16 Avian Social Relations, Social Cognition and Cooperation Thomas Bugnyar and Jorg J.M. Massen Social Relationships We recently encountered a group of wild common ravens, Corvus corax, gathering around a fresh deer carcass. As the body had not yet been opened the birds could access only the soft tissue close to the mouth. One bird quickly monopolised this part, aggres- sively defending it against feeding attempts by other ravens. The exceptions were two other birds: one of them repeatedly got full access and always carried off chunks of ‘meat; the other sat next to the feeding spot, signalling submissively to the monopo- lizing subject while taking a few bites. With time, more and more ravens gathered at the carcass and at some point two of them simultaneously challenged the monopolizing subject, pulling it off the carcass. These two birds proceeded to monopolise the feeding spot together until other ravens chased them. Such a scenario is very likely when ravens aggregate at patchily distributed but highly valued food sources (Heinrich, 1989). Whereas most are prevented from feeding by some powerful individuals — usually adult males — a few individuals seem to be allowed to feed. These ravens might be pair partners of the monopolizing males (in our example the bird that repeatedly carried off chunks of meat) or other affiliates like siblings or grown-up offspring (in our example, the bird that took a few bites). Monopolizing males may get displaced from the feeding spot, though, when two or more individuals work together. Ravens’ success during group foraging is thus not only determined by physical strength and resource holding potential, but also by their social relationships and support in conflicts. Social relationships play a key role in the daily life of many animals, Defined as content, quality and pattern of interactions over time (Hinde, 1976), they capture the outcome of iterated interactions between given individuals. When based on agonistic interactions, we refer to the relationship as dominance (Drews, 1993); when based on affiliative interactions, we tend to use the term social bond, which may include rela- tionships between kin as well as between non-related individuals such as pair partners or friends (Silk, 2002; Massen et al., 2010). Functionally, dominance relations regulate the access to limited resources, with ritualised displays functioning to prevent excessive fighting, whereas social bonds act as cooperative units, in which individuals support each other in various ways. Social relationships may vary considerably across individuals with respect to their direct benefit, e.g. the value of food or help in confiicts may be different when being16 Social Relations, Social Cognition and Cooperation 315 shared or given by a pair partner or a sibling. Likewise, the relationships between individuals may vary in the level of mutual tolerance and their stability over time (Cords & Aureli, 2000). Recognition of these differences in relationship quality between individuals may be critical for manoeuvring the social world, as it allows individu- als to predict others’ behaviour according to the situation (Kummer, 1978; Cheney et al., 1986). For instance, in our raven example above, the bird who repeatedly car- ried off food likely knew that her relationship to the monopolizing male allowed her to approach the food without hesitation, whereas the grown-up offspring could not be as sure, hence the submissive signalling. The other ravens probably noticed the privi- leged status of the pair-bonded female, preventing them from aggressively stealing food from her, ‘As dealing with different types of social relationships quickly gets complicated, in terms of the information that needs to be taken into account in which situation and context (Jolly, 1966; Humphrey, 1976), life in structured social groups has been con- sidered as one of the main driving forces for brain evolution (Byrne & Whiten, 1988). Specifically, it has been proposed that living socially selects for cognitive abilities such as conceptual knowledge, inferential reasoning, attribution and prospective cognition (Tomasello & Call, 1997; see Barrett & Henzi, 2005 for a different view). Support for these ideas comes from lab and field studies, which traditionally had the focus on pri- ‘mates and other social mammals (e.g. de Waal, 1982; Cheney & Seyfarth, 1990, 2007; Hare et al., 2000; de Waal & Tyack, 2003; Mulcahy & Call, 2006) but recently have included other taxonomic groups like birds (e.g. Emery & Clayton, 2004; Paz-y-Mifio et al., 2004; Bugnyar, 2013) and fish (e.g. Bshary et al., 2014). A critical assumption in social cognition is that individuals can use their knowledge about others to choose and/or recruit appropriate cooperation partners (Byrne & Whiten, 1988). Yet, several cooperative behaviours like raising young, sharing food, defending resources and forming alliances can also be accomplished on the basis of what are considered to be relatively simple mechanisms like a combination of decision rules, motivational states and learning (Brosnan & de Waal, 2003; Schino et al., 2007). Con- temporary researchers try to understand the cognition that underpins cooperation by discriminating among different factors that could influence cooperation such as equity in reward distribution, individual differences and social constraints (Noé, 2006). Coop- erative interactions with bonding partners, for instance, may require different cognitive control mechanisms than do cooperative interactions with non-bonded and/or unfamil- iar individuals (Massen et al., 2010): while the former may reflect an attitude towards a partner based on the partner's behaviour in short- or longer temporal succession (atti- tudinal reciprocity, de Waal, 2000; emotional book-keeping, Schino et al., 2007), the