Adult-adult social play in captive chimpanzees: is it indicative of positive animal welfare?

How individuals form social relationships with others is important for the welfare of both human and non-human animals. Studies have shown that close bonds increase the survival rate of infants and lead to longevity (Silk et al., 2003, Silk et al., 2010). A recent study of wild chimpanzees in Budongo forest, Uganda, found that grooming with bonding partners can have a stress reducing effect, as observed in the reduction of the urinary glucocorticoids level after grooming (Wittig et al., 2016). However, social relationships can sometimes have an opposite effect on animal welfare. Increasing social stress is one obvious example. Studies have reported an increase of physiological stress indicators after aggression and with increasing group size in wild chimpanzees (Markham et al., 2013, Wittig et al., 2015). Our previous studies showed that the hair cortisol level (as a physiological index of long-term stress) was higher in captive chimpanzees who received a higher level of aggression, suggesting that social situation affected their stress levels over long periods (Yamanashi et al., 2013, Yamanashi et al., 2016a). In wild populations of primates, animals have various tactics to reduce the costs of group living. One of these is fission. Another coping strategy is reducing tension and the risk of aggression by engaging in certain affiliative behaviours, such as social grooming and play. Especially in captive situations, where there is not much space for fission, coping by means of this active strategy is important. It has been observed that primates avoid aggression by engaging in affiliative behaviours. Studies have shown that the affiliative behaviours among great apes increased in spatial crowding situations (de Waal, 1989, Nieuwenhuijsen and de Waal, 1982, Ross et al., 2010, Videan and Fritz, 2007).

There are several types of affiliative social behaviours. Among those, social grooming is the most ubiquitous form of affiliative behaviour. Social grooming is considered to be used as a means to establish and maintain social bonds as well as to maintain hygiene. Studies have shown a correlation between grooming interchange and support for agonistic interactions in captive chimpanzees (Hemelrijk and Ek, 1991). Mitani (2009) reported that grooming reciprocity occurred between maternal brothers and dyads with strong social bonds. Studies of other primates likewise reported a link between social grooming and affiliative relationships, such as close kin, bond duration and co-feeding (Silk et al., 2006, Ventura et al., 2006). However, compared to the wealth of knowledge concerning social grooming, studies of social play, especially among adults, are scarce. This is because adult–adult social play is rare in non-human animals. For example, Matsusaka (2004) reported that, in a study of wild chimpanzees in Mahale Mountain National Park, Tanzania, only 12 out of 793 bouts of social play occurred between adults, whereas the other 781 bouts occurred including immature individuals. Although many previous studies classify social play and social grooming in the same category of affiliative behaviours, it is not clear whether these two behaviours can be considered identical. It is known that bonobos, a species that is closely related to chimpanzees, are more playful than chimpanzees. Palagi (2006) reported that the rate of social play of adult bonobos increased before feeding rather than control periods. Additionally, Palagi (2006) found that the rate of social play and co-feeding was positively correlated. Tension among group members can be high during the pre-feeding period due to the anticipation for food. Therefore, she concluded that social play between adult bonobos is related to reduce tension between affiliative pairs and hence to reduce aggression. However, the social traits of chimpanzees and bonobos are very different (Hare et al., 2012), and some studies of other mammals failed to find a relation between affiliative relationships and social play (Cordoni, 2009, Sharpe, 2005, Sharpe and Cherry, 2003). Therefore, it is important to understand the details of social play among adults and its relevance to other social behaviours.

