Switching spatial scale reveals dominance-dependent social foraging tactics in a wild primate

Introduction

Socially exploitative behaviours occur when individuals make use of the resources of competitors. A wide range of both theoretical and empirical studies over recent decades have highlighted such behaviours as a major potential cost of group living (Giraldeau & Dubois, 2008). Because resources such as food, mates, breeding territories, or safety from predation generally show variation in their distribution through space or time, individuals should benefit from gathering information about their local environment to improve decision-making (Valone, 1989; Valone, 2006; McNamara, Green & Olsson, 2006). However, when the personal collection of information requires search effort or risk taking, selection should favour the avoidance of these costs through the collection and use of social information, where individuals attend to the behaviours of others in a social group to exploit their efforts and knowledge (for review, see Valone & Templeton, 2002; Danchin et al., 2004; Rieucau & Giraldeau, 2011).

The dynamics of these interactions have been formalised as the producer-scrounger game (Barnard & Sibly, 1981; Barnard, 1984; Vickery et al., 1991). Supported by a wide range of empirical studies (e.g., Koops & Giraldeau, 1996; Mottley & Giraldeau, 2000; Morand-Ferron, Giraldeau & Lefebvre, 2007), the producer-scrounger game has emerged as the prevailing theoretical framework in which to study social foraging decisions (Vickery et al., 1991; Giraldeau & Caraco, 2000). In this game, producers actively search for resources, while scroungers instead rely on social information to exploit the discoveries of producers. The two tactics are considered mutually exclusive. Scrounging is thus under negative frequency-dependent selection since its success, being dependent on the efforts of producers, is determined by the relative frequencies of the two tactics within a group. This dynamic is expected to lead populations to an evolutionarily or behaviourally stable mix of producing and scrounging (Giraldeau & Dubois, 2008; Fawcett, Hamblin & Giraldeau, 2013). As such, scrounging behaviour has the potential to reduce the per capita rate of discovery of new resources (Vickery et al., 1991), which may act to reduce average individual fitness in a population (Coolen, Giraldeau & Vickery, 2007).

The basic producer-scrounger model assumes that an individual’s phenotype has no influence on its decision or ability to play either tactic. All individuals are essentially equivalent, and are expected to receive equal payoffs. However, many empirical studies have shown that an individual’s tactic choice may be strongly influenced or constrained by its phenotype (e.g., Beauchamp, 2001; Stahl et al., 2001; Di Bitetti & Janson, 2001; Kurvers et al., 2010). This has potentially important fitness implications, since theory predicts that phenotype-limited games may not reach an evolutionarily stable mix of strategies, resulting in differential payoffs across individuals (Parker, 1982).

Since scrounging behaviour represents the exploitation of another’s resource, one might expect it to be strongly influenced by social dominance. Specifically, the competitive advantage of high-ranking individuals should allow them to scrounge from others more easily (Parker, 1974; Maynard Smith & Parker, 1976; Hammerstein, 1981). Despite this expectation, empirical studies have not been unanimous. While some experiments have demonstrated a clear positive relationship between social dominance and scrounging behaviour (Stahl et al., 2001; Liker & Barta, 2002; Lendvai, Liker & Barta, 2006; McCormack, Jablonski & Brown, 2007), a number of other studies have not (Bugnyar & Kotrschal, 2002; Robinette Ha & Ha, 2003; Beauchamp, 2006; Teichroeb, White & Chapman, 2015). This conflict might be reconciled by considering more systematically the spatiotemporal distribution of resources faced by different taxa in both naturalistic and experimental settings. The competitive benefits of social dominance are expected to be associated with priority of access to resources, manifest as contest competition (Kaufmann, 1983; Łomnicki, 2009). Consistent with this, resource defence theory predicts that individuals should be more aggressive when defending a resource in accordance with both its value and how easily it can be defended (Grant, 1993; Grant & Guha, 1993; Robb & Grant, 1998). Empirical studies into dominance and resource defence have demonstrated higher foraging success for socially dominant individuals only when presented with limited food patches that are monopolisable (Theimer, 1987; Vahl et al., 2005).

