The Evolution of Urgency-Based and Functionally Referential Alarm Calls in Ground-Dwelling Species

(1)

University of Zurich

Zurich Open Repository and Archive

Winterthurerstr. 190 CH-8057 Zurich http://www.zora.uzh.ch

Year: 2009

The Evolution of Urgency-Based and Functionally Referential

Alarm Calls in Ground-Dwelling Species

Furrer, R D; Manser, M B

Furrer, R D; Manser, M B (2009). The Evolution of Urgency-Based and Functionally Referential Alarm Calls in Ground-Dwelling Species. American Naturalist, 173(3):400-410.

Postprint available at: http://www.zora.uzh.ch

Posted at the Zurich Open Repository and Archive, University of Zurich. http://www.zora.uzh.ch

Originally published at:

American Naturalist 2009, 173(3):400-410.

Furrer, R D; Manser, M B (2009). The Evolution of Urgency-Based and Functionally Referential Alarm Calls in Ground-Dwelling Species. American Naturalist, 173(3):400-410.

Postprint available at: http://www.zora.uzh.ch

Posted at the Zurich Open Repository and Archive, University of Zurich. http://www.zora.uzh.ch

Originally published at:

(2)

Alarm Calls in Ground-Dwelling Species

Abstract

A major evolutionary force driving functionally referential alarm calls is the need for different strategies to escape various predator types in complex structured habitats. In contrast, a single escape strategy appears to be sufficient in less-structured open habitats, and under such conditions urgency-dependent alarm calls may be favored. Nevertheless, some species, such as meerkats (Suricata suricatta), have evolved functionally referential alarm calls despite living in open areas, using only bolt-holes for retreat. To understand the evolution of different alarm call systems, we investigated the calls of sympatric Cape ground squirrels (Xerus inauris) and compared their antipredator and foraging behavior with that of meerkats. Cape ground squirrels emitted urgency-dependent alarm calls and responded to playbacks depending on urgency, not predator type. Vigilance behavior and habitat use differed between the two species. Meerkats roam widely to find prey and for efficient foraging depend on coordinated predator vigilance and escape behavior. As herbivores with smaller territories, Cape ground squirrels depend less on coordinated antipredator behavior, and urgency-dependent alarm calls encode all essential

information. We conclude that habitat complexity does not explain the evolution of functionally referential alarm calls in all species, and other constraints, such as the need to coordinate group movements to maintain foraging efficiency, could be more relevant.

(3)

vol. 173, no. 3 the american naturalist march 2009

The Evolution of Urgency-Based and Functionally Referential

Alarm Calls in Ground-Dwelling Species

Roman D. Furrer

*

and Marta B. Manser

Animal Behaviour, Zoological Institute, University of Zurich, Winterthurerstrasse 190, CH-8057 Zurich, Switzerland Submitted July 24, 2008; Accepted October 27, 2008; Electronically published January 27, 2009

abstract: A major evolutionary force driving functionally refer-ential alarm calls is the need for different strategies to escape various predator types in complex structured habitats. In contrast, a single escape strategy appears to be sufficient in less-structured open hab-itats, and under such conditions urgency-dependent alarm calls may be favored. Nevertheless, some species, such as meerkats (Suricata suricatta), have evolved functionally referential alarm calls despite living in open areas, using only bolt-holes for retreat. To understand the evolution of different alarm call systems, we investigated the calls of sympatric Cape ground squirrels (Xerus inauris) and compared their antipredator and foraging behavior with that of meerkats. Cape ground squirrels emitted urgency-dependent alarm calls and re-sponded to playbacks depending on urgency, not predator type. Vigi-lance behavior and habitat use differed between the two species. Meerkats roam widely to find prey and for efficient foraging depend on coordinated predator vigilance and escape behavior. As herbivores with smaller territories, Cape ground squirrels depend less on coor-dinated antipredator behavior, and urgency-dependent alarm calls encode all essential information. We conclude that habitat complexity does not explain the evolution of functionally referential alarm calls in all species, and other constraints, such as the need to coordinate group movements to maintain foraging efficiency, could be more relevant.

Keywords: vocal communication, alarm calls, functionally referential alarm calls, urgency-based alarm calls.

Introduction

Many species emit alarm calls when approached by pred-ators (Klump and Shalter 1984). Their acoustic structure can vary according to the context in which they are emitted (Marler 1955; Sherman 1977) and can encode information about external objects or events, the individual’s level of arousal (Macedonia and Evans 1993), or both (Manser 2001). In vervet monkeys (Cercopithecus aethiops), alarm calls serve as “functionally referential” signals, where the acoustic structure denotes the predator type and evokes * Corresponding author; e-mail: [email protected].

