6.  Coalition formation with the knowledge of paternal kinship

A coalition¹ is formed when one individual intervenes in an ongoing conflict between two opponents in order to support one of them. Since support in favour of one party is simultaneously targeting the other party, coalitions are triadic interactions involving a supporter, a recipient and a target. The adaptive function of altruistic behaviour such as sterile insect benefiting the family was a headache for Darwin (1859) when he realised that altruistic behaviour contradicted his theory of natural selection since an individual should only help another when it is likely to benefit from the support it provides. In other words, selection is not expected to favour individuals who reduce their own fitness to increase the fitness of others. Ever since there are two possible explanations for the function of support: (i) intervention is of altruistic or (ii) even of selfish origin. Altruistic behaviour is associated with costs to the altruist (i.e., time, energy, risk of injury and retaliation) and benefits to the recipient (i.e., access to a limited resource, suffering less injury in an attack). If coalition formation is altruistic, it is expected to be directed towards kin or reciprocal partners.

The theory of kin selection (ref. Maynard Smith 1964) predicts that individuals who support kin enhance their indirect fitness, because they share genes with the recipient of support (Hamilton 1964). Selection is expected to favour altruism among kin the closer individuals are related, the lower the costs to the altruist, and the greater the benefits to the recipient.

In reciprocal altruism, however, the altruistic act incurs costs and no immediate benefits, but the altruist receives future benefits from the recipient (Trivers 1971, Axelrod & Hamilton 1981). Since the benefits to the recipient ars expected to be greater than the costs to the altruist, both partners enhance their direct fitness over a longer period of time even when unrelated. Following Hemelrijk (1990a) repayment by the former recipient can be given in the same currency (reciprocity: e.g., support for support) or in different currencies (interchange: e.g., grooming for support). Both terms will be distinguished throughout the chapter.

Supporters are not altruists when they receive immediate benefits from their support of unrelated individuals (Wrangham 1982). Male chimpanzees act selfishly when supporting non-kin against certain opponents since their support can act to increase their own dominance rank (de Waal 1982). Thus co-operative (or mutualistic) behaviour may evolve among individuals with self-interests if they can reach their goals more effectively when co-operating, than acting alone (Brown 1983). Compared to a non-cooperative situation both recipient and supporter immediately benefit, although the benefit can be asymmetrically distributed among the partners (Axelrod 1984). For more details on the three theories see Widdig et al. (2000).

There is evidence for kin selection with respect to support among both male and female primates (e.g., Kurland 1977, Kaplan 1977, 1978, Massey 1977, Walters 1980, Silk 1982, Bernstein & Ehardt 1985a, Widdig et al. 2000, Silk et al. 2002), especially females support their kin at a high risk when they intervene against opponents which are higher-ranking than themselves (Datta 1983b,c, Bernstein & Ehardt 1985a, de Waal & Luttrell 1985). Recall that earlier studies only distinguished between maternal kin and maternal unrelated individuals, pooling paternal kin and non-kin. Therefore, nothing is known about the impact of paternal kinship on coalition formation.

Reciprocal altruism was first reported among pairs of maternally unrelated male baboons who alternate in enlisting each other to take over an oestrous female from a consort male (Packer 1977). Later studies questioned the extent of reciprocity involved in these coalitions (Bercovitch 1988, Noë 1989) since both partners were as likely to gain access to the consort female. Bercovitch (1988) suggested that coalitions are the outcome of co-operation between males pursuing self-interests.

De Waal (1982) showed that male chimpanzees intervene in conflicts in favour of unrelated males to increase their own dominance rank against the target of support. Additional studies have provided evidence of non-kin support in terms of co-operative interactions for both male (Noë 1992, Widdig et al. 2000) and female primates (Chapais et al. 1991, Silk et al. 2002).

While juvenile individuals benefit from support by their maternal kin in order to acquire the dominance rank predicted from birth order (reviewed in Chapais 1992), interventions among adults are more likely to stabilise the existing dominance hierarchy (ibid.), even though adults sometimes use coalitions to increase their dominance rank (e.g., Samuels et al. 1987). In addition, Chapais (1988b) showed using an experimental study on rank reversal, that any female, regardless of age, depends on allies to maintain her rank. Coalition formation is therefore important for the acquisition and maintenance of the rank structure in a group.

Recall, that female macaques stay in their natal group with their kin throughout their life, while males migrate and thus have reduced opportunities to support kin (Kaplan 1977, 1978, Bercovitch 1988, Silk 1992a). Studies on coalition have suggested a sex difference as interest of males and females in interventions seems to be different (e.g., Bond & Vinacke 1961, Kaplan 1977, de Waal 1984, Bernstein & Ehardt 1985a). Since rank relations among male primates are more unstable than in females, the interest of male partners change over time (de Waal 1985). Therefore, coalition formation among males is more changeable and dominance-orientated than among females, who tend to support the lower-ranking but familiar opponent more often (see also Hemelrijk & Ek 1991).

Having paternity data available when investigating decision making in coalition formation, interesting questions will arise. Given that an individual is going to intervene in a conflict between its maternal cousin and its paternal sibling, who should it support? Kin selection theory (Hamilton 1964) would predict that the intervener supports the opponent with whom its shares more genes assuming all others being equal. In other words, one would expect that the individual gives support to its paternal sibling while targeting its maternal cousin. However, other factors might influence this decision. As shown in the previous chapter, affiliation (as a measure of familiarity) is less among paternal half-siblings than among maternal cousins. In addition, maternal kinship reflects rank relations which are extremely important among female rhesus macaques. In removal experiments it was shown that female Japanese macaques clearly depend upon the support by their high-ranking maternal kin or other maternally unrelated individuals of high rank in order to stay high-ranking themselves (Chapais et al. 1991). Therefore, it is likely that constraints to an individuals’ own competitive ability also play an important role.

