Foraging opportunities drive interspecific associations between rock kestrels and desert baboons


  • Editor: Nigel Bennett

Andrew J. King, Institute of Zoology, Zoological Society of London, Regent's Park, NW1 4RY, London, UK


Interspecific associations can arise for varied reasons including reduced predation risk and improved foraging success. In the case of bird–primate associations, birds typically appear to follow primate groups to harvest insects flushed by primates' movements. However, while previous studies have linked temporal changes in bird–primate associations to environmental conditions, few have assessed the additional effects of bird activity patterns and primate group behaviour and none have disentangled their potentially interdependent effects. Here, we test the hypothesis that foraging opportunities can drive interspecific associations in a previously undescribed bird–primate association between rock kestrels Falco rupicolus and chacma baboons Papio ursinus in central Namibia. Data were collected from two baboon groups and associated kestrels using instantaneous scan sampling during full-day follows over a 7-month field period, and analysed using generalized linear mixed models. We found that kestrel associations with baboons vary with season, show diurnal cycles and are more frequent when the baboons are in open desert habitat, engaged in travel foraging and in a large group. These patterns are statistically independent and consistent with the hypothesis that the kestrel–baboon association is driven by the foraging opportunities acquired by the kestrels. As the baboons do not appear to gain any benefits nor incur any costs from the association, we conclude that the kestrels are likely to be commensal with the baboons.


Associations between the members of two different species can vary in their duration and frequency, and are thought to form because of the benefits provided to individual members of one or both associating species (Stensland, Angerbjorn & Berggren, 2003). These associations can be described in one of three ways depending upon how the benefits are distributed: (1) mutualism, where members of both species benefit; (2) parasitism, where members of one species benefits at the expense of the other; (3) commensalism, where members of one species benefit and the other is unaffected by the association. Benefits include many of the same reasons that single-species groups form, such as improved predator detection and avoidance (e.g. Morse, 1977; McGraw & Bshary, 2002; Teelen, 2007) or increased foraging efficiency (e.g. Peres, 1992; Buchanan-Smith, 1999; Bearzi, 2006).

Where associating species do not share the same predators, foraging benefits may be the most important driver of interspecific associations. Studies that have reported improved individual foraging efficiency as a result of interspecific association are plentiful within and across a variety of animal taxa (Ruggiero & Eves, 1998; Buchanan-Smith, 1999; Beisiegel, 2007), and often involve commensal relationships (Dickman, 1992; Schaefer & Fagan, 2006; Herring & Herring, 2007). In the case of bird–primate commensal relationships, bird species such as kites (Fontaine, 1980; Egler, 1991; Heymann, 1992), and woodcreepers and cuckoos (Boinski & Scott, 1988; Kuniy, de Morais & Gomes Jr, 2003; Hankerson, Dietz & Raboy, 2006) are all reported to associate with primate groups. Such insectivorous birds are thought to benefit from the disturbance created by the primates' movement through vegetation, allowing the birds to harvest flushed prey, while the primates receive no evident benefit (Boinski & Scott, 1988; Egler, 1991). Such studies are largely confined to the neotropics (although see Ruggiero & Eves, 1998; Seavy, Apodaca & Balcomb, 2001) and routinely describe trends in the temporal frequency of associations throughout the day and across seasons (e.g. Boinski & Scott, 1988; Ferrari, 1990). These patterns are thought to reflect environmental changes in insect availability (Rodrigues et al., 1994; Hankerson et al., 2006). However, relatively few studies have controlled for the effect of bird behaviour, for example diurnal foraging cycles, or examined the influences of primate group behaviour upon such bird–primate associations (cf. Boinski & Scott, 1988). Moreover, none have disentangled the potentially interdependent effects of these factors.

