Introduction

Urbanisation is a critical threat to biodiversity and is likely to continue to be so into the future (Davis 1976; Chase and Walsh 2006; McKinney 2006). Defined as the transformation of rural/natural-formed landscapes into urban ones through the complex interaction of various processes (e.g. spatial diffusion, physical planning and landscape geography; Antrop 2000), urbanisation results in fragmentation of the natural habitat and ultimately impacts the biodiversity and ecology of an area (Devictor et al. 2006; Shochat et al. 2006). Yet, some species can thrive in urban areas, and their ability to do so is dependent on species-specific traits, which enable them to track human-induced changes by evolving new characteristics or enhancing pre-existing adaptations (Sih et al. 2011).

The ability of a species to succeed in urban environments might depend on whether it can respond behaviourally to change (Watson 2009; Magle and Angeloni 2010), with species that display behavioural plasticity having a particular advantage (Sih et al. 2011). Such plasticity might manifest in anti-predator behaviour which is likely to decrease in species living in urban areas, either because animals living in urban areas habituate to the presence of people (the habituation hypothesis; Watson 2009; Magle and Angeloni 2010) or because they respond to the absence of natural predators (McKinney 2002; Shochat et al. 2006). It is often difficult to tease apart these effects because urbanisation generally decreases natural predator abundance. However, although natural predators might be extirpated through the process of urbanisation, domestic predators, such as dogs and cats, could be expected to be more common, thereby occupying the niche vacated by natural predators.

Habituation is defined as a decrease in behavioural response resulting from gradual de-sensitisation to repeated stimuli with time (Rankin et al. 2009) with little or no consequence for the individual. Habituation could result in a switch in the traditional trade-off between vigilance behaviour and foraging behaviour, resulting in individuals expending less time/energy on anti-predator behaviour and more time/energy on acquiring available food (Arenz and Leger 1999; Unck et al. 2009). Bowers and Breland (1996) found that gray squirrels Sciurus carolinensis living near humans did not trade-off between foraging (giving up density) and vigilance/refuge-seeking compared to squirrels living further away from humans.

A competing hypothesis maintains that animals in urban areas may be phylogenetically constrained to behave in a similar manner to those in natural environments (phylogenetic conservatism; Poisot et al. 2011) because they would perceive humans and domesticated animals in a similar way to natural predators (the risk-disturbance hypothesis; Frid and Dill 2002). For example, woodchucks Marmota monax showed similar levels of vigilance behaviour across an urbanisation gradient (Watson 2009). Alternately, species that cannot modify their behaviour (i.e. are inflexible) in response to urbanisation or are phylogenetically constrained to respond inappropriately are likely to become locally extinct (van der Ree and McCarthy 2005; McDonald et al. 2008; Magle and Angeloni 2010). For example, grey flycatcher Empidonax wrightii occupancy declined in noisy urban environments (in contrast to quieter ones) because it did not shift vocalisation frequency, whereas ash-throated flycatcher Myiarchus cinerascens occupancy did not differ between locations because it was able to shift vocalisations to a higher frequency (Francis et al. 2011). The differences in how species respond to urbanisation suggest the need for studies of more species, particularly to aid in the conservation of biodiversity in urban areas.

The Cape ground squirrel Xerus inauris is an ideal model for investigating behavioural correlates of urbanisation because it occurs in natural and urban areas, indicating that it can withstand the pressures of a range of environmental conditions. Cape ground squirrels are relatively large (650 g), highly social, diurnal sciurids that occupy open areas and use vigilance to detect predators (Unck et al. 2009). Vigilance is costly, however, since it reduces energy acquisition, so in response, vigilance decreases with increasing group size, thereby promoting more foraging behaviour of individuals (Edwards and Waterman 2011). Furthermore, feeding competition, particularly in winter, is known to influence female mortality and body mass (Waterman 2002).

