpredator encounters

Response code Explanation

Look briefly Look in the directionof threat for <3 s

Watch continuously Follow the threat, continuouslyobserving it until it has passed by

Move Run but not all the wayto a bolthole

Move to bolthole Run to a bolthole mouth and stopMove below Flee down a boltholeMobbing Gather around threat with

erect fur and tail

HOLLEN & MANSER: ALARM RESPONSES IN YOUNG MEERKATS 1347

or watched continuously were then classified as scanningthe surroundings or looking towards another individual.Looking towards another individual was defined as theyoung clearly focusing on an individual nearby insteadof continually moving its head to follow the threat. Re-sponses where young moved a short distance, moved toa bolthole, moved below ground or mobbed were classi-fied as first running up to another individual (within0.5 m) or running independently for shelter or towardsthreat. Finally, responses where young responded inde-pendently of other individuals, i.e. did not look towardsor run to another individual, were classified as adultlike(same response as adults) or non-adultlike (differentresponse to adults). To investigate whether the non-adultlike responses still resembled those of adults or ifthey differed considerably, we divided the non-adultlikeresponses into two categories: (1) looking responses (lookbriefly, watch continuously), and (2) moving responses(move, move to bolthole, move below ground). If theresponses of young fell into the same category as the adultresponses, i.e. either looking or moving, they were scoredas similar; otherwise, they were scored as different. We ana-lysed 323 responses from 48 young (26 females and 22males, age range 19e180 days) in 19 litters and 10 groups.

Playback Experiments

Recording methods and call selectionWe recorded alarm calls used for playback experiments

at a distance of 1e2 m from the caller, and at 44.1-kHzsampling frequency, using a Sony digital audio tape re-corder DAT-TCD D100 (frequency response: 20e20 000 Hz �1 dB) connected to a Sennheiser directionalmicrophone (K6 power module and ME66 recordinghead with a MZW66 pro windscreen; frequency response40e20 000 Hz �2.5 dB, Old Lyme, Connecticut, U.S.A.).Calls were uploaded on to a PC notebook and digitized(16-bits, 44.1 kHz) using the 24-bit U24 waveterminalUSB audio interface (Ego-sys, Seoul, Korea). Natural call se-quences with a high signal-to-noise ratio were then pre-pared with Cool Edit 2000 (Syntrillium SoftwareCorporation, Phoenix, Arizona, U.S.A.), and recordedback on to a DAT tape for use in the field.

Behavioural responses to alarm calls vary with predatortype and urgency level, so we used six alarm call types:aerial and terrestrial calls at medium and high urgencylevels and recruitment calls (elicited in response to snakesor deposits such as faecal, urine or hair samples of othermeerkats or predators) at low and high urgency levels (forspectrograms and details on call-specific responses, seeManser 2001; Manser et al. 2001). We used at least six exam-ples of each call type. To exclude any age-related differencesin acoustic structure, we used only calls from adults. Weused calls from both own group members and unfamiliarindividuals, because previous playback experiments haveshown that meerkats do not distinguish between them(Manser 1998). Bird song is frequently heard in the area,so we included three examples from two species, the white-browed sparrow weaver, Plocepasser mahali, and the Kala-hari scrub robin, Cercotrichas paean, as control stimuli.

Experimental protocolResponses to alarm calls without the presence of

a predator were investigated by playing back calls to11 groups. We conducted playbacks from first emer-gence ðX� SD ¼ 17� 2:4 daysÞ until young reached anage of 95 days. We tested 16 randomly chosen individ-uals (eight females, eight males in 13 litters and ninegroups) repeatedly over their development (typically ev-ery 2 weeks). These individuals received, on average, fiveplaybacks each (range 2e7). The low numbers of play-backs for some individuals were due to bad weather con-ditions, depredation or technical difficulties. Otherindividuals, randomly selected for each playback, weretested only once. Overall, we tested 25 females and 16males belonging to 23 litters in 146 playbacks (includ-ing 23 control playbacks). We conducted 14 alarm-callplaybacks (aerial and terrestrial medium urgency and re-cruitment high urgency) and six control playbacks oneight litters in seven groups during the first 2 weeks ofemergence (age range 24e30 days), when young stayedbehind at the sleeping burrow together with one or sev-eral babysitters. All but one litter received multiple play-backs but with different call types. During these first 2weeks, all pups in a litter typically stayed close togetherand showed the same response, so we obtained one re-sponse per litter. The remaining playbacks were con-ducted once young started foraging with the groupðX� SD ¼ 28� 2:6 daysÞ. Seven of the litters tested dur-ing the first 2 weeks of emergence were also tested againafter foraging with the group.