The present study has shown that adult female rhesus macaques are capable of discriminating their paternal kin from non-kin, but whether or not they treat their paternal kin preferentially is context-dependent (chapter 4). Here, we will focus on the question of whether paternal kin discrimination is expressed in coalition formations, a behaviour which can be of enormous cost to the intervener. As outlined above, coalition formation might result from kin selection, reciprocal altruism or selfishness depending upon the cost associated to the supporter. In some conflicts, females may intervene at high costs to themselves to increase their own dominance rank which is likely to increase their own reproductive success (cf. Silk 1987, van Noordwijk & van Schaik 1999). In other conflicts, females may intervene at high costs to themselves in favour of their kin, both maternal and paternal, but if self-interests conflict with kin support females may also intervene against their kin with whom they compete over rank (cf. Chapais [page 89↓] et al. 1994). The costs to intervene seem to differ depending upon the target of the support with costs defined as the risk of retaliation assuming that injuries have an effect on fitness (cf. Chapais 1992, 2001).

This chapter will focus on two issues: (i) it will re-evaluate the function of agonistic intervention by testing the three introduced functional explanations with the knowledge of paternity, i.e., when maternally unrelated individuals can be separated into paternal kin or non-kin and (ii) it will investigate whether paternal kinship has an impact on coalition formation. Recall from the Methods, that all coalitions observed over the 8 months study period, either collected ad libitum or in a focal protocol, were included in the data analyses, whenever a focal female (N=34) or an adult female (N=49) was intervening. Which sample (N=34 or 49) will be noted below. Both, female support either given to a recipient or against a target belonging toa certain kin and age category were analysed. Depending upon the question addressed, the kin and age categories were either maternal half-siblings, paternal half-siblings, non-kin or adult females in general. The third party in a coalition considered could be any other individual of group R including adult males.

Table 6.1 shows the correlation between individual attributes such as age, rank and number of kin and non-kin.

Table 6.6.1: Correlation between individual attributes

Spearman's correlation coefficient (r _s ) for all adult females (N=49) using age in years (Age), basic rank (Rank), and number of kin and non-kin. Abbreviations are as follows: number of maternal half-sisters (N MS), number of paternal half-sisters (N PS) and number of non-kin (N NK), pooling peer and non-peers in each kin category.

First and most important, rank and age did not correlate with each other which was expected given that daughters socially inherit the rank of their mother. Rank was neither related to the number of kin nor number of non-kin. Younger females had significantly more paternal half-[page 90↓]sisters than older females as older females were more likely to lose their peers (which tend to be paternal half-siblings, see chapter 3) with years due to death. In contrast, the older females had significantly more female non-kin than younger females.

Additionally, these individual attributes for all 49 adult females were correlated with variables of relevance for interventions such as the number of dyadic aggression initiated and received, the number of being the target or the recipient of support and the number of support given per individual female (see Table 6.2).

Table 6.6.2: Testing individual attributes with variables related to support

Spearman's correlation coefficient (r _s ) for all adult females (N=49) using age in years (Age), basic rank (Rank), the number of dyadic aggression initiated (Aggressor), the number of dyadic aggression received (Victim), the number of being the target of support (Target), the number of being the recipient of support (Recipient), the number of support given (Support).

Regarding age, the older the females were, the less they initiate and receive dyadic aggression. Older females also showed a trend of receiving less support. Regarding rank, the higher in rank (highest rank is 1, lowest 49) the more aggression was initiated and the less aggression was received by the female. High-ranking females were more engaged in coalition formation, both they gave and received more support than lower-ranking females.

The following table presents the correlation between a behaviour (such as grooming, support given or support against) and a dyadic attribute (such as rank distance, age distance or degree of relatedness) within dyads. Rank distance was calculated in absolute terms ranging from 1 to 48, with the rank difference between the highest- and the lowest-ranking adult female (N=49) being 48. Age distance was also measured in absolute terms (ranging from peers, i.e., 0 to the maximum age difference among adult females, i.e., 16 years). Degree of relatedness was calculated by incorporating all degrees of maternal and paternal relatedness. As an example, the question will be addressed whether an increased frequency in grooming within dyads is correlated with the rank differences of this dyad, etc. The Mantel R-test (Hemelrijk [page 91↓]1990a,b, de Vries 1993) testing reciprocity on the dyadic level was performed using 2000 permutations per test. Recall from the Methods, that this test investigates whether the sequence of preference among all dyads is correlating between the two matrices. Roughly spoken, for the tests on grooming the questions to answer would be (i) is the closest ranking dyad also grooming each other the most, while the most distant ranking dyad is also grooming each other the lowest (i) is the closest ageing dyad also grooming each other the most, while the most distant ageing dyad is also grooming each other the lowest and (iii) is the closest related dyad also grooming each other the most, while the most distant related dyad is also grooming each other the lowest (see Table 6.3).

Table 6.6.3: Testing the effect of dyadic attributes on grooming and support

Correlation between a dyadic attribute (such as rank distance, age distance or degree of relatedness) and a behaviour (such as grooming, support given or support against) based on a 49x49 matrix including all adult females (for details on the tests see above or Methods).

