Introduction

Comparisons between chimpanzees (Pan troglodytes) and bonobos (Pan paniscus) emphasize species-typical differences within the Pan genus (Furuichi and Ihobe 1994; Stanford 1998; Sannen et al. 2003; Palagi 2006; Heilbronner et al. 2008; Wobber et al. 2010; Hare et al. 2012; Hashimoto and Furuichi 2015). Traditional models of primate socioecology characterized primates with female dispersal as non-female-bonded and chimpanzee patterns generally follow these predictions (Wrangham 1980; Stumpf 2007; Lonsdorf et al. 2014). However, female bonobos are considered an exception to these patterns (Parish 1994, 1996; White 1996; Parish and De Waal 2000; Tokuyama and Furuichi 2016). Some studies suggest that female bonds in bonobos are overemphasized (Stanford 1998; Stevens et al. 2006), whereas female bonds in chimpanzees are overlooked (Wakefield 2008, 2013; Lehmann and Boesch 2009). Such evidence suggests that a revised model of Pan socioecology may better characterize social bonds in these species. Social network analysis provides an opportunity to re-evaluate the comparative bonds in each species using modern statistical methods designed to illustrate the structure of social interactions (Lusseau et al. 2008; Wey et al. 2008; Brent et al. 2011; Sueur et al. 2011).

The use of sociograms to visualize grooming relationships has a long history within primatology (Brent et al. 2011). However, newer mathematical modeling techniques combined with network analysis allow for the examination of a range of questions regarding individual network position and assortative tendencies within social networks (Lusseau and Newman 2004; Wey et al. 2008; Brent et al. 2011; Sueur et al. 2011). Previous work using social network analyses in captive chimpanzees indicates that grooming relationships are crucial to social cohesion (Kanngiesser et al. 2011). Furthermore, analytical techniques using mathematical removal of individuals indicates that wild-born individuals have a larger impact on social cohesion, despite similar social network positions (Leve et al. 2016). Because grooming is a predominant primate social interaction indicative of social bonds, primate social networks often focus on grooming bouts (Seyfarth 1977; Dunbar 1991; Wey et al. 2008; Fedurek and Dunbar 2009; Brent et al. 2011).

In the wild, association patterns in species with high fission–fusion dynamics is typically shaped by ecological pressures, although individual preferences may influence subgrouping patterns (Wrangham 1980; Chapman 1990; Symington 1990; Furuichi and Ihobe 1994; Chapman et al. 1995; Hashimoto et al. 2003; Ramos-Fernández 2005; Lehmann et al. 2007; Aureli et al. 2008; Mulavwa et al. 2008; Wakefield 2008; Lehmann and Boesch 2009; Wakefield 2013; Hashimoto and Furuichi 2015; Rodrigues 2017). Male chimpanzees exhibit strong association patterns and close social bonds, as would be expected based on their dispersal patterns (Wrangham 1980; Furuichi and Ihobe 1994; Itoh and Nishida 2007; Stumpf 2007; Hashimoto and Furuichi 2015; Surbeck et al. 2017). Wild male–male dyads of chimpanzees and bonobos spend equivalent times grooming and in proximity to one another (Furuichi and Ihobe 1994). Female chimpanzees exhibit comparatively weaker associations patterns than males due to foraging constraints and the burden of dependent offspring (Otali and Gilchrist 2005; Hashimoto and Furuichi 2015; Surbeck et al. 2017). Nonetheless, female chimpanzees vary in sociality across sites, and close female bonds emerge under favorable ecological conditions (Wakefield 2008, 2013; Lehmann and Boesch 2009; Hashimoto and Furuichi 2015). Larger party sizes among bonobos facilitates female bonding and coalition-building (White 1996, 1998; Hohmann and Fruth 2000; Mulavwa et al. 2008; Tokuyama and Furuichi 2016).

In captivity, animals may be less constrained by ecological limits on association patterns. In a large complex of enclosures that facilitated social choice, association and interaction rates in chimpanzees were correlated, suggesting social preferences (Clark 2011). With reduced time constraints, captive chimpanzees can devote a large proportion of their time to grooming (Majolo et al. 2008; Leve et al. 2016). Thus, examining grooming bouts in captive settings provides a way to examine assortative social bonding in the absence of ecological constraints. In captivity, female chimpanzees can form strong bonds (Baker and Smuts 1994), suggesting that weak relationships in the wild are due to ecological pressures. However, research in captive settings may also be affected by individual histories and management practices. For example, Stevens et al. (2006) cautions that the emphasis on strong female bonds in bonobos may be inflated by captive management decisions. Female–female affiliation is more pronounced in the process of forming new social bonds, and after time and familiarity, rates of female affiliative behavior decline (Stevens et al. 2006). Furthermore, individuals with atypical rearing histories are less extraverted and groom less than individuals who were mother-reared (Freeman and Ross 2014; Freeman et al. 2016).