Understanding what type of social relationship social play reflects is also important from the perspective of animal welfare assessment. Attention for positive animal welfare has been increasing, and play is sometimes considered as a positive indicator of welfare (Boissy et al., 2007, Held and Špinka, 2011). This is because play is often observed when animals are without chronic stress (Graham and Burghardt, 2010) and is often accompanied by signs of pleasure (Held and Špinka, 2011). In addition to such immediate benefits, play also has delayed benefits and is important for the socio-cognitive development of immature animals (Shimada and Sueur, 2014). A recent study showed that social play during juvenile periods correlated with future copulation behaviours in American minks (Ahloy Dallaire and Mason, 2017). Locomotor play facilitates motor skill acquisition in Assamese macaques (Berghänel et al., 2015). Therefore, increasing the level of play among immature animals can be important from the perspective of animal welfare. Nevertheless, it is controversial whether play, especially social play, which is the most prevalent form of play after maturation (Pellis and Iwaniuk, 2000), can be considered as a positive indicator of welfare (Blois-Heulin et al., 2015). Age, sex and species differences are often associated with the level of social play (Burghardt, 2005). In addition, social play of adult non-human primates is also frequent in contexts of heightened tension, such as pre-feeding time for bonobos (Palagi et al., 2006) and in contexts of group encounter for Verreaux’s sifaka (Antonacci et al., 2010). Furthermore, the above-mentioned study of immature Assamese macaques (Berghänel et al., 2015) also reported a negative correlation between physical development and locomotor play rate, which suggests energy-demanding aspects of play. Therefore, it is under debate how we should deal with social play from the perspective of animal welfare.

This study investigates the factors that influence social play in adult captive chimpanzees by using unique captive settings to discuss the function of social play. First, we describe the influence of age, sex and group formation on social play levels in chimpanzees living in mixed-sex and all-male groups. Then, we investigate the time distribution of social behaviours, the association among three social behaviours (social play, social grooming and aggressive interaction) in all-male groups of chimpanzees. We also examined the association between social play and abnormal and self-directed behaviours because these behaviours have been frequently used as indicators of animal welfare (Baker and Aureli, 1997, Duncan and Fraser, 1997). If social play used as a means to reduce social tensions among individuals, social play among adult chimpanzees can increase during times of heightened tension, such as pre-feeding time. Furthermore, we predicted that social play is higher in males than in females and can be increased in all-male groups with many adult males coexisting in a closed environment. If social play can be considered as a means to form and maintain social bond, social play can be an indicative of affiliative relationships similar to how mutual grooming was observed to be (Fedurek and Dunbar, 2009). We predicted that social play can be observed between dyads of affiliative social relationships with high levels of social grooming and low levels of aggressive interaction. Based on these findings, we will discuss the use of social play as an indicator of positive welfare.

The subjects are 37 adult chimpanzees (17 males and 20 females) living in Kumamoto Sanctuary, Wildlife Research Center, Kyoto University (KS). The subjects were divided into several social groups which include both all-male groups and mixed-sex groups (Table 1). In the case of males, we collected data from 11 all-male group chimpanzees and six mixed-sex chimpanzees. Two out of five mixed-sex groups have two males, while other groups have only one male in a group. However, one male from a mixed-sex group with two males moved to another zoo for breeding during the study (September 2014). Therefore, only data from the six males in the mixed-sex group was included. Three to six females lived with those males. The members of the all-male groups changed periodically as described in the Supplementary Table. All individuals were above nine years old, and the details of their profiles and social structures are supplied in the Supplementary 1. Males in the all-male groups of chimpanzees were biologically unrelated except for one dyad that had an uncle-nephew relationship.