DOI: 10.1086/596541

specific responses by receivers (Seyfarth and Cheney 1980; Seyfarth et al. 1980). On the other hand, the alarm calls of Belding’s ground squirrels (Spermophilus beldingi) ap-pear to denote different levels of response urgency (Rob-inson 1980). These vocalizations vary with speed and dis-tance of the approaching animal rather than predator type (Robinson 1981). Both functionally referential and urgency-dependent alarm call systems have been found in mammal and bird taxa (see Leavesley and Magrath 2005). The evolution of functionally referential alarm calls has been explained by the need for more than one adaptive way of escaping in response to predators with different hunting strategies (Marler 1967; Cheney and Seyfarth 1990; Macedonia and Evans 1993). Ring-tailed lemurs

(Le-mur catta) are semiarboreal and live in rather open

wood-land. In this complex habitat, they are hunted in different ways by different predators from which they escape using specific strategies. Therefore, it is crucial that information about the approaching predator type is encoded in the alarm call (Macedonia and Evans 1993). In the same way, the evolution of predator-specific calls in red squirrels (Tamiasciurus hudsonicus) has been explained (Greene and Meagher 1998). In contrast, species that live in relatively open habitats—for example, some ground-dwelling ro-dents—run to their burrows in response to any predator type (Blumstein and Armitage 1997a). For these species, information about the urgency to respond may be more important than that about predator type. However, this does not explain why meerkats have evolved functionally referential alarm calls, since they live like many ground-dwelling rodents on relatively open, unstructured plains and use several burrow systems to retreat (Doolan and MacDonald 1996; Manser and Bell 2004). Unlike other species that emit referential alarm calls, their only escape from predators is to retreat into a bolt-hole (Manser et al. 2001). Therefore, not only habitat structure in com-bination with different hunting techniques of predators but also other factors like the social complexity of a species might have an influence on the evolution of functionally

(4)

referential alarm calls (Blumstein and Armitage 1997b; Manser 2001).

In some arid areas of southern Africa, meerkats live sympatrically with Cape ground squirrels (Xerus inauris) and often share their sleeping burrows. Female Cape ground squirrels form social groups of one to three adult and up to nine subadult individuals, which are charac-terized by female philopatry and male-biased dispersal (Waterman 1997). Males form bands that can include up to 19 individuals (Waterman 1997), within which tem-porary subbands form (Waterman 1995). Cape ground squirrels feed on vegetation close to their burrow systems and are often vigilant when aboveground (Herzig-Straschil 1978). In situations of impending danger, they often emit a very high-pitched whistle that causes receivers to scan the area or to run toward the entrance of the burrow system (Herzig-Straschil 1978). Meerkats live in groups of up to 50 individuals, with mainly the dominant pair breed-ing and their offsprbreed-ing delaybreed-ing dispersal and helpbreed-ing to rear the pups (Doolan and MacDonald 1997; Clutton-Brock et al. 1999a). They roam widely in their large ter-ritories from one burrow system to another (Manser and Bell 2004) and forage as a cohesive group, often with a sentinel on duty (Doolan and MacDonald 1997). They have evolved a complex alarm call system with functionally referential alarm calls as well as nonspecific alarm calls (Manser 2001).

If Cape ground squirrels actually produce urgency-based alarm calls, as many ground-dwelling rodent species do (e.g., Owings and Virginia 1978; Robinson 1980; Blum-stein and Arnold 1995; BlumBlum-stein and Armitage 1997a; Le Roux et al. 2001; Randall and Rogovin 2002), then this environment would house two ground-dwelling, burrow-using small mammal species that face the same predation pressure but use different alarm call systems. If so, we might predict that interspecific differences in vigilance, antipredator behavior, and habitat use may provide in-sights into whether factors other than the need for different escape strategies may have an influence on the evolution of particular alarm call systems. In this study, we inves-tigated the alarm call production of Cape ground squirrels (“Alarm Call Production in Cape Ground Squirrels”), tested their responses to different alarm calls (“Responses to Playback of Alarm Calls in Cape Ground Squirrels”), and collected data on vigilance, antipredator behavior, and habitat use that could influence the evolution of alarm call systems in Cape ground squirrels and meerkats (“Be-havioral Comparisons between Cape Ground Squirrels and Meerkats”).

Study Site and Animals

Data were collected from July to December 2004 and from December 2005 to February 2006 on a wild population of

Cape ground squirrels living in the southern Kalahari Des-ert, approximately 30 km west of Van Zylsrus, South Africa (26⬚58⬘S, 21⬚49⬘E; for details of the study area see Russell et al. 2002). The study population consisted of 18 groups of Cape ground squirrels, ranging from three to 16 non-habituated individuals. Observations were made from dis-tances of 20–100 m, to avoid influencing their behavior. Experiments were conducted with the observer hidden from the subjects. No comprehensive survey on predation of the Cape ground squirrels has been undertaken, but Clutton-Brock and others (1999a) provide data on pre-dation of the sympatric meerkat. For behavioral compar-isons between Cape ground squirrels and meerkats, we also collected data on six fully habituated social groups of meerkats that allowed observers to follow them within a few meters (Clutton-Brock et al. 1998).

Alarm Call Production in Cape Ground Squirrels

Methods

Recordings of Alarm Calls. We recorded alarm calls during

naturally occurring predator encounters and during ma-nipulation experiments while Cape ground squirrels were foraging. The stimulus that elicited alarm calls and the distance between stimulus and caller were noted. Record-ings of vocalizations emitted by adults were analyzed. Since few alarm calls could be recorded during naturally oc-curring predator encounters, we conducted the following manipulation experiments to increase the sample of alarm vocalizations to known stimuli. (a) Terrestrial predator: a person walked toward the focal squirrels at a constant rate (cf. Davis 1991). (b) Aerial predator: two kites were used, one in the shape of a dark bird of prey with a wingspan of 1.5 m and the other in the shape and color of a martial eagle (Polemaetus bellicosus) with a wingspan of 2 m. These kites were flown toward the squirrels in such a way that the human operator could not be seen by them. (c) Snakes: a defrosted dead mole snake (Pseudaspis cana) and puff adder (Bitis arietans) or live Cape cobra (Naja nivea) in a transparent Plexiglas box (40 cm # 50 cm # 40 cm, with holes for oxygen flow) were presented adjacent to the burrow system of a social group of squirrels. The empty Plexiglas box was presented as a control to all six tested groups once before each trial, and the squirrels never re-sponded to it.