The aim of our study was to investigate whether anti-predator behaviour differed in Cape ground squirrels in three localities that differed in their degree of urbanisation. We evaluated two alternative hypotheses that might explain the responses of Cape ground squirrels to urbanisation, in particular the reduction in anti-predator behaviour in urban areas. If they were habituated to humans, we hypothesised that there would be a smaller perceived predation risk in a highly urbanised location, and predicted that squirrels would: (1) be less vigilant and forage more in urban areas because of decreased predation risk; (2) trade-off foraging in favour of flight/vigilance (regardless of group size) in non-urbanised areas (greater predation risk), but the trade-off would not be apparent in highly urbanised area; and (3) have shorter flight initiation distances (FID) due to habituation to a potential predator (i.e. a human observer approaching them) in highly urbanised areas. In contrast, if Cape ground squirrels perceived humans and domestic animals as predators (phylogenetic conservatism), locality (i.e. level of urbanisation) would not be a predictor of anti-predator behaviour, providing support for the risk-disturbance hypothesis.

Materials and methods

Study localities

We studied Cape ground squirrels in three South African localities: the Johannesburg Zoo (designated urban, Johannesburg, Gauteng Province, 26°9′58.95″ S, 28°2′17.47″ E); Krugersdorp Nature Reserve (designated peri-urban, Krugersdorp, Gauteng Province, 26°4′54″ S, 27°42′44″ E); and Madikwe Nature Reserve (designated natural, North West Province, 24°45′37.57″ S, 26°16′39.75″ E). The Johannesburg Zoo (81 ha), situated in a densely populated suburb, comprises a large open parkland, containing enclosures for several animal species. The ground squirrels here are free-roaming in the open parkland. The squirrels easily burrow into the enclosures and have access to the abundant food and water provided for the resident zoo animals. As large numbers of people pass through the zoo daily, ground squirrels regularly encounter humans. The landscape has been transformed and is not representative of the landscapes exploited by ground squirrels in their natural environment and natural predators are absent. However, feral cats are common. Krugersdorp Nature Reserve is a small nature reserve (1,500 ha) situated 50 km north-west of Johannesburg in the less urbanised outskirts where ground squirrels are comparatively less exposed to people. Although the public has daily access to this reserve, encounters with people are limited mainly to weekend visitors in vehicles (du Toit pers com). The reserve has several natural predators (including owls, raptors and jackal) of ground squirrels, and represented the transition between the urbanised and less urbanised locality. Lastly, Madikwe Game Reserve is a large reserve (75,000 ha) situated 200 km north-west of Johannesburg, where ground squirrels have limited exposure to people (game drives excepted). This locality is most representative of the natural habitat of ground squirrels in comparison to the other localities because it contains many more natural predators, including large mammalian carnivores (lions, spotted hyenas and leopards) and birds of prey (eagles and owls). All colonies in all three localities were located on sandy soils adjacent to grassland.

Burrow entrance number and population size

All observations were made in the austral summer (March and November) and winter (June–August) at all three localities in 2011. Within each locality, ground squirrels were studied for 7–10 days/season at several colonies (Table 1), depending on accessibility (e.g. proximity to roads and walkways and clear visibility for observations). We use the term “colony” for all individuals that we could count occupying a single burrow cluster, where a burrow cluster is defined as a group of burrow entrances separated from adjacent clusters by a distance vacant of burrows that is larger in size than the cluster area (Waterman 2002). Although we returned to the same colonies in both seasons, the selected colonies often shifted their position on the landscape, and were not located in exactly the same position seasonally. Fewer colonies were sampled at the Madikwe locality (Table 1) because it was often difficult to find ground squirrels outside their burrows and because of safety concerns of conducting observations in a reserve harbouring large predators; an armed ranger accompanied us during experiments in Madikwe Game Reserve. We recorded the number of burrow entrances at each colony within each locality. Due to occasional difficulties associated with the terrain, we were not able to ascertain colony density. Therefore, we calculated colony size by counting the number of individuals (adults and juveniles) visible above-ground from two to four censuses per colony, recorded in the morning when we arrived at the colonies. We then calculated population size for each locality by combining the colony size censuses for each season within each locality (averaged to determine the mean ± SE).