Calls were played back with a Sony DAT-TCD D100recorder connected to a Sony SRS-A60 loudspeaker (frequ-ency response 70e20 000 Hz) and broadcast at ampli-tudes of 54e62 dB, measured 1 m in front of thespeaker (Voltcraft 329 Sound Level Meter, Conrad Elec-tronic, Hirschau, Germany; accuracy �2 dB at 94 dB).This range of amplitudes corresponds to that observedfor calls given during naturally occurring predatorencounters. The duration of each playback was 3e20 s,depending on call type.

The focal pup was filmed with a Sony digital videocamera DCR-PC 120E for at least 20 s before the call wasplayed and until normal foraging behaviour was resumedafter the call. The loudspeaker was hidden behind vegeta-tion 5e10 m from the pup, and calls were played whenthe pup was at least 5 m from a bolthole or burrow system(after the first 2 weeks of emergence), at least 1 m from an-other individual and not engaged in vigilant behaviour.Calls were played only if there had not been a naturalpredator encounter during the preceding 20 min. To avoidhabituation, the number of playback experiments waskept below the natural rate of alarm calling, and at least3 days passed between successive playbacks in each group.A maximum of two playback experiments were conductedeach day, with at least 1 h in between.

Scan Sampling: Proximity to Adults

To investigate how the spatial relationship betweenyoung and adults changed during development, we

ANIMAL BEHAVIOUR, 72, 61348

collected data on 30 female young and 28 male young in15 litters and 11 groups during three periods, using a scan-sampling procedure (Martin & Bateson 1993). The first pe-riod included the first 3 weeks of young foraging with thegroup (age range 33e54 days; N ¼ 12 females, 8 males), thesecond period included week 6e8 (age range 70e93 days;N ¼ 9 females, 13 males) and the third period week 11e13 (age range 106e125 days; N ¼ 9 females, 7 males).Each individual was used as a subject only once. We col-lected scans on each individual during one morning forag-ing session, and every 15 min we recorded the distance tothe nearest adult (<0.5 m: accuracy 0.1 m, >0.5 m: accu-racy 0.5 m) on a Visor Pro handheld computer. On average,we obtained five scans per individual (range 2e6).

Data Analyses

Video analysisResponses to playback experiments were quantified

from videotapes using frame-by-frame analysis (12.5frames/s) in Microsoft Windows Movie Maker version5.1. We obtained the following measurements for bothyoung and adults: (1) first response (see above), (2) mostextreme response (see above), (3) latency to respond(time between onset of call and first response, hereafterreferred to as reaction time), (4) response duration (timebetween onset and end of response, which was defined asthe time when normal foraging behaviour was resumedand the individual did not return to vigilant behaviourwithin 30 s), (5) time spent scanning the surroundings(‘scanning time’) after the most extreme response, ifthis response involved movement (scanning time forlooking responses was included in response duration),(6) the distance (�2 m or >2 m) to nearest adult and(7) number of individuals within 1 m after the most ex-treme response.

Reaction time, response duration and scanning timewere extracted only from playbacks conducted afteryoung started foraging with the group. The responses ofyoung and adults may not be independent, so we in-cluded in the analyses only those playbacks in which wecould obtain measurements for both young and adults.Consequently, because we had incomplete data on adultmeasurements, sample size for each variable was reduced.Measurements on reaction time were extracted from 53playbacks in 10 groups, response duration from 45 play-backs in 11 groups and scanning time from 30 playbacks in10 groups.

Statistical analysesAll analyses were conducted using R for Microsoft

Windows version 2.0.1 (R Development Core Team2004) and the software packages nlme (http://www.r-pro-ject.org) and MASS (Venables & Ripley 2002). Where theassumptions of residual normality and variance homosce-dasticity were violated, we transformed continuous vari-ables with a natural logarithm. All tests were two tailedand based on type I sum of squares, thus controlling forpreceding terms in the model. We first calculated theinitial model including all explanatory variables and

appropriate interaction terms. Significance level was setat P < 0.05, and factors with a P value above 0.10 weresequentially dropped. We identified the best model bycomparing them using Akaike’s information criterion(Pinheiro & Bates 2000). Predator type and urgency levelwere also included as potential explanatory variables,but these results are not presented here (L.I. Hollen &M.B. Manser, unpublished data).

We separately analysed behavioural observations, play-back experiments and scans. Owing to the samplingprocedure of some variables and incomplete measure-ments, our data were not fully balanced and sample sizevaries with analyses.