Alarm calls were recorded at a sampling frequency of 44.1 kHz using a Marantz PMD670 solid-state recorder and Sennheiser ME 66/K6 directional microphones. Since pure ultrasonic alarm calls were described in Richardson’s ground squirrels recently (Wilson and Hare 2004), we also recorded alarm calls in some of the manipulation exper-iments at a sampling frequency of 192 kHz, using a Fostex

(5)

402 The American Naturalist

Table 1: Different alarm call types labeled to different contexts

Call type, level of urgency, and predator type

Calls analyzed Distance (m) Nonmoving predator: Lowest: Terrestrial 20 Variable (10–100) Aerial 2 Variable (1₅₀₎ Lower urgency: Low: Terrestrial 13 1₃₀ Aerial 8 1₁₀₀ Higher urgency: High: Terrestrial 16 !₃₀ Aerial 9 !₁₀₀

field memory recorder FR-2 and an Avisoft ultrasound gate CPVS P48 microphone. To increase the number of high-quality recordings, two to four microphones were placed on the predicted foraging area around the burrow system. The distance between the subject and the micro-phone did not exceed 10 m for the former and 5 m for the latter microphone type. To avoid the influence of in-dividual variation in alarm call production, recordings from at least 10 different groups were included for each vocalization type.

Categories of Calls Depending on Context. We categorized

calls according to predator type and urgency level. For aerial predators, stimuli within 100 m of the focal indi-viduals were classified as “close”; stimuli farther away were “far.” For terrestrial predators, stimuli within 30 m were classified as close, and stimuli 30 m or farther were far. Encounters with terrestrial and aerial predators at close distance were labeled as higher-urgency situations, and encounters with both predator types at far distance were labeled as lower-urgency situations.

Acoustic and Statistical Analysis. Spectrograms, displayed

with Flat Top window function, were generated using Avisoft-SAS LAB PRO software by conducting a 256 short-time Fourier transformation with a short-time resolution of 2.9 ms and a frequency range of 648 Hz. We measured 18 different acoustic parameters describing temporal and fre-quency dynamics, as well as bandwidth and entropy, from the spectrogram. Additionally, in calls recorded with the Avisoft CPVS P48 microphone, energy distribution along the frequency axis of the spectrograms was calculated. One call per group per manipulation experiment or predator encounter was analyzed in nonrepetitive alarm call types. In repetitively emitted alarm calls, five calls distributed over the total calling bout were analyzed, and the mean for each acoustic parameter was calculated.

Statistical analyses were conducted using SPSS 11.0. Four acoustic parameters that correlated highly (r1_0.8) with other parameters were excluded (on the basis of Spearman correlations). With the remaining 14 acoustic parameters, we conducted single-factor ANOVAs to reveal statistically significant differences—first, between calls elic-ited with different levels of urgency and, second, between calls evoked by different predator types. Significant pa-rameters were then entered into a discriminant function analysis (DFA) to determine classification probabilities of alarm calls produced. A DFA identifies linear combinations of predictor variables that best characterize the differences among groups and assigns each call to its appropriate group (correct assignment) or to another group (incorrect assignment). For external validation, we used a leave-one-out cross-validation procedure. The average percentage of

correct assignment found was compared with correct as-signment by chance (bootstrapping with 10,000 runs). To minimize a possible bias in our results due to the relatively low sample size of calls evoked by aerial predators (19) compared with those from terrestrial predators (49), we randomly chose equal-sized subsets of calls elicited by ter-restrial predators and repeated the DFA for predator type 10 times.

Results

Alarm Calls in Response to Terrestrial and Aerial Predators and Snakes. When detecting a stationary terrestrial

pred-ator, subjects emitted nonmoving predator alarm calls re-petitively, independent of distances between caller and stimulus. Nonmoving predator alarm calls were recorded for perched aerial predators only twice, and this call type was never produced toward a flying bird of prey or kite. When an approaching but distant aerial or terrestrial pred-ator was detected, subjects emitted one or a few lower-urgency alarm calls. Close predator encounters evoked a very short, higher-urgency alarm call, which was generally not repeated (table 1; fig. 1). In all of the six calls per call type analyzed, we found one to three weak frequency bands in ultrasonic frequency range. However, as 90%–95% of the total energy was found in the frequency range below 20 kHz, the ultrasonic component is unlikely to be of importance. Cape ground squirrels detected the mole snake (n p 10) and puff adder (n p 8) in all presenta-tions, and in six out of seven presentations the Cape cobra was detected. No vocalizations were recorded in any of the presentations using either microphone type.

Acoustic Structure Depending on Context. The single-factor

ANOVA revealed that seven acoustic parameters differed across urgency levels and that five parameters differed be-tween the predator types evoking the alarm calls (table 2).

(6)

Figure 1:Spectrograms of different alarm call types recorded at a sampling frequency of 44.1 kHz with the use of Marantz PMD670 solid-state recorders and Sennheiser ME 66/K6 directional microphones. Alarm call types: A, nonmoving predator; B, lower urgency; C, higher urgency.

