Group-specific expressions of co-feeding tolerance in bonobos and chimpanzees preclude dichotomous species generalizations

Summary Bonobos are typically portrayed as more socially tolerant than chimpanzees, yet the current evidence supporting such a species-level categorization is equivocal. Here, we used validated group-level co-feeding assays to systematically test expressions of social tolerance in sixteen groups of zoo- and sanctuary-housed bonobos and chimpanzees. We found that co-feeding tolerance substantially overlaps between the species, thus precluding categorical inference at the species level. Instead, marked differences were observed between groups, with some bonobo communities exhibiting higher social tolerance than chimpanzee communities, and vice versa. Moreover, considerable intergroup variation was found within species living in the same environment, which attests to Pan’s behavioral flexibility. Lastly, chimpanzees showed more tolerance in male-skewed communities, whereas bonobos responded less pronounced to sex-ratio variation. We conclude that the pervasive dichotomy between the tolerant bonobo and the belligerent chimpanzee requires quantitative nuance, and that accurate phylogenetic tracing of (human) social behavior warrants estimations of intraspecific group variation.


INTRODUCTION
Social tolerance-defined as the propensity to be in enduring close proximity to conspecifics in the presence of valuable resources 1,2 -lies at the heart of many fitness-affecting behaviors in both humans and other animals.4][5][6][7] As such, high levels of social tolerance have been suggested to be a defining characteristic of the human species, to the extent that humans have been depicted as ''ultra-social'' 8 and ''hyper-cooperative.'' 9 Considering chimpanzees' (Pan troglodytes) and bonobos' (Pan paniscus) close phylogenetic relatedness to humans and each other, 10 one could expect to find considerable behavioral similarities with humans and between these sibling species.6][17][18] While some have questioned this apparent dichotomy between the two Pan species, 19,20 most studies focus on their behavioral differences rather than addressing the ranges of behavioral overlap. 21,22n the realm of social tolerance, the dichotomy between bonobos and chimpanzees has been explicitly postulated and documented.Social tolerance is thought to enable bonobos to peacefully co-feed and cooperate, [21][22][23][24][25][26][27] while chimpanzees' intolerance prevents fruitful interactions and cooperation. 6,14,28,29However, these postulations are biased given that many studies compare single groups of the two species to draw general conclusions about interspecies differences and their impact on our understanding of human evolution. 19,30][36][37]

OPEN ACCESS
We compare several groups of each species in two different environments: zoos and sanctuaries.Traditionally, comparative research on apes has occurred in zoos where conditions are less natural yet arguably more similar between the two species than in the wild.For instance, across zoo settings, there may be more similarity in diet, food abundance, predation risk (absent), and territory size than in the wild.In sanctuaries, husbandry conditions can be considered less artificial compared to zoo conditions, as the apes typically live in larger groups, in more naturalistic and larger enclosures.Moreover, often several groups of the same species can be studied in the same location under similar withininstitution husbandry management. 38Apes may differ behaviorally between captive and sanctuary settings due to typically more natural living conditions and less visitor effects in sanctuaries compared to zoos. 38Hence, both settings are complementary relevant to assess species and/ or group variation in co-feeding tolerance.
For our study, we focus on two behavioral measures of co-feeding tolerance that can be experimentally tested at the group level in a standardized fashion: (1) the proportion of group members co-feeding in a pre-defined food zone in which the space measurements are controlled for group size 3 and, (2) the temporal fluctuation of this proportion, given that resource availability and related levels of competition are influenced by the co-feeding dynamics in previous time points. 37Where the first measure reflects an overall level of co-feeding tolerance, the second measure captures the speed by which group members choose to abandon the competitive context.We test the hypothesis that co-feeding tolerance in bonobos and chimpanzees does not differ at the species level, but rather at the group level.To identify possible drivers of group-level variation in co-feeding tolerance, we assessed the influences of known determinants of sociality in primates, namely group size, genetic relatedness, and group-averaged age. 39,40Lastly, we tested the effect of the groups' male/female ratio on their respective co-feeding tolerance, as bonobos are matriarchal while chimpanzees are patriarchal. 41
With increasing resource depletion (represented by scan number), however, their co-feeding tolerance progressed in different ways (LRT species|scan interaction: c 2 = 12.49, df = 1, p < 0.001), which, upon visual inspection, seemed mainly due to differences between the species in the zoo settings rather than in the sanctuary settings (Figure 2).Lastly, co-feeding tolerance was generally higher for the Pan species in the sanctuaries compared to the zoo setting (LRT: c 2 = 10.97,df = 1, p < 0.001; estimate GSE = À1.21G 0.30; 95% CI: À1.85 to À0.57), although this effect was not obviously different for the two species (see Figure 1).Note, here, that due to logistical constraints, different versions of the co-feeding tolerance test were used in the zoos (pasta plot) compared to the sanctuaries (peanut swing), 37 which precludes a direct interpretation in terms of Pan behaving differently in the respective settings (see STAR methods).