latter may require some form of mental book keeping; i.e., memory for others’ debits and credits in cooperative interactions (de Waal & Luttrell, 1988). Another pressing question concerns possible species differences in the cognitive capacities for cooper- ation, which might be predicted by the species’ socio-ecology and/or phylogeny and tested with a comparative approach (e.g. Fitch et al., 2010; Salwiczek et al., 2012) Comparison between species of distant taxonomic groups such as mammals and birds are particularly revealing, as they allow us to test if similar mechanisms are a result of316 ‘Thomas Bugnyar and Jorg J.M.Massen ‘convergence in response to similar environmental conditions rather than to phylogenetic descent (Emery & Clayton, 2004; MacLean et al., 2014), In this chapter, we will review what is known about the social cognition of birds. We will first focus on the social organisation, asking what kind of social structure may favour which type of social relationship and what birds know about their own and others” relationships. We will then review different forms of avian cooperation and discuss the use of social knowledge in different cooperative settings. Finally, we will discuss the possibility of other-regarding motivations underlying cooperative decisions. Social Organisation and Social Knowledge in Birds Pair Bonds as Main Units At first glance, birds appear to be very different from mammals with regard to their social organisation. Most birds are highly mobile and are seasonally variable in their social structure, e.g. during migration or between breeding and non-breeding periods (Lack, 1940). Furthermore, egg incubation and the lack of lactation leads to social monogamy as the far most common breeding strategy, ic. males and females pair up for raising young in about 80 per cent of all bird species (Cockburn, 2006). As pair members need to coordinate for offspring care, they should benefit from mechanisms that help them to predict the other's behaviour. These skills may be extrapolated to any social interaction and make monogamously breeding birds prone to cooperate in a vari- ety of contexts. In species with long-term monogamy, where pair partners stay together over sev- «ral breeding periods, pairs tend to increase in physiological synchronisation (Hirschen- hauser et al., 1999; Weif et al., 2009) and behavioural coordination the longer they stay together (Black, 1996). Furthermore, as species that form long-term pair bonds tend to have bigger brains relative to their close relatives with other breeding systems (Dun- bar & Shultz, 2007), it has been suggested that long-term pair bonding may represent a ‘qualitative shift from temporal social units to complex negotiated relationships that act ‘as social alliances also outside breeding (Emery et al., 2007; Dunbar & Shultz, 2007). Indeed, bonds in long-term monogamous birds show similar characteristics to those of the social bonds described for mammals (Emery, 2006; Emery et al., 2007; Scheiber et al., 2008; Fraser & Bugnyar, 2010a, 2010b, 2011). In many long-term monogamous species, social bonds can also be found in the non- reproductive life stages (e.g. as juveniles and sub-adults; Wanker et al., 1996; Scheiber et al., 2005; Loretto et al., 2012), or in the reproductive life stage but beyond the pair (Boucherie et al., 2016). Such bonds may concer siblings (birds originating from the same nest) or unrelated individuals from the same or different sex and age class (some- times referred to as ‘friends’ or allies), and are expressed in spatial association patterns and various forms of social support. In greylag geese Anser anser, for instance, female siblings tend to rest close to each other even as adults (Weif et al., 2008) and unrelated but bonded males may acquire high dominance status by supporting each other in fights16 Social Relations, Social Cognition and Cooperation 317 (Kotrschal et al., 2006). From an evolutionary perspective, the tendency to engage in bonds with conspecifics before and/or outside of reproduction may represent the first step of generalizing to social partners other than the pair (as has been suggested for primates, Shultz & Dunbar, 2007). Social Knowledge Given the importance of social relations in the daily life of many birds, an open ques- tion is how individuals process their own and others’ relationships. One possibility is that they act on the basis of innate predispositions such as always behaving aggres- sively towards smaller individuals and always submitting to larger individuals (Lack, 1940). Reliance on intrinsic factors such as body size (or age, sex, confidence) may ‘work for simple interactions like one-shot encounters at monopolised resources; how- ever, it hardly allows the flexibility required to deal with individualised relationships in structured groups (Barnard & Burk, 1979). Here, birds need to form associations and/or ‘mental representations about relationships, which enable them to deal with others by applying learned rules and/or by making inferences. Dominance relationships have formed a major part of work addressing avian social knowledge. Starting with the seminal work by Schjelderup-Ebbe (1935) on pecking order in hens, the default assumption is typically that individuals need to learn their posi- tions in a dominance rank hierarchy relative to those of others through direct interaction. If groups get large or unstable over time, however, interactions with every individual and assessment of their fighting abilities are time demanding and often impossible. One solution to the problem is to infer the rank of unfamiliar individuals by observing them interacting with individuals of known rank (Bond et al., 2003), following the principle that if A>B and B>C, then also A>C, Such transitive inference has been experimen- tally