In this study, we examine an association between rock kestrels Falco rupicolus and chacma baboons Papio ursinus on the edge of the Namib Desert, Namibia. During the associations, kestrels were seen to prey on the Orthopteran insects (grasshoppers, locusts and crickets) that fly into the air following their disturbance by the baboons. We test the hypothesis that the association arises from the foraging opportunities derived by kestrels. We test this hypothesis by first investigating the predicted effects of environmental conditions and kestrel activity. We then test for the predicted effects of specific primate group behaviours on the availability of kestrel prey items (and thus potential for associations). We describe our predictions for each of these effects in turn.

For environmental conditions, we know that Orthopteran species density increases in association with rainfall in arid regions (Noy-Meir, 1973; Belovsky & Slade, 1995). If kestrel–baboon associations reflect seasonal changes in insect availability, we would predict a seasonal pattern of kestrel–baboon association that reflects rainfall (prediction P1). Associations are also likely to change in accordance with kestrel activity patterns. Kestrels can show remarkable constancy from day to day in the temporal distribution of specific behaviours and of spatial movements that are a consequence of both prey activity patterns and environmental variables (Rijnsdorp, Daan & Dijkstra, 1981; Barnard, 1986). We therefore predicted associations to occur most frequently post-dawn and pre-sunset (P2) when kestrels traditionally forage (Rijnsdorp et al., 1981; Van Zyl, Jenkins & Allan, 1994).

Finally, in the case of primate group behaviour, the collective activities of the baboon groups may influence the likelihood of kestrel–baboon associations in three ways. First, habitat type (in which the baboons are observed foraging) may affect kestrel associations as a consequence of variation in Orthoptera prey densities (e.g. Babah & Sword, 2004) and differences in the opportunity for aerial prey detection and capture by the kestrels (e.g. Thiollay & Clobert, 1990). We therefore predict that kestrel–baboon associations will be most likely when baboons are in those habitats where both the density and detectability/catchability of Orthoptera prey species are highest, that is open, grassland habitat rather than closed, woodland habitat (P3). Second, specific activities of baboons may be more likely to disturb Orthopteran species, generating foraging opportunities for kestrels and thus promoting associations. We therefore test the prediction that kestrel–baboon associations will be more likely when the baboons are more active (travelling and travel foraging) than when they are sedentary (resting, stationary foraging), since that is when they are actively disturbing rocks and vegetation (P4). Third, we test the prediction that kestrels associate more when the baboons are in a large group rather than a small group, given that more individuals will cause more disturbances of Orthopteran species over a larger area (P5).

Materials and methods

Study site and subjects

The study subjects were wild rock kestrels present in the area and two groups of wild chacma baboons (group L, n=26 members; group J, n=54 members), previously habituated to the presence of human observers. The study was conducted at Tsaobis Leopard Park, (22°23′S 15°45′W) in the semi-desert Pro-Namib region of Namibia. The park landscape is dominated by mountains and ravines, which are fringed by steep rocky foothills and gravel and alluvial plains. To the north it is bordered by the ephemeral Swakop River. Annual rainfall is light and seasonal: mean±se=215±17 mm, n=36 years, with rains falling primarily in the late austral summer (January–March) (Fig. 1). However, these rains support a relatively diverse desert plant community (Cowlishaw & Davies, 1997). Typical vegetation found on the hills and plains includes perennial grasses and herbs, for example Aristida spp. and Petalidium variabile, with shrubs and dwarf trees, for example Catophractes alexandri, Acacia erubescens and especially Commiphora virgata. During the austral summer, baboons often forage on the vegetation and small invertebrate prey found in these open, rocky desert habitats (hereafter referred to as ‘open desert’). During the winter, as these foods die back following the rains, the baboons forage increasingly on the flowers, fruits and pods of the large shrubs and trees that grow in patches of riparian woodland along the Swakop River (hereafter referred to as ‘closed woodland’). These woodlands are supported by groundwater and dominated by Faidherbia albida, Prosopis glandulosa and Salvadora persica. (See Cowlishaw, 1997 for further details).

Figure 1.