Rank distance showed a significant negative correlation with both grooming and support given, implying that females groom and support each other more, the closer they are in rank. This result was expected as close maternal kin are of adjacent rank. Testing reciprocity between rank distance and support against also revealed significant results, but we should not consider this a relevant association, because of the very low correlation coefficient (cf. large sample sizes also tend to result in significant results without a relevant correlation coefficient). Age distance showed no relevant association with grooming, support given or support against which is not contradicting the importance of age proximity found in chapter 4, instead it supports the results of chapter 5, that the closest bonds were the one between mothers and daughters which are of distant age. As predicted from kin selection theory [page 92↓](Hamilton 1964), the degree of relatedness showed a significant positive correlation with both grooming and support given, implying that females groom and support each other more, the higher their degree of relatedness. No relevant association was found between degree of relatedness and support against.

From these finding one could predict that the degree of relatedness and rank distance are associated with each other, meaning that females with a small rank distance will also share a higher degree of relatedness (cf. de Waal 1991). A comparison between them revealed a significant negative correlation using the Mantel R-test with 2000 permutations (Table 6.4).

Table 6.6.4: Correlation between rank distance and the degree of relatedness

Correlation between rank distance and the degree of relatedness based on a 49x49 matrix including all adult females (for details see Table 6.3).

This result was expected from what we know about the social structure in macaques where daughters socially inherit their dominance rank of their mother and maternal half-siblings are of adjacent rank. As a consequence, it makes sense to control for rank distance when correlating a behaviour with the degree of relatedness. Therefore, the correlation between the degree of relatedness and a behaviour was re-done, while controlling for rank distance at the same time. The partial Kr-test (Hemelrijk 1990a,b, de Vries 1993) was performed using 2000 permutations per test. The partial Kr-test tests reciprocity on the individual level, in contrast to the Mantel R-test which tests reciprocity on the dyadic level (see Methods). So, the question is slightly different than tested above. A perfect correlation would occur if for each female the sequence of grooming preferences corresponds with the degree of relatedness among her partners even though one controls for rank distance (Table 6.5).

Tabl 6.6.5: Testing the effect of the degree of relatedness on grooming and support controlling for rank distance

Correlation between the degree of relatedness and a behaviour (such as grooming, support given or support against) while controlling for rank distance based on a 49x49 matrix including all adult females (for details on the tests see Methods).

When controlled for rank distance, but now tested on the individual level, the degree of relatedness was still associated with grooming and support given, while support against was of no relevant association (see above). In other words, rank distance alone was not explaining the correlation between the degree of relatedness and grooming or support, respectively.

Data were analysed to investigate the distribution of support either given to or against a maternal, paternal half-siblings or non-kin controlling for age proximity as in chapter 4 (see Table 6.6). All coalitions where a focal female (N=34) was intervening either in favour or against a certain kin and age category were considered. As outlined in the Methods, two different procedures were used to compare whether adult females intervened more on behalf of a particular kin and age category. The number of observed interventions was either divided (i) by the number of potential partners available in a particular kin and age category (hereafter: availability) or (ii) by the number of opportunities to intervene on behalf of particular kin and age categories (hereafter: opportunities) (cf. Silk et al. 2002). The number of available partners was based on the number of individuals of a particular kin and age categories present during the study period. The number of opportunities to intervene on behalf of a potential recipient were derived from the number of non-silent dyadic conflicts (see Methods) in which this potential recipient was involved.

Table 6.6.6: Testing kin selection theory with respect to support

Cell values are mean proportions (± SD) of the focal females who intervened either in favour of a recipient or against a target. Data analysis was restricted to interventions where the focal females (N=34) choose a maternal half-sisters, non-peer (MS-NP), a paternal half-sisters either a peer (PS-P) or a non-peer (PS-NP) or a non-kin either a peer (NK-P) or a non-peer (NK-NP) either as a target or a recipient. N is the number of individuals tested (N). Proportions were derived from two different procedures (i) observed support of a kin and age category with respect to number of available partner in this kin and age category (Support given _ava or Support against _ava) and (ii) observed support of a kin and age category with respect to number of opportunities (Support given _opp or Support against _opp, see above).

As evident from Table 6.6, focal females intervened most often in conflicts in favour of a maternal half-sibling (being the recipient) but also against a maternal half-sibling (being the target). No clear pattern was visible for all other kin groups. All data will be tested below.

Testing age proximity controlling for kinship

In order to investigate whether patterns of coalition include a peer effect, unrelated peers were compared with unrelated non-peers (Table 6.7).

Table 6.6.7: Testing age proximity controlling for kinship

Data from Table 6.6 were tested using the Wilcoxon-test because some data did not meet the assumptions of parametric tests. In order to keep results comparable all tests on the same data set (Table 6.7-6.9) were tested with non-parametric tests (see Methods). P’ represents the Dunn-Šidák correction for multiple testing including all five different tests for the same behaviour in Table 6.7-6.9 (for more details see Methods).

Unrelated females preferred to support their peers more often than their non-peers, but they did not differ in targeting peers or non-peers. Targeting peers as often as non-peers, females either do not discriminate between both age categories in this specific context or data may also reflect competition among peers.

Testing maternal vs. paternal half-siblings and non-kin controlling for age proximity

Maternal half-siblings were compared with paternal half-siblings and non-kin, all being non-peers (Table 6.8).

Table 6.6.8: Testing maternal vs. paternal half-siblings and non-kin controlling for age proximity

Data from Table 6.6 were tested using the Wilcoxon-test. See Table 6.7 for more details.

Maternal half-siblings supported each other more often than either paternal half-siblings or non-kin, but maternal half-siblings were also targeting each other equally often than either paternal half-siblings or non-kin. In other words, females preferred to help their maternal half-siblings, but they also directed support against their maternal half-siblings.

Testing paternal half-siblings vs. non-kin controlling for age proximity

Table 6.9 shows the comparison between paternal half-siblings and non-kin being either peers or non-peers.