shown for various avian taxa using operant designs (e.g. corvids: Bond et al., 2003, 2010; Lazareva et al., 2004; Mikolasch et al., 2013; geese: Weil et al., 2010; Weil & Scheiber, 2013; pigeons: von Fersen et al., 1991; chicken: Daisley et al., 2009). Fur- thermore, its practical application in competitive encounters with unknown individuals has been demonstrated in Pinyon jays Gymnorhinus cyanocephalus (Paz-y-Mifio et al., 2004). As expected, the birds’ social complexity also seems positively correlated with the performance in transitive inference tests (MacLean et al., 2008; Bond et al., 2010), In long-term monogamous species, social bonds have a strong effect on dominance. Notably, bonded females rise in rank and acquire a similar dominance status as their partner (Lorenz, 1931; Gwinner, 1964). This phenomenon of rank dependence is prob- ably caused by the mutual social support of bonded individuals, who tend to help each other in conflicts and to show-off their relationships post-conflict. The latter may involve species-specific behaviours like triumphing in greylag geese (Fischer, 1965) and bill twining in rooks Corvus frugilegus (Seed et al. 2007). These behaviours may facilitate third-party recognition in others, i.e. learning about the relationship among individuals from a bystander perspective (Emery et al., 2007). Indirect support for the possibility that birds may recognise third-party relationships comes from a physiological study on318 ‘Thomas Bugnyar and Jorg J.M.Massen greylag geese, showing different heart-rate modulations in bystanders that watch social and non-social events (Wascher et al., 2008). Third-party recognition is considered to be cognitively sophisticated as it involves comprehending tertiary relationships (Tomasello & Call, 1997), which in turn may be the basis for abilities like inferential reasoning and attribution. Notably, animals capa- ble of third-party recognition may take into account that their relationships with specific individuals are conditional on the presence or absence of other individuals, allowing flexibility in social interactions (Jolly 1966). Seminal work by Cheney and colleagues (Cheney et al., 1995) revealed that yellow baboons, Papio cynocephalus, do under- stand the relationship between others, and later work showed that chacma baboons, P hamadryas ursinus, keep track of both the rank- and kin- relations of others simulta- neously (Bergman et al., 2003). Apart from primate species, this form of third-party recognition has so far only been shown in one other social mammal species; i.e., spotted hyenas, Crocuta crocuta (Engh et al., 2005). Third-party recognition in birds has been investigated experimentally in captive ravens by exposing individuals to playbacks of simulated interactions between a dom- inant and subordinate conspecific (Massen et al., 2014a). The simulated sequences always consisted of the same number and type of calls but differed in who was giving the dominant and submissive vocalisation, respectively: the played pattern either followed the existing dominance rank hierarchy in the group (dominant displays towards subor- dinate) or it was reversed (subordinate displays towards dominant; compare Bergman et al., 2003). The listeners clearly differentiated between these conditions and behaved as if the latter playback sequence violated their expectation about the existing rank hier- archy: in comparison to their reaction to expected interactions, they showed increased activity and stress-related behaviours to a simulated rank reversal. Note that these differ- ent behavioural responses could not be explained by potential difference in call modula- tions between dominants and subordinates. The ravens could also detect the difference when the playbacks featured members of their neighbouring group with which they never had any physical contact, but unlike their responses to apparent rank reversals within their own group, they responded to simulated rank reversals of neighbours with a change in vocalisation and attention patterns. These findings suggest that ravens come to represent mentally the dominance relationships of others not only via direct encoun- ters but also by mere observation ‘These findings on third-party recognition fit with observations on third-party inter- ventions in wild ravens (Massen et al., 2014b). Such interventions concern not only coalitionary support in conflicts but also separation attempts in others’ affiliative inter- actions like grooming and playing (Braun & Bugnyar, 2012). In the latter case, there is a clear effect of bonding status on who intervenes and who becomes the target of intervention: interveners are strongly bonded and they selectively target ravens that are in the process of becoming bonded themselves. Most interventions just result in the ter- mination of the affiliative interaction of others and rarely involve signs of competition for a specific bonding partner such as the replacement of one of the affiliating birds by the intervener. Instead, these interventions probably represent attempts to prevent others from bonding (Massen et al., 2014b). Functionally, this makes sense as bonded ravens16 Social Relations, Social Cognition and Cooperation 319 rise in rank (Gwinner, 1964; Braun & Bugnyar, 2012). However, as there seems to be hardly any direct benefit for the intervener in separating affiliating ravens, from a cogni- tive perspective, such interventions may require the representation of others’ affiliative relationships and, possibly, an understanding that bonded individuals tend to cooperate. Cooperation in Birds ‘A major benefit of avian social bonds is that individuals may act as cooperative units not only for reproduction but also in a variety of other situations (Clutton-Brock, 2009) and thereby can achieve goals that