 Mean±se monthly rainfall in the study region (Karibib) between 1961 and 1997. Study period is from June to December 2005.

Data collection

The two baboon groups were followed on foot from dawn to dusk from June to December 2005 (121 days for group L; 75 days for group J). Detailed behavioural and ecological data were collected on the baboons and the occurrence of kestrel–baboon associations. However, individual kestrels were not identifiable, and their behaviour and activities outside of observed associations were not known.

Data were collected by scan sampling at 30-min intervals throughout the day, with the first scan beginning 30 min after the baboon group had left their morning sleeping site. A total of 3951 scans across both baboon groups were obtained. Baboon groups travelled large distances each day (range 1.8–9.8 km, average ±se=5.9±0.06, n=180 days), and each group occupied adjacent home ranges that overlapped (A. J. King et al., unpubl. data). Kestrel densities at the site are unknown, but elsewhere in their range can be found at up to 24.3 pairs per 100 km2 (Eastern Cape, South Africa: Van Zyl, 1999). Given the more arid nature of the Tsaobis environment, it is likely that kestrel densities are substantially lower in our locality.

Data on the presence or absence of kestrel–baboon associations were recorded at each scan. Associations occurred whenever one or more kestrels were present within the area covered by the baboon group. For all such observations kestrels were seen to be ‘flight-hunting’ and/or ‘perching’ (Rijnsdorp et al., 1981), which involved bouts of hovering and short flights, and sitting in trees or shrubs with a view of the ground and the baboons, respectively. These behaviours are known to be the most effective method of catching prey by kestrels (Rijnsdorp et al., 1981).

Data on the baboon group's habitat and activity were also collected at each scan. Habitat was categorized as either (1) open desert or (2) closed woodland. Baboon group activity was divided into four speed-related binary response variables (i.e. ≤50% group members engaged in behaviour vs. >50% engaged in behaviour): (1) travelling; (2) travel foraging; (3) stationary foraging; (4) resting. Travelling was defined as the rapid locomotion of individuals, travel foraging as the slow locomotion of individuals while searching, manipulating and ingesting food material. Stationary foraging describes searching, manipulating and ingesting food material while remaining in one location. Resting describes the baboons' sedentary state in which they were not travelling or foraging and included grooming and sleeping.

Statistical analysis

To assess the variables influencing the kestrel–baboon association, we used a generalized linear mixed model (GLMM) with binomial error structure and a logit link function implemented in MLwiN (Rasbash et al., 2004). We used association, non-association as a binary response term, and fit ‘scan number’ and ‘observation day’ as a random effect to take account of repeated measures within and across days, respectively (Browne, Goldstein & Rasbash, 2001). The following categorical variables were entered as fixed effects: habitat (open desert, closed woodland), activity (four variables), group identity (large, small) and time of year (June–September, termed early winter, September–December termed late winter; see Fig. 1). Time of day was fitted as a continuous variable (centred at 12:00 h midday). All fixed effects were entered and dropped sequentially until only those that explained significant variation remained (minimal model, e.g. Sokal & Rohlf, 1995). Each dropped term was then put back into the model to obtain their level of non-significance, and check that significant terms had not been wrongly excluded. Biologically relevant two-way interactions were also tested, but did not contribute significantly to the explanatory power of the model and are not discussed further. The significance of effects was tested using the Wald statistic, evaluated against the χ2 distribution.


During baboon-group scans, we observed one to four kestrels associating with groups (median: one kestrel) for an average 3.5% of all scan observations during our study period (June–December 2005). During these associations the kestrels appeared to follow the baboons and monitor their activities, frequently catching Orthopteran prey ‘on the wing’ as they were disturbed by the baboons (see Fig. 2).

Figure 2.

 Schematic representation of the foraging opportunities derived by kestrels from their association with baboon groups. Flight hunting includes the short flights in between bouts of hovering. Perched kestrel (left hand side) and baboon (lower right hand side) are from the original photograph; all other elements are graphic representations through time. Scheme adapted from the behaviour categories and diagram published by Rijnsdorp et al. (1981).