Established in 2007, KS was the first chimpanzee sanctuary in Japan (it was renamed from Chimpanzee Sanctuary Uto (CSU) in 2011 when the institution was passed from Sanwa Kagaku Research Institute onto Kyoto University). For more information, see (Morimura et al., 2010). KS accommodates ex-laboratory chimpanzees and chimpanzees that are considered surplus in Japanese zoos. It promotes the social life of chimpanzees. Three types of social groups have formed within KS: all-male groups; one-male and multi-female groups; and multi-male, multi-female groups. The members of the all-male groups are changed periodically to provide social stimulation and prevent escalated aggression, especially directed toward immigrant individuals. All individuals had access to both indoor and outdoor enclosures, most of which are cage style (i.e. a building with a roof), although one outdoor enclosure (approximately 270 m² in area) has no roof. The outdoor cages range in size from approximately 70 m² in area and 5.4 m in height to about 120 m² in area and 12 m in height. All these outdoor cages are connected to other cages. During wintertime, some of the outdoor cages were covered with plastic sheeting so that the chimpanzees could avoid cold temperatures. Passages that connect several cages within KS were introduced, totalling 150 m in length, and groups of chimpanzees could access these passages for exploration in turn. They had free access to water at any time, and regular meals (consisting mainly of fruit, vegetables and monkey pellets) were provided three times per day. Additionally, routine feeding enrichment (e.g. juice feeders, puzzle feeders, browsing opportunities and food concealed in boxes or newspapers) were changed daily. Other types of environmental enrichment were also provided. For example, fire hoses, ropes, hammocks, climbing structures, and substrate materials were installed and natural vegetation was planted to increase the physical environment’s complexity. Moreover, spaces were available for the chimpanzees to escape from rain, strong sunlight and cold, and they were provided with comfortable bedding materials for day- and night-time sleep. Materials that they could manipulate freely were also provided, such as toys, buoys and sacks. For additional details on environmental enrichment, see (Kumamoto Sanctuary). In addition to the lifelong care of these chimpanzees and bonobos, non-invasive research (cognitive, behavioural, endocrinological and genetic) is conducted at KS (Kano et al., 2015, Krupenye et al., 2016, Morimura and Mori, 2010, Yamanashi et al., 2016a, Yamanashi et al., 2016b).

Data were collected between June and July 2014 (summer), and between December 2014 and March 2015 (winter), respectively. The total observation time was 367 h (229 h for the all-male groups and 138 h for the mixed-sex groups). All data were collected by YY, using the iOS application “ISBOapp” (Ogura, 2013) and a notebook to complement. Table 2 shows details of the social play patterns and definitions of behaviours. The definitions were based on Palagi and Paoli (2007), Goodall (1989), Baker and Aureli (1997), Birkett and Newton-Fisher (2011), Walsh et al. (1982) and Nishida et al. (2010). Between 9:00 h and 16:00 h, YY conducted 30 min of focal observation of 17 male chimpanzees in a randomly assigned order and recorded behaviours of the chimpanzees in their enclosures (Martin and Bateson, 2007). At least 20 h (20–21 h in all-male groups and 21.5–24.5 h in mixed-sex groups) of focal animal sampling were conducted for each male individual. During the observations, YY recorded social grooming, social play, self-directed behaviours, abnormal behaviours and other general behaviours (forage, rest, move, and other behaviours, not used for this study) every 30 s and all aggressive interactions and social play that occurred within the groups. Although we tried to quantify the ranks of the individuals by the direction of pant grunting (Nishida et al., 2010, Noe et al., 1980), we could identify only the two highest- and two lowest-ranking individuals (Yamanashi et al., under review). The group members directed pant grunts toward the alpha male, and the lowest-ranking individuals directed pant grunts toward several individuals. However, there were many blank relationships and other individuals did not pant-grunt at each other. Thus, we were unable to quantify the rank of every individual. YY stopped making focal observations and began recording aggressive interactions when a scream or play sound was heard or when there were any other signs of aggression or social play. YY recorded the aggressors and receivers of each aggressive interaction. To record social play, YY checked whether the same dyads continued social play 30 s after first noticing the start of the play and repeated to check after every 30 s until YY saw that the chimpanzees did not engage in social play at all, as social play often continued for relatively a long period. We considered the observed behaviours as two separate bouts if no social play was seen for the duration of two recording times and it started again from the next recording time. Social play seldom occurred in mixed-sex groups, hence it was often missed by the time-sampling approach. To compensate for this difficulty, we collected the social play data in two different ways (focal animal sampling and all-occurrence sampling), and as a result we were able to compare the rate of play both within and across groups and sexes. In order to check the validity of the data collected, we checked the matrix correlation between the social play data obtained by means of focal sampling and the data obtained by means of all-occurrence sampling in the all-male groups (11 males). We obtained a significant correlation between the two data sets (Goodman–Kruskal gamma = 0.800, p < 0.001, UCINET 6). All the data were collected by a single observer, therefore running the risk of some social play observations being overlooked. Nevertheless, this positive correlation implies that our observations were valid.