In the DFA conducted to test the classification probabilities of calls of varying urgency levels, the first discriminant function explained 97.9% of the total variance, and calls showed a 95.6% correct assignment, compared with 33% expected by chance (bootstrapping;P!_.0001), to the ap-propriate contexts (fig. 2). In the DFA conducted to classify calls on the basis of predator type (aerial vs. terrestrial), calls yielded a correct assignment of only 66.4%Ⳳ 4.1% (meanⳲ SD over 10 repeats), compared with 50% ex-pected by chance (bootstrapping;P p .26).

Responses to Playback of Alarm Calls in Cape Ground Squirrels

Methods

Experimental Procedure. To investigate whether receivers

extract specific information from distinctive alarm call types and show different behavioral responses, we con-ducted playbacks of alarm calls (lower urgency, higher urgency, nonmoving predator) as well as a “growl” that was emitted during aggressive interactions between con-specifics (table 3). In total, 43 calls or call sequences with a high signal-to-noise ratio were used in 50 playbacks to 18 different groups. Each alarm call type was tested once per group. Playbacks of higher- and lower-urgency alarm calls consisted of only one call, whereas for nonmoving predator alarms a sequence of several calls lasting 10 s was

played back. Playbacks were never conducted using calls recorded from the same group as the subject. Calls were played from a Marantz PMD670 solid-state recorder con-nected to a Sony Walkman SR A60 speaker. The volume of the calls was adjusted to the amplitude observed for calls given during naturally occurring predator encounters. Behavioral responses of a single subject were filmed with a Sony DCR-PCI20E digital video camera. Playbacks were conducted while subjects were foraging some meters away from the nearest bolt-hole (6.57Ⳳ 3.59 m, mean Ⳳ SD). Only one adult or subadult squirrel that was located within 7–15 m of the hidden speaker was selected as the subject. Only one playback experiment was performed per day per group, followed by at least 5 days without playback to avoid habituation. We analyzed the first behavioral display since the immediate response is crucial, dividing the re-sponse into two categories: running or nonrunning. “Run-ning” included running toward closest shelter, running to next burrow or bolt-hole entrance, and running into bur-row or bolt-hole. “Nonrunning” consisted of being vigilant quadrupedally or being on guard on the hind legs.

Statistical Analyses. To test the influence of alarm call type

on the frequencies of specific first responses, we conducted a logistic regression. To test whether the predator type evoking the alarm vocalizations influenced the subject’s first response in playbacks of lower- and higher-urgency alarm calls, we conducted Fisher’s exact tests.

(7)

Table 2: Single-factor ANOVAs of acoustic parameters

explaining the variation in calls (F value and P value) depending on the level of urgency and the predator type Acoustic parameter F value P value Level of urgency:

Duration 515.43 !_.000

Freq. max. mean 92.36 !_.000

Bandw. end 69.95 !_.000

Bandw. mean 67.75 !_.000

Peak freq. start 29.01 !_.000

Entro. max 4.23 .019

Quart. 75 4.05 .022

Entro. end 2.73 .072

Freq. max. max. 2.63 .080

Freq. max. start 1.76 .180

Entro. mean 1.72 .186

Bandw. start 1.24 .294

Ratio to max. 1.18 .313

Entro. start .89 .415

Predator type:

Peak freq. start 7.13 .010

Entro. end 5.42 .023 Duration 5.26 .025 Entro. max. 5.19 .026 Entro. mean 4.48 .038 Quart. 75 2.90 .093 Entro. start 2.03 .159 Bandw. end .71 .403 Bandw. mean .67 .416

Freq. max. mean .51 .480

Freq. max. max. .49 .485

Bandw. start .44 .511

Ratio to max. .08 .779

Freq. max. start .08 .786

Note: Freq. p frequency; max. p maximum; bandw. p bandwith; entro. p entropy; quart. p quartile.

Results

Subjects always responded within the first 1.5 s of a play-back and performed up to three distinctive behaviors in succession until relaxation. The first behavioral response varied between nonmoving predator, lower-urgency, and higher-urgency alarm calls (binary logistic regression, , , ; fig. 3). In the control

play-2

x p5.903 df p 2 P p .015

backs of growls, subjects never ran but did become vigilant, and in one case the subject did not respond. In playbacks of both lower- and higher-urgency alarm calls, the predator type evoking the alarm calls did not influence the subject’s immediate response (running or nonrunning; Fisher’s ex-act test, lower urgency,P p .64; higher urgency,P1_.9).

Behavioral Comparisons between Cape Ground Squirrels and Meerkats

Methods

We used scans to compare vigilance behavior, head po-sition, distance to nearest neighbor, and distance to nearest shelter between Cape ground squirrels and meerkats. Six groups of both species with no pups foraging were followed on two separate mornings. Scans were conducted every 10 min during 3 h of foraging, which began after the groups left their sleeping burrows. Scans on nonhabituated Cape ground squirrels were carried out with binoculars (Bausch and Lomb, Legacy field 11, 8 # 40) from a vehicle or the top of a sand dune, from distances that appeared not to influence the natural behavior pattern of the sub-jects. Scans on habituated meerkats were conducted by following them closely. In groups with more than eight subadult or adult meerkats, eight individually marked in-dividuals were randomly selected as subjects.