Table 1 The population characteristics and sampling parameters of Cape ground squirrels in three localities in two seasons
Full size table

Behavioural observations

We recorded the above-ground behaviour of 30 adults (≥20 cm body length) per season per locality (n = 180), sampling both sexes, which were distinguished on the appearance of external genitalia (males have large scrotal sacs and reproductively active females have prominent nipples); the colonies sampled were usually mixed sexed or sexes occurred in close proximity. Observations were made between 1000 and 1200 or 1500 and 1700 hours (once per day), coinciding with the times when most squirrels were active above-ground in our study populations. All observations were made at a distance of 2–10 m (either on foot at the zoo since no vehicles are allowed on zoo premises or from a vehicle in the other localities) from focal animals using 8 × 42 binoculars. The number of individuals of each sex sampled is provided in Table 1. Behavioural observations totalled 45 h (30 individuals per season per locality observed for 15 min each).

Upon locating a colony, the squirrels were given 5–10 min to habituate to our presence and return to their normal activities. Thereafter, using continuous focal animal sampling, we pseudo-randomly selected an adult individual (alternating between squirrels on the left or right of a group) and recorded the number of times (frequency) each behaviour (five behavioural states; Table 2) occurred in a 15-min period or unless the individual disappeared from view (see below). We recorded frequency rather than duration because the transition between behaviours (e.g. from foraging to travelling) was not always immediately apparent. Thus, frequency was a more conservative measure to avoid over-estimating one behaviour in favour of another. After observing one individual, we waited 2–3 min before selecting a subsequent focal individual that was located at a different part of the colony, so as to minimise sampling bias. In instances where this was not possible (e.g. where colony size did not allow for random selection), the entire colony was sampled. We attempted to differentiate between colony members by noting the sex and physical appearance (e.g. scars) of each individual, although we realise physical appearance is not a reliable method for identifying individuals in larger colonies.

Table 2 Behaviours that were recorded for Cape ground squirrels at three localities
Full size table

During the focal sampling, an individual disappearing from view (which occurred for 15 individuals; 8 % in total) was treated differently depending on the circumstances. When the colony size was very large (e.g. 25 in the peri-urban area in summer; Table 1), making it difficult to differentiate between individuals, the sampling for that individual was terminated, and another individual was selected for study. In a situation where the colony size was small enough to differentiate between individuals (e.g. two in the urban area in winter; Table 1), the stopwatch was stopped and restarted if the individual reappeared at the same burrow entrance within 10 s and observations continued for that individual until 15 min in total had been recorded. If the squirrel did not reappear after this time, the sampling for the individual was terminated and another focal individual was chosen.

Anti-predator behaviour (distance from the burrow and flight initiation distance)

We studied two aspects of anti-predator behaviour: (1) the maximum distance of squirrels from a burrow entrance within the colony at any time during the focal sample (real predation risk); and (2) the FID of each squirrel (perceived predation risk), defined as the distance at which an animal begins to flee from a perceived predator (Bonenfant and Kramer 1996). We used the fact that squirrels might be habituated to people as a mechanism to test the FID when approached by a person. During behavioural sampling of focal individuals, we estimated the distance of the focal individual from the closest burrow entrance using the average 20-cm body length of the squirrel (Skinner and Chimimba 2005) as a guide (e.g. two body lengths from burrow = about 40 cm); we noted the distance whenever the focal squirrel moved away from an entrance and recorded the maximum distance from the entrance during focal sampling.

We determined FID after behavioural observations (see above). We rotated through the colonies in a locality, first completing the behavioural observations and returning later in the day to test FID. We measured FID using a laser range finder (SPECTRA Precision laser HD150; accuracy, ±2 mm up to 150 m). To ascertain the FID of a squirrel, we randomly selected a focal individual and an observer walked toward the squirrel from a starting position of approximately 60 m away, pointed the range finder on the ground just in front of the squirrel and recorded the distance when the squirrel either fled down a burrow (81 % of observations) or away from the observer. These observations were made on 10 different squirrels at each of the three localities per season (n = 60).

