Conceived and designed the experiments: BTH. Performed the experiments: BTH. Analyzed the data: BTH MAS JEM. Contributed reagents/materials/analysis tools: BTH MAS JEM. Wrote the paper: BTH MAS JEM.
The authors have declared that no competing interests exist.
Animal groups typically contain individuals with varying degrees of genetic relatedness, and this variation in kinship has a major influence on patterns of aggression and affiliative behaviors. This link between kinship and social behavior underlies socioecological models which have been developed to explain how and why different types of animal societies evolve. We tested if kinship and age-sex class homophily in two groups of ring-tailed coatis (
Kinship plays a key role in shaping animal societies, particularly in group-living species. In many societies, animals preferentially associate with kin, direct affiliative behaviors toward close relatives, and support close kin during agonistic interactions
In addition to the key influence that genetic relatedness has on social organization, other factors including feeding ecology, competitive regime, and demography are also important. For example, socioecological models of resource distribution and feeding competition have been used to predict the presence and degree of nepotism in primate groups
Ring-tailed coatis (
In the current study, we use social network analyses to determine the degree to which genetic relatedness influences the structure of three types of coati social networks: 1) spatial association, 2) grooming, and 3) aggressive interactions. Social network analysis allows for the quantification of multi-actor interactions, which provides a more realistic depiction of animal societies than traditional dyadic measures. With the use of these statistical and descriptive methods, it is possible to determine the degree to which kinship, age, and sex contribute to behavioral interactions between individuals and shape social structure. Although previous evidence suggested that kinship is not a major factor driving patterns of ring-tailed coati aggression, this hypothesis has never been explicitly tested. With the use of genetic markers we calculated relatedness between individuals and confirmed the identity of mother-offspring pairs. Here we test two non-mutually exclusive hypotheses: 1) Kinship explains the structure of coati social networks and 2) Age-sex class homophily explains the structure of coati social networks. Our predictions for each behavioral dimension are as follows:
Relatedness will be a significant predictor of association network structure. This pattern should hold between and among age and sex class categories.
Age and/or sex class homophily will be a significant predictor of the association network structure with higher association coefficients between members of the same age and sex class.
Relatedness will be a significant predictor of grooming network structure, with higher relatedness resulting in increased grooming rates. In particular, we predict that mother-offspring pairs should groom each other more frequently than more distantly related dyads.
Age and/or sex class homophily will be a significant predictor of the grooming network structure with more interactions between members of the same age and sex classes.
Relatedness will be a significant predictor of aggression network structure, with individuals directing less aggression towards related individuals
Alternately, because coati dominance hierarchies have previously been reported to be ‘age structured’ we predict that relatedness could have little effect on aggressive networks
Finally, because polyadic agonistic interactions can influence the structure of dyadic aggression networks, we conducted an additional analysis to determine the degree to which adult female coalitionary support for juveniles is shaped by kinship.
Adult females should preferentially support their offspring and/or closely related juveniles in agonistic interactions.
Adult females should support all juveniles during agonistic interactions, regardless of the degree of relatedness between the adult female and juvenile
This study complied with all institutional, national and ASAB / ABS guidelines for animal welfare. Local permission was granted from APN (Argentina National Park service) and animal handling procedures were approved by the SUNY Stony Brook Institutional Animal Use and Care Committee (IACUC# 20021175).