Here, we compare grooming networks in captive chimpanzees and bonobos. Based on data emphasizing cross-sex bonding in chimpanzees and bonobos (Furuichi 1997; Hohmann et al. 1999; Lehmann and Boesch 2005; Walz 2008), and flexibility in social bonds across captive and wild conditions (Baker and Smuts 1994; Furuichi 1997; Stevens et al. 2006; Wakefield 2008, 2013; Hashimoto and Furuichi 2015) we predict that (1) male–female bonds will be the strongest in both species (i.e., that within-sex grooming relationships will be stronger than between-sex grooming relationships), and (2) males and females will exhibit no significant sex differences in eigenvector centrality, strength, or clustering coefficient in either species. To account for idiosyncratic factors that may structure captive social networks (Stevens et al. 2006; Kanngiesser et al. 2011; Leve et al. 2016), we further examine the impact of origin (captive or wild-born), kinship, and group tenure.

Methods

Ethical note

All research was reviewed and approved by the Columbus Zoo and Aquarium, the North Carolina Zoo, and the Chimpanzee Species Survival Plan. This research adheres to the American Society of Primatologists and International Primatological Society’s ethical guidelines. We have no conflicts of interests to declare.

Subjects and housing, Columbus Zoo

Data were collected on the bonobos at the Columbus Zoo, in Columbus, Ohio, from June to August 2013 during zoo hours from 0900 to 1900 h (Fig. 1). Fifteen hours of data were collected on 11 subjects, for a total of 165 h. Focal data was collected on subadult and adult individuals, ranging in age from 7 to 34 years old. Subjects included five females and six males (Table 1). LA, SU, JI, and TO were wild-born, and were together in a European zoo before transferring to the Columbus Zoo in 1990.

Fig. 1
figure 1

Bonobos (Pan paniscus) grooming at the Columbus Zoo (Photo by Michelle Rodrigues)

Full size image
Table 1 Focal subjects at the Columbus Zoo, June–August 2013
Full size table

The bonobos were observed from the public viewing areas. The public exhibits include two indoor enclosures (each 54.8 m2) and one large outdoor enclosure (2647.7 m2: Walz 2008; Fortunato 2009; Boose et al. 2013; Brand and Marchant 2015). The door between the two indoor enclosures could be opened to permit one subgroup access to both or closed to separate bonobos into two separate subgroups. The indoor enclosures had climbing structures and woodwool for nesting, and the outdoor enclosure include naturalistic features and substrates including grass, a waterfall, trees/shrubs, woodwool for nesting, and a climbing structure. Additionally, bonobos were housed in two off-exhibit indoor enclosures and two off-exhibit outdoor enclosures. All enclosures were interconnected via a series of doors and runways. The keepers managed the bonobos to allow for fission–fusion social groups (Walz 2008; Fortunato 2009; Boose et al. 2013; Brand and Marchant 2015). To form subgroups, bonobos are given access to each other either to choose who to nest with at night, or in the morning, given access to each other and indoor/outdoor enclosures, and allowed to choose which individuals/enclosures to associate with. Subgroups were often kept stable for 2–3 days, and then allowed to repeat the procedure. During this study, males BI and DO were not allowed access to each other to prevent severe aggression, but all other dyads were allowed opportunities to associate. Bonobos were fed fruit, vegetables, and chow in the mornings and evenings upon entering a new enclosure, with occasional supplemental feeding during the day. Bonobos were typically in the mornings (between 0700 and 0900) and evenings (between 1800 and 1900).