Some researchers have suggested that the integration of producer-scrounger and resource defence theory might elucidate an interesting relationship between socially exploitative behaviour and dominance (Barta & Giraldeau, 1998; Giraldeau & Dubois, 2008). Specifically, dominant individuals should benefit disproportionately if they can use their competitive advantage to ensure that only they can use social information effectively. Two studies (Barta & Giraldeau, 1998; Lee et al., 2016) have explored this hypothesis by modelling the effects of between-individual asymmetries in competitive ability on producer-scrounger dynamics in a group. They found that when social rank conferred no competitive advantage to an individual—that is, resources were not monopolisable—groups converged on basic producer-scrounger equilibria in which all individuals behave equivalently and receive equal payoffs. In contrast, when individuals could use their social rank to gain a competitive advantage in monopolising a resource, scrounging behaviour was strongly associated with dominance, and dominant individuals achieved the highest payoffs (Barta & Giraldeau, 1998; Lee et al., 2016). The driving force behind this pattern was the fact that scrounging behaviour requires that a competitor is joined at a resource in space and time, forging a causal link between the degree to which contest competition acts and constraints on an individual’s ability to use social information (Lee et al., 2016). However, to date there has been no attempt to test these predictions empirically, either in the laboratory or under naturalistic conditions.

In this study, we explore a key prediction generated by the unification of producer-scrounger and resource defence theories, namely that there should be a strong link between social dominance and the scrounger tactic only when resources are monopolisable. We did this by studying the natural social foraging decisions made by wild chacma baboons (Papio ursinus) across two spatial scales that are expected to differ in the degree to which dominant individuals can monopolise food. At the first spatial scale—the ‘patch’—resource clumps were too large to be monopolised independently, while at the second—the ‘sub-patch’—resource clumps were smaller and monopolisation was possible (see ‘Methods’ for further details on how these spatial scales were defined). Because our focus was on naturalistic behaviour, we did not manipulate the information available to individuals while foraging to manufacture a situation where joining a competitor always represented the exclusive use of social information, which could accurately be termed scrounging (Vickery et al., 1991). Rather, we consider the observable joining behaviours of individuals as they foraged, representing competitive interactions fundamental to the predictions of producer-scrounger theory (Lee et al., 2016). In this way, we explore the competitive constraints that social rank may impose on an individual’s ability to use social information through joining behaviour, and the relation of such constraints to resource monopolisability in a naturalistic setting in which information use is expected to be important. We make the following three predictions: (1) joining behaviour should show a strong positive association with dominance rank at the smaller, sub-patch level but not at the larger, patch level; (2) individuals should join those to whom they are dominant at the sub-patch level, but there should be no systematic asymmetry in dominance when joining occurs at the patch level; and (3) joining should be associated with competitive exclusion (i.e., resource monopolisation) at the sub-patch level, but not at the patch level.

Materials and Methods

Study site and species

Fieldwork was conducted at Tsaobis Nature Park, Namibia (22°23′S, 15°45′E), during two three-month periods between August and October in 2012 and 2013. Two groups of chacma baboons, hereafter referred to as troop ‘J’ (group size and compositions: N_J,2012 = 54, adult females = 16, adult males = 15, juveniles = 23; N_J,2013 = 58, adult females = 18, adult males = 9, juveniles = 31) and troop ‘L’ (N_L,2012 = 51, adult females = 18, adult males = 6, juveniles = 27; N_L,2013 = 62, adult females = 19, adult males = 11, juveniles = 32), were the focus of study. All baboons were individually recognisable and habituated to the presence of observers at close proximity. Each group was followed daily from dawn until dusk (see Huchard et al., 2009 for further information). For each year, data were collected for all individual baboons >6 months of age (the age at which young baboons begin to forage independently of their mother) at the start of the study period, resulting in a total sample of 101 individuals (2012: 54 adults, 43 juveniles; 2013: 50 adults, 41 juveniles). Differences in the sample of individuals across years were due to death, emigration, or passing the minimum age threshold.

Chacma baboons are an ideal model system for our study, since they live in large, stable social groups in which linear dominance hierarchies are clear (Altmann & Altmann, 1973), individuals generally feed at the same time (King & Cowlishaw, 2009), the use of social information while foraging has been demonstrated in field-based experiments (Carter, Torrents Ticó & Cowlishaw, 2016), and socially exploitative foraging interactions are common (King, Isaac & Cowlishaw, 2009; Marshall et al., 2012). Furthermore, our study troops spent approximately 80% of their foraging time during the study period in a riparian woodland environment, characterised by large trees including Faidherbia albida, Salvadora persica, Acacia erioloba, Acacia tortilis, and Prosopis glandulosa. Within this feeding environment, we defined two spatial scales between which the ability of dominant individuals to monopolise food were predicted to differ: the patch and the sub-patch.