Behavioral data were collected in Iguazu National Park, Argentina (54°W, 26°S), between July 2002 and December 2004. A total of 150 coatis were captured in 32×10×12 inch Tomahawk or similar traps, immobilized with Ketamine and Xylazine and fitted with unique combinations of multicolored ear tags for individual identification (Rototag ear tags, Dalton Co.). Data from two neighboring coati groups (PQ and PSG) for two study years (2003 and 2004) were used in this study (N = 65 individuals). These two groups were socially segregated, and individuals rarely interacted with members of other groups. All coatis in the two social groups were individually recognizable due to their ear tags except for young juveniles which had not yet been tagged (juveniles were typically tagged when 4 months old). Coati groups were well habituated to the presence of human observers and we were able to follow habituated individuals within 2 m without disturbing them. Coati groups were comprised of adult females (24 months of age or older), subadults (12–24 months of age), juveniles (2–12 months of age), and one adult male (generally 36 months or older). Adult males disperse from their natal groups at 2 years of age, while females remain in their groups. Although group composition changed from year to year, group membership was relatively stable during the two study periods
All agonistic interactions were recorded ad libitum by the author, or by field assistants trained for at least 2 months
Ten second individual focal samples were recorded to determine levels of association between individuals
When individuals were captured, a small plug of skin tissue was punched out during ear tagging and the tissue was stored in 10% DMSO saline solution. DNA purification was carried out using a Qiagen Bio-Sprint 96 workstation following the protocol for DNA extraction from animal tissues as supplied by the manufacturer. All individuals were genotyped at 15 previously developed microsatellite loci which averaged 4.2 alleles per locus (range 2–7)
Three social networks were built for each group for each year based on matrices of: 1) association, 2) grooming interactions, and 3) aggressive interactions. In the association network a connection, or tie, existed between any two individuals who were observed associated as defined above. These ties were weighted based on that dyad's halfweight coefficient, which is a commonly used measure of association. The halfweight coefficient is essentially a corrected ratio that accounts for differences in sighting frequency or sampling effort by comparing the number of times individuals were seen together to the number of times they were seen in total and is calculated as X/X+0.5(Ya+Yb)+Yab where X = number of times individuals a and b were observed together, Ya is the number of observations where a was observed without b, Yb is the number of observations in which b was observed without a, and Yab is the number of observations a and b were both observed, but in separate groups
We used multiple regression quadratic assignment procedures (MRQAP) with the double semi-partialing permutation method to determine what factors influenced social structure in the coati groups
We investigated general differences between age-sex classes with respect to grooming and aggression network measures. Three measures of centrality were calculated for each individual in each network: in-strength, out-strength, and eigenvector centrality. In-strength centrality (also known as weighted in-degree) is defined as the sum of the weights of all incoming ties, where as out-strength (weighted out-degree) is the sum of the weights of all outgoing ties. Eigenvector centrality is the corresponding eigenvalue of the first eigenvector of a given matrix and is a measure of both direct and indirect connectedness
Patterns of coalitionary support were assessed by comparing the number of observed cases of mother-offspring coalitionary support to the predicted number if females randomly aided all juveniles. We also tested whether adult females preferentially supported closely related juveniles (i.e. not just offspring). We compared the average degree of relatedness between juveniles and the adult females that supported them to the average pair-wise relatedness of social group members using ANOVA tests carried out in JMP 5.1.
The mother-offspring matrices were significant predictors of the association and grooming networks in all group-years (
Groups: | PQ 2003 | PQ 2004 | PSG 2003 | PSG 2004 | ||||
Association | slope | P | slope | P | slope | P | slope | P |
Mother-offspring *** | 0.338 | 0.002* | - | - | 0.207 | 0.005* | 0.241 | 0.001* |
Relatedness | 0.044 | 0.325 | - | - | 0.068 | 0.216 | 0.033 | 0.292 |
Sex | 0.014 | 0.374 | - | - | 0.009 | 0.430 | 0.084 | 0.043* |
Adult | 0.071 | 0.205 | - | - | 0.014 | 0.433 | 0.013 | 0.420 |
Subadult | 0.127 | 0.075 | - | - | - | - | 0.205 | 0.001* |
Juvenile *** | 0.682 | 0.002* | - | - | 0.835 | 0.001* | 0.372 | 0.001* |
|
||||||||
Mother-offspring *** | 0.568 | 0.001* | 0.689 | 0.001* | 0.569 | 0.001* | 0.527 | 0.001* |
Relatedness | 0.063 | 0.179 | −0.012 | 0.298 | −0.017 | 0.440 | −0.031 | 0.208 |
Sex | 0.034 | 0.283 | 0.009 | 0.324 | 0.081 | 0.179 | 0.056 | 0.054 |
Adult *** | 0.077 | 0.125 | 0.226 | 0.001* | 0.252 | 0.007* | 0.390 | 0.001* |
Subadult | 0.121 | 0.049* | - | - | - | - | 0.027 | 0.134 |
Juvenile | −0.129 | 0.028* | −0.037 | 0.071 | −0.020 | 0.421 | −0.065 | 0.088 |
|
||||||||
Mother-offspring | −0.069 | 0.191 | −0.024 | 0.260 | −0.024 | 0.424 | −0.026 | 0.334 |
Relatedness | 0.058 | 0.233 | −0.012 | 0.407 | 0.095 | 0.248 | 0.047 | 0.175 |
Sex | 0.150 | 0.004* | −0.064 | 0.076 | −0.221 | 0.015* | 0.046 | 0.110 |
Adult | −0.014 | 0.490 | 0.064 | 0.073* | 0.017 | 0.419 | −0.013 | 0.455 |
Subadult | −0.066 | 0.049* | - | - | - | - | 0.003 | 0.377 |
Juvenile | −0.070 | 0.246 | −0.024 | 0.357 | 0.116 | 0.184 | −0.052 | 0.268 |
Age class categories represent age homophily. No subadults were present in the PQ 2004 and PSG 2003 groups. Significant predictor variables (P<0.05) for individual group-years = *. Variables that were significant in at least three out of four group years = ***.
