Save

Joint tail and vocal alarm signals of gray squirrels (Sciurus carolinensis)

In: Behaviour
Authors:
Thaddeus R. McRaeDepartment of Biology, University of Miami, 215 Cox Science Center, 1301 Memorial Drive, Coral Gables, FL 33124-0421, USA

Search for other papers by Thaddeus R. McRae in
https://brill.com/search?f_0=author&q_0=Thaddeus+R.+McRae
Google Scholar
PubMed
Close
and
Steven M. GreenDepartment of Biology, University of Miami, 215 Cox Science Center, 1301 Memorial Drive, Coral Gables, FL 33124-0421, USA

Search for other papers by Steven M. Green in
https://brill.com/search?f_0=author&q_0=Steven+M.+Green
Google Scholar
PubMed
Close
Full Access

Threat-specific vocalizations have been observed in primates and ground squirrels, but their contemporaneous usage with visible signals has not been experimentally analyzed for association with threat type. Here we examine the eastern gray squirrel, an arboreal squirrel that uses both vocal and tail signals as alarms. Squirrels were presented with cat and hawk models simulating natural terrestrial or aerial predator attacks and also with control objects that do not resemble predators but approach in a similar manner. Individuals responded with tail signals (twitches and flags) and vocalizations (kuks, quaas and moans), but only flags and moans are associated with predator type. Moans were elicited primarily by aerial stimuli and flags by terrestrial stimuli. Eastern gray squirrels use an alarm-signaling system in which signals in each modality potentially are associated with particular attributes of a threat or may be general alarms. Terrestrially-approaching stimuli yielded vocal and tail alarm signals regardless of whether the stimulus resembled a predator. With aerially-approaching stimuli, however, quaas were used more often when the stimulus resembled an aerial predator than when it did not. An approaching object’s physical appearance may therefore affect squirrels’ responses to aerial, but not terrestrial, objects. When the stimuli resembled real predators approaching in the natural manner (terrestrially or aerially), both tail flags and vocal moans were associated with predator type, so we also considered moans and flags together. The presence and absence of moans and flags in an alarm signaling bout yields a higher statistical index of predictive association as to whether the threat is aerial or terrestrial than does either component alone.

Abstract

Threat-specific vocalizations have been observed in primates and ground squirrels, but their contemporaneous usage with visible signals has not been experimentally analyzed for association with threat type. Here we examine the eastern gray squirrel, an arboreal squirrel that uses both vocal and tail signals as alarms. Squirrels were presented with cat and hawk models simulating natural terrestrial or aerial predator attacks and also with control objects that do not resemble predators but approach in a similar manner. Individuals responded with tail signals (twitches and flags) and vocalizations (kuks, quaas and moans), but only flags and moans are associated with predator type. Moans were elicited primarily by aerial stimuli and flags by terrestrial stimuli. Eastern gray squirrels use an alarm-signaling system in which signals in each modality potentially are associated with particular attributes of a threat or may be general alarms. Terrestrially-approaching stimuli yielded vocal and tail alarm signals regardless of whether the stimulus resembled a predator. With aerially-approaching stimuli, however, quaas were used more often when the stimulus resembled an aerial predator than when it did not. An approaching object’s physical appearance may therefore affect squirrels’ responses to aerial, but not terrestrial, objects. When the stimuli resembled real predators approaching in the natural manner (terrestrially or aerially), both tail flags and vocal moans were associated with predator type, so we also considered moans and flags together. The presence and absence of moans and flags in an alarm signaling bout yields a higher statistical index of predictive association as to whether the threat is aerial or terrestrial than does either component alone.

1. Introduction

Functionally referential signals are employed as if designating an external object or other stimulus, although they need not be truly referential, and no implication of mental representation is implied (Seyfarth et al., 1980a; Marler et al., 1992). Such signals are (1) associated with some external stimulus and (2) elicit responses similar to those elicited by the signal-eliciting stimulus even when only the signal itself is presented (Marler et al., 1992). This functionality requires some degree of association between signal and stimulus. Aside from studies of human communication, most experimental studies of putative functionally referential signals focus on alarm calls, beginning with the seminal experiments by Seyfarth et al. (1980a, b) verifying that vervet monkeys (Cercopithecus aethiops) in Kenya use functionally referential alarms as described by Struhsaker (1967).

Alarm calling systems of some ground squirrels were initially described as predator-specific and perhaps functionally referential (Owings & Virginia, 1978; Davis, 1984). Later studies, however, suggested that in some species the calls differed in conjunction with speed of approach or distance at first detection (Owings & Leger, 1980; Macedonia & Evans, 1993; Warkentin et al., 2001; Sloan et al., 2005), hence differing, for example, between raptors approaching through the air and carnivores approaching on the ground. Because distant raptors can elicit the putative terrestrial predator alarm call or suddenly appearing carnivores the putative aerial predator alarm call, these alarm calling systems were subsequently labeled as urgency-based rather than functionally referential systems (Macedonia & Evans, 1993). A threat-specific signal based on manner of approach does not imply it is not functionally referential. That call still is associated with a definitive attribute of the threat, and could potentially elicit responses specific to threats sharing that attribute. A call could be associated primarily with a predator’s shape, structure, and color (e.g., raptor versus mammal) or its manner of approach (altitude, speed, etc.). In many cases, it remains unclear whether previously described predator-specific alarm calls are elicited by a predator’s physical appearance or its manner of approach (e.g., fast/slow or aerially/terrestrially). A notable exception is the study by Robinson (1981) of Belding’s ground squirrels. Alarm calls are elicited in Belding’s ground squirrels by walking great egrets but not by flying great egrets and flying raptors elicit alarms more frequently than flying omnivores (Robinson, 1981), suggesting that both predator appearance and manner of approach affect alarm signals in some sciurids.