The patch represents the scale traditionally used in foraging theory and ecology, and is defined as a spatially discrete unit of a food resource (Wiens, 1976). Here, we refined this definition such that the operational definition of a patch was a single tree or shrub, or a collection of conspecific trees or shrubs growing together with a continuous canopy separated by no more than 1 m (median surface area = 156 m², interquartile range = 28–237 m²; n = 59; see Marshall et al., 2012 for further details). In contrast, the sub-patch was defined as the area in a patch within which an individual could feed without travelling (i.e., within arm’s reach of a stationary baboon, approximately 2.25 m²). This is equivalent to the ‘feeding station’ scale that has received some attention in the foraging literature (Kotliar & Wiens, 1990; see Searle, Hobbs & Shipley, 2005 for a review). Given the large size of patches compared with sub-patches, dominant individuals should be able to competitively exclude subordinate others more easily at the latter scale.

Results

Dominance and social foraging decisions at different spatial scales

At the patch scale, joining behaviour was common regardless of social rank (Table 1; Fig. 1A), consistent with our predictions. Although there was some increase in joining with social dominance in adults, even the lowest ranked individuals entered occupied patches approximately 55% (95% confidence intervals: 41% and 67%) of the time. While joining was in general much less common at the sub-patch scale, there was a strong positive relationship in adults between dominance and joining behaviour (twice that at the patch scale) that was consistent with our predictions (Table 1; Fig. 1B). The lowest ranked adults had around a 1% probability (95% confidence intervals: 0% and 2%) of joining when entering a new sub-patch, while mid-ranked and top-ranked adults did so approximately 4% (95% confidence intervals: 3% and 6%) and 11% (95% confidence intervals: 8% and 17%) of the time, respectively.

Table 1:

Factors predicting the probability of joining behaviour and competitive exclusion associated with joining.

Response	N	Fixed effect	β	s.e.	χ2	p
Probability of joining	6,911	Intercept	−0.05	0.26
		Spatial scale (Sub-patch)	−4.51	0.28
		Rank	1.10	0.38
		Age class (Juvenile)	1.57	0.30
		Troop (L)	0.20	0.15
		Year (2013)	0.27	0.12
		Spatial scale (Sub-patch) * Rank	1.18	0.37	10.30	0.001
		Rank * Age class (Juvenile)	−1.68	0.54	9.03	0.003
		Spatial scale (Sub-patch) * Age class (Juvenile)	−0.68	0.21	10.40	0.001
Probability of competitive exclusion	385	Intercept	−2.44	0.47
		Spatial scale (Sub-patch)	4.55	0.65	164.27	<0.001
		Age class (Juvenile)	−1.61	0.47	14.06	<0.001
		Troop	−0.89	0.44

DOI: 10.7717/peerj.3462/table-1

Notes:

Model reference categories: Spatial scale (Patch), Age class (Adult), Troop (J), Year (2012).

As predicted, the effects of social rank on the probability of joining were weaker in juveniles, and this held across both spatial scales such that there was in general no relationship between dominance and joining frequency in juveniles (Table 1; Fig. 1). In addition, juveniles were on average more likely than adults to join at the patch scale, but this primarily reflected low-ranking juveniles joining much more frequently than similarly ranked adults when entering a new patch. This effect was somewhat weakened at the sub-patch scale: low-ranked juveniles joined more frequently when entering a new sub-patch than similarly ranked adults but the pattern was reversed for high-ranked juveniles.

Combined, our results reflect support for the three possible two-way interactions between spatial scale, dominance rank, and age class (Table 1). Our lack of support for a three-way interaction between these effects (likelihood ratio test: χ² = 0.70, p = 0.41) reflects the fact that there was some increase in joining with social rank in adults at both spatial scales. This meant that a stronger effect of dominance in adults at the sub-patch versus patch level, combined with no clear relationship between joining frequency and social rank in juveniles at either spatial scale, was captured by a general (i.e., not age class-specific) increase in joining with dominance at the sub-patch level and a general (i.e., not spatial scale-specific) decrease in the effect of dominance on joining in juveniles.

Social constraints on joining behaviour at different spatial scales

At the patch scale, there was no evidence that adult individuals joined others systematically higher or lower ranked than themselves compared with the rank asymmetries expected if individuals joined others randomly with respect to their rank difference (Fig. 2A, N_patch,adult = 313). Although adults were less likely to join individuals dominant to them at this spatial scale, this could be explained under joining behaviour that was random with respect to rank difference by the finding in our first analysis that the frequency of joining behaviour increased slightly with increasing rank (Fig. 1A).