cannot (so easily) be reached alone. Here, we will describe forms of naturally occurring cooperation in the setting of resource acquisition and defence, with a focus on the specific roles individuals play when cooperating, i.e. whether they act apart together, take similar and/or complementary roles. We will then discuss experimental work on cooperation in birds and the possible underlying cognitive mechanisms. Third, we will consider reciprocal altruism as an evolutionary explanation of cooperation and discuss the possible proximate mechanisms underlying such reciprocity. And finally, we will examine the possibility of altruistic motivations underlying cooperative behaviour. In all parts we will specifically focus on the role of social relationships. Cooperation for Resource Acquisition and Defence Many bird species prey on other animals for food. They can hunt these other animals either solo or in cooperation with others. Cooperative hunting can be simple acting apart together, as, for example, in herons and egrets where an increase in group size increases individuals’ success rates of capturing small aquatic animals (Krebs, 1974; Scott, 1984; Cézilly et al., 1990). This form of cooperation relies on increased individual benefits, can be relatively anonymous, does not necessarily require any form of social knowledge and whether it is truly cooperation is arguable. From a socio-cognitive point of view, cooperative hunting becomes interesting when it involves coordination in, for example, turn-taking and even complementary roles. In this scenario, individual recognition and close social bonding might be paramount, since it increases trust in, and predictability of, the partners an individual relies on while hunting. Such complex hunting tactics have been observed among breeding pairs in various raptors or raptor-like binds (e.g Hector, 1986; Hannah, 2005); moreover, it has been described for cooperatively breed- ing family groups (¢.g., Thiollay, 1991; Bowman, 2003), and for some corvid species that show advanced social networks also in non-breeding groups (e.g. common ravens: Hendricks & Schlang, 1998). Mostly, these hunts concern tandem attacks, where two birds alternate their attacks. However, the group hunts of Harris Hawks, Parabuteo unicinetus (Bednarz, 1988), and brown-necked ravens, Corvus ruficollis (Yosef & ‘Yosef, 2010) in which individuals take up different roles, requires more elaborate ‘coordination, and is comparable to the cooperative hunting techniques of several social carnivores, other social mammals and between some fish (e.g. Stander, 1992; Boesch, 2002; Bshary et al., 2006).320 ‘Thomas Bugnyar and Jorg J.M.Massen Cooperation among birds does not only involve attacking, but also defence strate- ‘gies when being attacked, either against predators or in defending territories from con- specifics, Cooperative territorial defence ranges from the duetting of many songbirds (reviewed in Dahlin & Benedict, 2014), to physical aggression towards intruders by the two pair mates separately (c.g. Quinard & Cézilly, 2012) or together (e.g. Hall & Peters, 2008), to cooperative Iek defence of unrelated males (e.g. Foster, 1981), to highly coordinated tandem attacks (e.g. Bossema & Benus, 1985). Cooperative anti-predator behaviour comprises individually motivated participation that leads to large mobs (i.. acting apart together) or mutual participation only (Ostreiher, 2003). Moreover, mobs can consist of breeding pairs that are helped by their neighbours, be it conspecific (c.g. Grabowska-Zhang et al., 2012) or heterospecific (Krams & Krama, 2002), of family groups of cooperative breeding species (e.g. Arnold, 2000; Griesser & Ekman, 2005) ot of complex aggregations with increased tolerance in those mobs (e.g. American crows, Corvus brachyrhynchos: Cornell et al.,2012). Such a variety of behaviours may reflect a range of underlying cognitive mechanisms. However, these have rarely been investigated (but see below for examples). Experimental Studies Examining Cooperation in Birds Studies on cognitive aspects of cooperation are strongly influenced by methods derived from primates, which are focused on how individuals interact during foraging. Manip- ulations of the set-up involve the accessibility of food, so that it can be reached with or without help from others, the distribution of food, so that it can or cannot be shared with ease and the role and availability of partners, so that different strategies are possible (No8, 2006). Depending on the paradigm, individuals are thus tested for their under- standing of when cooperation is necessary (whether a goal can be achieved alone), how cooperation works (who is taking which role) and who is an effective coopera- tion partner. Other critical questions concern the ability to cope with delayed benefits in reciprocal cooperation, which may be mediated by variability in cognitive ability or motivation. ‘A commonly-used paradigm in mammals to study a species’ cooperative problem solving capacities is the loose-string paradigm (Hirata, 2003, see Figure 16.1), where ‘two individuals need to simultaneously pull two ends of a rope to gain access to a platform with rewards, whereas when only one individual pulls, the rope will become unthreaded without moving the platform. In the handful of studies to date on birds, the birds were able to coordinate their actions to achieve cooperation. Moreover, in four species the degree of inter-individual tolerance shown in daily life predicted the dyads’ success in the task, supporting the assumption that tolerance and social bonding ‘might be important for avian cooperation (rooks, Seed et al., 2008; but see Scheid & Noé, 2010; African grey parrots Psittacus erithacus, Péron et al., 2011; ravens, Massen etal., 2015a; Asakawa-Haas et al., 2016; keas, Nestor