Subjects and housing, North Carolina Zoo

Data were collected on chimpanzees at the North Carolina Zoo, in Asheboro, North Carolina, from May to July 2014 (Fig. 2). Data were collected during zoo hours from 0900 to 1700. Chimpanzees were housed in two social groups, the socially “typical” and socially “atypical” groups. Although the keepers usually managed them in a modified fission–fusion manner, with some females allowed to move between the two groups, the groups were kept stable during the study period to allow new introduction GE time to settle into the socially typical group. Although data were collected on both groups, data are only presented here for the socially typical group, which exhibit species-typical grooming and social behavior. This group included five females and three males (Table 2). A total of 159 h of data were collected on subjects in this group. Twenty hours of data were collected for MA, AM, GE, JO, SO, and LA, and 19.5 h of data were collected on RU and RY due to the resumption of moving RU between groups.

Fig. 2
figure 2

Chimpanzees (Pan troglodytes) grooming at the North Carolina Zoo (Photo by Michelle Rodrigues)

Full size image
Table 2 Chimpanzee focal subjects at the North Carolina Zoo, May–July 2014
Full size table

The two groups were shifted between a large, outdoor enclosure (4195 m2) with public viewing, and indoor enclosures that were off-exhibit. Chimpanzees were fed fruit, vegetables, and chow in the morning (between 0800 and 0900) and late afternoon (between 0400 and 0500) when shifted between enclosures. The large outdoor enclosure included naturalistic features including grass, rocks, trees/shrubs, and a large climbing structure.

Behavioral data collection

Focal follows were conducted for 30 min, with 2-min instantaneous recording of activity and nearest neighbors/social partners, and all-occurrence recording of social behaviors (Altmann 1974). At the beginning of each focal period, the following was recorded: (1) enclosure of the focal animal, (2) identity of all other individuals in the enclosure, (3) sexual swelling of focal females. Activity categories were feed, travel, rest, social, other, or out of sight. During instantaneous recording of social behaviors, social behavior and all social partners were recorded. During instantaneous recording of other behaviors, the specific other behavior (self-directed behavior, object manipulation, observer-directed behavior, visitor-directed behavior) was recorded. Duration of all-occurrence social behavior was recorded to the nearest minute, and partner(s) and initiator, recipient, or mutual recorded for relevant behaviors. Affiliative social behaviors in the behavioral catalogue included grooming, playing, embracing, huddling, kissing, touching, nodding, nursing, carrying, affiliative vocalization, genital–genital rubbing, present, genital manipulation, and copulation. We defined a grooming bout as a continuous period of grooming by one individual (the initiator) directed toward another individual (the recipient). Any changes in initiator/recipient or activity changes longer than 10 s were considered termination of the bout (Dunbar 1976; Henzi et al. 1997; Manson et al. 2004). Treating changes in initiator/recipient as separate bouts allow for analyzes of reciprocity and parceling (Manson et al. 2004). Agonistic social behaviors in the behavioral catalogue include displacing, chasing, displaying, harassing, fighting, distressing vocalizations, and fear-grimacing. Focal follows were conducted according to a random, pre-set list whenever possible, with effort made to sample individuals equally across morning and afternoon hours.

Our sample sizes (bonobos N = 11; chimpanzees N = 8) are typical of most studies studying captive groups. Typical group sizes for captive studies of chimpanzees and bonobos social behavior range from approximately 6–22 individuals per group (Vervaecke et al. 2000; Palagi 2006; Stevens et al. 2006; Fedurek and Dunbar 2009; Clark 2011; Kanngiesser et al. 2011; Schel et al. 2013; Massen and Koski 2014; Leve et al. 2016). In a large data set of grooming matrices from 44 sites, group sizes had a median of 11, with an interquartile range of 8–16 individuals (Voelkl et al. 2011). Similarly, our hours of focal data per individual (bonobos 15 h/individual; chimpanzees 19.5–20 h per individual) are comparable to other studies. Recent studies of captive chimpanzee social networks report hours of observation per individual ranging from 3.63 to 27.41 h/individual (Clark 2011; Kanngiesser et al. 2011; Schel et al. 2013; Leve et al. 2016).

SOCPROG analysis

Grooming can be treated as either a state or an event (Altmann 1974). Studies constructing social networks based on grooming data use frequency of grooming bouts (Kanngiesser et al. 2011), proportions based on duration of grooming time relative to observation time (Leve et al. 2016), or proportions based on duration of dyadic grooming time relative group averages (Schel et al. 2013). Following the example of Kanngiesser et al. (2011), we chose to use frequency of grooming bouts. This method preserved initiator/recipient data so that we could test for grooming asymmetry. Such tests are essential to ensure that asymmetries are not overlooked before treating data as reciprocal. However, to ensure that our social networks are comparable to those based on duration, we performed Mantel Z-test correlations comparing frequency and duration with 1000 permutations (Whitehead 2009). Grooming bouts in bonobos had a mean duration of 2.57 min (range, 1–27 min). Grooming bouts in chimpanzees had a mean duration of 2.85 min (range, 1–30 min). Frequency and duration were significantly correlated in both bonobos (Z = 0.98126, p < 0.0001) and chimpanzees (Z = 0.93808, p < 0.0001).