Trade-off between feeding and flight/vigilance

To investigate whether ground squirrels trade-off between anti-predator (responding to potential threats) and foraging (feeding on a preferred food source) behaviour, we followed the same procedure described for FID. However, 12–14 apple slices (a preferred food source; personal observation) of equal size (2- to 3-cm pieces) were scattered on the ground close to where the squirrels were initially observed foraging. The observer walked back to the start position (60 m away from the squirrels) and waited for the squirrels to approach and start eating the apple. The observer then randomly selected one individual in the group and walked toward it, again recording how close the squirrel could be approached before it disappeared/fled. This procedure provided an indication of the limits of tolerance of a potential threat (human) and whether or not a potential threat alone was enough to disrupt feeding of the squirrels on a preferred food source. In 83 % of cases, squirrels dropped the apple, while the remaining 17 % ran to a burrow carrying the slice. This procedure was repeated ten times on different squirrels from each of the three localities per season (n = 60). FID with apple was always studied after the FID without apple, and each was conducted on different days.

Data analysis

All analyses were conducted using Statistica 7.1 (Statsoft Inc, www.statsoft.com) and Prism version 4 (GraphPad, www.graphpad.com). We first tested the data set for normality (homogeneity of variance and linearity) and subsequently square-rooted transformed the behavioural data to meet the assumptions of normality. We then compared three population parameters (colony size, burrow entrance number, maximum distance from a burrow entrance during observations) between localities in both summer and winter using a two-way ANOVA. We next ran a variance components test using expected mean squares on the behavioural data set to determine whether the three population parameters (colony size, burrow entrance number and maximum distance from a burrow entrance during observations; co-variates) and colony affiliation (random factor) predicted the outcome of the behaviours scored. We included colony affiliation as a random factor in our analyses because our sampling strategy (not marking animals) could have resulted in a small number of animals being re-sampled in a season. None of these factors was a significant predictor of behaviour (p > 0.05) and were not considered further in the analysis.

Of the five behaviours, only three (foraging, vigilance and travel) occurred frequently. Therefore, we compared the frequencies of only these three behaviours among the localities for both seasons (fixed factors) using a general linear model (GLM) with a multivariate design. Sex was included as a fixed factor and the three behaviours were multiple dependents. To test whether or not ground squirrels traded-off between foraging and vigilance, we ran linear regressions between these two behaviours for individuals in each locality in summer and winter. We then compared the slopes of the regressions per locality in each season using analysis of covariance (ANCOVA).

Finally, we analysed the flight initiation distance of squirrels using a GLM in which locality, season and treatment (without and with apple) were fixed factors. Because different colonies/individuals were tested in each replicate, treatment was included as a fixed effect. We excluded sex since it was not a significant predictor of behaviour (p > 0.05).

Where appropriate, Fisher’s HSD post hoc comparisons were performed to establish specific differences among the levels within the fixed factors. All tests were two-tailed and the model-level significance was set at α = 0.05.

Results

Burrow entrance number and population size

The size of the populations sampled was largest in the peri-urban locality in summer and lowest at the same locality in winter (Table 1). Population size remained relatively stable in the urban locality, whereas population size in the natural locality showed a significant decrease in winter (Table 1). Despite locality and seasonal differences in population sizes, the numbers of burrow entrances at each locality were similar for both seasons, except the natural locality had a significantly greater number of burrow entrances in summer than in winter (Table 1).

Behavioural observations

Comparisons of foraging, vigilance and travel behaviour made using a GLM with a multivariate design indicated that locality (F 6, 332 = 2.98; p = 0.008), season (F 3, 166 = 2.17; p = 0.043) and locality × season (F 6, 332 = 2.17; p = 0.046) were significant predictors of behaviour. Post hoc tests revealed that foraging was significantly greater in the peri-urban locality in summer and the urban locality in winter than the other localities/seasons (Fig. 1). Vigilance was most common in the natural locality in both seasons, although vigilance was also common in the peri-urban locality in summer (Fig. 1). In all three localities, travelling was significantly greater in summer than in winter and was greatest overall in the natural locality in summer (Fig. 1). The following did not affect behaviour: sex (F 3, 166 = 0.24; p = 0.869); locality × sex (F 6, 332 = 0.34; p = 0.913); and season × sex (F 3, 166 = 0.19; p = 0.903).