Behavioural observations. Unless otherwise stated, factorsinfluencing the responses of young were analysed witha mixed-effects logistic regression model procedure char-acterized by a binomial error structure and logit linkfunction. Mixed models allow both fixed and randomeffects to be incorporated (Pinheiro & Bates 2000; Ven-ables & Ripley 2002). The models were fitted with penal-ized quasi likelihood estimation (PQL, glmmPQLfunction), which is a log likelihood estimation methodfor generalized models implemented in R (Breslow & Clay-ton 1993; Venables & Ripley 2002). We controlled for re-peated sampling on young within different litters andgroups by fitting individual identity nested within litterand group identity as a random term. Age, sex andhelper-to-pup ratio were fitted as fixed effects. Individualsover 3 months old are more or less independent, so wecalculated helper-to-pup ratios as the number of groupmembers over 3 months old to the number under 3months old. We compared the responses of young andadults by using chi-square tests.

Playback experiments. To investigate factors influencingthe reaction time, response duration and scanning time ofyoung, we used a linear mixed-effects model procedurefitted with residual maximum likelihood estimation(REML, lme function; Venables & Ripley 2002). Groupidentity was fitted as a random term and age, sex,helper-to-pup ratio, the number of individuals within1 m after the most extreme response, and adult time asfixed effects. Unless otherwise stated, we also used a linearmixed-effects model procedure to compare the responsesof young and adults. Playback situation nested withingroup identity was then fitted as a random term. Thisway, we controlled for repeated sampling within groupsand dependencies between young and adults withineach playback. Age class (young or adult) was fitted asa fixed effect. To investigate whether the proximity toadults influenced the likelihood of young looking towardsothers, we used a mixed-effects logistic regression modelprocedure with group identity fitted as a random termand age and distance to adults as fixed effects.

Scans. We used a linear mixed-effects model procedure(REML) with litter identity fitted as a random term andage, sex and helper-to-pup ratio as fixed effects. Scans for

HOLLEN & MANSER: ALARM RESPONSES IN YOUNG MEERKATS 1349

each individual were pooled and the mean distance toadults was used in the analysis.

RESULTS

Responses to Naturally Occurring Alarm Calls

Behavioural observations showed that young and adultsdid not differ in the likelihood of responding (responseslisted in Table 1) to alarm calls (young: 89.4%; adults:90.2%; Yates’ corrected chi-square test: c2

1 ¼ 0:02,P ¼ 0.87). However, in contrast to adults, young often re-sponded to alarm calls uttered in response to nondanger-ous birds (positive response, young: 88%; adults: 22%;c2

1 ¼ 18:76, P < 0.001) but ignored those given in re-sponse to aerial predators far away (negative response,young: 48%; adults: 9%; c2

1 ¼ 7:16, P ¼ 0.007). However,the lack of response to aerial alarm calls never seemed toinvolve a great risk, because adults looked up only brieflyin response to the same calls.

Young foraging with the group, especially before an ageof 90 days, commonly ran to the nearest individual (42%,N ¼ 159) but looked less often towards other individuals(19%, N ¼ 139) when hearing an alarm call. The probabil-ity of responding independently of others increased withthe age of young (running: F1,124 ¼ 11.73, P ¼ 0.0008;scan: F1,105 ¼ 4.78, P ¼ 0.03; Fig. 1a), but was not influ-enced by sex or helper-to-pup ratio (range of P values0.26e0.65).

Although a large proportion (67%, N ¼ 235) of the inde-pendent responses were already adultlike, 82% (N ¼ 33) ofthe non-adultlike responses (not responding excluded)differed considerably from those of adults (significantlydifferent from the hypothesized value of 50%, binomialtest: P < 0.001). Young often moved to shelter whenadults looked up only briefly (19/27 cases) or looked upbriefly when adults moved (8/27). The probability ofshowing adultlike responses increased as young grew older(F1,197 ¼ 12.07, P ¼ 0.0006; Fig. 1b), and 87% of the non-adultlike responses were shown before an age of 90 days.Independent of age (age*sex: F1,197 ¼ 0.43, P ¼ 0.51), fe-male young were more likely to show adultlike responsesthan were male young (F1,197 ¼ 5.21, P ¼ 0.02; Fig. 1b).Helper-to-pup ratio did not influence the probability ofshowing adultlike responses (F1,196 ¼ 0.31, P ¼ 0.58), butlooking towards other individuals as a first response influ-enced the subsequent behaviour of young. Although only23 of 119 most extreme responses were non-adultlike,74% of them were shown by young responding indepen-dently instead of looking towards other individuals (bino-mial test: P ¼ 0.04).