In courtship signaling, multiple sensory modalities often work in concert to simultaneously produce signals associated with more than one characteristic of the signaler, or communicate more effectively, than one modality alone (Otovic & Partan, 2009; Uhl & Elias, 2011). Studies of alarm signals in terrestrial vertebrates have focused on single modalities, examining vocalizations or tail movements (Hennessy et al., 1981; Swaisgood et al., 2003), but not both modalities together. Arboreal squirrels frequently use both vocalizations and tail signals as alarms (Partan et al., 2009). Whether used together or separately, both tail signals and vocal signals by eastern gray squirrels (Sciurus carolinensis) elicit alarm calls and antipredator behavior in conspecifics, although the response is strongest to visual and auditory stimuli issued together (Partan et al., 2009).

We experimentally investigated the unimodal and multimodal alarm signals of eastern gray squirrels to determine whether they are associated with the physical appearance or manner of approach of a stimulus. Eastern gray squirrels are primarily arboreal, frequently forage on the ground, and are not territorial, thereby providing a comparison to studies of alarm-specificity in semi-terrestrial primates and terrestrial ground squirrels and to the arboreal, but territorial, red squirrel (Tamiasciuris hudsonicus). Eastern gray squirrels in this study occupy habitat structurally similar to that of vervets, consisting of grassy expanses with scattered trees (primarily live oaks (Quercus virginiana) and several palm species) in a subtropical area where there is foliage year-round. Playback experiments confirm that eastern gray squirrels in this population respond to conspecific alarm calls by increasing scanning and decreasing foraging activities (McRae, 2012).

In studies of animal communication the term ‘information’ has been used in a number of different ways, perhaps leading to the wide variety of opinions on the utility of the term and concept in the study of animal behavior, e.g., Searcy & Nowicki (2006), Owren et al. (2010), and Stegmann (2013). In the context of this study of alarm signals in the gray squirrel, the degree of association between signal and stimulus can be considered an index of information content. Used in this sense, the term ‘information’ refers simply to the statistical relationship that constrains the best identification of a stimulus given only the signal, and the term does not imply anything about the cognitive process of the signaler or any receiver that may respond to the signal. Although the term ‘information’ has been used to refer to statistical measures of association, including the index of predictive association that we employ here, we will use the more neutral term ‘stimulus-signal association’.

We first test whether gray squirrel acoustic and visual alarms are associated with predator type, namely models of a terrestrial cat or an aerial hawk. If a signal is associated with predator type, that association could be based on physical appearance of the stimulus object (cat or hawk) or on its manner of approach (terrestrial or aerial). To control for physical resemblance to any particular predator and directly test for effects of manner of approach, we then examine whether each signal type in a squirrel’s response is associated with a terrestrial or aerial control object that does not resemble a predator. Finally, manner of approach is controlled and resemblance to a predator is examined by comparing signals used in response to a model cat versus a terrestrial ball, and also in response to a model flying hawk versus an aerially thrown ball. Taken together, these experiments disambiguate: (1) the effects of the combination of physical appearance and manner of approach as in natural predator encounters; (2) the effects of manner of approach while controlling physical appearance; and (3) the effects of physical appearance while controlling manner of approach.

After separately testing the occurrence of each type of vocal and tail signal for any predator-specificity as unimodal signals, we examine predator-specificity of these vocal and tail signals used together, thus determining if such multimodal signals exhibit greater stimulus association relative to unimodal signals. This study is the first to consider alarms in more than one modality while testing for predator-specific alarm signals.

2. Material and methods

2.1. Location and subjects

Gray squirrels are ubiquitous in the urban areas of the eastern United States and have been introduced to cities around the world (Huynh et al., 2011). We conducted experiments on a wild urban population of eastern gray squirrels within their native range on the University of Miami campus in Coral Gables, FL, USA (80°16.732′W, 25°43.393′N), between 29 January 2010 and 13 July 2011. Forty-eight squirrels were captured using Havahart live traps. Each captured squirrel was weighed, sexed, fitted with sequentially numbered Monel ear tags (style 1005-1 from National Band and Tag Company, color-coded with Testors enamel paint to enhance individual identifiability from a distance), and released at the location of capture. We selected three different colors of paint from a palette of six, providing 120 unique combinations. Animal welfare issues were considered and the study protocol was approved by the University of Miami IACUC under protocol number 08-118.