notabilis, Schwing et al., 2016). Furthermore, these data corroborate findings in other social species like marmosets, Callithrix jachus (Werdenich & Huber, 2002), chimpanzees (Melis et al., 2006; Suchak et al., 2014), bonobos, Pan paniscus (Hare et al., 2007) and hyenas (Drea & Carter,16 Social Relations, Social Cognition and Cooperation 321 Figure 16.1 Two ravens cooperating in the loose-string paradigm. Only when they simultaneously pull the rope the platform will move towards the wire-mesh, granting them access to the rewards on the platform, When only one raven pulls the rope, the rope will become unthreaded and the possibility to move the platform will be lost. Drawing adapted from Massen et al. (2015a) by Nadja Kavcik-Graumann, 2009), suggesting that tolerance is an important driving force for cooperation in general Whether or not birds cooperate in experimental set-ups also seems to be affected by how the rewards are distributed. In ravens and Keas, at least, the probability that two birds cooperated successfully (.e. simultaneously pull on string) depended on the reward distribution in the previous successful trial for the same two birds: when the reward distribution was equal, the probability that they would cooperate again was high, whereas when the reward distribution was unequal the under-benefitting bird stopped cooperating (Schwing et al., 2016; Massen et al., 2015a). These results correspond to evidence that capuchin monkeys, Cebus apella and chimpanzees also base their coop- erative decisions on whether they received a reward with the same partner or not in the previous trial (Mendres & de Waal, 2000; Engelmann et al., 2015). These patterns322 ‘Thomas Bugnyar and Jorg J.M.Massen Figure 162 Two keas cooperating in a seesaw set-up. One kea should sit on the perch atthe right cend for the seesaw to open a box containing rewards on the other end. However, when it leaves the perch, the box will automatically close again. The other kea only needs to take the rewards ‘out of the box. Drawing adapted from Tebbich et al. (1996) by Nadja Kavcik-Graumann. hint towards inequity aversion, a trait shown in ravens and crows (Wascher & Bugnyar, 2013), several primate species (e.g. Brosnan & de Waal, 2003; Brosnan et al., 2005) and dogs (Range et al., 2009). Alternatively, the decisions to cooperate might simply depend ‘on whether an animal obtained a reward in the previous trial although such a scenario seems unlikely as the strong relationship between inter-individual tolerance and coop- ‘ration indicates the importance of partner identity (Brosnan & de Waal, 2014), Nevertheless, most of the birds tested in the loose-string paradigm appeared not to consider or understand the need for a partner in this task, as they did not pass controls in which they had to wait for their partner before they could commence pulling, or inhibit pulling while they were alone (Seed et al., 2008; Schwing et al., 2016; Massen et al., 2015a). This stands in contrast to data from chimpanzees (Melis et al., 2006), elephants, Elephas maximus (Plotnik et al., 2011) and coral trout, Plectropomus leopardus (Vail et al., 2014), all of which seem to know when and with whom to cooperate. It is pos- sible that the social organisation of long-term monogamous birds with a single partner does not require the same kind of understanding of cooperation as found in primates, which form short- and long-term relationships with a varying number of individuals (Seed et al., 2007). Conversely, however, not all primates tested seem to understand the need for a partner in these cooperation experiments (Fady, 1972; Petit et al., 1992; Chalmeau et al., 1997; Visalberghi et al., 2000). Furthermore, an animal's failure in the delay task could also be due to methodological issues like different times required to ‘wait for the partner (Bugnyar, 2008) or a lack of sufficient exposure to the task. Indeed, the data from rooks and African grey parrots were not conclusive (Scheid & Noé, 2010; Péron et al., 2011). Moreover, in a different paradigm (Figure 16.2) with complementary roles, rather than similar roles, keas did appear to understand the need for a partner, and dominant birds aggressively manipulated subordinates to press a lever, which enabled themselves to collect food (Tebbich et al., 1996). Consequently, not only social affil- iation but also social dominance might play an important role in cooperation among birds.16 Social Relations, Social Cognition and Cooperation 323 Reciprocity in Repeated Interactions of Birds The reciprocation of costly favours or actions is considered to be one of the main evo- lutionary explanations for the observed altruism among unrelated individuals (Trivers, 1971). These costly favours can be exchanged for similar favours, or interchanged for other services (Hemelrijk & Ek, 1991). The tit-fortat rule (begin with behaviour that benefits the partner and subsequently copy the partner’s behaviour: Axelrod & Hamil- ton, 1981) is one of the simplest behavioural rules that result in reciprocal altruism. As remembering large numbers of interactions with different individuals may be challeng- ing (Stevens & Hauser, 2004), decisions to cooperate or to defect that depend on active scorekeeping of the value and amount of what has been given and received by whom (cf. calculated reciprocity, de Waal and Luttrell, 1988) seem unlikely to explain observed reciprocal relationships, maybe even in humans. Consequently, a distinction should be made between complex cognitive mechanisms (like memory, individual recognition, fature planning, qualitative and quantitative calculations) that result in reciprocal altru- ism, and mechanisms that are mediated by, for example, generalised arousal states (cf. generalized reciprocity, Hamilton & Taborsky, 2005) and emotions towards specific