Vigilance Behavior and Head Position. Behavioral activities

conducted by Cape ground squirrels and meerkats away from their burrow systems were categorized and quantified during scans. The following activities were observed in both species: (a) nonvigilant behavior, including resting, grooming, allogrooming, moving, foraging, feeding, fight-ing, play-fightfight-ing, play, and different marking behaviors; and (b) vigilant behavior, comprising being on guard (standing on hind legs) and being vigilant quadrupedally (scanning the surroundings). In addition, raised guard (guarding from an elevated position) was conducted by meerkats only. To determine to what degree individuals of both species are able to scan the surrounding area dur-ing their normal activity patterns, we collected data on their head positions. Head position was divided into three categories depending on which direction the head was pointing: downward, horizontally, and upward.

Interme-diate positions rarely occurred and were excluded from the analysis.

Distance to Nearest Neighbor and to Shelter. Distances to

the nearest neighbor and nearest shelter are factors that might influence vigilance behavior. To estimate the dis-tance to the nearest neighbor in Cape ground squirrels as accurately as possible, we previously placed a grid of wooden sticks (at 5-m intervals) around the burrow and on the predicted foraging ground. This allowed distances to be estimated through binoculars. To determine the dis-tance to the nearest shelter, two disdis-tance zones around the burrow system were marked with wooden sticks. The first line of sticks encompassed a range of 0–3 m to the nearest shelter, and the second circle enclosed a range of 3–10 m to shelter. Scattered bolt-holes away from the burrow

(8)

sys-Figure 2:Discriminant function scores for the call types nonmoving predator (nmp), lower urgency (lu), and higher urgency (hu) yielded 95.6% correct assignment to context-dependent urgency to respond.

tem were marked with sticks separately. During scans, the distances of all focal individuals to the nearest shelter were recorded in four categories (0–3 m, 3–10 m, 10–20 m, and1_{20 m).}

Locations of the focal meerkats were marked with wooden sticks at the moment the scan was conducted. Afterward the distance between the two nearest sticks, which indicates the distance to the nearest neighbor during the scan, was measured using a range finder (Bushnell). To determine the distance to the nearest shelter, the lo-cations of the focal meerkats were marked with wooden sticks during scans. The surrounding area was later searched for bolt-holes or burrow systems. The distance between the location of the focal individual and the shelter was measured using the range finder.

Statistical Analysis. Scans were pooled per group. When

data were normally distributed, t-tests of means from the six observed groups per species were performed (Sokal and Rohlf 1995). Mann-Whitney U-tests were conducted when data were not normally distributed.

Results

Cape ground squirrels and meerkats (bothn p 6groups) did not differ in their proportion of time spent vigilant

(Cape ground squirrels, 18.2% Ⳳ 7.3%, mean Ⳳ SD; meerkats, 19.8% Ⳳ 7.9%; t p 0.43 df p 10 P p .68, , ). The position of the head varied between Cape ground squirrels and meerkats for head pointing downward (t p⫺11.65 df p 10 P, , !_.001), horizontally (_{t p 11.60},

, ), and upward ( , ; fig.

df p 10 P!_.001 _{U p 3 P p .015}

4A). The mean distance to the nearest neighbor tended to be farther in Cape ground squirrels (4.8 Ⳳ 0.99 m, mean Ⳳ SD) than in meerkats (3.8 Ⳳ 0.5 m; t p 2.2, , ). Proportions for staying in different df p 10 P p .052

distance categories from nearest shelter differed between Cape ground squirrels and meerkats for three distance categories (0–3 m, U!_{0.001 P p .002}, ; 10–20 m, _U!

, ; 1_{20 m,} _, _{) and were}

0.001 P p .002 U!_{0.001 P p .002}

nearly equal for the distance category 3–10 m (U p 17, ; fig. 4B). Cape ground squirrels were usually

P p .94

found in the vicinity of their burrow or bolt-holes. They were rarely located more than 10 meters away from the nearest shelter. Meerkat locations were more evenly dis-tributed among the distance categories.

Discussion

Alarm Calls of Cape Ground Squirrels. Cape ground

squir-rel alarm calls contained information about the level of urgency to respond but were not functionally referential

(9)

Table 3: Number of playback experiments per call type in different groups

and the level of urgency and predator type that elicited the call Call type, level of urgency, and

predator type No. playbacks No. different calls No. groups per call type Nonmoving predator: Lowest: Terrestrial 11 10 11 Lower urgency: Low: Terrestrial 11 8 16 Aerial 6 6 Higher urgency: High: Terrestrial 10 10 14 Aerial 6 6 Growla ₆ ₃ ₆

a_{Social call: level of urgency and predator type are not applicable.}

Figure 3:Proportion (%) of shown first behavioral responses to playbacks of nonmoving predator (nmp), lower-urgency (lu), and higher-urgency (hu) alarm calls as well as a growl.

as in the sympatric-living meerkats (Manser 2001). The acoustic structure of the three different alarm call types emitted varied according to the risk the caller faced, and receivers responded to playbacks of the different alarm call types with specific adaptive responses. Subjects mainly re-sponded to playbacks of nonmoving predator alarm calls with vigilance as a first response. As the urgency to respond increased from lower- to higher-urgency alarm calls, sub-jects more often ran immediately for shelter. In contrast to meerkats (Manser et al. 2001), responses did not differ depending on the predator type evoking the calls used in the playbacks, which supports the acoustic analysis that no information about the predator type is encoded.