Fig. 1
figure 1

Frequency of foraging, vigilance and travel behaviour in Cape ground squirrels in three localities in summer and winter. Values are given as mean (+SE) occurrences per 15 min. Bars with the same alphabets for each behaviour are not significantly different (post hoc tests)

Full size image

Trade-off between feeding and flight/vigilance

Regressions between foraging and vigilance in each locality (Fig. 2) showed that in summer, there was a weakly negative relationship between these behaviours in the natural locality (R 2 = 0.29; F 1, 28 = 11.70; p = 0.002), a weakly positive relationship in the urban locality (R 2 = 0.17; F 1, 28 = 5.64; p = 0.025) and no relationship in the peri-urban locality (R 2 = 0.02; F 1, 28 = 0.53; p = 0.476). The slopes generated for these localities were significantly different from each other (F 2, 84 = 9.24; p < 0.001; ANCOVA). In winter, there was a strongly negative relationship between foraging and vigilance in the natural locality (R 2 = 0.57; F 1, 28 = 36.74; p < 0.001) and again a weakly positive relationship in the urban locality (R 2 = 0.33; F 1, 28 = 14.03; p = 0.001) and no relationship in the peri-urban locality (R 2 = 0.10; F 1, 28 = 3.12; p = 0.089). The slopes generated for these localities were also significantly different from each other (F 2, 84 = 29.19; p < 0.001; ANCOVA).

Fig. 2
figure 2

The relationship between vigilance and foraging in Cape ground squirrels in three localities in a summer and b winter. The points indicate frequency and lines are generated from linear regression analyses: dotted lines urban, dashed lines peri-urban and solid line natural. R 2 values for the regression lines: summer–urban = 0.17, peri-urban = 0.02, natural = 0.29, winter–urban = 0.33, peri-urban = 0.10 and natural = 0.57

Full size image

Anti-predator behaviour (distance from the burrow entrance and flight initiation distance)

We used two techniques to measure anti-predator response. Firstly, our estimates of the distance of focal squirrels from a burrow revealed that most were active at or within 2 m of a burrow entrance. The comparison between localities/season showed that squirrels in the natural locality were closest to a burrow entrance in both seasons whereas the peri-urban squirrels were active up to 5 m away from a burrow entrance in winter (Table 1).

Secondly, we calculated the FID with and without the provision of apple. This design enabled us to investigate differences in responses for natural vs. preferred introduced food. Locality (F 2, 108 = 16.26, p < 0.000), season (F 1, 108 = 11.79, p = 0.001) and treatment (F 1, 108 = 8.96, p = 0.003) were all significant predictors of FID, as were the interactions of locality × season (F 2, 108 = 8.24, p < 0.001) and locality × treatment (F 2, 108 = 4.13, p = 0.019), but not season × treatment (F 1, 108 = 3.51, p = 0.064). Post hoc tests revealed that FID values, regardless of treatment (with or without apples), were significantly greater in winter in the natural locality compared to the urban and peri-urban localities (Fig. 3). In summer, squirrels from urban and peri-urban localities showed significantly shorter FID values in both treatments, whereas the natural population had the greatest FID when no apple was present. However, FID decreased significantly in the natural locality compared to that of the other localities/treatment when apple was provided (Fig. 3). In winter, the urban locality again had the lowest FID values in both treatments, although this was not significantly different from the peri-urban locality when apple was provided. Squirrels in the peri-urban locality showed greater FID values when apple was absent, and the natural population had the greatest FID values in our study (mean of 29.2 m) when apple was absent, which decreased significantly when apple was present but still remained high (mean of 15.3 m; Fig. 3).