Responses to Playback Experiments

The results of the playback experiments corroboratedobservational data in suggesting that young and adults areequally likely to respond to alarm calls. Both young andadults responded to all alarm call playbacks (N ¼ 123) butto none of the control (birdsong) playbacks (N ¼ 23).However, during the first 2 weeks of emergence, when

playbacks were conducted on young and adults(NY ¼ NA ¼ 14) staying behind at the sleeping burrow, allalarm calls caused the majority of young (79%) to moveeither below ground or closer to the entrance. In contrast,the majority of adults (79%) looked up only briefly in re-sponse to the same alarm calls (Yates’ corrected chi-squaretest: c2

1 ¼ 7:00, P ¼ 0.008).When foraging with the group, young commonly ran to

the nearest individual but seldom looked towards others(running: 50%; looking: 23%, N ¼ 120), supporting the re-sults of observational data. Young staying further awayfrom adults were more likely to look towards other indi-viduals before responding than were young staying inclose proximity (F1,62 ¼ 7.12, P ¼ 0.01). Again, most(82%, N ¼ 38) of the independent responses were alreadyadultlike.

Despite a strong influence of adult reaction time(F1,36 ¼ 34.79, P < 0.001), young reacted more slowlythan did adults (F1,52 ¼ 6.17, P ¼ 0.02; Fig. 2a). Femaleyoung tended to react faster than did male young(F1,36 ¼ 2.95, P ¼ 0.09). However, when the paired adultreaction times were controlled for, this difference disap-peared (F1,36 ¼ 1.74, P ¼ 0.19). Compared to adults, youngalso resumed foraging faster (F1,44 ¼ 17.97, P < 0.001;

Pred

icte

d p

roba

bili

ty

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

30 60 90 120 150 180

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

30 60 90 120 150 180

Age in days since birth

(a)

(b)

Figure 1. Fitted logistic regression curves with the predicted proba-

bility of (a) the response being independent of (:: running; -:scanning the surroundings) and dependent on (C) other group

members and (b) female (:) and male (-) young showing adultlike

responses. Probabilities were calculated from young first foraging

with the group (30 days) until reaching subadult age (180 days).

ANIMAL BEHAVIOUR, 72, 61350

(a)

(b)

(c)

Rea

ctio

n t

ime

(s)

Res

pon

se d

ura

tion

(s)

Scan

nin

g ti

me

(s)

1

2

3

4

10

20

30

40

50

10

20

30

Pup

Pup

Pup

Adult

Adult

Adult

∗

∗∗∗

∗∗∗

53 53

45 45

3030

Figure 2. Differences in (a) reaction time, (b) response duration and

(c) scanning time between young and adults after playbacks of

alarm calls. Numbers indicate number of playbacks. Values are

back-transformed means � 95% confidence intervals. *P < 0.05;***P < 0.001.

Fig. 2b) and spent less time scanning the surroundings(F1,29 ¼ 17.9, P < 0.001; Fig. 2c). Female young resumedforaging more quickly than did male young (duration:F1,30 ¼ 5.52, P ¼ 0.03). Response duration was not influ-enced by adult response duration (F1,27 ¼ 1.14, P ¼ 0.30),or the number of individuals within 1 m after the most ex-treme response (F1,30 ¼ 2.09, P ¼ 0.16). In contrast, youngincreased the time spent scanning with increasing adultscanning time (F1,16 ¼ 11.00, P ¼ 0.004), and tended toscan longer with fewer individuals around (F1,16 ¼ 3.69,P ¼ 0.07). Again, when the paired adult scanning timeswere controlled, this tendency disappeared (F1,16 ¼ 2.17,P ¼ 0.16). Reaction time, response duration and scanningtime did not change during the first 3 months of ageand were not influenced by helper-to-pup ratio (range ofP values: 0.30e0.83).

Proximity to Adults

The distance between young and adults increased withthe age of young (F1,13 ¼ 15.07, P ¼ 0.002; Fig. 3). Inde-pendent of age (age*sex: F1,41 ¼ 0.78, P ¼ 0.38), femaleyoung stayed closer to adults than did male young(F1,41 ¼ 5.65, P ¼ 0.02; Fig. 3). Helper-to-pup ratio didnot influence the distance between young and adults(F1,12 ¼ 0.22, P ¼ 0.64).