Behavioral Comparisons between Cape Ground Squirrels and Meerkats. Cape ground squirrels and meerkats were both

vigilant for approximately 20% of the time spent foraging, but they differed in several aspects of their foraging and antipredator behavior. Foraging Cape ground squirrels of-ten picked up their food items from the ground and fed while squatting on their haunches with their heads point-ing horizontally, enablpoint-ing them to scan their surroundpoint-ings while feeding. In contrast, the heads of meerkats were commonly pointing toward the ground and their eyes were frequently underneath the surface of the soil when digging for prey, preventing them from seeing far (Manser et al. 2001). Cape ground squirrels and meerkats were located

(10)

Figure 4:A, Percentages of three distinct head positions (mean⫹ SD; downward, horizontal, upward) in Cape ground squirrels and meerkats. B, Percentages of time spent in different distance categories from nearest shelter (mean_{⫹ SD) in Cape ground squirrels and meerkats.}

within 5 m of the nearest neighbor, on average. This high group cohesion enables both species to profit from com-munal antipredator behavior such as alarm calling. Al-though both species retreat to bolt-holes and burrow sys-tems in dangerous situations, distances to the nearest shelter differed between species. Cape ground squirrels

foraged close to bolt-holes and burrows and quickly tra-versed areas without shelters, whereas foraging meerkats were regularly encountered quite far away from shelter in open areas. This difference is probably caused by the fact that Cape ground squirrels feed on vegetable matter such as grasses, seeds, leaves, and roots (Herzig-Straschil 1978)

(11)

that are usually available close to their shelter. Meerkats prey on mobile invertebrates and vertebrates (Doolan and MacDonald 1996), which they have to search for through-out their relatively large territories. Their coordinated an-tipredator system, with a sentinel emitting alarm calls about approaching predators (Clutton-Brock et al. 1999b), enables foraging group members to substantially decrease their individual vigilance (Manser 1999).

The Evolution of Alarm Calls. As for many

ground-dwelling species living in relatively open, little-structured habitats (see Blumstein 2007), the main response of Cape ground squirrels to an approaching predator is to retreat to a bolt-hole or burrow. In dangerous situations they always run back to burrows or bolt-holes and wait there until the predator moves off. As they use the same exten-sive burrow system with several entrances over long pe-riods (Herzig-Straschil 1978) and rarely forage far away from shelter, the costs of retreating to the burrow and waiting there for prolonged periods may not be that high. Hence, information regarding the degree of risk, which is encoded in an urgency-dependent alarm call system, seems to be sufficient for Cape ground squirrels. Although the meerkats’ final retreat from predators ends by moving to a bolt-hole or burrow, some graded responses occur within this escape strategy. On hearing terrestrial alarm calls, they frequently leave the dangerous area as a group. Moving away from ambushing predators, such as jackals, that often watch meerkats for a long time (Manser et al. 2001) seems to allow individuals to resume foraging earlier than waiting until the predator leaves (Manser 1998). Meerkats need to roam widely to find food, and they frequently change sleeping burrows in their territory (Manser and Bell 2004). Meerkats, but not Cape ground squirrels, have evolved cooperative vigilance, with sentinel duties decreasing the individuals’ time spent vigilant (Manser 1999), likely in-creasing their foraging efficiency (as in pied babblers; Hol-le´n et al. 2008). A meerkat losing contact with the rest of the group due to taking a different escape direction than its group mates may experience severe costs, as single meerkats and small groups suffer much higher predation than larger groups (Clutton-Brock et al. 1999a). Conse-quently, they depend on strong group cohesion and the specific information about predator type encoded in alarm calls to coordinate their escape direction.

Habitat structure has been suggested to be the main factor driving the evolution of functionally referential alarm calls. In mammals, complex habitat structure may result in escape strategies specific to predator type (Mac-edonia and Evans 1993), which consequently require func-tionally referential alarm calls. For birds it has been argued that open vegetation allows them to spot a predator from a distance and enables different adaptive escapes, whereas

in dense vegetation only the level of urgency is important (Evans 1997). Habitat structure, however, does not explain why meerkats evolved functionally referential calls while the sympatric-living Cape ground squirrels did not. The comparison of the antipredator behavior of these two spe-cies suggests instead that group coordination may explain why meerkats evolved functionally referential alarm calls, allowing them to increase their foraging efficiency sub-stantially in an open habitat. This may also explain the evolution of functionally referential alarm calls in the group-foraging dwarf mongoose (Beynon and Rasa 1989), whereas the solitary-foraging yellow mongoose use urgency-dependent alarm calls not related to predator type (Le Roux 2007). Yellow mongoose, as a member of the herpestide family, are phylogenetically closely related to meerkats and dwarf mongoose and live sympatrically with meerkats. They mainly forage solitarily but share their sleeping burrows with other group members. Information on level of response urgency for group members close to shelter in this species may be sufficient, as they do not need specific responses to coordinate group movement while foraging. Whether group coordination may also ex-plain the evolution of functionally referential alarm calls in Gunnison’s prairie dogs (Cynomys gunnisoni), another ground-dwelling species, is not clear. They emit different alarm calls in response to aerial and terrestrial predators (Placer and Slobodchikoff 2000, 2001), with receivers showing adaptive subtle differences to calls elicited by dif-ferent predators (Kiriazis and Slobodchikoff 2006). In marmot species (Marmota spp.), the alarm call repertoire size varies with the social complexity (Blumstein and Ar-mitage 1997b), with more socially complex species pro-ducing more alarm call types (Blumstein 2003). However, evidence that social complexity triggers the evolution of functional reference is lacking, as none of these marmots species emit functionally referential alarm calls (Blumstein 2007).