We chose to construct a weighted network focused on grooming frequency because such networks are more salient to the animals’ social lives. Weighted networks are better suited to the study of social bonds, because they preserve data on their variation (Kerth et al. 2011; Voelkl et al. 2011). Unweighted (binary) networks are easier to analyze, and there are additional analyses that can be done upon them [see Newman (2001), Kanngiesser et al. (2011)]. However, transforming to unweighted networks requires setting a threshold beneath which weaker grooming relationships are treated as absent (Voelkl et al. 2011). Such transformation requires setting arbitrary thresholds and excludes valuable information (Voelkl et al. 2011).

Frequencies can be used directly without controlling for observation time if observation time is equal (Altmann 1974). We recognize that in the case of the bonobos, time spent with each conspecific was unequal due to the fission–fusion management. However, in most cases (with the exception of keepers separating DO and BI), this was due to individuals’ own social choices. Controlling grooming data based on time spent in the same subgroup would weight proportionally higher grooming rates for dyads that spent little time together over dyads that chose to spend more time together.

We formatted grooming data from all-occurrence records for analysis in SOCPROG 2.7 (Whitehead 2009). Grooming was entered in dyadic mode where the initiator is entered first and the recipient was entered second. Each change in initiator/recipient was thus considered a new grooming bout. For mutual grooming, the bout was entered twice (A grooms B; B grooms A). For triadic grooming, each partnership was treated as a separate dyad (A grooms B, B grooms C). Across a large set of grooming networks, re-sampling networks based on small samples (< 100 records), were subject to greater disturbances than those based on large samples (< 2000: Voelkl et al. 2011). In dyadic mode, our sample of grooming events (bonobos: 1403 events; chimpanzees: 1103), meets the standards for a “reasonable” sample of greater than 1000 records (Voelkl et al. 2011). Supplementary data was included with individual ID code and sex. All data were analyzed as interactions, and sampling period was set as daily.

Grooming data were initially analyzed asymmetrically to assess reciprocity. Dietz’s R-test was used to assess absolute reciprocity and Hemelrijk’s Kr-test was used to assess relative reciprocity (Hemelrijk 1990a, 1990b). For these tests, the null hypothesis is that grooming is unidirectional, whereas the alternative hypothesis is that grooming is reciprocal. High correlations and low p values support the alternative hypothesis. One thousand permutations were run. Both bonobos (Dietz r = 0.751, p < 0.001; Hemelrijk r = 0.735, p < 0.001; N = 11) and chimpanzees (Dietz r = 0.737, p < 0.001; Hemelrijk r = 0.503, p < 0.001; N = 8) exhibited grooming reciprocity. Because grooming was reciprocal for both sexes, all further analyses were analyzed symmetrically.

Mantel tests were used to assess if grooming rates between classes (male–female) and within classes (female–female, male–male) were similar (Schnell et al. 1985). The null hypothesis is that there are no differences in grooming rates between and within classes, and is supported if 0.025 < p < 0.975 for two-tailed tests (Whitehead 2009; Pace et al. 2012). If within-class grooming rates are higher, t will be positive and p > 0.975 (Whitehead 2009). If between-class grooming rates are higher, t will be negative and p < 0.025.

Network analysis statistics were generated for each individual (Wey et al. 2008; Whitehead 2009). We chose to focus on three network measures: eigenvector centrality, strength, and clustering coefficient. Eigenvector centrality is a measure how well associated the actor and their associates are associated (Newman 2004; Whitehead 2009), thus providing information about the actor’s position in the social network. Strength is the sum of the associations of the actor with other individuals (Barrat et al. 2004; Whitehead 2009) and thus provides information about the actor’s individual social connection to conspecifics. Clustering coefficient indicates how well the actor’s associates are themselves associated (Holme et al. 2007; Whitehead 2009), thus indicating the actor’s propensity to associate within a clique. Next, a sociogram based on these social networks was generated within SOCPROG, with modifications to differentiate male–male, male–female, and female–female dyads.