2.2. Experimental stimuli

To simulate encounters with terrestrial and aerial threats, five kinds of moving stimuli were presented to squirrels — three approaching terrestrially and two approaching aerially (Figure 1). Stimulus presentations were initiated when a focal squirrel subject was on the ground, therefore equally susceptible to aerial and terrestrial threats. The terrestrially approaching stimuli consisted of a battery-powered radio-controlled (RC) wheeled base (a modified Traxxas eRevo mini) remotely driven toward the focal squirrel under three conditions: (1) carrying a realistic model cat approximately 35 cm in body length, resembling a common terrestrial predator, the domestic cat (Felis catus); (2) carrying a red ball approximately 30 cm in diameter, a predator-sized object that does not resemble a predator; or (3) carrying nothing as a control to distinguish effects of terrestrial motion and the electric motor alone from effects specific to the model cat or ball. The two aerially approaching stimuli were thrown so as to pass overhead of the focal squirrel and consisted of: (1) a customized, Styrofoam glider 35 cm in length painted and shaped to resemble a common aerial predator, the adult Cooper’s hawk (Accipiter cooperii); and (2) a ball approximately 30 cm in diameter, identical to that used on the motorized base but thrown through the air. The model hawk and ball were launched at approximately 5 m/s ground speed (gauged from our video recordings) and the motorized models were driven at this same speed. This rapid direct approach without pause was designed to simulate the final stage of a predatory attack, namely the rush by a domestic cat or the swoop of a hawk. The presumed historical felid predator of gray squirrels in this area is the bobcat (Lynx rufus). Although it shares some features with our model, the bobcat is much larger and has a pale reddish brown coat (Lariviere & Walton, 1997) that probably differs somewhat from the neutral gray of our model even to a squirrel’s dichromatic vision. Domestic cats and Cooper’s hawks prey on adult gray squirrels (Meng, 1959; McRae, 2012) and are present in the study area. The moving stimuli therefore resemble terrestrial and aerial predators, each approaching in the appropriate manner, and a novel object, the ball, that could approach in either manner. Based on our incidental observations of natural encounters between squirrels on the ground and hawks or cats, responses to the model predators and real predators do not differ. Whether the predator is a cat or hawk, a typical response involves the squirrel running at least partway up the nearest tree and orienting its head toward the predator, often also tail signaling and alarm calling.

Figure 1.
Figure 1.

The five objects (A–E) used as moving stimuli. Scale bars represent 10 cm, either horizontal distance at ground level in overhead views or vertical distance from the point in the scene at the lower end of the scale bar in other views. A and B show the two stimuli designed to resemble real predators. A1 and A2 show the model cat mounted on the radio-controlled motorized base, viewed from the front (A1) and a 3/4 front view (A2). B1 and B2 show the model hawk glider, viewed from below (B1) and above (B2). C and D show the ball used as a novel object, where the same object can approach either terrestrially or aerially. C1 and C2 show the ball mounted on the base, viewed from a 3/4 front view (C1) and from above (C2). D shows the ball alone, which was thrown by hand in the same manner as the hawk glider. E shows the base alone, which was included to test for specific effects of the ball and cat model. This figure is published in colour in the online edition of this journal, which can be accessed via http://booksandjournals.brillonline.com/content/journals/1568539x.

Citation: Behaviour 151, 10 (2014) ; 10.1163/1568539X-00003194

2.3. Experimental design

In our balanced-order design, each individual squirrel was assigned a unique, predetermined sequence of the five stimulus varieties based on its ear tag number. These sequences were ordered so that the first stimulus presented to each consecutive squirrel rotated through the five stimuli. Once a squirrel subject was seen on the ground exhibiting low-vigilance behavior (foraging, caching, or grooming), audio and video recording began while its assigned stimulus was prepared and sent toward the squirrel. Stimuli were sent directly toward the focal squirrel from an initial distance of 10 m at approximately 5 m/s. All squirrels were within 10 m of a tree when the stimulus was presented. We recorded sound using two Sennheiser ME67 microphones covered with black WindCutter windscreens in either 16-bit linear PCM format at a 48 kHz sampling rate on a Marantz PMD660 or in 24-bit linear PCM format at a 96 kHz sampling rate on a Marantz PMD661. We used a Panasonic HDC-SD10 camcorder at 1080p 30 fps in AVCHD format for video.

Stimuli were in an opaque plastic bin (Rubbermaid Roughneck Storage Box, 68.1 l) when not in use, and the bin screened the stimulus from the focal squirrel’s view until the moment it was sent toward the squirrel. Each trial consisted of one stimulus presentation and lasted from the initial motion of the stimulus until 2 min after the squirrel was silent and had resumed low-vigilance behavior or until the squirrel left the area, whichever occurred first. No squirrel received more than two presentation trials in one day, and if it received two trials in a day they were separated by at least 45 min. For the 15 squirrels tested twice on the same day, there is no association in the incidence of vocalizations (Fisher’s exact two-tailed p=1) or tail signals (Fisher’s exact two-tailed p=1) with the first compared to the second stimulus presentation, so there is no evidence of habituation or sensitization to presentations on the same day.

2.4. Analysis of vocalizations

Squirrel vocalizations were analyzed using Raven Pro 1.4 (The Cornell Lab of Ornithology Bioacoustics Research Program, 2011). Spectrograms were produced using a Hanning window of 1024 samples and each recording was scanned visually on the computer monitor and by ear for squirrel vocalizations. Each was classified as a kuk, quaa, or moan after Lishak (1984). No other vocalization types were found.

2.5. Analysis of tail signals

Each focal squirrel’s use of tail signals was scored using Sony Vegas Pro 9 or Sony Vegas Movie Studio HD 11 software to play back the video recordings (Sony, 2009, 2011). Sony Vegas enables video to be slowed or viewed frame-by-frame (0.033 s resolution) while showing the cumulative duration, thereby permitting accurate identification of tail signals.