Between-group variation
At the group level, the Pan communities differed substantially from each other in their co-feeding propensities (LRT on ''Group'': c 2 = 81.53,df = 1, p < 0.001; Figure 3).Similarly, the progression of co-feeding tolerance over experimental time was highly group specific (LRT group|scan interaction: c 2 = 226.45,df = 15, p < 0.001; Figure 4).Importantly, in these group-level models (see STAR methods), ''species'' did not have any additional explanatory power (c 2 = 0, df = 1, p = 1).Furthermore, while controlling for the following group-level metrics: group size, average relatedness, age, and sex ratio (see below), ''group'' remained a highly important factor explaining variation in co-feeding tolerance (c 2 = 57.25,df = 1, p < 0.001).The group-level metrics themselves did not significantly affect co-feeding tolerance (all p > 0.14), except for ''sex ratio,'' which affected co-feeding tolerance differentially for bonobos and chimpanzees (LRT interaction: c 2 = 10.34,df = 1, p < 0.002).Co-feeding tolerance increased more for chimpanzees than for bonobos with a higher proportion of males in the group (Figure 5).Caution with interpreting the group-level effects is warranted, however, given the relatively large ranges of respective estimates resulting from the model stability check (Figure S1; see STAR methods).
Table S1 (zoo settings) and Table S2 (sanctuaries) provide statistical details on all group contrasts (separate for the two settings in light of the significant main effect of ''setting'').Interestingly, in some facilities and sanctuaries where several Pan groups were housed, we found substantial within-species differences in co-feeding tolerance despite their similar facility setups, husbandry styles, and other potentially latent determinants like the weather and (the absence of) visitors (see supplementary materials).

DISCUSSION
By systematically comparing a relatively large sample of independent Pan groups in two settings (zoos and sanctuaries), we show that cofeeding tolerance is not universally more pronounced in bonobos compared to chimpanzees, 14 but is in fact a group-specific characteristic in the Pan species.The Pan groups in our sample differed substantially with respect to how many group members could benefit from food resources simultaneously.Moreover, groups differed in their strategies of co-feeding: whereas in some groups the apes were only together for one or two scans, others were characterized by continuous co-feeding until the food was depleted.These results put earlier findings on single (or fewer) groups into perspective: bonobos as a species are neither more tolerant with respect to co-feeding than chimpanzees ( 14 cf; 42,43 ), nor does the opposite view hold. 35,446][47] Here, we explicitly note that we have tested captive Pan populations (in zoos and sanctuaries) and that we do not know to what extent our findings would generalize to their wild counterparts.A recent study in lemurs 48 found higher levels of co-feeding tolerance in captive versus wild groups, whereas another study reported the opposite pattern for common marmosets. 49Given that a complex interplay between ecological and social conditions (e.g., food availability and magnitudes of interdependence 50,51 ), including potential cultural differences, 52,53 likely shapes wild Pan's expressions of co-feeding tolerance, this question remains outstanding for future research.
To investigate which mechanisms might be involved in creating the observed group differences in co-feeding tolerance, we assessed the effect of a group-level index central to the social organization of the Pan species: sex ratio. 41The proportion of males in the group was more positively related to co-feeding tolerance in chimpanzees than bonobos, which may be explained by the male-biased dominance system in chimpanzees. 54Overall, though, there may be more socio-demographic factors (e.g., number of immatures) that could affect co-feeding tolerance in the Pan species. 55However, the sample size at the group level (n = 16) did not allow for incorporating more variables to estimate, and thus for identifying robust group-level determinants of co-feeding tolerance, which is a key avenue to explore in the future. 30We also note that our conclusions are limited to the co-feeding contexts created by our paradigms-our findings neither negate nor preclude the possibility of species differences in other measures of tolerance, e.g., in the context of intergroup encounters. 16,18As such, our results are specifically relevant for revealing the breadth of co-feeding tolerance in bonobos and chimpanzees, including their overlap, 35 which warrants further systematic studies in other domains of social tolerance. 3t the proximate level, our findings are relevant for understanding the possible causes of behavior in the Pan species more generally.If cofeeding tolerance reflects sociality in terms of daily interaction patterns more generally (see Cronin et al. and van Leeuwen et al. 34,36 ), then other social behavior like social learning, cooperation, or resource sharing could substantially correlate with tolerance (e.g., see van Boekholt et al. and Cantor and Whitehead 56,57 ).In agreement with this, a recent study found marked differences in expressions of prosociality between groups of chimpanzees living in the same sanctuary, with the groups highest in prosociality also being the highest in co-feeding tolerance. 36imilarly, in Japanese macaques, substantial group differences in social tolerance were found, 58 which positively correlated with cooperative success on an experimental task. 5Such intergroup variation in sociality has been documented in other species as well, for instance, vervet monkeys, 33,59 meerkats, 60 and whales. 61These indications of inter-group variation in social behavior caution against the use of single group studies for broad species inference. 30For instance, finding that one group of chimpanzees behaves relatively indifferent to benefitting others in a group context, 9 bears little relevance to characterizing their species-specific ways of behaving when other groups seem relatively prosocial with a similar experimental design. 36Rather, the aggregation of these (independent) tests could cumulatively paint a picture of the species-specific range of behaving (also see DeTroy et al. 55 ), which in turn could be used to compare against ranges of other species, like in the current study.
At the ultimate level, our findings point toward group-specific social dynamics as a main determinant of social tolerance.The proportion of group members that could benefit from the valuable resources, including their progression over time, varied substantially from group to group, which indicates that in both bonobos and chimpanzees the local group climate determines in large part how resources will distribute over the group members. 55Tolerance is likely to be a result of group living, 62 with the magnitude of its expression being shaped by the fitness benefits of the behaviors it facilitates, like cooperation 16 and social learning. 4Such feedback loops may be influenced by (the threat of) intergroup competition, spurring increased social tolerance, affiliation, and cooperation within groups. 63,64As such, it would have been valid to assume that chimpanzees and not bonobos show higher levels of social tolerance, given that their evolutionary environment of adaptation most likely comprised higher levels of intergroup conflict. 16While such intergroup encounters turn out to be markedly more peaceful in bonobos compared to chimpanzees, 12,16 the findings of the current study reveal similar underlying propensities with respect to tolerance around valuable resources when environmental conditions (e.g., food availability and distribution) and possible effects of group-level metrics (e.g., group size, sex ratio) 65 are controlled for.This inconsistency may further attest to Pan's large behavioral flexibility to adapt to various environmental and social circumstances. 66,67aken together, the current study shows that bonobos and chimpanzees in zoo and sanctuary settings behave tolerantly toward group members in a group-specific, rather than a species-specific manner.Future multi-group research is warranted to test whether this finding extends beyond tolerance in the competitive context of limited food resources, as measured in the current study, to other contexts in which tolerance could play a pivotal role, like during daily interaction patterns (e.g., for social learning, cooperation, or reconciliation) and/or intergroup encounters.In conjunction, such assessments can identify the scope of adaptive potential within the panins and may provide leads as to the selection pressures shaping the form and function of group-level interaction styles for panins and possibly also for hominins.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:   Pan groups differ substantially in their co-feeding strategies (y axis) when resources are depleting over experimental time (x axis).Some groups stably co-fed with a relatively low proportion of group members in the resource zone until the resources were depleted (e.g., Bonobo_F1 and Chimp_C1), whereas others co-fed with almost the entire group from the beginning after which the proportion decreased steadily (e.g., Chimp_C4).Filled circles represent the predicted values (y axis); the error bars represent the 95% confidence intervals.The colors are group specific and match the colors for the same groups in Figure 3.