partners (Brosnan & de Waal, 2002; Schino et al., 2007). Moreover, studies on reci- procity emphasise the importance of the time frame of reciprocation (Schino et al., 2007). Most studies thus far report reciprocal exchange patterns that lack short-term contingency. Consequently, most, if not all, species may not take into account what is given and received from a particular individual on the short-term (in the previous inter- actions), but nonetheless reach and maintain contingent reciprocal relationships on the long-term (over several consecutive interactions). Decisions to cooperate or not may be mediated by emotions that reflect the long-term properties of the relationship with a par- ticular individual. These emotions can serve as an intervening variable that is affected by all (and thus not only the last) previous interactions and subsequently affects future interactions (cf. emotional book-keeping, Schino et al., 2007). ‘An emotionally-mediated mechanism of cooperation based on partner selection seems a feasible strategy to elicit cooperation among group members (Massen et al., 2010), although it works only with individuals with whom there is sufficient emotional experience (Aureli et al., 2008) and sufficient previous interactions to create a distin- guishable emotion towards that individual. Indeed, male chaffinches Fringilla coelebs only start initiating mobbing after being in a multi-species aggregation for longer than a week after migration (Krams & Krama, 2002). If these chaffinches rely on a caleu- lated tit-for-tat strategy, the cooperation in mobbing should be initiated immediately after the first mobbing act of another individual; an emotionally mediated mechanism, in contrast, would require to develop positively or negatively valenced states about par- ticular individuals, and therefore, might explain the slightly longer period. Similarly to chaffinches, great tits are more likely to join in reciprocal predator defence with neigh- bours from the previous year than with new neighbours (Grabowska-Zhang et al., 2012). Furthermore, ex- and interchange-pattems of preening and support in conflicts of ravens are long-term, rather than contingent on short-term events, and are more frequent among individuals that share a good relationship (Fraser & Bugnyar, 2012).324 ‘Thomas Bugnyar and Jorg J.M.Massen Even though most patterns of reciprocity seem long-term, this does not mean that recent events might not change the attitude of individuals towards the individuals with whom they normally reciprocate (ef. attitudinal reciprocity, Brosnan & de Waal, 2002). For example, experimental manipulations simulating defection of neighbours in nest defence do decrease subsequent help for those neighbours in nest defence in both red-winged blackbirds (4gelaius phoeniceus, Olendort et al., 2004) and pied flycatcher (Ficedula hypoleuca, Krams et al., 2008). In their experiment (Figure 16.3), Krams and colleagues (2008) placed a stuffed tawny owl, Strix aluco, next to the nestbox of one of three neighbouring pairs of pied flycatchers while the birds from one of the neighbouring nestboxes were secretly caught, preventing them from assisting in any mobbing against the owl. In contrast, the other neighbouring birds did assist the pair that had the stuffed owl placed next to their nestbox by mobbing the potential predator. It is possible that cooperative mobbing in this species is based on a tit-for-tat strategy as in later repetitions with all birds present, the previous victims assisted the pair that had helped them when the owl was at their nest, but did not help the pair that had ‘defected’ when the owl was at their nest (Krams et al., 2008). Likewise, ‘dear enemy" relationships between hooded warblers, Wilsonia citrina, can be changed based on recent events as males respond more aggressively to neighbours than to strangers after simulated intrusions of these individuals, suggesting some sort of retaliation (Godard, 1993). Moreover, changes in response can even be based on indirect experiences (cf. indirect reciprocity, Nowak & Sigmund, 1998), since male song sparrows, Melospiza ‘melodia, will react aggressively towards neighbours having heard that neighbour apparently intruding (via playback) into a third bird's territory (Akgay et al., 2010). Such results could be explained by scorekeeping through (episodic-like) memory of past events and/or may rest primarily on changes in attitudes and emotions that are coupled with learning (Russel & Wright, 2009; but see Wheatcroft & Krams, 2009). More controlled experiments are therefore needed. To study reciprocal altruism in controlled settings, a common paradigm is the Iterated Prisoner's Dilemma. Obtaining stable cooperation based on a tit-for-tat strategy has, however, proven difficult in two-player experiments where individuals can choose to cooperate or defect (e.g. in touch-screen or key pressing experiments in starlings, ‘Sturnus vulgaris, Reboreda & Kacelnik, 1993; and pigeons, Columba livia: Green et al., 1995; Hall, 2003). This might be because the birds (and other animals) discount the future and, therefore, cannot weigh the long-term benefits of cooperating in relation to the short-term smaller benefits of defecting. When the time frame of reward delivery is controlled, blue jays, Cyanocitta cristata, can reach stable cooperation with each other in an Iterated Prisoner’s Dilemma game (Stephens et al., 2002). Moreover, such recip- rocal cooperation seems to be enhanced by long-term social bonds (e.g. zebra finches, Taeniopygia guttata: St-Pierre et al., 2009). A similar two-player two-choice set-up was used to investigate cooperation in ravens and African grey parrots, albeit with a human experimenter as partner rather than a conspecific. The ravens failed to reach a stable tit-for-tat strategy, whereas