Our comparison between the sympatric-living Cape ground squirrels and meerkats shows that the evolution of functionally referential alarm calls explained by the di-versity of adaptive escape strategies (Cheney and Seyfarth 1990; Macedonia and Evans 1993) is not based on differ-ences in habitat structure in all species. Cape ground squir-rels and meerkats are exposed to the same predators and use the same shelters for cover. However, they differ in how they utilize their territory, most probably as a result of their different diets. Because of selection pressure to increase their foraging efficiency, meerkats appear to have evolved additional antipredator behaviors, with their sen-tinel system and alarm calls conveying predator specificity and level of urgency (Manser et al. 2002). Even though our results are based on two species only, the comparison to other ground-dwelling species supports our finding that

(12)

habitat structure does not explain the evolution of func-tionally referential alarm calls in all species. Furthermore, for meerkats and ground squirrels, we find examples of closely related species that have evolved functionally ref-erential (e.g., red squirrels, prairie dogs) or urgency-based (e.g., yellow mongoose) alarm calls. This suggests that we can exclude the influence of phylogenetic constraints, for example, on cognitive abilities in the evolution of func-tionally referential alarm calls. We conclude that while in some species different adaptive escape strategies due to habitat structure may trigger the evolution of predator-specific calls, other constraints, such as the coordination of group movements for increased foraging efficiency, should also be considered when investigating the evolution of functionally referential alarm calls.

Acknowledgments

We are indebted to T. Clutton-Brock for allowing us to work on the Cape ground squirrel and meerkat population of the Kalahari Meerkat Project. We are also grateful to all students and volunteers who contributed help through-out this study. Thanks to E. Greene and B. Ko¨nig for general advice and S. English, H. Hedworth, L. Holle´n, N. Jordan, H. Kunc, C. Mu¨ller, and three referees for their comments that greatly improved the manuscript. We also thank W. Blanckenhorn, A. Le Roux, and C. Mu¨ller for statistical assistance. This project was funded by a grant given to M.B.M. from the Swiss National Science Foun-dation, SNF-Fo¨rderprofessur 631-066129.

Literature Cited

Beynon, P., and A. E. Rasa. 1989. Do dwarf mongooses have a lan-guage? warning vocalizations transmit complex information. South African Journal of Science 85:447–450.

Blumstein, D. T. 2003. Social complexity but not the acoustic en-vironment is responsible for the evolution of complex alarm com-munication. Pages 41–49 in R. Ramousse, D. Allaine´, and M. Le Berre, eds. Adaptive strategies and diversity in marmots. Inter-national Marmot Network, Lyon.

———. 2007. The evolution, function, and meaning of marmot alarm communication. Advances in the Study of Behavior 37:371– 401.

Blumstein, D. T., and K. B. Armitage. 1997a. Alarm calling in yellow-bellied marmots. I. The meaning of situationally variable alarm calls. Animal Behaviour 53:143–171.

———. 1997b. Does sociality drive the evolution of communicative complexity? a comparative test with ground-dwelling sciurid alarm calls. American Naturalist 150:179–200.

Blumstein, D. T., and W. Arnold. 1995. Situational specificity in Alpine-marmot alarm communication. Ethology 100:1–13. Cheney, D. L., and R. M. Seyfarth. 1990. How monkeys see the world:

inside the mind of another species. University of Chicago Press, Chicago.

Clutton-Brock, T. H., D. Gaynor, R. Kansky, A. D. C. MacColl, G. McIlrath, P. Chadwick, P. N. M. Brotherton, J. M. O’Riain, M.

Manser, and J. D. Skinner. 1998. Costs of cooperative behaviour in suricates (Suricata suricatta). Proceedings of the Royal Society B: Biological Sciences 265:185–190.

Clutton-Brock, T. H., D. Gaynor, G. McIlrath, A. D. C. MacColl, R. Kansky, P. Chadwick, M. Manser, J. D. Skinner, and P. N. M. Brotherton. 1999a. Predation, group size and mortality in a co-operative mongoose, Suricata suricatta. Journal of Animal Ecology 68:672–683.

Clutton-Brock, T. H., J. M. O’Riain, P. N. M. Brotherton, R. Kansky, A. S. Griffin, and M. Manser. 1999b. Selfish sentinels in cooperative mammals. Science 284:1640–1644.

Davis, R. O. 1991. Semantical communication in antipredator alarm calls. Pages 275–312 in Natural history of eastern California and high-altitude research. University of California, White Mountain Research Station, Los Angeles.

Doolan, S. P., and D. W. MacDonald. 1996. Diet and foraging be-haviour of group-living meerkats, Suricata suricatta, in the south-western Kalahari. Journal of Zoology (London) 239:697–716. ———. 1997. Breeding and juvenile survival among slender-tailed

meerkats (Suricata suricatta) in the south-western Kalahari: eco-logical and social influences. Journal of Zoology (London) 242: 309–327.

Evans, C. S. 1997. Referential signals. Perspectives in Ethology 12: 99–143.