Hemelrijk’s r test (Hemelrijk 1990a, b) was used to conduct matrix correlations between grooming relationships and kinship. An unweighted kinship matrix was generated with ‘0’ signifying no relationship, and ‘1’ signifying a kin relationship, and correlated with a weighted matrix of grooming interactions. Matrix correlations were conducted in SOCPROG with 1000 permutations, with two-tailed p values. Alpha was set at p < 0.05.

Statistical analysis

For all tests conducted within SOCPROG, hypothesis testing and alpha was set according to the recommendations provided by Whitehead (2009). Additional analyses for within-species comparisons of social network statistics were conducted in SPSS 18. Two-tailed t tests and Levene’s tests for equal variances were conducted for within-species comparisons of sex differences and wild vs. captive origin. All means are presented as mean + SE. Additionally, Pearson’s correlations were conducted to test the relationships between group tenure and social network statistics. All tests were two-tailed and alpha was set at p < 0.05. To avoid reducing statistical power, instead of conducting Bonferroni adjustments for multiple testing, we instead provide a measure of effect size (Cohen’s D) for t tests to guide interpretations (Nakagawa 2004). We suggest following the standard interpretation of d = 0.2 as small effect, d = 0.5 as a medium effect, and d = 0.8 as large effect (Cohen 1988).

Results

Grooming between and within sexes

Bonobos did not differ in dyadic grooming between or within the sexes (t = − 0.070, r = − 0.009, two-tailed p = 0.942, Fig.  3 ). Chimpanzees also did not differ in dyadic grooming between or within sexes (t = − 0.495, r = − 0.0939, two-tailed p = 0.6205; Fig. 4).

Fig. 3
figure 3

Bonobo sociogram. Females (SU, AN, GI, LA, UN) are in red, males (DO, GA, BI, MA, TO) are in blue. Female–female relationships are indicated by red lines, female–male relationships are indicated by purple lines, and male–male relationships are indicated by blue lines. Line thickness indicates dyadic grooming rates

Full size image
Fig. 4
figure 4

Chimpanzee sociogram. Females (AM, MA, RU, RY, GE) are in red, males (SO, LA, JO) are in blue. Female–female relationships are indicated by red lines, female–male relationships are indicated by purple lines, and male–male relationships are indicated by blue lines. Line thickness indicates dyadic grooming rates

Full size image

Sex differences in social network position

Bonobos did not exhibit sex differences in eigenvector centrality (female mean 0.206 ± 0.052; males mean 0.312 ± 0.073; t = − 1.132, df = 9, p = 0.287; d = − 0.701; Fig. 5). Variances did not significantly differ between the sexes (F = 0.988; p = 0.362). Bonobos also did not exhibit sex differences in strength (female mean, 224.00 ± 23.13; male mean, 281.00 ± 43.75; t = − 1.082; df = 9, p = 0.307; d = − 0.677: Fig. 6). Variances did not significantly differ between the sexes (F = 2.167; p = 0.175). Finally, bonobos did not exhibit sex differences in clustering (female mean = 0.118 ± 0.035; male mean = 0.108 ± 0.028; t = 0.219; df = 9; p = 0.831; d = 0.132: Fig. 7). Variances did not significantly differ between the sexes (F = 0.487; p = 0.503).

Fig. 5
figure 5

Eigenvector centrality in captive bonobos at the Columbus Zoo. Females are red, males are blue. Patterning indicates wild-born individuals, whereas solid bars indicate captive-born individuals. Individual and sex are indicated below each column

Full size image
Fig. 6
figure 6

Strength in captive bonobos at the Columbus Zoo. Females are red, males are blue. Patterning indicates wild-born individuals, whereas solid bars indicate captive-born individuals. Individual and sex are indicated below each column

Full size image
Fig. 7
figure 7

Clustering coefficient in captive bonobos at the Columbus Zoo. Females are red, males are blue. Patterning indicates wild-born individuals, whereas solid bars indicate captive-born individuals. Individual and sex are indicated below each column