DISCUSSION

Although young meerkats were as likely as adults torespond to alarm calls, responses differed from those ofadults in a number of ways. Unlike adults, young oftenresponded to nondangerous stimuli and ran for shelterwhen not necessary. Young also reacted more slowly andresumed foraging earlier than did adults after playbacks ofalarm calls. However, as young grew older, their responsesalso became increasingly adultlike. Behavioural observa-tions showed that the greatest change towards adultlikebehaviour occurred before reaching independence at 3months of age. For unknown reasons, changes in reactiontime, response duration and scanning time seem to occurlater. Although these results may indicate that experience isneeded to adjust the responses of young, correct responseswere shown early in development. One probable explana-tion is that, despite giving parental care, thus providing theopportunity for the pups to learn how to avoid predators,adult meerkats engage in little active defence. Thus, youngmeerkats may be under equally high pressure as young inspecies without parental care to acquire correct alarm callresponses rapidly (e.g. Miller & Blaich 1986; Goth 2001),especially because predation is the major source of pupmortality (Clutton-Brock et al. 1999b).

Despite abundant evidence that animals are capable ofimproving their responses as a result of experience(reviewed in Griffin et al. 2000), these developmentalchanges could be the result of maturation rather than ex-perience. However, the development of antipredator be-haviour is likely to be a complex process that relies onan interaction between maturational processes and learn-ing, so it may be inappropriate to discard the role of either

HOLLEN & MANSER: ALARM RESPONSES IN YOUNG MEERKATS 1351

one of these processes. For example, newly emerged pupsignored playbacks of birdsong but behaved as though theydid not discriminate between different alarm calls andentered a burrow in response to almost all playbacks.Furthermore, young foraging with the group often movedto shelter when adults only looked up briefly. These resultsmight suggest that young are predisposed to recognize fea-tures of alarm calls from other irrelevant sounds, but withtime they learn to discriminate between the differentalarm calls and associate them with the threat that theypose (see also Davies et al. 2004). However, a simple mat-urational change in the discrimination threshold couldalso explain these results. The optimal threshold will de-pend on the costs of treating a nonthreatening signal asan alarm or ignoring a true alarm (Reeve 1989; Shermanet al. 1997). The cost of running below ground duringthe first few weeks of emergence is unlikely to be high be-cause pups remain near their sleeping burrow, but doingso may increase survival because young escape visual de-tection by the predator. Depending on the type of preda-tor, foraging young may also be safer if they are alreadyat shelter when predators approach. Therefore, lowerthresholds for moving could be selected in young becauseof their high predation risk.

Our playback experiments also showed that youngresumed foraging earlier and spent less time scanningthe surroundings than did adults. Many studies haveshown that juvenile mammals are commonly less vigilantthan adults despite their greater risk of predation (e.g.Arenz & Leger 2000). A higher nutritional demand in juve-niles (e.g. Arenz & Leger 2000), and thus strong selectionon intense foraging, may lead young to rely on the vigi-lance of other individuals (Loughry & McDonough1989). In meerkats, factors affecting daily weight gainare likely to be under strong selection because daily weightgain is positively related to survival throughout the first

Age in days since birth

14012010080604020

Ln m

ean

dis

tan

ce t

o ad

ult

2.5

2

1.5

1

0.5

0

–0.5

–1

–1.5

Figure 3. Logarithm of the mean distance to adults as a function of

age and sex with fitted regression lines for male and female young

separately :,ee: male; -,d: female, (Rmale2 ¼ 0.35, Rfemale

2 ¼ 0.44).

year of life (Clutton-Brock et al. 2001). The differencesfound between young and adults may therefore simply re-flect differences in nutritional requirements, but theycould also be because young are unable to assess the riskassociated with the alarm calls.

In addition to associative learning, other learningmechanisms such as simple habituation and/or observa-tional learning (Moore 2004) may be responsible for someof the observed changes in pup behaviour. First, a selectivehabituation process, where responses to frequent nonpre-dator stimuli diminish and responses to infrequent preda-tor stimuli are maintained, could underlie the lack ofresponses to alarm calls given by adults in response tobirds. Second, although young do not simply copy the be-haviour of adults, they occasionally seem to make use ofsocial information provided indirectly from the responsesof other group members. Both the time to react and thetime allocated to vigilance after playbacks were stronglyinfluenced by adult behaviour. In addition, even thoughthe number of adults present did not influence the behav-iour of young, very young individuals in particular oftenran to the nearest individual. This response is similar tothat of infant vervet and squirrel monkeys, Saimiri sciur-eus, which commonly run to their mothers when alarmed(Seyfarth & Cheney 1986; McCowan et al. 2001). Lookingtowards other individuals was less common, but also sim-ilar to responses of vervet monkeys (Seyfarth & Cheney1986); young doing so were more likely to show adultlikeresponses than young that responded independently.