Tail signals were defined as any cyclical tail movement while a squirrel’s feet were in the same location (i.e., the squirrel was not shifting its body, walking, jumping, or running). We further categorized tail signals based on the arc of the caudal half of the tail during one cycle. The angle was estimated relative to the starting position of the tail. A classification criterion of 45° was used because the overall shape of movement differed qualitatively between signals greater or less than 45° in amplitude. Movements less than 45° are labeled ‘twitches’ and movements greater than 45° ‘flags’. Two independent observers watching a randomly selected video took instantaneous samples (N=38) (Altmann, 1974) of tail signaling behavior at one second intervals and classified 29 twitches and flags with complete agreement (Cohen’s κ=1) (Cohen, 1960; Kaufman & Rosenthal, 2009).

Twitches look like a wave running through the tail, which remains mostly parallel to either the substrate or the squirrel’s body with most of the movement along the dorso-ventral axis. Twitches are occasionally used with the tail raised over the back when the squirrel is on a horizontal tree limb. More often, the tail is held parallel to the gripped surface of the tree when the squirrel is on the trunk or a large branch.

Flags differ qualitatively from twitches. Rather than remaining generally parallel to the substrate or squirrel’s body, as with twitches, tail flags make a whipping motion with the tail’s tip curving back toward its base as the tail changes direction. Flags also frequently include movement in the dorso-ventral and lateral axes simultaneously that, in conjunction with their greater amplitude, produces large conspicuous movements. The path of a flagged tail’s tip varies widely, describing arcs, figure eights, circles, and various squiggles. In contrast, twitches are more controlled, so the tail tip path usually forms a simple short arc.

2.6. Statistical analyses

Each type of vocal (kuk, quaa, moan) and tail (twitch, flag) signal was scored as present or absent in each trial. Contingency tables were constructed from these presence and absence data. Fisher’s exact test probabilities were calculated for 2 × 2 contingency tables and larger tables were analyzed using the analogous Fisher–Freeman–Halton exact tests for r × c contingency tables. We use the term significant association to indicate statistical significance (p0.05) of the results of these tests of independence or association and independent may be used if there is no such significant association. Lack of independence may be conceived of as interaction as well as of association. When significant associations were found, the asymmetrical index of predictive association (λB) was also calculated to measure the average overall decrease of error rate in predicting the nature of the eliciting stimulus given knowledge of the type of signal in a squirrel’s response. Because all tests were of a priori hypotheses predicated on the experimental design, no Bonferroni corrections are employed. Statistical analyses were performed using JMP Pro, Version 9 (SAS Institute).

3. Results

3.1. Alarm signals observed

Squirrels used kuks, quaas and moans as well as tail twitches and flags during stimulus presentations and incidental observations of natural predator encounters, suggesting that the experimental stimuli provoke similar vocal and tail responses as natural threats. Twitches were also occasionally observed in foraging squirrels, especially when they descended a tree trunk, despite the absence of any apparent source of danger. Squirrels vocalized in response to 15% of the 156 stimulus presentation trials and used tail signals in response to 69% of the presentations. With the exception of tail twitches, these alarm signals were only observed during natural predator encounters or experimental trials. During experimental trials, the squirrel’s flight always began before the model passed over the squirrel’s initial location. Focal squirrels typically oriented briefly toward the approaching stimulus and then rapidly ran partway up the nearest tree before stopping and orienting their face toward the stimulus, followed by any tail signaling, vocalizing, and additional travel.

3.2. Experimental results

3.2.1. Model cat approaching terrestrially vs. model hawk approaching aerially

Squirrels did not often vocalize to any type of model predator stimulus (Figure 2A). Squirrels responded to approach of the model cat and model hawk with kuk and quaa vocalizations that were used independently of predator type (Fisher’s exact tests: kuk use, p=0.721; quaa use, p=0.280). The model hawk elicited moans more often (24.1%) than encounters with the model cat (2.9%) (Fisher’s exact test p=0.019).

Figure 2.
Figure 2.

Proportion of all trials that elicited vocal signals. Some signaling bouts contained multiple call types and thus contributed to more than one category of call. (A) model cat (N=34) and model hawk trials (N=29); (B) terrestrial ball (N=29) and aerial ball (N=33) trials; (C) model cat (N=34) and terrestrial ball (N=29) trials; (D) model hawk (N=29) and aerial ball (N=33) trials. An asterisk denotes significant association (p0.05) of elicited signal with stimulus.

Citation: Behaviour 151, 10 (2014) ; 10.1163/1568539X-00003194

Squirrels’ tail twitches and flags differed in their pattern of statistical association with predator (Figure 3A). Tail twitch usage is not significantly associated with predator type (Fisher’s exact test, p=0.788). In contrast, the model cat elicited flags more often (44%) than the model hawk (17%) (Fisher’s exact test p=0.031).

Figure 3.
Figure 3.

Proportion of all trials that elicited tail signals. Some signaling bouts contained multiple tail signal types and thus contributed to more than one category of tail signal. (A) Model cat (N=34) and model hawk trials (N=29); (B) terrestrial ball (N=29) and aerial ball (N=33) trials; (C) model cat (N=34) and terrestrial ball (N=29) trials; (D) model hawk (N=29) and aerial ball (N=33) trials. An asterisk denotes significant association (p0.05) of elicited signal with stimulus.