Full size image

Chimpanzees did not exhibit sex differences in eigenvector centrality (female mean = 0.290 ± 0.081; male mean = 0.313; t = − 0.147, df = 6; p = 0.866; d = 0.100 Fig. 8). Variances did not significantly differ between the sexes (F = 2.553; p = 0.161). Chimpanzees also did not exhibit sex differences in strength (female mean = 275.70 ± 61.49; male mean = 152.86; t = -0.176, df = 6; p = 0.888; d = 0.116; Fig. 9). Variances did not significantly differ between the sexes (F = 1.1033; p = 0.334). Chimpanzees did not exhibit sex differences in clustering (female mean 0.1700 ± 0.021; male mean 0.1733 ± 0.054; t = − 0.069, df = 6, p = 0.947; d = 0.045; Fig. 10). Variances did not significantly differ between the sexes (F = 2.116, p = 0.196).

Fig. 8
figure 8

Eigenvector centrality for chimpanzees at the North Carolina Zoo. Females are red, males are blue. Patterning indicates wild-born individuals, whereas solid bars indicate captive-born individuals. Individual and sex are indicated below each column

Full size image
Fig. 9
figure 9

Strength for chimpanzees at the North Carolina Zoo. Females are red, males are blue. Patterning indicates wild-born individuals, whereas solid bars indicate captive-born individuals. Individual and sex are indicated below each column

Full size image
Fig. 10
figure 10

Clustering coefficient for chimpanzees at the North Carolina Zoo. Females are red, males are blue. Patterning indicates wild-born individuals, whereas solid bars indicate captive-born individuals. Individual and sex are indicated below each column

Full size image

Wild versus captive origin

Wild-born bonobos exhibited significantly greater values for eigenvector centrality (wild-born mean 0.3925 ± 0.09105; captive-mean 0.1900 ± 0.0310; t = − 2.592, df = 9, p = 0.29, d = 1.431; Fig. 5) and strength (wild-born mean 324.25 ± 55.08; captive-born mean 215.57 ± 15.82; t = − 2.401; df = 9, p = 0.040; d = 1.304; Fig. 6). No significant differences were observed in clustering between wild-born and captive-born bonobos, though the p value is marginally above our significance threshold (wild-born mean 0.1650 ± 0.0375; captive-born mean: 0.0829 ± 0.0182; t = − 2.239; df = 9 p = 0.052; d = 1.301, Fig. 7). Variances did not significantly vary between captive-born and wild-born individuals (centrality F = 1.897; p = 0.202; strength: F = 2.953, p = 0.120; clustering F = 1.908; p = 0.201).

Captive chimpanzees did not exhibit significant differences in centrality (wild-born mean 0.2700 ± 0.1700; captive-born mean 0.3083 ± 0.0868; t = 0.216, df = 6, p = 0.836; d = 0.284, Fig. 8), strength (wild-born mean 272.50 ± 173.50; captive-born mean 288.83 ± 72.19, t = 0.105, df = 6, p = 0.920; d = 0.076, Fig. 9), or clustering (wild-born mean 0.1600 ± 0.060; captive-born mean 0.1750 ± 0.2513; t = 0.278, df = 6, p = 0.790, d = 0.202; Fig. 10). No differences in variances were found in centrality (F = 0.015, p = 0.906) strength (F = 0.294, p = 0.607) or clustering (F = 0.298, p = 0.605).

Group tenure and kinship

In bonobos, kinship was not significantly correlated with grooming relationships (Hemelrijk’s r = 0.180, p = 0.209, N = 11). Kinship was also not significantly correlated with grooming relationship in chimpanzees (Hemelrijk’s r = 0.274, p = 0.160, N = 8).

In bonobos, group tenure significantly correlated with strength (Pearson’s r = 0.608; N = 11, p − 0 = 0.47) but not centrality (r = 0.493, N = 11, p = 0.123) or clustering (Pearson’s r = 0.488, N = 11, p = 0.128). In chimpanzees, group tenure did not significantly correlate with centrality (Pearson’s r = − 0.411, N = 8, p = 0.312), strength (Pearson’s r = − 0.272, N = 8, p = 0.514, or clustering (Pearson’s r = − 0.108, N = 9, p = 0.798).