Adult females groomed each other more often than predicted (in three out of four group-years) but this age homophily pattern was not seen in the grooming behavior of juveniles and subadults (
The nodes represent individual coatis and the thickness of the lines between nodes is proportionate to the number of interactions between those individuals. Circles: females; squares: males; green: adults; yellow: subadults; purple: juveniles.
Age-class | N | Eigenvector | Out degree strength | In degree strength |
Adult female | 10 | 0.622±0.288 | 89.497±64.325 | 44.305±27.843 |
Adult male | 2 | 0.549±0.203 | 35.874±19.543 | 74.471±53.047 |
Subadult female | 4 | 0.373±0.159 | 35.224±21.114 | 21.448±11.069 |
Subadult male | 2 | 0.080±0.010 | 9.653±2.275 | 4.826±2.275 |
Juvenile female | 25 | 0.224±0.169 | 6.188±10.380 | 22.061±27.605 |
Juvenile male | 28 | 0.236±0.179 | 11.709±14.909 | 23.387±21.371 |
N = number of individuals. Adult females groomed others the most, while adult males received the most grooming. Values were averaged across groups and years.
Males were more central in the aggression network than females (average eigenvector centrality males = 52.974±25.128 SD, females = 23.839±15.698) and directed more aggression than females (average normalized out strength degree males = 47.134±52.683, females = 16.910±22.192). Adult males were particularly aggressive, and directed more aggression than other age-sex classes (
The nodes represent individual coatis and the thickness of the lines between nodes is proportionate to the number of interactions between those individuals. Circles: females; squares: males; green: adults; yellow: subadults; purple: juveniles.
Age-class | N | Eigenvector | Out degree strength | In degree strength |
Adult female | 10 | 0.431±0.190 | 22.419±31.460 | 42.066±30.989 |
Adult male | 2 | 0.685±0.373 | 105.399±87.312 | 26.820±28.192 |
Subadult female | 4 | 0.440±0.373 | 10.859±14.522 | 63.935±70.177 |
Subadult male | 2 | 0.739±0.004 | 15.255±6.780 | 39.226±3.698 |
Juvenile female | 25 | 0.266±0.144 | 13.906±13.351 | 19.989±23.071 |
Juvenile male | 28 | 0.423±0.251 | 41.087±43.901 | 24.957±26.723 |
N = number of individuals. Adult and juvenile males directed the most aggression to others, while adult females and subadults received the most aggression. Values were averaged across groups and years.
A total of 37 cases of adult female coalitionary support of juveniles were recorded in the two main study groups during 2003–2004. If adult females (3–5 per group-year) randomly gave support to all juveniles, it was expected that juveniles would be supported by their mother in 8 incidences. We found that mothers supported their offspring twice as much as random (16 cases), but a larger proportion of adult female support for juveniles was from non-mothers (57%). No between group differences were found in the proportion of cases where adult females supported their offspring (PQ = 42%, PSG = 44%). Adult females also did not preferentially support closely related juveniles; the average degree of relatedness between juveniles and the adult females that supported them was not statistically different from the average degree of relatedness between individuals in the social group (ANOVA tests; PQ: F1,798 = 0.028, P = 0.866; PSG: F1,647 = 0.057, P = 0.391).