The results of the GLMM are consistent with the hypothesis that the kestrel–baboon associations are driven by the foraging opportunities acquired by the kestrels. First, kestrel–baboon associations were shown to reflect seasonal changes in insect availability, being more common in the early austral winter compared with the late austral winter (Table 1; Fig. 3a) as predicted (P1). Second, kestrel–baboon associations were more common in those periods when kestrels traditionally forage, with peaks in the early morning and late afternoon (Table 1; Fig. 3b), as predicted (P2). Three aspects of baboon group behaviour were also shown to influence the likelihood of kestrel–baboon associations, in support of predictions P3, P4 and P5. In the case of habitat (P3), kestrels were significantly more likely to associate with baboons in open desert habitat compared with closed woodland habitat (Table 1; Fig. 4a). Kestrel–baboon associations were also most frequent when the baboons were travel foraging (Table 1; Fig. 4b), and for the large baboon group compared with the small group (Table 1; Fig. 4c) (P4, P5, respectively). We found that variation in the numbers of baboons engaged in other activities (travelling, resting and stationary foraging) had no significant effect upon probability of kestrel associations (Table 1), although the directions of these trends were in the anticipated directions, that is negative where the majority of baboons were resting, and positive when a majority were travelling or stationary foraging.

Table 1.   Factors affecting the probability of kestrel–baboon associations, controlling for repeated scan observations within and across days (each entered as a random effects)
 Effect (se)d.f.WaldP
  1. Table shows parameter estimates (effect), standard errors (se), statistical values (Wald statistic) and significance evaluated against a χ2 distribution (P) based on 3951 scan observations of two baboon groups. Estimates for non-significant terms were obtained from adding terms individually to the minimal model

Model term
Time of year 139.34<0.001
 Early winter3.781 (0.603)   
 Late winter0.000 (0.000)   
Time of day 327.59<0.001
 Hour−0.351 (0.077)   
 Hour20.029 (0.011)   
 Hour30.012 (0.003)   
Habitat 147.63<0.001
 Open desert3.022 (0.438)   
 Closed woodland0.000 (0.000)   
Baboons travel foraging 118.18<0.001
 ≤0.5 group0.000 (0.000)   
 >0.5 group2.602 (0.113)   
Baboon group identity 17.740.005
 Small0.000 (0.000)   
 Large0.681 (0.245)   
Scan observation (random term)6.299 (0.629)   
Observation day (random term)0.570 (0.245)   
Constant−10.11 (0.773)   
Non-significant terms
Baboons stationary foraging 10.8330.361
 ≤0.5 group0.000 (0.000)   
 >0.5 group0.297 (0.325)   
Baboons travelling 10.8060.369
 ≤0.5 group0.000 (0.000)   
 >0.5 group0.316 (0.352)   
Baboons resting 10.7010.402
 ≤0.5 group0.000 (0.000)   
 >0.5 group−0.807 (0.964)   
Figure 3.

 The fitted values (±se) for the effect of (a) season, (b) time of day on the probability of kestrel–baboon associations. Values are obtained from the model parameters given in Table 1, controlling for the effect of all other significant terms and for the influence of repeated observations within and across days

Figure 4.

 The fitted values (±se) for the effect of (a) habitat type (b) proportion of baboon group travel foraging, and (c) baboon group identity, on the probability of kestrel–baboon associations. Values are obtained from the model parameters given in Table 1, controlling for the effect of all other significant terms and for the influence of repeated observations within and across days.

Hence, our results show that although kestrels were present in only 3.5% of all baboon-group scans, they were present in 22% of baboon-group scans observed during the first 6 h of the day in the early austral winter period, when the baboons were travel foraging in open desert habitat (n=238).