Table 6.6.9: Testing paternal half-siblings vs. non-kin controlling for age proximity

Data from Table 6.6 were tested using the Wilcoxon-test. P_e refers to the exact P-value as this test has a sample size of only N=15 (see Mundry & Fischer 1998 or Methods). See Table 6.7 for more details.

Testing peers, females did not support their paternal half-siblings more often than unrelated females and they also did not differ in targeting paternal half-siblings and non-kin. In other words, no evidence of paternal kin discrimination emerged among peers with respect to coalition formation. Testing non-peers, there seems an indirect indication for paternal kin discrimination as paternal half-siblings tended to targeting each other less often than non-kin.

Reciprocity can arise among both kin and non-kin (Trivers 1971). However, reciprocity among kin is expected to be evolved through kin selection, not reciprocal altruism, and reciprocity among non-kin might be evolved through reciprocal altruism if associated with a cost to the supporter. The tests below will use the matrix correlation method (Hemelrijk 1990a,b, de Vries 1993) as outlined in the Methods.

The following two question will be analysed: (i) is there reciprocity for the same behaviour on the individual level, i.e., a perfect correlation would be if female A grooms female B most often, female C second most often and female D third most often and if female A then also receives grooming in this sequence (most often from female B, second most often from female C, third most often from female C) and (ii) is there interchange for different behaviour on the individual level, i.e., a perfect correlation would be if female A grooms female B most often, female C second most often and female D third most often and if female A then also receives support in this sequence (most often from female B, second most often from female C, third most often from female C). The goal of this section is to analyse the extent of reciprocity or interchange on two levels: (i) on all adult females (N=49) and (ii) on all adult females when controlling for maternal half-siblings, paternal half-siblings or non-kin (N=49).

Reciprocity for the same behaviour among all adult females

Reciprocity for the same behaviour (grooming for grooming, support given for support given and support against for support against) was analysed using all ad libitum data involving an adult female (N=49) as supporter by correlating the initiator-receiver matrix with its transposed form (Table 6.10). In other words, the matrix was diagonally split and correlated with the mirror image taken around the diagonal so called transposed matrix. The K_r-test was performed to investigates reciprocity on the individual level (see Methods) and the number of permutations per test was 2000 (Hemelrijk 1990a,b). The question we ask here is therefore: assume that female A gave most of her grooming to female B and second most to female C, a [page 97↓]perfect correlation would means, that female A would also receive most of her grooming from female B, and second most from female C (more details in Methods).

Tabelle 6.6.10: Reciprocity for the same behaviour among all adult females

Tests for reciprocity are based on a 49x49 matrix including all adult females (for details on the tests see Methods).

As evident from this table, grooming and support given were based on reciprocity, implying that the sequence of preferred grooming partners of female A corresponds with the sequence of grooming received by female A. However, no such correlation was found for support against, but this result was expected as a low-ranking female will hardly ever intervene against high-ranking females, even though high-ranking females often intervene against them.

Reciprocity for the same behaviour among half-siblings and non-kin

As maternal and paternal half-siblings share on average the same degree of relatedness assuming costs and benefits being equal we would expect the same extent of reciprocity among them. Reciprocity for the same behaviour was separately analysed for maternal half-siblings, paternal half-siblings or non-kin by correlating the 1^st with the transposed of the 2^nd matrix. The number of permutations per test was 2000 and the partial K_r-test (Hemelrijk 1990a,b) was used to test reciprocity on the individual level while controlling for one of the three kinship categories (Table 6.11).

[page 98↓]

Table 6.6.11: Reciprocity for the same behaviour among half-siblings and non-kin

Tests for reciprocity among maternal half-siblings (MS), paternal half-siblings (PS) and non-kin (NK) based on a 49x49 matrix including only adult females who have at least one maternal half-sister, paternal half-sister or non-kin (for details on the tests see Methods).

Repeating the analyses from Table 6.10, but controlling for the three kin categories, the results reveal the same trend. Independent of whether we test kin or non-kin, grooming and support given was based on reciprocity. Among non-kin, the effect of reciprocity in grooming was strong, but reciprocity in support given should only be interpreted as a trend, as the correlation coefficient (+0.23) is quite small (see text below Table 6.3). As maternal half-siblings are always of adjacent rank, but paternal half-siblings and non-kin can vary in their rank distance, the fact, that reciprocity was found in all three groups may imply that reciprocity is not due to differences in rank.

Interchange for different behaviours among all adult females

Interchange for the different behaviours (grooming for support given, grooming for support against, support given for support against) was analysed for all adult females by correlating the 1^st with the transposed of the 2^nd matrix (Table 6.12). The number of permutations per test was 2000 and the K_r-test (Hemelrijk 1990a,b) was used to test reciprocity on the individual level described above.

[page 99↓]

Table 6.6.12: Interchange for different behaviours among all adult females

Correlation between behaviours based on a 49x49 matrix including all adult females (for details on the tests see Methods).

Although, all tests reveal significant results, we should only consider the first one as a relevant association (see text below Table 6.3). However, testing grooming for support, the results showed that while female A groomed female B more than female C, female A also received more support from female B than female C.

Interchange for different behaviour among half-siblings and non-kin

Interchange for different behaviours was analysed separately for maternal half-siblings, paternal half-siblings or non-kin by correlating the 1^st with the transposed of the 2^nd matrix (Table 6.13). The number of permutations per test was 2000 and the partial K_r-test (Hemelrijk 1990a,b) was used to test reciprocity on the individual level.