Discussion

We hypothesized that male–female dyads (between-sex grooming) would have stronger grooming relationships than male–male or female–female dyads (within-sex grooming), and that there would be no sex differences in eigenvector centrality, strength, or clustering coefficients. Our hypotheses were partially supported. Contrary to our predictions, neither species exhibited differences in grooming relationships between or within the sexes. However, in accordance with our predictions, neither species exhibited sex differences in eigenvector centrality, strength, or clustering coefficients. However, in bonobos, wild-born individuals exhibited significantly greater centrality and strength, and group tenure was significantly correlated with strength. In chimpanzees, social network positions did not vary with the idiosyncratic factors we examined. These results suggest that when the foraging constraints that shape wild chimpanzees’ social lives are removed, captive chimpanzee and bonobo grooming relationships lack the sex differences expected in the wild. Both species exhibit variation in grooming dyads that may be better explained by individual variation than sex. These findings suggest that flexibility for social networks may have evolved in conjunction with fission–fusion social dynamics. However, the patterns that emerge in captivity may be structured based on abnormal histories, and this should be considered when testing evolutionary hypotheses.

The impact of idiosyncratic factors on social networks

Idiosyncratic factors may play a crucial role in structuring social relationships, particularly in captive environments. We found that wild-born bonobos had strong social network positions, and that group tenure was associated with strength, indicating that these individuals serve as the social core of the group. Our results are similar to those found in a previous study of captive chimpanzees (Leve et al. 2016). However, it is unclear whether these findings emerged because wild-born individuals have greater social competence than captive-born individuals. If so, this suggests the artificial nature of captivity alters the development of chimpanzee sociality in ways that qualitatively differ from wild Pan sociality. Alternatively, wild-born individuals may hold social positions analogous to resident wild animals. In our study, the wild-born individuals were also the oldest, and had resided in the group the longest. Thus, we cannot truly separate out wild-born identity from group tenure or age.

Wild-born individuals may play a crucial role in captive social networks, as theoretical experiments indicate that removing wild-born individuals has a larger impact on social cohesion than removing captive-born individuals (Leve et al. 2016). However, why this occurs is unclear. In a study of captive chimpanzees, social network statistics between wild-born and captive-born individuals were similar, suggesting that the impact of origin on social networks may be subtle (Leve et al. 2016). Research on great ape re-introductions suggests that wild-born individuals have the best reintroduction outcomes, presumably due to greater ecological competence due to early social learning (Beck et al. 2007). However, early socialization in wild-born individuals also provides a social advantage in terms of learning species-typical social behaviors. In our study, the bonobos with the highest eigenvector centrality and strength were all wild-born individuals. Previous research in this bonobo population indicates that wild-born individuals engage in less hair-plucking, a socially learned behavior that may indicate social stress (Brand and Marchant 2015). While the reason for this is unclear, this supports the hypothesis that wild-born individuals differ behaviorally from captive-born individuals. In our study, wild-born chimpanzees were not more central than captive-born individuals. However, the chimpanzee group contained fewer wild-born individuals, as well as individuals with mixed rearing histories such as peer-raising with later social integration. In the bonobos, group tenure cannot be separated from origin, because the original group residents were the wild-born individuals. Additionally, these individuals were also the oldest. Given that wild-born individuals are likely to be the oldest in captive groups, the influence of wild-born individuals on captive social networks may be due to age and residence. In wild populations, males are resident their entire lives, and older resident females may wield social influence (Wrangham 1980; Goodall 1986; Nishida 1988; Furuichi 1997; White 1998; Riedel et al. 2011; Hashimoto and Furuichi 2015; Moscovice et al. 2017). It is possible that in wild groups, the central position of wild-born individuals is due to occupying an analogous position combination of age and residence.

Within captive populations, developmental differences based on social experience, such as origin and rearing history, may be influential. Chimpanzees with a high degree of exposure to human interaction in early rearing and limited exposure to other chimpanzees exhibit lower rates of grooming and sexual behaviors (Freeman and Ross 2014). Such atypical rearing histories can directly influence personality, as chimpanzees with limited exposure to conspecifics early in life are less extraverted (Freeman et al. 2016). In this study, all captive-born bonobos were mother-reared, while captive-born chimpanzees had mixed rearing histories. These prevalent differences in early life experience contributes to the methodological challenges in studying sociality across captive species. Additionally, there were other idiosyncratic factors that we were not able to test. Research on captive chimpanzees indicates that chimpanzees with similar personality traits preferentially associate (Massen and Koski 2014). Furthermore, personality differences may affect social networks both indirectly. Personality differences may drive individual friendships and individual patterns in affiliation and activity level (Massen and Koski 2014; Martin and Suarez 2017). In addition to structuring close dyadic relationships, individuals high in agreeableness, openness, or extraversion may be more likely to become central in grooming networks.