Between the all-male groups and mixed-sex groups, several differences were observed in husbandry routine in addition to group formation. Before 11:00 h every morning, 11 all-male group chimpanzees were divided into smaller groups that consisted of two to three individuals. At approximately 11:00 h, food was scattered in the enclosures and two social groups that consisted of four to seven individuals were formed until 15:00 h. The groups were fed evening meals and separated again into small social groups after 15:00 h. The members of each small social unit were changed daily. Data was collected between 11:00 h and 15:00 h, therefore observed behaviours from the all-male groups were recorded when the chimpanzees were divided into two social groups. The mixed-sex groups were also separated into smaller groups (one–seven individuals per group) during the night. Larger social groups, which consisted of all members of each social group, were formed in the morning. At approximately 13:00 h, food was scattered in the enclosures, and sometimes the enclosures used by each group would be changed afterwards. The groups were fed evening meals at approximately 15:30 h–16:00 h and separated again into small groups afterwards. The data were collected between 9:00 h and 16:00 h for the mixed-sex groups.

To analyse the effects of sex, age and group compositions on the level of social play, we used the data obtained from all-male and mixed-sex groups. To check the effects of sex on the occurrence of social play, we calculated the expected value for the occurrence of social play for each sex combination (male-male, male-female and female-female). The expected value was calculated by dividing the number of play occurrences by the number of possible sex combinations.

We focused on the all-male groups to analyse the timing of social behaviours and the relationship amongst three social behaviours. This was done because a sufficient amount of social play data from all-male group chimpanzees was obtained. We divided social grooming into two categories (mutual social grooming and unilateral social grooming) for analysis (Nishida et al., 2010). This is because previous studies often considered mutual grooming to be an indicator of a strong social bond, as mutual grooming requires both individuals’ active engagement in the behaviour (Fedurek and Dunbar, 2009), while unilateral grooming does not. We considered the differences and analysed each behaviour separately.

We used grooming equality index and the rate of mutual social grooming to check the relationship among social behaviours (social play, aggressive interactions and social grooming). Grooming equality index was calculated based on Mitani (2009) as follows;

where Gab is the amount of grooming that individual a gave to individual b, Gba is the amount of grooming that individual b gave to individual a, and Ga ⇔ Gb is the total amount of grooming between individual a and b. This measure’s strength lies in quantifying whether grooming between pairs of chimpanzees was balanced or skewed. Therefore, the lower the index, the more the direction of unilateral social grooming was skewed without reciprocation. The index ranges from 0 to 1. Further, we also used the rate of mutual grooming because the chimpanzees in our subject spent considerable time for mutual grooming and the grooming equality index overlook the level of mutual social grooming between pairs of animals. The dyad level rate of mutual grooming was calculated by summing the rate of mutual grooming in each individual focal time. The dyad-level rate of social interactions was used because we wanted to investigate the type of social relationship the social play implies.

We calculated one social matrix for each social behaviour by calculating the sum of the social interactions for each dyad. Because the group formation changed daily in the all-male groups, we adjusted the social matrices using the rate of the same groups by dividing the matrices by the total number of observational sessions of the same group. Missing values were accounted for using statistical procedures described in the next section. We did not discuss the relationship between individual rates of social behaviours because these rates can be affected by factors such as group size and individual situations. Instead, we focussed on the quality of relationship in this study.

We used Spearman’s rank correlation test to check the relationship between age and level of play. We analysed the effects of age separately for all-male group males, mixed-sex group males and females because the absolute levels of play differ significantly between group and sex (data shown in the result section). We used the ‘coin’ package to perform Spearman’s rank correlation test (Hothorn et al., 2006, Hothorn et al., 2008). We used a proportion test to test the differences in the number of social play and aggressive interactions among the different groups. We used Generalized Linear Mixed Model (GLMM) to test the differences in the rate of social play and social grooming between the different groups. To test the effects of the group, we included group ID as explanatory factor. We also included individual ID and season (summer or winter) as random factors, as there was substantial difference in the absolute level of play between different seasons (i.e. they played less in summer). We used the glmer function of the ‘lme4’ package and gamma distribution with inverse link function.