Table 6.6.13: Interchange for different behaviour among half-siblings and non-kin

Tests for interchange among maternal half-siblings (MS), paternal half-siblings (PS) and non-kin (NK) based on a 49x49 matrix including only adult females who have at least one maternal half-sister, paternal half-sister or non-kin (for details on the tests see Methods).

The trend stays the same when kin categories were tested separately for interchange. Only grooming was clearly associated with support given among maternal and paternal half-[page 100↓]siblings. This trend was less pronounced among non-kin given the smaller correlation coefficient (cf. again Text below Table 6.3). However, independent of whether we test kin or non-kin, there seemed to be (at least) a trend of interchanging grooming for support which is unlikely to be influenced by rank distance given the rank difference among maternal half-siblings, paternal half-siblings and non-kin as explained above.

Intervening is one of the best examples that a behaviour can be costly to the actor, as the supporter invests at least time and energy, but also risks injuries and retaliation, especially by intervening against higher-ranking targets. Since support of one party is always directed against the other party (de Waal 1992), a coalition is a triadic interaction involving a target, a recipient and a supporter. Here an attempt was made to simultaneously analyse all three participants with respect to the cost of interventions, measured as the risk of retaliation when the supporter intervenes either against a higher-ranking or lower-ranking individual (cf. Widdig et al. 2000).

If female A is the highest-ranking, female B the middle-ranking and female C the lowest-ranking individual within a triad, one of six possible rank patterns can be observed in each triad. In the notation of a rank pattern (e.g., A vs. B + C, hereafter: ABC), the first letter is assigned to the target, the second to the recipient and the third to the supporter (Box 6.1).

Box 6.1: The six rank pattern and their risk of retaliation

Six rank patterns with A as the highest-ranking, B as the middle-ranking and C as the lowest-ranking male within a triad (A>B>C, reading: target vs. recipient + supporter). The rank patterns are distinguished in terms of potential risk of retaliation. High risk of retaliation is assumed by intervening against a higher-ranking target. Low risk (if any) of retaliation is assumed by intervening against a lower-ranking target.

These six rank patterns are assumed to be associated with a different risk to the intervener (Bernstein & Ehardt 1985a). In Box 6.1 the rank patterns on the left side are likely to include a high risk of retaliation (Ehardt & Bernstein 1992) as the supporter intervenes against a higher-ranking target. These rank patterns are interpreted to destabilise the existing hierarchy. [page 101↓]Rank patterns on the right side are likely to include a low (if any) risk of retaliation as the supporter intervenes against a higher-ranking target. In contrast, these rank patterns are thought to stabilise the existing hierarchy either the supporters’ or the recipients’ rank (cf. Chapais 1992). The following analysis aims to investigate whether all six rank patterns are equally likely to occur. In other words, if the risk of retaliation by a higher-ranking target is minimum (or close to zero) all rank patterns are expected with equal frequencies.

If the risk of retaliation by a higher-ranking target is minimum or close to zero, no risk should be involved in intervening and all rank patterns may be predicted with equal frequency. However, if some rank patterns involve a higher risk of retaliation than other rank patterns, kin selection theory would predict that individuals should risk more when helping their kin than non-kin (Hamilton 1964).

All coalitions in which an adult female (N=49) was intervening in favour of a maternal half-sibling, a paternal half-sibling or a non-kin were presented (see Fig. 6.1-6.3). The third party in this coalition could be any other individual of group R including adult males.

	Fig. 6.6.1: Support given to maternal half-siblings. The distribution of the six rank patterns of support given by adult females (N=49) to their maternal half-siblings (MS). The white bars are the observed distribution, the blue bars are the expected distribution.

The risky rank patterns are the first three starting from the left side, because the intervener is lower-ranking than the target. In contrast, the three rank patterns of the right are of low cost to the intervener (see above). Support given to maternal half-siblings was observed more often than expected when the target was the highest-ranking individual within the triad (rank pattern ABC and ACB). Both rank patterns are risky, they are mainly used against adult males.

	Fig. 6.6.2: Support given to paternal half-siblings. The distribution of the six rank patterns of support given by adult females (N=49) to their paternal half-siblings (PS). Details in Fig. 6.1.

Similar to maternal half-siblings, but with a very limited amount of data, the risky rank pattern ABC also tended to be observed more often than expected in support given to paternal half-siblings. In addition, two rank patterns involving no risk (CAB and CBA) tended to be observed more often than expected. Note, that the total number of dyads for paternal half-siblings is much smaller, therefore the likelihood of being involved in a conflict is smaller than the number of opportunities to receive support by a paternal half-siblings.

	Fig. 6.6.3: Support given to non-kin. The distribution of the six rank patterns of support given by adult females (N=49) to non-kin (NK). Details in Fig. 6.1.

The rank patterns found for support given to non-kin were similar as found among paternal half-siblings, that is, two rank patterns without risk (CAB and CBA) were more often observed than expected. One exception should be noted here in comparison to paternal half-siblings. Support given to non-kin was not more at risk than expected.

In sum, support given to non-kin was less associated with a risk of retaliation, but support given to maternal half-siblings is more often associated with a risk of retaliation. However, [page 103↓]support given to paternal half-siblings seems to include a combination of rank patterns found among maternal half-siblings (risky pattern) and non-kin (no risky pattern). The results of this analysis suggest that supporters pursue self-interests when intervening in conflicts, as competition over rank with the recipient of support seems to be present both in kin and non-kin support.

The data indicate that kin support, reciprocal support and co-operative support were involved in coalition formation among adult female rhesus macaques, although it appears that the extent differs depending upon the costs involved in intervening. All three theories will be discussed in more detail now.