Figure 1 .
Figure 1.Co-feeding for the Pan species across the different settings Bonobos and chimpanzees (x axis) do not differ in their co-feeding propensities measured as the proportion of group members present during a standardized cofeeding experiment.The left panel depicts the sanctuary apes, the right panel the zoo apes.Filled circles represent the predicted values (y axis); the error bars represent the 95% confidence intervals.

Figure 2 .
Figure 2. Co-feeding progression over time for both Pan speciesIn zoo settings (right panel), bonobos and chimpanzees differ in their co-feeding strategies (y axis) when resources are depleting over experimental time (x axis), whereas in sanctuary settings (left panel), their strategies do not obviously differ.Filled circles represent the predicted values; the error bars represent the 95% confidence intervals.

Figure 3 .
Figure 3. Predicted co-feeding proportions per group Co-feeding tolerance is group specific in the Pan species in sanctuary settings (upper panel) and zoo settings (lower panel).Bonobos are depicted in the left panels, chimpanzees on the right.Filled circles represent the predicted values (y axis); the error bars represent the 95% confidence intervals.

Figure 4 .
Figure 4. Predicted co-feeding progressions over time per group Pan groups differ substantially in their co-feeding strategies (y axis) when resources are depleting over experimental time (x axis).Some groups stably co-fed with a relatively low proportion of group members in the resource zone until the resources were depleted (e.g., Bonobo_F1 and Chimp_C1), whereas others co-fed with almost the entire group from the beginning after which the proportion decreased steadily (e.g., Chimp_C4).Filled circles represent the predicted values (y axis); the error bars represent the 95% confidence intervals.The colors are group specific and match the colors for the same groups in Figure3.

d
EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS d METHOD DETAILS B Experimental measures B Experimental procedure and coding d QUANTIFICATION AND STATISTICAL ANALYSIS

TABLE
d RESOURCE AVAILABILITY B Lead contact B Materials availability B Data and code availability