Citation: Behaviour 151, 10 (2014) ; 10.1163/1568539X-00003194

3.2.2. Novel object approaching terrestrially vs. aerially

Kuks were used more often in response to the terrestrial than the aerial ball (Fisher’s exact test, p=0.021). Similarly, quaas were used in 17.2% of encounters with the terrestrial ball and never in encounters with its aerial presentation and therefore also show a significant association with the manner of approach (Fisher’s exact test, p=0.018). Moans were uttered in only two trials, both times in response to the aerial ball, and therefore show no significant association with manner of approach (Fisher’s exact test, p=0.494) (Figure 2B).

In contrast to the association exhibited by vocal signals, tail signals used in response to the terrestrial versus aerial ball show no such result (Figure 3B). As with the model predators, squirrels used twitches independently of whether the ball approached terrestrially or aerially (Fisher’s exact test, p=0.168). Flags were also independent of whether the ball approached terrestrially or aerially (Fisher’s exact test, p=0.126).

3.2.3. Model predator vs. novel object approaching in the same manner

3.2.3.1. Model cat vs. terrestrial ball

Squirrels use of kuks shows no significant association with the type of terrestrial object (Fisher’s exact test, p=0.318) (Figure 2C). Similarly, quaas are not significantly associated with the type of terrestrial object (Fisher’s exact test, p=0.453). Moans were used in only a single alarm calling bout and so cannot demonstrate a statistical association with the type of terrestrial object (Fisher’s exact test, p=1).

Tail signals were used frequently on presentation of either terrestrial stimulus (Figure 3C). Twitches show no significant association with the type of terrestrial object (Fisher’s exact test, p=0.564) (Figure 3C). Flags are also independent of the type of terrestrial object (Fisher’s exact test, p=0.208).

3.2.3.2. Model hawk vs. aerial ball

Kuks show no association with aerial stimuli (Fisher’s exact test, p=0.089) (Figure 2D). In contrast, quaas are significantly associated with the type of aerial object (Fisher’s exact test, p=0.008). Squirrels used quaas exclusively in response to the model hawk, which resembles an actual predator, and never used quaas in response to the aerial ball that is similar in size and manner of approach to the model hawk, but differs in morphology (Figures 1, 2D). Moans are independent of aerial object (Fisher’s exact test, p=0.070) (Figure 2D). Twitches and flags are also independent of aerial object (Fisher’s exact test, p=0.794 and p=0.091, respectively) (Figure 3D).

3.2.3.3. Terrestrial stimuli null control

Although the ball or the hawk model were presented aerially without additional experimental paraphernalia, the model cat and the terrestrial ball were always mounted on a RC base for their terrestrial approach. To test for effects of the base itself on the use of vocal and tail signals, trials with the model cat (N=34) on its base were compared to those with base alone (N=31). Kuks were used in 11.8% of encounters with the model cat and 3.2% of encounters with the base alone, quaas were used in 8.8% of encounters with the model cat and 3.2% of encounters with the base alone, and moans were used in 2.9% of encounters with the model cat and not at all in encounters with the base alone, revealing no significant association of vocal signal with type of stimulus approaching terrestrially (Fisher’s exact tests: kuks, p=0.358; quaas, p=0.615; moans, p=1).

Twitches were used in 70.6% of encounters with the model cat and 61.3% of encounters with the base alone. Flags were used in 44.1% of encounters with the model cat and 35.5% of encounters with the base alone. Twitches and flags were also independent of whether the terrestrially approaching object was the cat on the base or the base alone (Fisher’s exact test, twitches, p=0.446; flags, p=0.613).

3.3. Summary of unimodal signal use

Eastern gray squirrels use one type of tail movement and one type of vocalization as a predator-specific alarm signal. These two signals, vocal moans and tail flags, are differentially employed in response to experimentally presented hawk and cat model predator stimuli. Moans and flags show this association only if both manner of approach and also physical appearance differ. When separately testing the role of a predator’s manner of approach and its physical appearance, three additional significant associations were found (Table 1). In combination with the predator-specificity of moans and flags, all these associations suggest three additional conclusions.

Table 1.

Summary of comparisons for all five stimulus pairs.

Table 1.

First, squirrels are more likely to vocalize (but not using moans) in response to terrestrially versus aerially approaching threats. Second, resemblance to a predator affects squirrels’ alarm responses to aerial threats, but not to terrestrial threats. Third, in response to aerial threats, squirrels used quaas only if the stimulus resembled a hawk.

3.4. Predictability of stimulus based on alarms

The signal-stimulus associations described above limit our ability to identify the nature of the stimulus given only the signal. Because error reduction in identifiability is measured by the index of predictive association, we on occasion use terms related to ‘prediction’ in characterizing our results. The predictability of the stimulus given only the alarms is an upper limit on the predictability of any threat-specific response to different alarms. Moans reduce the error in correctly predicting predator type (cat vs. hawk) by about 21% (λB=0.207). Flags reduce the error in correctly predicting predator type by 17.2% (λB=0.172). Flags reduce error in predicting manner of approach by 38.5% (λB=0.385). Kuks and quaas reduce error in correctly predicting the manner of approach by about 21% (λB=0.207) (kuks) and 17.2% (λB=0.172) (quaas).