To test the effects of the time of day on the rate of social grooming and social play, we used a Generalized Linear Model. We included eight time-categories (periods of 30 min between 11:00 h and 15:00 h) and individual ID as explanatory factors. We subsequently used the glm function and gamma distribution with inverse link function. We did not use GLMM to avoid convergent errors. We compared the models with and without the explanatory parameters mentioned above based on the likelihood ratio test with approximate chi-squared distribution (Kubo, 2012). We conducted a test for equality of proportion to assess the differences in the frequency of aggressive interaction and social play across the day. Subsequently, we also conducted post-hoc pairwise comparison of proportions and p values were adjusted with the holm methods. We used a partial Kr matrix correlation tests to analyse the relationship between social behaviours (grooming, play and aggressive interactions). The socio-matrixes missed certain values, as two individuals had never been together with three individuals. Therefore, we used a partial Kr test to account for the missing values while using all available data (Hemelrijk, 1990a). The number of permutation was set at 10000. We further examined the correlation between social play and abnormal and self-directed behaviours using Spearman’s rank test. All statistical testing was conducted with R. 3.4.1 (R Development Core Team, 2011), except for the partial Kr test, which was analysed using a special ad-in created by Hemelrijk and downloaded through her website (Hemelrijk, 1990a, Hemelrijk, 1990b). The level of significance was set at α = 0.05 and we considered it to be marginally significant when 0.05 < α < 0.1.

We found a significant positive correlation between age and rate of play in all-male group chimpanzees (Fig. 1: Z = 2.40, n = 11, p = 0.0165). We did not find such a correlation in mixed-sex males (Z = −0.130, n = 6, p = 0.897) or females (Z = 1.15, n = 21, p = 0.251). Overall, social play increased in chimpanzees of all-male groups. The rate of play was higher in all-male groups, as we observed only 29 play bouts in mixed-sex groups during 138 h of observation, while we observed 384 play bouts during 229 h of observation of the all-male groups (χ² = 166.4, df = 1, p < 0.001). We obtained the same results from comparisons with social play data obtained from focal animal sampling of male chimpanzees and found that the rate of social play was higher in all-male groups (Fig. 2-a: est. = 59.5, SE = 2.60, t = 22.9, p < 0.001). We also found that the rate of aggressive interactions was higher in all-male groups, as we observed only 25 aggression bouts in mixed-sex groups during 138 h of observation while 151 aggression bouts were observed in all-male groups during 229 h of observation (Fig. 2-c: χ² = 40.4, df = 1, p < 0.001). We did not find significant differences in social grooming rate (including both mutual and unilateral social grooming) between male chimpanzees from mixed-sex and all-male groups (Fig. 2-b: est. = 1.63, SE = 1.59, t = 1.03, p = 0.305) (Fig. 2).

Among the 29 play bouts observed in mixed-sex groups, 22 play bouts were between males and females. Only 6 play bouts were between females (Table 3).

The distribution of unilateral social grooming and social play across the day showed similar patterns. Both behaviours increased toward the middle of the day, dropped once in the early afternoon and increased again toward pre-feeding time (Table 4; unilateral social grooming: likelihood ratio test, ΔD = 16.9, p = 0.0159; social play: likelihood ratio test, ΔD = 41,7, p = 0.0452). However, we did not find significant variation in mutual social grooming across the day (Table 4; likelihood ratio test, ΔD = 20.9, p = 0.235). Social play was also significant when we tested the differences in the number of social play bouts across the day (χ² = 71.0, df = 7, p < 0.001). In addition to pre-feeding time, we found another peak in the middle of the day. The results of the post-hoc pairwise comparisons were shown in Supplementary 2. Aggressive interactions varied significantly across the day (Fig. 3: χ² = 15.3, df = 7, p = 0.0318), but post-hoc test did not find any significant difference in any pairwise comparison (Supplementary 2, p > 0.43). The frequency of aggressive interactions was relatively higher during pre-feeding time, although the peak of aggressive interactions was also observed between12:00 h and 12:30 h.