Greene, E., and T. Meagher. 1998. Red squirrels, Tamiasciurus hud-sonicus, produce predator-class specific alarm calls. Animal Be-haviour 55:511–518.

Herzig-Straschil, B. 1978. On the biology of Xerus inauris (Zim-mermann 1780) (Rodentia. Sciuridae). Zeitschrift fu¨r Sa¨ugetier-kunde 43:262–278.

Holle´n, L. I., M. B. V. Bell, and A. N. Radford. 2008. Cooperative sentinel calling? foragers gain increased biomass intake. Current Biology 18:576–579.

Kiriazis, J., and C. N. Slobodchikoff. 2006. Perceptual specificity in the alarm calls of Gunnison’s prairie dogs. Behavioural Processes 73:29–35.

Klump, G. M., and M. D. Shalter. 1984. Acoustic behaviour of birds and mammals in the predator context. I. Factors affecting the structure of alarm signals. II. The functional significance and evo-lution of alarm signals. Zeitschrift fu¨r Tierpsychologie 66:189–226. Leavesley, A. J., and R. D. Magrath. 2005. Communicating about danger: urgency alarm calling in a bird. Animal Behaviour 70:365– 373.

Le Roux, A. 2007. Communication in the yellow mongoose, Cynictis penicillata. PhD diss. Stellenbosch University.

Le Roux, A., T. B. Jackson, and M. I. Cherry. 2001. Does Brants’ whistling rat (Parotomys brantsii) use an urgency-based alarm sys-tem in reaction to aerial and terrestrial predators? Behaviour 138: 757–773.

Macedonia, J. M., and C. S. Evans. 1993. Variation among mam-malian alarm call systems and the problem of meaning in animal signals. Ethology 93:177–197.

Manser, M. B. 1998. The evolution of auditory communication in suricates, Suricata suricatta. PhD diss. University of Cambridge. ———. 1999. Response of foraging group members to sentinel calls

in suricates, Suricata suricatta. Proceedings of the Royal Society B: Biological Sciences 266:1013–1019.

———. 2001. The acoustic structure of suricates’ alarm calls varies with predator type and the level of response urgency. Proceedings of the Royal Society B: Biological Sciences 268:2315–2324.

(13)

Manser, M. B., and M. B. Bell. 2004. Spatial representation of shelter locations in meerkats, Suricata suricatta. Animal Behaviour 68: 151–157.

Manser, M. B., M. B. Bell, and L. B. Fletcher. 2001. The information that receivers extract from alarm calls in suricates. Proceedings of the Royal Society B: Biological Sciences 268:2485–2491. Manser, M. B., R. M. Seyfarth, and D. L. Cheney. 2002. Suricate

alarm calls signal predator class and urgency. Trends in Cognitive Sciences 6:55–57.

Marler, P. 1955. Characteristics of some animal calls. Nature 176: 6–8.

———. 1967. Animal communication signals. Science 157:768–784. Owings, D. H., and R. A. Virginia. 1978. Alarm calls of California ground squirrels (Spermophilus beecheyi). Zeitschrift fu¨r Tierpsy-chologie 46:58–70.

Placer, J., and C. N. Slobodchikoff. 2000. A fuzzy-neural system for identification of species-specific alarm calls of Gunnison’s prairie dogs. Behavioural Processes 52:1–9.

———. 2001. Developing new metrics for the investigation of animal vocalizations. Intelligent Automation and Soft Computing 7:249– 258.

Randall, J. A., and K. A. Rogovin. 2002. Variation in and meaning of alarm calls in a social desert rodent Rhombomys opimus. Ethol-ogy 108:513–527.

Robinson, S. R. 1980. Antipredator behavior and predator recogni-tion in Belding’s ground squirrels. Animal Behaviour 28:840–852.

———. 1981. Alarm communication in Belding’s ground squirrels. Zeitschrift fu¨r Tierpsychologie 56:150–168.

Russell, A. F., T. H. Clutton-Brock, P. N. M. Brotherton, L. L. Sharpe, G. M. McIlrath, F. D. Dalerum, E. Z. Cameron, and J. A. Barnard. 2002. Factors affecting pup growth and survival in co-operatively breeding meerkats Suricata suricatta. Journal of Animal Ecology 71:700–709.

Seyfarth, R. M., and D. L. Cheney. 1980. The ontogeny of vervet monkey alarm calling behavior: a preliminary report. Zeitschrift fu¨r Tierpsychologie 54:37–56.

Seyfarth, R. M., D. L. Cheney, and P. M. Marler. 1980. Vervet monkey alarm calls: semantic communication in a free-ranging primate. Animal Behaviour 28:1070–1094.

Sherman, P. W. 1977. Nepotism and evolution of alarm calls. Science 197:1246–1253.

Sokal, R. R., and F. J. Rohlf. 1995. Biometry. 3rd ed. W. H. Freeman, New York.

Waterman, J. M. 1995. The social-organization of the Cape ground-squirrel (Xerus inauris, Rodentia, Sciuridae). Ethology 101:130– 147.

———. 1997. Why do male Cape ground squirrels live in groups? Animal Behaviour 53:809–817.

Wilson, D. R., and J. F. Hare. 2004. Ground squirrel uses ultrasonic alarms. Nature 430:523.

Associate Editor: Thomas N. Sherratt Editor: Michael C. Whitlock