3.5. Multimodal specificity

When considering each signaling modality separately, flags (a tail signal) and moans (a vocal signal) were the two signal types exhibiting an association with whether the predator model was a cat or hawk and each reduced the error in correctly predicting the eliciting predator stimulus. Tail and vocal signals are often combined in an alarm signaling bout. The presence and absence of flags and moans in such a multimodal alarm signal could either reduce or increase the predictive power of the multimodal signal relative to flags and moans considered separately. When examining the presence or absence of flags and moans in the same stimulus trial, there is a significant association of signals with predator type (Table 2: Fisher–Freeman–Halton exact test, df = 3, p<0.0001). In response to the model cat, flags without moans were used 1.6 times more often than expected by chance and in response to the model hawk, only 0.25 times as often as expected by chance. Moans without flags were used exclusively in response to the model hawk, never in response to the model cat. The potentially contradictory signal containing moans and flags was rare, observed only once in response to each model predator. Simultaneously considering moans and flags reduces the prediction error by about 52% (λB=0.517), reducing the error rate three times more than flags alone (λB=0.172) and two and a half times more than moans alone (λB=0.207).

Table 2.

Multimodal analysis of predator-specific signals.

Table 2.

4. Discussion

Eastern gray squirrels have an alarm system with different degrees of specificity of tail signals. Twitches are general; they have no significant stimulus-signal association with predator appearance, manner of approach, or their combinations. Flags have a moderate stimulus-signal association with predator type when both appearance and manner of approach differ (cat vs. hawk), Flags do not, however, have any stimulus-signal association with the manner of approach when the stimulus object does not resemble a known predator. Vocalizations used as alarm signals also show mixed specificity. Kuks and quaas are partially specific. They show a stimulus-signal association with a threat’s manner of approach, but not with predator appearance, except that, in response to aerial threats, quaas are used only when the threat resembles a hawk. Moans are specific, showing a strong stimulus-signal association with the combination of appearance and manner of approach, i.e., predator type.

These results highlight the importance of considering multiple modalities when investigating animal communication. Vocal signals considered alone cannot reveal the inherent complexity of their actual usage. By examining both modalities, we determined that multimodal alarm signals are more strongly associated with predators than either modality considered alone.

The criterion for predator-specific vocalizations is usually near-exclusive use of an acoustic call type in the context of a specific predator type (Seyfarth et al., 1980b; Macedonia & Evans, 1993) rather than a level of predictive association. In eastern gray squirrels, moans fit this exclusivity criterion well.

Although flags are associated with terrestrial predators, 29% of the presentations in which squirrels used flags were of aerial stimuli. If the presence of flags were used to identify the nature of a terrestrial threat, there would be a 29% error rate in prediction. The use of flags in response to some aerial stimuli may result from a squirrel classifying aerial stimuli as terrestrial threats because aerially launched stimuli come to rest on the ground, after the squirrel flees to a tree but usually before it starts signaling.

Production specificity of alarms could be due to predator type or response urgency (Leger et al., 1980; Owings & Leger, 1980; Robinson, 1981; Furrer & Manser, 2009) although urgency covaries with predator type in some species (Leger & Owings, 1978; Macedonia & Evans, 1993). Varying urgency may result in varying stress or arousal levels. Eastern gray squirrel tail signals are somewhat associated with a predator’s attributes, but may also signal stress or arousal, with twitches signaling moderate levels and flags higher levels. Fox squirrels (Sciurus niger) in California used flags when frustrated by a previously solvable puzzle (M. Delgado, pers. comm.), and gray squirrels use tail signals during agonistic conspecific interactions (Barkalow & Shorten, 1973; Steele & Koprowski, 2003; pers. obs.). If tail signals primarily reflect stress or arousal in gray squirrels, then flags should be used equally often for equally dangerous predators. Although predation and attack rates are unknown, there is no reason to believe that aerial predators are less dangerous than terrestrial predators, as both raptors and cats have been observed attacking squirrels in the study population (pers. obs.).

Tail flags are very conspicuous; they readily drew our attention to previously unobserved squirrels. A squirrel flagging in response to a terrestrial threat is announcing its own location not only to conspecifics (Partan et al., 2010) but also to predators. An arboreal tail flagging squirrel is safe from a terrestrial predator, and can signal its location with impunity. Flags may advertise a squirrel’s location to terrestrial predators, perhaps functioning to discourage predators from lingering in an area where their prey has spotted them and reached safe refuge (Fitzgibbon, 1994; Zuberbühler et al., 1999; Digweed & Rendall, 2009). California ground squirrels tail flag in response to rattlesnakes, and these tail signals cause rattlesnakes to reduce attack frequency and to leave the area sooner than in the absence of tail flags (Barbour & Clark, 2012).

Squirrels faced with an aerial threat might suppress tail signaling that reveals their location to a visually oriented predator capable of attacking them in a tree. Because raptors often attack in a stoop from above, grounded raptors may not pose a significant threat relative to those in flight or in trees, freeing squirrels to use the conspicuous flags. This risk differential may explain the lower specificity of flags relative to moans.