Young could, however, gather helpful cues from otherindividuals without intentionally seeking them. First, theresult that young reacted more slowly than adults afterplayback of an alarm call suggests that young may wait foradults to react before reacting themselves. Second, youngstaying further away from adults were more likely to looktowards others than were young staying closer. Finally,female young, which remained in closer contact to othergroup members than did male young, tended to reactfaster and were also more likely to show adultlike re-sponses. This result does not necessarily mean thatfemales are generally better at responding, especiallybecause the reaction time was strongly influenced byadult time, but young females may be more likely togather cues from others. However, the difference indistance to adults between male and female young wasfairly small, so whether this has implications for theability to show adultlike responses remains to be thor-oughly investigated.

Our results may also be compatible with a differentperspective, formulated for both mammals (Owings &Loughry 1985; Hersek & Owings 1994; Hoffman et al.1999; Hanson & Coss 2001) and birds (Platzen & Magrath2005), suggesting that, instead of viewing the responsesof young as imperfect versions of adult responses, youngshould be seen as adopting an optimal response for their de-velopmental stage. The idea seems to be widely applicable,and it would not be surprising that selection may favour re-sponses that increase juvenile survival at their current stagein development, given their high vulnerability. However,whether the behaviour of meerkat young represents age-adaptive changes or not remains an open question, because

ANIMAL BEHAVIOUR, 72, 61352

our results seem equally consistent with the idea of youngshowing imperfect versions of adult behaviour.

Although we cannot rule out the possibility that youngcould be responding directly to predators or nondangerousbirds rather than to any call uttered by other groupmembers during natural observations, this conclusionseems unlikely where predators are concerned. First, youngindividuals mostly showed a response before detecting thepredator themselves. Second, the results of our playbackexperiments, in which the presence of a predator wascontrolled for, corroborated those of natural observations.However, even with playback experiments, we cannotdiscount that young may occasionally respond to cuesfrom other group members rather than the alarm callsthemselves.

To conclude, we suggest that the observed differences inalarm call responses between young and adult meerkatscould be caused by several mechanisms (and their in-teractions), including maturation, nonassociative learningand associative learning. Furthermore, although youngmay have the capacity to acquire appropriate alarm callresponses independently, our results indicate that theymay gain from using cues available from other groupmembers. Perhaps as a result of spending less time in closeproximity to adults as they grow older, adultlike andindependent responses emerge over development (Sey-farth & Cheney 1980; McCowan et al. 2001). Early inlife, when young forage close to adults, it might be easierand safer for them to run to the nearest individual. Forolder individuals foraging further away from adults, thisresponse would be inefficient and they may be better offresponding independently. These developmental trendsobserved in young meerkats show parallels with findingsin nonhuman primates, ground squirrels and birds (Sey-farth & Cheney 1986; Mateo 1996a,b; Hanson & Coss2001; Platzen & Magrath 2005), indicating that predationrisk leads not only to convergence in alarm call systems(Macedonia & Evans 1993; Evans 1997), but also to similardevelopmental trajectories.

Acknowledgments

We are indebted to Tim Clutton-Brock for scientific adviceand for allowing us to work on the meerkat population ofthe Kalahari Meerkat Project. We are also grateful to Mrand Mrs Hennie Kotze for the permission to work on theirland, to Johan Du Toit and Martin Haupt at the MammalResearch Institute, University of Pretoria, for logisticalsupport, and to all students and volunteers contributinghelp throughout this study. Thanks to Hansjoerg Kunc,Alan McElligott and three anonymous referees for helpfulcomments on the manuscript. This project was funded bya grant given to M.B.M. from the Swiss National ScienceFoundations, SNF-Forderprofessur No. 631-066129.

References

Arenz, C. L. & Leger, D. W. 2000. Antipredator vigilance of juvenile

and adult thirteen-lined ground squirrels and the role of nutritionalneed. Animal Behaviour, 59, 535e541.

Berger, J., Swenson, J. E. & Persson, I.-L. 2001. Recolonizing carni-

vores and naive prey: conservation lessons from Pleistocene ex-

tinctions. Science, 291, 1036e1039.

Berglund, E., Eriksson, M. & Westerlund, M. 2005. Communica-

tive skills in relation to gender, birth order, childcare and socioeco-nomic status in 18-month-old children. Scandinavian Journal of

Psychology, 46, 485e491.

Breslow, N. E. & Clayton, D. G. 1993. Approximate inference in

generalized linear mixed models. Journal of the American Statistical

Association, 88, 9e25.