At the sub-patch scale, adult individuals joined individuals dominant to themselves in 9% of join events, and so were less likely to do so than would have been expected if individuals joined others randomly with respect to rank difference (Fig. 2B; N_{sub-patch, adult} = 154) and given the fact that high ranked individuals are much more likely than low ranked individuals to join others at this spatial scale (Fig. 1B). We thus found support for our prediction that adult individuals would consistently join others lower ranked than themselves at the sub-patch level.

Comparisons between observed and random joining behaviour for juveniles were broadly similar to those for adults. At the patch scale, the rank asymmetries at observed join events did not deviate from expectations under joining behaviour random with respect to rank difference (Fig. 2C; N_{patch,juvenile} = 349). At the sub-patch scale, we observed juveniles joining others to whom they were subordinate in 49% of cases. Although this shows that juveniles were frequently able to join others dominant to themselves, they nonetheless did so less often than expected under random joining (Fig. 2D; N_{sub-patch, juvenile} = 147). Despite the fact that our first analysis found no relationship between dominance and joining frequency for juveniles at the sub-patch level (Fig. 1B), the random distribution at this spatial scale reflects the fact that juveniles are usually below average in rank.

Discussion

We provide empirical evidence that joining behaviour should be more strongly related to social rank when the competitive asymmetries associated with dominance are stronger, in support of previous theoretical predictions (Barta & Giraldeau, 1998; Lee et al., 2016). We show that changes in resource monopolisability can mediate this shift in competitive asymmetry through changes in competitive exclusion at different spatial scales. At the larger patch scale, adults could join others regardless of any differences in their social ranks, and did so frequently. At the smaller sub-patch scale, joining was a rarer event—likely reflecting the higher finder’s share at this spatial scale (Vickery et al., 1991)—but also represented a more exclusive tactic, almost non-existent in the lowest-ranking individuals but increasing in probability with social dominance. When resources are monopolisable, socially subordinate individuals may thus be constrained in their ability to exploit available social information when its use requires joining behaviour, as is assumed by producer-scrounger theory. Since the size of a resource is expected to influence its economic defensibility (Grant, 1993), our study supports calls to unify producer-scrounger and resource defence theory in order to better understand the relationship between dominance and socially exploitative behaviours (Barta & Giraldeau, 1998; Dubois, Giraldeau & Grant, 2003; Dubois & Giraldeau, 2005; Giraldeau & Dubois, 2008). However, our study also highlights two areas in which our understanding of the evolutionary ecology of animal information use is still lacking, requiring further theoretical and empirical developments.

First, we showed a strong increase in the frequency of joining at higher ranks for adults at the sub-patch scale, where single individuals could use their dominance to exclude competitors. However, the pattern we observed was weaker than that predicted by Barta & Giraldeau (1998), who suggested a complete absence of scrounging in all but the most dominant individuals when resources could be effectively monopolised. We found that even middle- and low-ranking adults can scrounge from others, provided that the others they join are even lower ranked than themselves. The failure of the producer-scrounger model to predict this pattern likely reflects two of its assumptions. Specifically, the model is built such that any individual playing scrounger can access all the discoveries of others, since (1) resources are assumed to be so rare that they are discovered one-at-a-time and (2) scroungers can access perfect social information and so detect each discovery (Vickery et al., 1991; Giraldeau & Caraco, 2000; cf. Ohtsuka & Toquenaga, 2009). When resources can be monopolised, these conditions mean that only the highest-ranked individuals will benefit from scrounging behaviour. However, in many groups of social foragers multiple patches can be discovered at the same time (and simultaneous discoveries or options will likely be the norm for many other types of resource too). As such, scrounging individuals will be unable to access all resource discoveries even if they have perfect social information (Ohtsuka & Toquenaga, 2009; Afshar & Giraldeau, 2014). Furthermore, since individuals within social groups are unlikely to be in close proximity at all times (Krause & Ruxton, 2002; for this study population, see Cowlishaw, 1999; Castles et al., 2014), scrounging individuals are unlikely to possess perfect social information regarding all discoveries occurring at the group level (Barta, Flynn & Giraldeau, 1997; Hirsch, 2007), and any temporal and energetic costs associated with scrounging may vary both within and between individuals.