at least one African grey parrot did succeed (Di Lascio et al., 2013; Péron et al., 2014), differences that might be explained by varying social bonds between the humans and the birds: a relatively anonymous experimenter in the case of the ravens versus well-known animal trainers in the case of the parrot.16 Social Relations, Social Cognition and Cooperation 325 Figure 16.3 Experimental set-up showing reciprocity in pied-flycatchers. Inthe initial phase (a), 1 stuffed owl is placed next to the focal pair's nest-box. Whereas the focal pair is helped with ‘mobbing the potential predator by one neighbouring couple, the other neighbouring couple cannot as they were secretly caught. When in the test phase (b), two stuffed owls are placed next to the nest-box of both neighbouring couples, the focal pair chooses to help mobbing that pair that helped them before and not the pair that did not help them. Drawing adapted from Wheatcroft & Price (2008) by Nadja Kaveik-Graumann. Other Regard: ‘Altruistic’ Motives Birds commonly show other-regarding behaviour in natural situations by sharing food with others. Specifically of interest here is when this food sharing is also initiated by the individual that holds the resource (i.e. active food sharing), rather than when two individuals are simply feeding together on one food resource (Stevens & Gilby, 2004),326 ‘Thomas Bugnyar and Jorg J.M. Massen or when the second individual is allowed to steal a piece (e.g. tolerated theft: Blurton- Jones, 1984; or harassment avoidance: Stevens & Gilby, 2004). In mammals, such active sharing is mainly shown between parents and offspring but rarely among adults (e.g. Jaeggi & van Schaik, 2011; Range et al., 2015). In birds, active sharing also happens ‘most often in the context of reproduction, either in courtship or when parents either feed their offspring or feed the incubating partner. However, this active sharing can also be observed outside reproductive contexts and with unrelated others, which may benefit group cohesion (e.g. Reyer 1986; Thiollay, 1991), status signalling (e.g. Carlisle & Zahavi, 1986; Scheid ct al., 2008; Duque & Stevens, 2016) or social bonding (e.g. Bugnyar & Kotrschal, 2002; von Bayern et al., 2007). Depending on the function, active sharing may be based on different motives (Duque & Stevens, 2016). ‘Another form of other regard is the consolation of distressed others. This bystander affiliation offered to victims of inter-individual aggression has been observed in rooks (Seed et al., 2007), common ravens (Fraser & Bugnyar, 2010b) and jackdaws (Logan et al., 2013). In rooks, this was only observed between mated pairs, whereas among ravens this can extend to other valuable relationships like kin and friends (Seed et al., 2007; Fraser & Bugnyar, 2010b). Although originally called consolation (de Waal & Roosmalen, 1979), the motivation behind this behaviour remains unclear (Fraser et al., 2009). It may reflect empathy for the victim but could also be explained by purely selfish motives like protection from redirected aggression. From a cognitive point, other-regarding behaviours are of interest because of the assumed link to attribution skills like empathy and sympathy as well as trust (Decety etal., 2016). However, as mentioned above, selfish motivations are powerful alternatives (e.g, sharing food to avoid harassment or third-party affiliation for self-protection) and need to be ruled out experimentally. The latter has been proven extremely difficult for mammals (ef. Ben-Ami Bartal et al., 2011; and Silberberg et al., 2014; and e.g. Liebal et al., 2014), and has not been investigated in birds yet. Experimental studies on other-regard have thus far been focused largely on primates and their comparison to humans in explanations of the evolution of human altruism. The most used paradigm is the dyadic pro-social choice paradigm in which an individual has the choice between pulling a tray that delivers food to a partner vs. a tray that has no food for the partner. While the primate data are inconclusive thus far (for a review see Cronin, 2012), they support the Cooperative Breeding Hypothesis, which assumes that pro-sociality has evolved as a consequence of the specific reproductive socio-ecology of humans (Hrdy, 2009), and can thus be only found in species with a similar repro- ductive system like the cooperatively breeding marmosets (Burkart et al., 2007, 2014), So far, however, the Cooperative Breeding Hypothesis has only been tested in primates (Thornton & McAuliffe, 2015). Given the comparable levels of social cognition of pri- mates and corvids (Emery & Clayton, 2004; Giintiitktin & Bugnyar, 2016) and possibly other birds, and the prevalence of cooperative breeding in birds (Riehl, 2013), compar- ative studies on pro-sociality in birds would be particularly informative when asking questions about the evolution of pro-sociality. There are only a handful of studies that have examined other regard or pro-sociality ‘experimentally in birds using different set-ups, and so far also here the results are inconclusive. Experimental studies on common ravens failed to show overall pro-social16 Social Relations, Social Cognition and Cooperation 327 behaviour in three different paradigms (Di Lascio et al., 2013; Massen et al., 2015b; Lambert et al., 2017). Jackdaws, in contrast do seem to choose pro-socially, albeit dependent on enhancement cues of the recipient, which reflected their interest in the reward (Schwab et al., 2012). Furthermore, there seems to be some evidence for sharing in grey parrots (Péron et al., 2013), albeit based on a small sample size. Finally, a recent study does seem to support the Cooperative Breeding Hypothesis in birds, since it showed high levels of proactive prosociality in the cooperative breeding Azure winged magpie, Cyanopica