Partan et al. (2009, 2010) found that eastern gray squirrels responded more strongly to a robotic squirrel that tail signaled and broadcast alarm calls than to either modality alone, suggesting a reinforcement function for multimodal alarm signals. The results of our study showing that flags and moans considered together better permit identification of threat type than either modality alone suggest that using multiple modalities may function to reinforce the signal. For general alarm signals (kuks, quaas and tail twitches), the vocalizations (kuks and quaas) may function as the primary signal of a predator’s presence because twitches are used in many other contexts. Twitches may function to make vocal alarms more noticeable, advertising the signaler’s location or potentially its identity.

Meerkats (Suricata suricatta) use alarms during which predator type and urgency are associated with different parameters of the same vocal signals (Manser, 2001). Tufted capuchins (Cebus apella nigritus) have aerial and terrestrial alarms but terrestrial-alarm rate is also associated with risk (Wheeler, 2010). Given the lack of predator-specificity of kuks and quaas and the limited specificity of flags, it may be that gray squirrel alarms are associated with urgency as well. Studies manipulating the urgency of predator stimuli are needed to clarify the role of predator type versus response urgency in driving the production specificity of vocal and tail signals observed in this present study. Additional work is also needed to clarify whether species with predator-specific alarms classify predators by physical appearance, manner of approach, or both.

Gray squirrels apparently have some, but not all, of their alarm signals, associated with predator type and multimodal alarms are more strongly associated with threats than unimodal alarms. This mixed specificity in unimodal and multimodal signals enables squirrels to manipulate the specificity of their alarms and also their own risk of being detected by predators.

Acknowledgements

This study was funded in part by awards from the Jay M. Savage Graduate Research Support Fund and the Kushlan Graduate Research Fund to T.R.M. and general research support to S.M.G. through the University of Miami Department of Biology, as well as a Graduate Research Fellowship to T.R.M. from the College of Arts and Sciences at the University of Miami in Coral Gables, FL, USA.

References

  • Altmann J. (1974). Observational study of behavior: sampling methods. — Behaviour 49: 227-267.

  • Barbour M.A., Clark R.W. (2012). Ground squirrel tail-flag displays alter both predatory strike and ambush site selection behaviours of rattlesnakes. — Proc. Roy. Soc. Lond. B: Biol. Sci. 279: 3827-3833.

    • Search Google Scholar
    • Export Citation
  • Barkalow F.S., Shorten M. (1973). The world of the gray squirrel, 1st edn.J.B. Lippincott, New York, NY.

  • Cohen J. (1960). A coefficient of agreement for nominal scales. — Educ. Psychol. Measurem. 20: 37-46.

  • Davis L.S. (1984). Alarm calling in Richardson’s ground squirrels (Spermophilus richardsonii). — Z. Tierpsychol. 66: 152-164.

  • Digweed S., Rendall D. (2009). Predator-associated vocalizations in North American red squirrels (Tamiasciurus hudsonicus): to whom are alarm calls addressed and how do they function?Ethology 115: 1190-1199.

    • Search Google Scholar
    • Export Citation
  • Fitzgibbon C.D. (1994). The costs and benefits of predator inspection behaviour in Thomson’s gazelles. — Behav. Ecol. Sociobiol. 34: 139-148.

    • Search Google Scholar
    • Export Citation
  • Furrer R.D., Manser M.B. (2009). The evolution of urgency-based and functionally referential alarm calls in ground-dwelling species. — Am. Nat. 173: 400-410.

    • Search Google Scholar
    • Export Citation
  • Hennessy D.F., Owings D.H., Rowe M.P., Coss R.G., Leger D.W. (1981). The information afforded by a variable signal: constraints on snake-elicited tail flagging by California ground squirrels. — Behaviour 78: 188-226.

    • Search Google Scholar
    • Export Citation
  • Huynh H., Bertolino S., Koprowski J., Williams G., Thompson C., McAlpine D. (2011). The even darker side of the eastern gray squirrel (Sciurus carolinensis): a review of global introductions, invasion biology, and pest management strategies. — Julius-Kühn-Archiv 432: 31.

    • Search Google Scholar
    • Export Citation
  • Kaufman K.B., Rosenthal R. (2009). Can you believe my eyes? The importance of interobserver reliability statistics in observations of animal behaviour. — Anim. Behav. 78: 1487-1491.

    • Search Google Scholar
    • Export Citation
  • Lariviere S., Walton L.W. (1997). Lynx rufus. — Mamm. Spec. 563: 1-8.

  • Leger D.W., Owings D.H. (1978). Responses to alarm calls by California ground squirrels: effects of call structure and maternal status. — Behav. Ecol. Sociobiol. 3: 177-186.

    • Search Google Scholar
    • Export Citation
  • Leger D.W., Owings D.H., Gelfand D.L. (1980). Single-note vocalizations of California ground squirrels: graded signals and situation-specificity of predator and socially evoked calls. — Z. Tierpsychol. 52: 227-246.

    • Search Google Scholar
    • Export Citation
  • Lishak R.S. (1984). Alarm vocalizations of adult gray squirrels. — J. Mammal. 65: 681-684.

  • Macedonia J.M., Evans C.S. (1993). Variation among mammalian alarm call systems and the problem of meaning in animal signals. — Ethology 93: 177-197.

    • Search Google Scholar
    • Export Citation
  • Manser M.B. (2001). The acoustic structure of suricates’ alarm calls varies with predator type and the level of response urgency. — Proc. Roy. Soc. Lond. B: Biol. Sci. 268: 2315-2324.