Fig. 3
figure 3

FID values recorded in Cape ground squirrels when apple was absent (open squares) and when apple was provided (closed squares) at three localities in a summer and b winter. Values are given as mean (±SE) metres. Points with the same alphabets in each season are not significantly different (post hoc tests)

Full size image

Discussion

The aim of this study was to establish whether anti-predator behaviour of Cape ground squirrels differed in three localities, characterised by different levels of urbanisation. Our results provided support for the habituation hypothesis but not phylogenetic conservatism (risk-disturbance hypothesis). The first prediction that ground squirrels in an urban locality would be less vigilant and would forage more because of decreased predation risk was partially supported. As expected, squirrels in the natural locality were more vigilant in both seasons than in both the peri-urban and urban localities, although vigilance in the peri-urban area did increase in summer. Unck et al. (2009) found that ground squirrels in areas of high predation risk invested more time in energetically costly, high quality vigilance behaviour than individuals in low-predation risk areas; high quality vigilance referred to exclusive vigilance without a foraging component. Since the natural locality in our study presumably contained many more natural predators than either the peri-urban or urban localities, squirrels there predictably showed the greatest levels of vigilance.

However, squirrels in the urban locality did not show the highest levels of foraging behaviour, except for a slight increase in winter. Foraging was highest in the peri-urban locality in summer. This is most likely due to greater availability of food rather than greater energy requirements for reproduction, since Cape ground squirrels generally show a peak in breeding during winter (Herzig-Straschil 1978; Waterman 1996). Nonetheless, we would still expect squirrels in the urban locality to show higher levels of foraging because of a reduction in vigilance behaviour. The lack of differences in foraging could possibly indicate similar feeding/energy requirements (phylogenetic conservatism) in Cape ground squirrels in the three localities. Another possibility is that squirrels in the urban locality did not show increased levels of foraging because high protein/calorie food was readily available (due to provision by zoo staff), reducing foraging time. Day et al. (2005) noted that, for most species, foraging increases in response to lower food availability and that individuals are unlikely to show increased foraging effort in response to greater food availability. Although we did not quantify the differences in food quality between sites, it is likely that Cape ground squirrels in urbanised areas did not need to increase foraging because of the greater food availability.

Our second prediction that squirrels in the natural locality with greater predation risk would trade-off foraging in favour of vigilance was supported, particularly in winter, whereas this trade-off was not evident in the urban locality. Since vigilance is costly in Cape ground squirrels (Edwards and Waterman 2011), any reduction in this behaviour, as a consequence of lower predations risk, should be advantageous (Magle and Angeloni 2010), which explains why squirrels in the urban and peri-urban localities did not decrease foraging behaviour in order to increase vigilance. Although Cape ground squirrels in the lower predation risk areas displayed less vigilance, it is likely that the quality of vigilance was the same as that shown in the natural locality. Unck et al. (2009) showed that ground squirrels experiencing low predation risk always invested in high quality (exclusive) vigilance. Similarly, in the low predation risk areas of our study, ground squirrels might still invest in high quality vigilance, but at a lower frequency, thereby reducing the overall costs associated with exclusive vigilance.

Vigilance behaviour is also influenced by group size in numerous rodents (e.g. Columbian ground squirrels Spermophilus columbianus, Fairbanks and Dobson 2007), including Cape ground squirrels (Edwards and Waterman 2011). However, colony sizes were similar in our study to the range reported in the literature for a whole colony (6–11; Nel 1975; Herzig-Straschil 1978; Bennett et al. 1984), and we did not find that colony size was a significant predictor of behaviour.

All the colonies in our study consisted of mixed sex groups, with a slightly biased female sex ratio. Since males usually form sub-bands, the occurrence of males with females, indicates undispersed males, which can remain in their natal groups for at least 19 months (Waterman 1995). However, the number of females was generally greater than reported in the literature (two to three adult females, Waterman 1995; Pettitt et al. 2008; van der Merwe and Brown 2008; one to six, Hillegass et al. 2008), while the number of undispersed males was greater or similar to that reported in the literature for all male colonies (one to three, Waterman 1995; one to two adult males, Hillegass et al. 2008).