Clutton-Brock, T. H. 1991. The Evolution of Parental Care. Princeton,

New Jersey: Princeton University Press.

Clutton-Brock, T. H., MacColl, A., Chadwick, P., Gaynor, D.,Kansky, R. & Skinner, J. D. 1999a. Reproduction and survival

of suricates (Suricata suricatta) in the southern Kalahari. AfricanJournal of Ecology, 37, 69e80.

Clutton-Brock, T. H., Gaynor, D., McIlrath, G. M., MacColl,A. D. C., Kansky, R., Chadwick, P., Manser, M., Skinner,J. D. & Brotherton, P. N. M. 1999b. Predation, group size and

mortality in a cooperative mongoose, Suricata suricatta. Journal

of Animal Ecology, 68, 672e683.

Clutton-Brock, T. H., Russell, A. F., Sharpe, L. L., Brotherton,P. N. M., McIlrath, G. M., White, S. & Cameron, E. Z. 2001.Effects of helpers on juvenile development and survival in

meerkats. Science, 293, 2446e2449.

Cook, M., Mineka, S., Wolkenstein, B. & Laitsch, K. 1985. Obser-

vational conditioning of snake fear in unrelated rhesus monkeys.

Journal of Abnormal Psychology, 94, 591e610.

Davies, N. B., Madden, J. R. & Butchart, S. H. M. 2004. Learning

fine-tunes a specific response of nestlings to the parental alarm

calls of their own species. Proceedings of the Royal Society of LondonSeries B, 271, 2297e2304.

Doolan, S. P. & Macdonald, D. W. 1997. Breeding and juvenilesurvival among slender-tailed meerkats Suricata suricatta in

the south-western Kalahari: ecological and social influences.

Journal of Zoology, 242, 309e327.

Doolan, S. P. & Macdonald, D. W. 1999. Co-operative rearing

by slender-tailed meerkats (Suricata suricatta) in the southern

Kalahari. Ethology, 105, 851e866.

Evans, C. S. 1997. Referential signals. Perspectives in Ethology, 12,

99e143.

Galsworthy, M., Dionne, G., Dale, P. & Plomin, R. 2000. Sex dif-

ferences in early verbal and non-verbal cognitive development.Developmental Science, 3, 206e215.

Gouzoules, H. & Gouzoules, S. 1989. Design features and develop-mental modification of pigtail macaque, Macaca nemestrina, ago-

nistic screams. Animal Behaviour, 37, 383e401.

Griffin, A. S., Blumstein, D. T. & Evans, C. 2000. Training captive-

bred or translocated animals to avoid predators. Conservation

Biology, 14, 1317e1326.

Griffin, A. S. 2004. Social learning about predators: a review and

prospectus. Learning and Behavior, 32, 131e140.

Goth, A. 2001. Innate predator-recognition in Australian brush-

turkey (Alectura lathami, Megapodiidae) hatchlings. Behaviour,

138, 117e136.

Hanson, M. T. & Coss, R. G. 2001. Age differences in the response

of California ground squirrels (Spermophilus beecheyi) to conspe-cific alarm calls. Ethology, 107, 259e275.

Hauser, M. D. 1988. How infant vervet monkeys learn to recognize

starling alarm calls: the role of experience. Behaviour, 105, 187e201.

Hersek, M. J. & Owings, D. H. 1994. Tail flagging by young Cali-fornia ground-squirrels, Spermophilus beecheyi: age-specific partic-

ipation in a tonic communicative system. Animal Behaviour, 48,

803e811.

HOLLEN & MANSER: ALARM RESPONSES IN YOUNG MEERKATS 1353

Hodge, S. J. 2003. The evolution of cooperation in the communal

breeding banded mongoose. Ph.D. thesis, University of

Cambridge.

Hoffman, C. M., Flory, G. S. & Alberts, J. R. 1999. Ontogenetic ad-

aptation and learning: a developmental constraint in learning fora thermal reinforcer. Developmental Psychobiology, 34, 73e86.

Impekoven, M. 1976. Responses of laughing gull chicks (Larusatricilla) to parental attraction calls and alarm calls, and effects of

prenatal auditory experience on responsiveness to such calls.

Behaviour, 56, 250e278.

Lima, S. L. & Dill, L. M. 1990. Behavioral decisions made under the

risk of predation: a review and prospectus. Canadian Journal of

Zoology, 68, 619e640.

Loughry, W. J. & McDonough, C. M. 1989. Calling and vigilance in

California ground squirrels: age, sex and seasonal differences in re-sponses to calls. American Midland Naturalist, 121, 312e321.