eyanus (Horn ct al., 2016). Nevertheless, more species with different social systems should be tested in the same paradigm before we can draw strong conclusions. Conclusion and Outlook Social relationships both in humans and non-human animals reflect an abstract con- cept made by us humans on the basis of observed repeated interactions between given individuals. Whether or not animals that form social relationships also represent those relationships as mental constructs is a hot topic of research. It is only recently that birds have been included in this debate and that the cognitive basis of their social life has become of interest. Social bonds of birds seem to be qualitatively similar to those of mammals, notably in long-term monogamous species. Yet, their affiliative social networks tend to be small, with the pair bond being the main unit. Importantly, the social organisation of many species shows strong temporal dynamics with respect to flock formation and composi- tion, whereby dominance relationships are often influenced by social bonds. We ate just beginning to understand how non-breeder groups are structured and how the relation- ships between their members compare to those in stable family groups. Tentatively, the reliance on bonds coupled with the high dynamics in group formation pose cognitive challenges resulting in abilities like third-part understanding and inferential reasoning, comparable to those in socially complex mammalian systems (e.g. primates, Tomasello & Call, 1997; elephants, McComb et al., 2001; hyenas, Holekamp et al., 2007; dolphins, Mann et al., 2000). Social bonding can create significant benefits when the individuals involved cooper- ate. Cooperation among birds reflects different levels of complexity, from anonymously acting-apart-together, to synchronisation, to turn-taking, to highly coordinated coop- eration with complementary roles. These levels may correspond to differences in the underlying cognitive mechanisms, ranging from the focus on a physical problem to paying attention to others’ actions, to attributing roles and understanding others’ intentions. Whatever level of complexity among non-anonymously cooperating birds, social bonds seem paramount in the choice of cooperative partner and in the eff ciency of cooperation. Consequently, birds tend to cooperate most with pair-partners, family members, neighbours and/or otherwise affiliated conspecifics. Experimental examination of cooperation among birds is still in its infancy, but the results thus far suggest that social bonding and social tolerance are important facilitators of ‘cooperation.328 Thomas Bugnyar and Jorg JM. Massen From an ultimate perspective, reciprocity may be a driving mechanism to assure shared benefits in cooperation. Most reciprocal patterns are, however, not contingent ‘ona short-term basis, but rather on a long-term basis, suggesting some sort of emo- tional mediation on the proximate level. Some emotional experience with a partner ie., some sort of social bond, is therefore necessary and corroborated by observations in the wild. Nonetheless, birds may also be sensitive for short-term changes, and experi- mental manipulations simulating defection of one partner have resulted in subsequent retaliation-like behaviour of the other partner. These results indicate that calculations based on a mental score-keeping of others” debits and credits are possible. In turn, sta- ble tit-for-tat strategies in artificial experiments, albeit difficult to achieve, are positively influenced by long term affiliative relationships. Hence, the mental and emotional pro- cesses underlying reciprocity seem to be tightly interlinked ‘What motivates cooperative decisions of birds in the first place remains a relatively understudied topic. Many birds share food and not only in the context of reproduc- tion, suggesting some sort of general pro-social tendency. So far, however, very few experiments have targeted this question and results are mixed, preventing any informed conclusions. Despite the huge interest in bird behaviour, open questions remain about their social lives and the cognitive challenges involved (Silk et al., 2014). Fortunately, recent tech- nological advances have created better opportunities to track birds over small and/or larger distances (Wikelski et al., 2007) and allow gathering detailed machine-sensed data of their social dynamics (Krause et al., 2013). We are therefore looking forward to the data generated from these increasingly apt tools, as they will provide a much clearer picture of the social organisation of bird populations. Over the last decades research in many behavioural fields (this book), but also neuro- logical work (Giintirkiin & Bugnyar, 2016), have shown that birds might be interesting model species to study the evolution of intelligence. With the social brain hypothesis (Byrne & Whiten, 1988; Dunbar, 1998) in mind, we therefore anticipate much more experimental work examining the social knowledge of birds, their ability to cooperate and the motivations behind this cooperation. We would like to encourage experiments that vary the inter-individual characteristics (e.g. sex, rank, familiarity and relationship quality) of the birds involved. Furthermore, we advocate a comparative approach, not only between birds (as a group in general) and other classes of animals, but also between different taxonomic groups within the birds. Comparing the socio-cognitive abilities of species with different socio-ecological backgrounds will allow us to test hypotheses about the evolution of (avian) social intelligence, References Aksay, C., Reed, V. A., Campbell, E., Templeton, C. N. and Beecher, M. D. (2010). Indirect reciprocity: song sparrows distrust aggressive neighbours based on eavesdropping. Animal Behaviour, 80, 1041-1047.