Kin selection

Maternal half-siblings supported each other more often than either paternal half-siblings or non-kin, but maternal half-siblings were targeting each other equally often than paternal half-siblings and non-kin. In other words, females preferred to help their maternal half-siblings, but they also directed support against their maternal half-siblings which may be a result of competition amongst them. Chapais (1995) pointed out that kin can be both associate and competitor which may lead to conflicts of interests (cf. Chapais et al. 1994). Competition may also explain, why unrelated peers preferred to support each other more than unrelated non-peers, while both groups did not differ in targeting each other. Recall that the present study found evidence for paternal kin discrimination in affiliation, but not aggression (chapter 4). With respect to coalition formation, paternal half-siblings did not support each other more often than non-kin neither among peer nor among non-peers. However, an indirect indication of paternal kin discrimination might be that testing non-peers, paternal half-siblings were targeting each other less often than non-kin (see below).

It should be noted, that the number of paternal half-siblings available per adult female was much smaller than the number of non-kin per adult female, resulting in fewer conflicts involving paternal half-siblings and therefore less opportunities given to support them. Infants born in close age proximity have a large number of paternal half-siblings as shown in chapter 3, but the number of paternal half-siblings will be reduced as older individuals will die or be trapped. Therefore, the number of paternal half-siblings available as adults is much smaller. Coalition formation is a behaviour not as frequently to observe as affiliation or dyadic aggression and it also requires a preceding dyadic aggression in order to have an opportunity [page 104↓]to intervene. This, in addition to the small number of paternal half-siblings available, might have resulted in the small number of observed support in favour of a paternal half-sibling.

Evidence that maternal kin support each other more than non-kin is available for both male and female primates (e.g., Kurland 1977, Kaplan 1977, 1978, Massey 1977, Walters 1980, Silk 1982, Bernstein & Ehardt 1985a, reviewed by Chapais 1992). In addition it was shown, that close kin support each other more than distant kin (e.g., Massey 1977, Kaplan 1978, Petit & Thierry 1994, Widdig et al. 2000, Silk et al. 2002). What not has been studied in much details is, that even maternal half-siblings do both: they give support towards each other, but, on the other hand, they also direct support against each other which might be a by-product of sharing continuously close spatial proximity (cf. Silk et al. 2002). Another explanation for this finding is suggested for Japanese macaques, where females act nepotistically more often towards those maternal kin with whom they do not compete for dominance (Chapais et al. 1994). Nepotism seems to prevail when it does not conflict with the supporter’s self-interest. Therefore, kin bias does not necessarily imply kin selection (i.e., altruistic act), when it cannot be excluded that the supporter benefits from kin support (see co-operation below).

To date, no data are available investigating the impact of paternal kinship on coalition formation. Silk et al. (2002) used peerage as a proxy to test paternal kin discrimination among female baboons, but they did not found evidence for a peer-effect with respect to support. In order to asses whether females discriminate their paternal kin from non-kin in the context of coalition formation, the present study tested support given to paternal half-siblings in comparison to both maternal half-siblings and non-kin. Consistent with measurements of affiliation and aggression there appears a bias of supporting maternal half-siblings more often than paternal half-siblings which again needs to be discussed as possible by-product of the attraction to similar ranking females (Waal 1991, Kapsalis & Berman 1996a,b, Chapais et al. 1991, 1994). Paternal kin discrimination found for affiliation could not be explained by rank difference as paternal half-siblings did not differ in mean rank from non-kin both among peers and non-peers (chapter 4). In contrast, the mean rank difference among maternal half-siblings should be much smaller than among both paternal half-siblings and non-kin (unpubl. data), as maternal half-siblings are of adjacent rank, while the two others can vary from being close in rank upon to the maximum rank difference possible (i.e., one individual can be high-, the other can be low-ranking). As a general rule, there seems a higher risk involved by supporting a paternal half-sibling or a non-kin than supporting a maternal half-sibling assuming that rank distance within a coalition is relevant. It would therefore make sense that no paternal kin [page 105↓]discrimination emerged in support given, but indirect evidence for paternal kin discrimination emerged from avoiding to intervene against paternal half-siblings more often against non-kin.

It should also be pointed out, that the effect of paternal half-siblings targeting each other less often than non-kin was restricted when non-peers (including multiple ages) were compared. In contrast, paternal kin discrimination was stronger among peers (same age) investigating affiliation and no evidence for paternal kin discrimination emerged for both peers and non-peers investigating aggression (chapter 4).

To sum, in contrast to other behaviours measured, coalition formation seems to give no clear evidence in favour of paternal kin discrimination as paternal half-siblings do not give more support towards each other than non-kin. This suggests that earlier studies which lacked paternity data did probably not bias their data by combining paternal and non-kin with respect to coalition formation. On the other hand, paternal half-siblings were targeting each other less often than non-kin indicating that avoidance to intervene against a paternal half-sibling may also reflect the ability to discriminate them from non-kin. Absence of an effect does not mean it does not exist, and as we saw before, paternal kin discrimination seems to be context-dependent. More detailed data would be required than just the 8-month study period to clarify indeed the impact of paternal kinship on coalition formation.

Reciprocal altruism

If reciprocity is the basis for coalition formation, females should support those from whom they received support in the past and refuse to support those who have failed to return support (Trivers 1971). The data of the present study revealed that grooming and support given was based on reciprocity, but reciprocity was not found for support against. The latter result was expected as low-ranking females will hardly ever intervene against high-ranking females, even though high-ranking females often target low-ranking females which is probably due to the risk of retaliation by higher-ranking females. Interestingly, the same trend was found when testing maternal half-sibling, paternal half-sibling or non-kin separately, implying that reciprocity cannot be restricted to kin, even though the trend was less pronounced for non-kin. As maternal half-siblings are always of adjacent rank, but paternal half-siblings and non-kin can vary in their rank distance, the fact, that reciprocity was found in all three groups suggests that reciprocity is not due to differences in rank. Interchange was found in grooming for support given and this trend was still evident when maternal half-sibling, paternal half-sibling and non-kin were tested separately. Again, interchange is suggested for both kin and non-kin, but the trend was less pronounced among non-kin. In addition, rank distance is unlikely to influence the findings (see above).