Under the conditions of simultaneous discoveries and imperfect access to social information, the most dominant individuals will be unable to monopolise all resource discoveries, regardless of the economic defensibility of single resource patches. Instead, as shown here, the difference in rank between individuals across a hierarchy should play an important role in mediating scrounging behaviour. This finding is consistent with the predictions of a recent game theoretic model (Lee et al., 2016), which proposes that if the highest ranking animals in a group do not detect a particular patch discovery, and/or are occupied at another discovery, then middle and lower ranking animals can benefit from scrounging, provided that the producers from whom they are scrounging are lower ranking than themselves. Our result that joining at the sub-patch level still increases with social rank likely reflects the fact that higher ranked individuals have more competitors who are subordinate to them. The most dominant individual should thus be unconstrained in its ability to act upon opportunities to scrounge, and constraints should increase down the dominance hierarchy. Furthermore, theory predicts that if more dominant individuals are more effective at monopolising other’s resource discoveries, then they should continue to scrounge even under conditions that drive lower-ranked individuals to switch to the producer tactic (e.g., when the finder’s share is large; Lee et al., 2016). Individuals may also benefit from positioning themselves so as to maximise scrounging opportunities (Barta, Flynn & Giraldeau, 1997; Di Bitetti & Janson, 2001; Hirsch, 2007). Since dominant individuals may be better able to secure more central positions in the group, this may improve their ability to detect, and increase their proximity to, the discoveries of others (Barta, Flynn & Giraldeau, 1997; King, Isaac & Cowlishaw, 2009), and further reinforce the effects of dominance on scrounging behaviour.

Second, we showed that the relationship between social dominance and joining is strongly influenced by age class. In stark contrast to adults, juvenile baboons showed no general relationship between social dominance and joining, particularly at the sub-patch level. While an adult baboon’s ability to join competitors at this spatial scale depended strongly on their relative dominance, juveniles were less constrained by their social status. Indeed, juvenile behaviour accounted for almost all instances where a subordinate joined a higher-ranking individual at the sub-patch level. There are two likely explanations for this pattern. Firstly, rank acquisition in chacma baboons is mediated primarily through maternal reinforcement. Specifically, a mother will use aggressive behaviours to establish the dominance of her developing offspring over others subordinate to her (Cheney, 1977; Holekamp & Smale, 1991; Lea et al., 2014). Consequently, social rank during early life might be particularly sensitive to context, such that social interactions between juveniles may only reflect rank differences between their mothers in the presence of the dominant mother. In the absence of the dominant mother, older but ‘subordinate’ juveniles might use their larger size to join younger, smaller, but more ‘dominant’ competitors. Secondly, there is evidence that juvenile baboons, like the juveniles of several other primate species (Janson, 1985), are more frequently tolerated at feeding sites than adults. The presence of co-foraging juveniles may impose only a minimal direct cost to adults, but permitting close kin access to resources may provide inclusive fitness benefits. In particular, father-offspring relationships in chacma baboons afford juveniles access to high-quality feeding sites (Huchard et al., 2013). Such toleration may mean that, in addition to better access to monopolisable resources, low-ranked juveniles may not be constrained in their ability to use social information in the same way that similarly ranked adults will be. Our findings do not provide strong evidence in support of either tolerance or juvenile rank instability in disrupting the expected positive relationship between social rank and joining behaviour when resources are monopolisable. However, we might expect juvenile rank stability to disrupt rank asymmetries but still to involve competitive exclusion when resources are monopolisable, while toleration of juveniles should result in reduced levels of competitive exclusion. The fact that the probability of competitive exclusion was reduced for juveniles compared with adults at the sub-patch scale might suggest that toleration is playing a bigger role than juvenile rank instability in our observations.

Our study demonstrates the way in which the monopolisability of resources may drive social constraints on a subordinate individual’s ability to use joining behaviour to access them. We also show that such constraints may be relaxed in juveniles. Since resources generally show some uncertainty in their distribution through space and time, information use is likely to play a key role in resource acquisition. Our study illustrates how competitive processes associated with dominance might facilitate or constrain an individual’s ability to benefit from collecting social information when its use requires joining behaviour, as in the producer-scrounger framework (Barta & Giraldeau, 1998; Lee et al., 2016). An important step in future research will be to develop frameworks that simultaneously consider how resource distributions underpin (1) the strength and type of competition between individuals, (2) the benefits of collecting information socially versus personally, and (3) the rate at which such information becomes out-dated. This approach will elucidate the environmental conditions that should generate interdependencies between contest competition and social information use by highlighting when social information use should be dependent on joining behaviour at a resource. These insights will allow better characterisation of the ways in which competition can modulate relationships between an individual’s ability to use information and its access to resources, and the implications of such modulations for the dynamics of natural populations.

Supplemental Information