Squirrels in the urban locality had shorter flight distances, which supported our third prediction. Moreover, as expected, squirrels in the natural locality were closer to a burrow entrance during observations and fled/retreated into their burrows at greater distances from an approaching human than squirrels at both the peri-urban and urban localities. This response was most apparent in winter, particularly when apple was not provided. The placement of apple decreased the FID twofold in squirrels in the peri-urban locality in winter and the natural locality in both seasons, indicating that squirrels here were willing to risk more for high calorie food. Similarly, European rabbits Oryctolagus cuniculus increased grazing behaviour when the nitrogen concentration of their food was experimentally increased, even when predation risk was also experimentally increased (Bakker et al. 2005). Interestingly, in summer, squirrels in the natural locality showed similar low FIDs to those in other localities, suggesting riskier behaviour for high-quality food. Squirrels in the urban locality were unaffected by the provision of apple, again supporting our expectation of their habituation to people and their regular access to available high calorie food.

Ground squirrels in the natural locality displayed behaviours indicative of greater predation risk in winter than in summer, such as greater FID and a trade-off between foraging and vigilance. Although we did not observe greater numbers of predators in winter than summer (but our presence could have deterred predators) or any environmental changes which are known to influence predation risk (e.g. the physical landscape; van der Merwe and Brown 2008), our results do suggest that the squirrels were risk averse in winter. There were no changes in anti-predator behaviour in the urban localities seasonally and there were mixed responses in the peri-urban locality.

The ability of a species to alter its behaviour in response to changing environments (particularly human induced changes) affects its success in an altered landscape (Sih et al. 2011). Urban areas are environmentally modified habitats that have characteristically high levels of disturbance (Jokimäki and Kaisanlahti-Jokimäki 2012), and it is expected that those individuals with traits that show greater behavioural plasticity, rapidly modifying their behaviours to the novel stimuli presented in this novel environment, would survive and persist in urban areas (Sih et al. 2011; Tuomainen and Candolin 2011). Our results demonstrate differences in the behaviour of Cape ground squirrels in the three localities with different levels of exposure to people. In particular, differences in anti-predator behaviour (less vigilance, shorter FID in the urban compared to natural locality) indicates firstly that the squirrels have become habituated to human presence (i.e. supporting the habituation hypothesis; Watson 2009; Magle and Angeloni 2010) and secondly that this species demonstrates a flexible behavioural response to urbanisation. The differences in anti-predator behaviour were most evident between the urban and natural population, with the peri-urban population displaying mixed responses. Our study does not uncouple the different factors that could have led to habituation of the Cape ground squirrels to urban localities, however, because human presence and predator presence are inversely related at our study localities, although domesticated predators would pose a risk in the urban locality.

Blair (2001) proposed the term ‘urban adapter’ to describe species that historically occurred in a heterogeneous environment and had evolved the flexibility to utilise a variety of habitat types later. The Cape ground squirrel historically occurred in numerous biomes (e.g. Nama-Karoo and Succulent Karoo) throughout South Africa (Skinner and Chimimba 2005) and our study indicates that they are ‘urban adapters’, exploiting novel resources in a highly urbanised locality. Even the natural population of squirrels reduced their FID when presented with a novel, palatable food resource (apple), suggesting an ability to exploit emergent food resources (Charmantier et al. 2008) that is fully developed in the urban population. Another explanation could be the development of a coping style (e.g. lowered stress response) that is adaptive to the prevailing environment, as suggested for European blackbirds Turdus merula (Partecke et al. 2006). This could influence the anti-predator behavioural phenotype, favouring squirrels with an increasing level of “tameness” in urban habitats. Future studies should tease apart the effects of acquired vs. evolved flexibility as driving factors in the expression of anti-predator behaviour in ground squirrels. This could be accomplished by studying Cape ground squirrels in a range of different urban and natural environments. Our study considered only three localities, each representative of a different level of urbanisation, such that our results could reflect idiosynchronies of the populations themselves, rather than general patterns related to urbanisation and predation risk.

In conclusion, understanding how species in urbanised areas modify their behaviours allows for the identification of urban areas that are capable of supporting urban adapters, which could be beneficial for conservation purposes (Hamer and McDonnell 2008). While urbanisation has a negative impact on numerous species (Davis 1976; Chase and Walsh 2006; McKinney 2006), Cape ground squirrels in South Africa are capable of living in urban environments, enabling their persistence in the face of rapid human induced environmental change.