    • Search Google Scholar
    • Export Citation
  • Marler P., Evans C.S., Hauser M.D. (1992). Animal signals: motivational, referential, or both? — In: Nonverbal vocal communication: comparative and developmental approaches. Studies in Emotion and Social Interaction ( Papousek H., Jürgens U., eds). Cambridge University Press, New York, NY, p.  66-86.

    • Search Google Scholar
    • Export Citation
  • McRae T.R. (2012). Predator-specificity of multimodal alarm signals in the eastern gray squirrel (Sciurus carolinensis). — PhD dissertation, University of Miami, Coral Gables, FL.

  • Meng H. (1959). Food habits of nesting Cooper’s hawks and goshawks in New York and Pennsylvania. — Wilson Bull. 71: 169-174.

  • Otovic P., Partan S.R. (2009). Multimodal signalling in animals. — In: The new encyclopedia of neuroscience ( Squire L.R., Editor-in-Chief). Academic Press, Oxford, p.  1095-1105.

    • Search Google Scholar
    • Export Citation
  • Owings D.H., Leger D.W. (1980). Chatter vocalizations of California ground squirrels: predator- and social-role specificity. — Z. Tierpsychol. 184: 163-184.

    • Search Google Scholar
    • Export Citation
  • Owings D.H., Virginia R.A. (1978). Alarm calls of California ground squirrels (Spermophilus beecheyi). — Z. Tierpsychol. 46: 58-70.

  • Owren M.J., Rendall D., Ryan M.J. (2010). Redefining animal signaling: influence versus information in communication. — Biol. Philos. 25: 755-780.

    • Search Google Scholar
    • Export Citation
  • Partan S.R., Larco C.P., Owens M.J. (2009). Wild tree squirrels respond with multisensory enhancement to conspecific robot alarm behaviour. — Anim. Behav. 77: 1127-1135.

    • Search Google Scholar
    • Export Citation
  • Partan S.R., Fulmer A.G., Gounard M.A.M., Redmond J.E. (2010). Multimodal alarm behavior in urban and rural gray squirrels studied by means of observation and a mechanical robot. — Curr. Zool. 56: 313-326.

    • Search Google Scholar
    • Export Citation
  • Robinson S.R. (1981). Alarm communication in Belding’s ground squirrels. — Z. Tierpsychol. 56: 150-168.

  • Searcy W.A., Nowicki S. (2006). Signal interception and the use of soft song in aggressive interactions. — Ethology 112: 865-872.

  • Seyfarth R.M., Cheney D.L., Marler P. (1980a). Monkey responses to three different alarm calls: evidence of predator classification and semantic communication. — Science 210: 801-803.

    • Search Google Scholar
    • Export Citation
  • Seyfarth R.M., Cheney D.L., Marler P.M. (1980b). Vervet monkey alarm calls — semantic communication in a free-ranging primate. — Anim. Behav. 28: 1070-1094.

    • Search Google Scholar
    • Export Citation
  • Sloan J.L., Wilson D.R., Hare J.F. (2005). Functional morphology of Richardson’s ground squirrel, Spermophilus richardsonii, alarm calls: the meaning of chirps, whistles and chucks. — Anim. Behav. 70: 937-944.

    • Search Google Scholar
    • Export Citation
  • Steele M.A., Koprowski J.L. (2003). North American tree squirrels, 1st edn.Smithsonian Books, Washington, DC.

  • Stegmann U. (2013). Animal communication theory — information and influence. — Cambridge University Press, Cambridge.

  • Struhsaker T.T. (1967). Auditory communication among vervet monkeys (Cercopithecus aethiops). — In: Social communication among primates, 1st edn. ( Altmann S.A., ed.). The University of Chicago Press, Chicago, IL, p.  281-324.

    • Search Google Scholar
    • Export Citation
  • Swaisgood R.R., Rowe M.P., Owings D.H. (2003). Antipredator responses of California ground squirrels to rattlesnakes and rattling sounds: the roles of sex, reproductive parity, and offspring age in assessment and decision-making rules. — Behav. Ecol. Sociobiol. 55: 22-31.

    • Search Google Scholar
    • Export Citation
  • Uhl G., Elias D.O. (2011). Communication. — In: Spider behaviour: flexibility and versatility, 1st edn. ( Herberstein M.E., ed.). Cambridge University Press, Cambridge, p.  127-189.

    • Search Google Scholar
    • Export Citation
  • Warkentin K.J., Keeley A.T.H., Hare J.F. (2001). Repetitive calls of juvenile Richardson’s ground squirrels (Spermophilus richardsonii) communicate response urgency. — Can. J. Zool. 79: 569-573.

    • Search Google Scholar
    • Export Citation
  • Wheeler B.C. (2010). Production and perception of situationally variable alarm calls in wild tufted capuchin monkeys (Cebus apella nigritus). — Behav. Ecol. Sociobiol. 64: 989-1000.

    • Search Google Scholar
    • Export Citation
  • Zuberbühler K., Jenny D., Bshary R. (1999). The predator deterrence function of primate alarm calls. — Ethology 105: 477-490.

Content Metrics

All Time Past Year Past 30 Days
Abstract Views 5726 652 118
Full Text Views 552 46 16
PDF Views & Downloads 197 102 39