The two study groups varied in the degree to which individual coatis were related to each other, which likely arose from the distinct origins of the two groups. The PQ group was founded by a single adult female and her offspring in 2001, while the PSG group formed when five adult females split off from a larger group in late 2002
The nodes represent individual coatis and the thickness of the lines between nodes is proportionate to the degree of relatedness between those individuals. The bottommost adult female is JW, who was not closely related to the other adult group members. Circles: females; squares: males; green: adults; yellow: subadults; purple: juveniles.
Grooming and association matrices were shaped by mother-offspring associative behavior in all four group-years. This result demonstrates the strong link between genetic relatedness and associative behaviors in ring-tailed coatis. Interestingly, pairwise relatedness values were not a significant predictor of grooming and association when mother-offspring pairs were included in the MRQAP regressions. While juveniles were more often associated with other juveniles, they did not preferentially associate with their closest juvenile relatives (i.e. same age maternal siblings). These results are consistent with the idea that affiliative behaviors in coatis are strongly shaped by mother-offspring relations, while other kinship categories are less important (full and half siblings, aunts, etc.).
In general, age class homophily was a weak predictor of grooming network structure. Subadults and juveniles rarely groomed within their age class, but grooming among adults was a significant variable determining grooming network structure in three out of four group-years (
Adult females frequently groomed each other, even though foraging adult females were not always within close proximity. These patterns contrast strongly with juveniles and subadults who rarely groomed within their age class, but generally associated with their same age class during foraging
Aggression network structure was not explained by kinship, or mother-offspring pairs. Indeed, it appears that none of the tested parameters reliably predicted the structure of aggression networks. Although no age or sex class homophily variables were significant predictors of the overall dominance network structure, a closer comparison of the direction of interactions within and between age-sex classes demonstrated clear patterns. Not surprisingly, adult males were particularly aggressive (
The patterns of coalitionary support between adult females and juveniles are only marginally consistent with the hypothesis that adult females preferentially support their offspring during aggressive interactions. Juveniles were supported by adult females that were not their mothers during more than half of these coalitionary interactions (57%). It is plausible that patterns of coalitionary support found in ring-tailed coatis could have arisen due to inclusive fitness benefits, with adult females supporting closely related juveniles in addition to their own offspring. In the PQ group, where all adult females were closely related to all juveniles, inclusive fitness benefits could have easily led adult females to support all juveniles. On the other hand, there was variability in the degree of relatedness among group members. Even if most individuals were closely related, it was predicted that females should preferentially support their offspring during aggressive conflicts with their siblings, parents, and aunts. We found little evidence to support these patterns. In the PSG group, adult females were more distantly related to each other than females in the PQ group. The one adult female (JW) which was more distantly related to all other adult females still came to the aid of non-offspring juveniles (N = 5). These patterns indicate that close kinship is not necessarily a prerequisite for coalitionary aid in this species. Our result that relatedness had no discernable effect on patterns of aggression is in stark contrast to most studies of aggression and dominance in social animals
In some species, group augmentation has been posited as a hypothesis for female tolerance of juvenile aggressive behavior. Clutton-Brock and colleagues
The unusual age-based aggression patterns in ring-tailed coatis appear to fall outside the purview of widely used socio-ecological and kinship based models of animal behavior. No previously published model of animal behavior would have predicted that adult females should direct aggression towards subadults regardless of their kinship ties, while coming to the aid of non-offspring juveniles. This coalitionary support provided by adult females is the major reason why juvenile coatis were able to feed in small patchy resources without being excluded by larger individuals, and thus provided a major fitness benefit to the youngest, most vulnerable age-class. Although previous studies have documented adult females preferentially aiding younger offspring in other species
We thank Yamil Di Blanco, Santiago Escobar, Carolina Ferrari, Mauro Tommone, Fermino Silva and Viviana Munoz for help and assistance during the course of the field work. Charles Janson provided essential support and guidance to BTH before, during, and after the field work. We thank Frank Hailer, Emily Latch, Nancy Rotzel, and especially Mirian Tsuchiya-Jerep for help with the genetics portion of this study. This paper benefited tremendously thanks to comments on earlier drafts by Meg Crofoot, Janet Mann, Jane Waterman and two anonymous reviewers.