Rock kestrel associations with chacma baboons are more common in the early austral winter, show diurnal cycles and are more frequent when baboons are in open desert habitat, engaged in travel foraging and in a large group. These patterns support the hypothesis that kestrels associate with baboons because they attain foraging opportunities from doing so. The results are also consistent with our impressions that these are not chance events. Rather, when kestrels are present they appear to actively accompany the baboons, hovering above them and frequently swooping down to capture insects that have been flushed into the air. Other studies of bird–primate associations have similarly reported seasonal and diurnal patterns, and a tendency for associations to occur in conjunction with particular habitats and activities (e.g. Boinski & Scott, 1988; Zhang & Wang, 2000). These studies have also proposed that such associations arise from the improved foraging efficiency of the birds concerned, suggesting that this may be a relatively common pattern in bird–primate interactions. However, ours is the first study to demonstrate the statistical independence of many of these effects. We discuss each of our results in turn.

We have shown that kestrel–baboon associations are more than 50 times more likely to occur in the early winter period of our study, compared with the late winter (Fig. 3a). This pattern of association is consistent with the seasonality of breeding in locust and grasshopper species in response to rainfall in this desert habitat (e.g. Hunter & Elder, 1999). We have also shown a diurnal cycle of kestrel–baboon association independent of baboon activity cycles. This diurnal pattern is a likely consequence of the predicted activity cycles of kestrels, which are known to most actively forage in the early morning and late afternoon. We have also shown that variations in primate behaviour can clearly influence kestrel associations. The location of baboon groups is important. We have shown that despite baboons spending approximately equal proportion of time in one of two habitats, open desert and closed woodland, kestrels were more than 20 times more likely to associate with baboons groups in open desert habitat (Fig. 4a). This finding is consistent with the predicted increase in Orthoptera prey densities in this habitat type and/or improved detection and capture as a consequence of habitat structure. But perhaps most importantly, we have shown that kestrel–baboon associations vary predictably with primate collective behaviours and group size. Associations were more than two-and-a-half times more likely where at least 50% of a baboon group were engaged in travel foraging (Fig. 4b). Since the periods when baboons were engaged in other behaviours (resting, stationary foraging and travelling) had no significant effect, it is evident that the active and continued disturbance of vegetation and substrates during travel foraging is a critical element in the emergence of kestrel–baboon associations. In the case of group size, this is the first study to document a primate group-identity effect, which we interpret as a consequence of the respective sizes of our two baboon groups: a larger group size means more baboons to disturb Orthopteran prey. However, a larger range of group sizes is needed to illustrate a group size effect with certainty.

There are several areas where further research would be fruitful. One topic that would benefit from further investigation is the independent behaviour of the kestrels. For example, how does the foraging success of kestrels vary in the presence and absence of the baboons? The fact that pale chanting goshawks Melierax canorus have also been seen to associate with baboons under similar conditions (mid-morning in the early austral winter, while the baboons were travel foraging in open desert habitat), and show similar foraging behaviour (capturing on the wing the Orthopterans disturbed by the passage of the baboons) (A. J. King & G. Cowlishaw pers. obs.), indicates that the potential for increased foraging opportunity may be substantial. Further refinements of our predictions are also possible. For example Van Zyl et al. (1994) suggested that kestrels may undertake relatively long-distance seasonal movements to track spatial variation in insect abundance. If the number of kestrels in our locality declined in the late austral winter as a result of such movements, it might contribute to the reduced frequency of kestrel–baboon associations recorded during this period. The breeding behaviour kestrels at this study location are also unknown, although Namibian nest records indicate rock kestrels typically breed November–January elsewhere in Namibia (Southern African Ornithological Society, Van Zyl et al., 1994). Such information would allow us to test predictions about kestrel–baboon associations in the context of kestrel breeding biology. It would also be instructive to ask how kestrels locate their baboon groups. This may seem a relatively straightforward question, but our finding that kestrels more commonly associate with the larger of our two baboon groups might be partially explained by the fact that larger groups are more conspicuous. Bigger groups are not usually noisier, but when foraging do spread out over larger areas.