Fission–fusion dynamics and ecological constraints

In wild contexts, ecological constraints are the most crucial factor in structuring social networks. In the wild, chimpanzees exhibit sex-segregated association patterns, with females ranging in smaller subgroups, range over smaller distances, and often remain within core areas (Wrangham 1980; Symington 1990; Chapman et al. 1995; Matsumoto-Oda et al. 1998; Itoh and Nishida 2007; Kahlenberg et al. 2008; Riedel et al. 2011). The constraints of scramble competition, in conjunction with the need to protect and accommodate offspring, may limit female chimpanzee’s social opportunities (Chapman et al. 1995; Otali and Gilchrist 2005; Pontzer and Wrangham 2006). However, there is nonetheless variation in ecological constraints across chimpanzee populations (Lehmann and Boesch 2005, 2009; Pruetz and Bertolani 2009; Hashimoto and Furuichi 2015). Female West African chimpanzees are generally more gregarious, and low-ranking females only decrease subgroup size when fruit is scarce (Lehmann and Boesch 2005, 2008; Riedel et al. 2011). In the extreme environment of Fongoli, Senegal, the community travels more cohesively, and subgroup sizes are larger than those typical at forested sites, although they decrease during the dry season (Pruetz and Bertolani 2009). Contrastingly, terrestrial herbaceous vegetation in bonobo habitats facilitates larger, mixed-sex subgroups (White and Wrangham 1988; Furuichi 1997; Mulavwa et al. 2008). Thus, wild female chimpanzees may be constrained in their ability to form and maintain social bonds compared to wild bonobos.

The lack of assortative grooming based on sex compared to patterns in wild populations supports our hypothesis that when ecological constraints are removed, both Pan species exhibit plasticity in social bonding. Although individual histories may be altered by captive management practices, behavioral plasticity may factor into the propensity for social flexibility inherent in social systems with high levels of fission–fusion social dynamics. Data on social networks in wild Grevy’s zebras and onagers suggests the flexibility of fission–fusion social dynamics facilitates rapid responses to changing environments (Gersick and Rubenstein 2017). Flexible subgrouping patterns facilitate rapid adjustment to environmental challenges, includes anthropogenic pressures (Hockings et al. 2012; Schaffner et al. 2012; Rodrigues 2017). In captive environments, species with high fission–fusion dynamics may flexibly respond to captive management changes through social means.

Limitations

Our study’s conclusions may be limited by small sample sizes and variability of rearing history common to captive studies. However, these conclusions could be strengthened by further studies incorporating more captive chimpanzee and bonobo groups. Grooming data is frequently collected in captive studies (Kanngiesser et al. 2011; Sueur et al. 2011; Leve et al. 2016), but different methodologies may limit direct comparisons. However, the issues of variable histories may be unavoidable in studying captive groups. To understand how rearing history impacts social networks, more research should focus on identifying how to control for these confounds. Additionally, differences in housing, diet, and zoo protocol could create micro-environmental variation that affect social behavior. This is especially true for differences in husbandry and housing that facilitate or limit fission–fusion dynamics. Because the bonobos in this study were managed in a fission–fusion setting, while the chimpanzees were not, this may have differentially affected group dynamics. Despite these limitations, captive environments are the best option for testing questions about the propensity for species-typical differences in the absence of the ecological constraints, although the influence of captive management must be considered. Such research complements data from wild populations across varying ecological contexts.

Conclusions

Although traditional models of Pan socioecology emphasize the role in sex in structuring in social bonds, we found that male and female chimpanzees and bonobos exhibit similarities in grooming networks in captive settings. Our findings indicate that idiosyncratic factors, such as wild origin and group tenure, play a greater role than sex in structuring captive grooming networks. Such individual factors may be due to the abnormal nature of captivity; however, in captive populations wild origin and group tenure may be analogous to residence in wild populations. Our findings suggest that in the absence of ecological constraints, captive chimpanzees and bonobos exhibit behavioral plasticity in the formation of social bonds. Such plasticity may be part of the social flexibility inherent in species with high degree of fission–fusion dynamics retained from the common Pan ancestor. Although results from small, captive populations should be considered with caution, these findings illustrate how research on captive social networks can reveal dynamics regulating social bonding and behavioral plasticity in closely related species.