McCowan, B., Franceschini, N. V. & Vicino, G. A. 2001. Age differ-ences and developmental trends in alarm peep responses by squir-

rel monkeys (Saimiri sciureus). American Journal of Primatology, 53,

19e31.

Macedonia, J. M. & Evans, C. S. 1993. Variation among mammalian

alarm call systems and the problem of meaning in animal signals.

Ethology, 93, 177e197.

Manser, M. B. 1998. The evolution of auditory communication in

suricates, Suricata suricatta. Ph.D. thesis, University of Cambridge.

Manser, M. B. 2001. The acoustic structure of suricates’ alarm

calls varies with predator type and the level of response ur-gency. Proceedings of the Royal Society of London Series B, 268,

2315e2324.

Manser, M. B., Bell, M. B. & Fletcher, L. B. 2001. The information

that receivers extract from alarm calls in suricates. Proceedings of

the Royal Society of London Series B, 268, 2485e2491.

Martin, P. & Bateson, P. 1993. Measuring Behaviour: An Introductory

Guide. second edn. Cambridge: Cambridge University Press.

Mateo, J. M. 1996a. The development of alarm-call response behav-

iour in free-living juvenile Belding’s ground squirrels. Animal

Behaviour, 52, 489e505.

Mateo, J. M. 1996b. Early auditory experience and the ontogeny of

alarm-call discrimination in Belding’s ground squirrels (Spermophi-lus beldingi). Journal of Comparative Psychology, 110, 115e124.

Mateo, J. M. & Holmes, W. G. 1997. Development of alarm-call

responses in Belding’s ground squirrels: the role of dams. AnimalBehaviour, 54, 509e524.

Miller, D. B. & Blaich, C. F. 1986. Alarm call responsivity of mallard

ducklings. 3. Acoustic features affecting behavioral inhibition.

Developmental Psychobiology, 19, 291e301.

Moore, B. R. 2004. The evolution of learning. Biological Reviews, 79,

301e335.

Owings, D. H. & Loughry, W. J. 1985. Variation in snake elicited

jump-yipping by black-tailed prairie dogs: ontogeny and snakespecificity. Journal of Comparative Ethology, 70, 177e200.

Pinheiro, J. C. & Bates, D. M. 2000. Mixed-Effects Models in S andS-plus. New York: Springer.

Platzen, D. & Magrath, R. D. 2004. Parental alarm calls suppressnestling vocalization. Proceedings of the Royal Society of London,

Series B, 271, 1271e1276.

Platzen, D. & Magrath, R. D. 2005. Adaptive differences in re-

sponse to two types of parental alarm call in altricial nestlings. Pro-

ceedings of the Royal Society of London Series B, 272, 1101e1106.

R Development Core Team. 2004. R: a Language and Environment

For Statistical Computing. Vienna: R Foundation for Statistical

Computing.

Reeve, H. K. 1989. The evolution of conspecific acceptance thresh-

olds. American Naturalist, 133, 407e435.

Russell, A. F., Clutton-Brock, T. H., Brotherton, P. N. M., Sharpe,L. L., McIlrath, G. M., Dalerum, F. D., Cameron, E. Z. &Barnard, J. A. 2002. Factors affecting pup growth and survival

in co-operatively breeding meerkats Suricata suricatta. Journal

of Animal Ecology, 71, 700e709.

Seyfarth, R. M. & Cheney, D. L. 1980. The ontogeny of vervet

monkey alarm calling behavior: a preliminary report. Journal of

Comparative Ethology, 54, 37e56.

Seyfarth, R. M. & Cheney, D. L. 1986. Vocal development in vervet

monkeys. Animal Behaviour, 34, 1640e1658.

Seyfarth, R. M. & Cheney, D. L. 1997. Some features of vocal de-

velopment in nonhuman primates. In: Social Influences on Vocal

Development (Ed. by C. T. Snowdon & M. Hausberger), pp.249e273. Cambridge: Cambridge University Press.

Sherman, P. W., Reeve, H. K. & Pfennig, D. W. 1997. Recognitionsystems. In: Behavioural Ecology: An Evolutionary Approach, fourth

edn. (Ed. by J. R. Krebs & N. B. Davies), pp. 69e96. Oxford:

Blackwell Science.

Venables, W. N. & Ripley, B. D. 2002. Modern Applied Statistics with

S. fourth edn. New York: Springer.

Yamaguchi, A. 2001. Sex differences in vocal learning in birds.

Nature, 411, 257e258.