When dyadic attributes were tested with the behaviours measured, females groomed and supported each other more, the less their absolute rank difference and the higher their degree of relatedness was. From these finding one could expect that the degree of relatedness and rank distance are associated with each other, meaning that females with a small rank distance will also share a higher degree of relatedness. However, even when controlled for rank distance, females still groom and support each other more, the higher their degree of relatedness.

Direct evidence for reciprocal altruism is difficult to assess as many proposed examples involve questionable assumptions concerning costs and benefits of the partners (reviewed in Koenig 1988, Wilkinson 1988, Ligon 1991). For example, there must be a direct fitness cost to the donor that is less than the fitness benefit received when the act is reciprocated (Koenig 1988). The only convincing evidence (cf. Ligon 1991) comes from vampire bats, Desmodus rotundus, which share food by regurgitation of blood on the basis of reciprocity with both close kin and familiar non-kin (Wilkinson 1984). While the donor’s cost is the loss of weight and time by regurgitating some blood, this will save the recipients from starvation (ibid).

The classical example for reciprocal altruism in primates was reported among unrelated male baboons who alternate in enlisting each other to take over an oestrous female from a consort male (Packer 1977), but later studies questioned the extent of reciprocity involved in these coalitions (Bercovitch 1988, Noë 1989) since both partners were as likely to gain access to the consort female. Ever since several studies reported reciprocal support in several primate species (e.g., Hunte & Horrocks 1987, de Waal & Luttrell 1988, Watts 1997, Hemelrijk & Ek 1991) and others, in addition, suggested interchange between grooming for support (e.g., Seyfarth & Cheney 1984, Silk 1992b, Hemelrijk 1994, Vervaecke et al. 2000). Interchange seems only to occur in groups with a linear hierarchy (Henzi & Barrett 1999) and reciprocity was reported only to be evident when the hierarchy is stable (de Waal 1978). Chapais et al. (1995) tested the affiliation-for-support hypothesis via experimentally induced rank reversals among Japanese macaques. However, mutual selfishness rather than reciprocal altruism provided a better explanation as groomer and supporter seem both to immediately benefit (see co-operation below).

A problem, related to studies of reciprocity and interchange was pointed out by Hemelrijk (1990a,b), is that an association between grooming given for the support received may in fact be a by-product of the correlation with other variables. Testing Seyfarth’s data (1976, 1980) on vervet monkeys and baboons with the use of the matrix partial correlation test, Hemelrijk (1990b) found that the correlation in vervet monkeys, but not in baboons, was indeed a by-[page 107↓]product of rank as high-ranking female gave more grooming and support than low-ranking females. Hemelrijk (1990a) concluded that in order to detect a genuine relationship between two variables, other variables should be kept constant. This problem may have influence the present data, too, but by analysing maternal half-sibling, paternal half-sibling or non-kin separately, it was controlled for rank distance (see above).

Following the theory of reciprocal altruism, a reciprocal relationship should be immune against cheating (Trivers 1971). However, refusal to reciprocate need not end reciprocal relationships (see Noë 1990). Reciprocal relationships among male baboons did not break off when one partner refused to join a coalition against a consort male (Bercovitch 1988). This suggested to Bercovitch that coalitions are the outcome of co-operation between males pursuing self-interests. Females in the present study did not show a trend of cheating as (i) they reciprocate support given within dyads, i.e., when female A supported female B, female B also supported female A and (ii) there was no interchange between support given and against, i.e., when female A supported female B, female B did not targeting female A.

Reciprocity was found for both kin and non-kin, although reciprocity constraints might also be more relaxed among kin than among non-kin (Chapais 2001). However, it seems difficult to distinguish between bilateral altruism among kin produced by kin selection or produced by reciprocal altruism, as the former does not require to be exchanged bilaterally. In other words, reciprocity among kin may simply be a by-product of the symmetry in relatedness within a kin dyad (cf. Chapais 2001). Ideally, unilaterally altruistic behaviour should be investigated which is testable on adult-immature dyads as immature primates still receive more support than they are able to provide (ibid.). However, the present study was focused on adult females. Bilateral support might also evolve as a by-product when both partners target the same individual for dominance and should therefore be interpreted in terms of self-interest even when observed among kin (Prud’homme & Chapais 1996, see below).

Co-operation

Testing whether the concept of co-operation can also explain coalition formation, data were analysed with respect to interventions being associated with a risk of retaliation when intervening against higher-ranking targets or without (or low) risk when the supporter was higher-ranking than the target which is likely to stabilise the existing hierarchy. The results suggest that a supporter sometimes also seems to benefit from intervening depending upon the kin relationship between the supporter and the recipient of support. Support given to maternal half-siblings was found to be most often associated with a risk of retaliation, targeting higher-ranking individuals, most often adult males. Non-kin support, in contrast, was most often [page 108↓]found without risk as females seem to stabilise the existing hierarchy through non-kin support. Support given to paternal half-siblings seemed to be a combination of rank patterns most often found among maternal half-siblings (risky patterns) and non-kin (non-risky patterns). Therefore, the present study suggests that supporters pursue self-interests in some interventions both kin and non-kin support.