Our findings suggest that foraging opportunities for kestrels lead them to associate with baboons. An increase in foraging efficiency for at least one species is a commonly cited reason for the evolution of associations between species (Hino, 1998; Ruggiero & Eves, 1998; Rehg, 2006). Nevertheless, there are other possibilities. The most important alternative explanation is defence against predators. The predator-defence hypothesis requires that the associating species share common predators, so that they can benefit from each other's anti-predator behaviour (Zuberbuhler, 2000; Fichtel, 2004; Rainey, Zuberbuhler & Slater, 2004). However, in this case, baboons and kestrels are at risk from different types of predator: leopards and larger raptors, respectively (Cowlishaw, 1994; Petty et al., 2003). Although large raptors may prey on baboons in other areas of Africa, no predation has ever been observed in this population, and raptor attacks to date have been in the context of nest defence only (A. J. King & G. Cowlishaw pers. obs.). (See also Cheney et al., 2004 for a recent discussion of raptor-baboon predation). Similarly, leopards do not predate kestrels (Hayward et al., 2006). Consequently, predator defence is unlikely to be a factor for either species.

Do the baboons receive any foraging benefits or costs from the kestrels? We were unable to discern any foraging benefits, and the costs may similarly be limited. When the baboons are travel foraging in the open desert habitat, they are not only feeding from grasses, herbs and dwarf trees, but also turning over rocks and feeding on a variety of small invertebrate prey. These will opportunistically include Orthopteran species, which the baboons occasionally capture in mid-air as they take flight following their disturbance. However, such prey items make up such a small proportion of the baboon diet, and the kestrels only capture those Orthopterans that the baboons have already missed or ignored. Similarly, the baboons do not forage on other types of kestrel prey, such as lizards and mice. It is therefore unlikely that kestrels pose a foraging cost to the baboons. In addition, there have been no observations of aggressive conflict between the kestrels and baboons, unlike that reported in another (very unusual) record of bird–primate interaction at Tsaobis involving black kites Milvus migrans competing with baboons for access to a fresh klipspringer Oreotragus oreotragus carcass (see Davies & Cowlishaw, 1996). Rather, the baboons consistently ignore the kestrels. Overall, the evidence suggests that the baboons experience no cost, and gain no ascertainable benefit from the accompanying kestrels.

In conclusion, our study suggests that rock kestrels associate with desert baboons in order to prey upon the Orthopteran species flushed by baboons. Future work examining differences in the prey capture rates of kestrels when foraging alone and in the presence of baboons will establish whether such associations lead to improved foraging performance. Baboons do not appear to gain any benefit or incur any costs from this association. The kestrel–baboon association therefore appears to be a commensal relationship. Subsequent studies of bird–primate associations might usefully consider the possibility that bird species not only rely on primates to flush potential prey, but vary their frequency of associations dependent on the collective activities (and consequent flushing achievements) of their commensals.


We thank Fay Clark, Elise Huchard, Nick Isaac, Richard Pettifor and Volker Sommer for their help and support in this work, Anthony van Zyl for answering our questions on rock kestrel behaviour, Jocelyn Hacker for the photo of kestrels and baboons used in Fig. 2, and the anonymous reviewers for their comments on the paper. Thanks also to the entire Tsaobis baboon project 2005 field team for assistance with data collection. This research was funded by a Natural Environment Research Council (NERC) studentship to A.J.K., while G.C. was supported by a NERC Advanced Fellowship. This study complies with the guidelines set out by the Namibian Ministry of Environment and Tourism (MET), and was carried out in affiliation with the Gobabeb Training and Research Centre. We are grateful to MET, the Ministry of Lands and Resettlement, Tsaobis Leopard Park and the Swart family for research permission. This paper is a contribution to the ZSL Institute of Zoology's Tsaobis Baboon Project.