Recall from the Introduction that female and male primates differ in their distribution of support. Male rank is likely to change over time, which may change their interests (de Waal 1985). Coalition formation among males was therefore suggested to be more dominance-orientated than among females, who tend to support their lower-ranking kin at high risk to themselves. Comparing non-kin support among male bonnet (Silk 1993, Table 8), male Barbary macaques (Widdig et al. 2000, Fig. 3) with the present data on female rhesus macaques (Fig. 6.3, this chapter) gives additional evidence that males tend to be (even) more conservative than females. Fig. 6.4 summarises three studies.

	Fig. 6.6.4: Percentage of support given to non-kin (NK) across the three studies of macaques

Note the six rank patterns are arranged as introduced in Box 6.1, with the first three on the left side representing risky interventions in terms of retaliation and the three remaining rank patterns representing low (if any) risk interventions. As evident from this figure, the distribution of interventions comparing both male studies are very similar, but differ from the female’s pattern. Female rhesus show higher percentage of risky intervention even in favour of non-kin than both male bonnet and Barbary macaques, but as pointed out before, rhesus females tend to support their non-kin using no-risk interventions, while supporting kin using more often risky interventions. However, even in no-risk interventions male and female macaques differ. Males mainly gave support when they were themselves the highest-ranking individual within the triad (CBA and BCA), while females mainly gave support when ranking between the recipient and the target (CAB). How to explain the last pattern just mentioned for [page 109↓]females? Using removal experiments Chapais et al. (1991) showed for Japanese macaques that unrelated females from two dominant matrilines dependent upon supporting each other in order to maintain their dominance over females from low-ranking matrilines, which is exactly the pattern (CAB) just described. Due to the selfish interests of the supporters, Chapais et al. (1991) concluded that co-operation provides a better explanation of non-kin interventions than reciprocal altruism.

Male chimpanzees frequently intervene against higher-ranking targets to improve their rank (de Waal 1982). Male macaques, in contrast, rarely form coalitions that put them at risk of retaliation. Silk (1993) concluded from her study that male bonnet macaques do not use coalitions to increase in rank, because only after rising in rank they intervene more often in conflicts. Colvin (1983) suggested that conflicts between higher-ranking opponents may stay dyadic, because lower-ranking individuals are inhibited from interfering. This seems especially likely for female macaques as dominance hierarchy can be predicted nearly 100% by the birth order (this study) and is very stable over time.

One of the most significant predictions testing kin altruism is that females should intervene on behalf of kin against higher-ranking targets at higher rates than supporting distant kin or non-kin. Females intervene on behalf of their kin, but rarely on behalf of non-kin, against adult males and females ranking higher than themselves (Kurland 1977, Kaplan 1977, Silk 1982, Datta 1983b, Lee 1983, de Waal & Luttrell 1985, this study). However, it cannot be concluded that supporting kin is always associated with a net cost which is the assumption of an altruistic act (cf. Silk 2002), therefore I agree with Chapais (2001) that distinguishing between altruistic nepotism evolved via kin selection and mutualistic nepotism evolved via natural selection is difficult at least for interventions with a questionable cost. In other words, assuming that kin are as competent as non-kin to provide support, the greater availability and familiarity due to close proximity of kin might generate nepotism by natural selection, regardless of the fact that they are genetically related. A female that faces a choice between co-operating with kin or non-kin, gains the same direct fitness, but in co-operating with a kin, she additionally gains indirect fitness, too. Therefore, it pays more to co-operate with kin. Kin bias in coalition formation may therefore be expressed as a joint operation of natural and kin selection.

Female rhesus macaques gave support most often to maternal half-siblings. In addition, unrelated peers supported each other more often than unrelated non-peers. Females did not support their paternal half-siblings more often than non-kin, but results may indicate indirect [page 110↓]evidence for paternal kin discrimination as females were targeting their paternal half-siblings less often than non-kin. This finding might be due to the fact, that paternal half-siblings can be very different in rank, while maternal half-sibling are of adjacent rank, implying that a low-ranking female cannot provide actual help to its paternal half-siblings, but may risk a higher probability of retaliation. As a compromise, females may instead selectively avoid to target their paternal half-siblings, suggesting that constraints to an individuals’ own competitive ability play an important role in coalition formation. Females also supported their maternal half-siblings at risk against adult males which are higher-ranking than themselves, while females supported non-kin without (or low) risk, mainly to stabilise the existing hierarchy. Paternal half-siblings tended to show an intermediate pattern. The results of this chapter also suggest that non-kin showed at least a trend in reciprocity and interchange, but also a high proportion of low (if any) cost interventions which can be explained by mutualism. Kin, on the other hand, showed a stronger trend in reciprocity and interchange and also provided a higher proportion of costly interventions towards each other than non-kin. However, at least low cost interventions by kin can also be explained by mutualism, as the cost to the intervener seems to be neglectable. It should also be concluded that patterns of coalition formation using paternity data can still be explained to a similar extent by kin selection, reciprocal altruism and co-operation compared to studies where paternity is unknown. In order words, lacking paternity data might not have such an important influence investigating coalition formation than shown for affiliation or dyadic aggression, but this assertion has be to proven by future studies.

Footnotes and Endnotes

¹ In the literature several terms have been used to describe the same behaviour, such as coalition formation, support, helping, agonistic aiding, interventions, alliances, fight interference etc. These terms will be used as synonyms throughout the chapter (cf. de Waal & Harcourt 1992). While a coalition describes any support observed between two individuals, an alliance, in contrast, describes a enduring relationship with respect to support (see de Waal & Harcourt 1992).