There was a negative correlation between the rate of aggressive interactions and mutual grooming (Fig. 4-a: τ_Kr = −0.70, p = 0.0002), and a similar trend was found between aggressive interactions and grooming equality index (Fig. 4-b: τ_Kr = −0.18, p = 0.061). There was no correlation between the rate of aggressive interactions and social play (Fig. 4-c: τ_Kr = −0.095, p = 0.25), and grooming equality index and social play (Fig. 4-d: τ_Kr = −0.11, p = 0.18). The rate of mutual social grooming and social play showed a negative relationship (Fig. 4-e: τ_Kr = −0.45, p = 0.0035). In a few cases, social play (Fig. 4-f) itself lead to aggressive interactions (6 out of 151 aggression bouts).

There was no significant correlation between social play and abnormal behaviour (Z = 1.06, n = 11, p = 0.288), and between social play and self-directed behaviour (Z = −0.144, n = 11, p = 0.886).

Social play increased in all-male group chimpanzees compared to those in mixed-sex groups with fewer males. Males are more playful than females, as most play bouts included at least one male. Females seldom engaged in social play without males, despite the fact that there were many females among our subjects. Previous studies on sex differences in immature chimpanzees (Lonsdorf et al., 2014, Mendoza-Granados and Sommer, 1995) and other mammals (Pellis, 2002) similarly reported that males are more playful than females. Our study reveals that this sex difference is consistent in adulthood.

Additionally, consistent with previous studies of chimpanzees and bonobos (Palagi, 2006, Palagi et al., 2004), social play and unilateral social grooming increased before feeding, when tension is heightened. There seems to be a similar rhythm of behaviours in both unilateral social play and social grooming. They increased toward the middle of the day, decreased once and then increased again toward pre-feeding time. The reason why social play and grooming increased in the middle of the day may be because the caregivers always came to check the chimpanzees and provided enrichment between 13:00 h and 14:00 h. Aggressive interactions occurred across the day, but it slightly increased before feeding. Although social play and unilateral social grooming showed similar patterns, the variation of mutual social grooming across the day was not significant and the peak of mutual social grooming occurred earlier than that of unilateral social grooming and social play. Considering that the peak of unilateral social grooming, social play and aggressive interactions occurred during pre-feeding time, mutual social grooming may have a different function.

Interestingly, the individual rate of social play increased with age in all-male group chimpanzees. Some suggest that the rate of play decreases according to age (Graham and Burghardt, 2010) but these studies often include subjects with ages occurring from infancy to adolescence (Bloomsmith et al., 1994). Our study is unique because it included individuals after the adolescent stage, between the ages of 9 and 44. More data is needed to support this finding, but it is possible that older chimpanzees may use social play as a tactic to reduce social tensions.

Although the time distribution was similar for unilateral social grooming and social play, mutual social grooming and social play imply totally different social relationships. Mutual social grooming reflected an affiliative relationship, as there was a negative relationship with aggressive interactions. The finding of an association between mutual social grooming and affiliative relationships is consistent with studies of another captive group of chimpanzees, which demonstrated a correlation between the rate of mutual social grooming and proximity and reported that dyads with kin relationship engaged in mutual social grooming more frequently than those without such a relationship (Fedurek and Dunbar, 2009). Such a negative relationship with aggressive interactions was also observed in grooming equality index, but it was weaker than in the rate of mutual grooming. Chimpanzees spent considerable time in mutual grooming. Machanda et al. (2014) did not find significant relationship between mutual grooming and association index, but they found that individuals who avoided each other were less likely to engage in mutual grooming. This means that wild chimpanzees can use “fission” strategy to avoid individuals with a low relationship quality. In captive conditions, chimpanzees often have to tolerate such individuals in the same enclosure. Such situation might influence the differences in the results of the studies conducted in wild and captive conditions.

However, social play occurred not only between dyads without aggressive interactions, but also between those with aggressive interactions. Furthermore, mutual social grooming and social play showed a negative correlation. Therefore, mutual social grooming implies an affiliative relationship, but that is not always the case for social play. We can distinguish social play from other behaviours, namely, as behaviours including play panting or play laughing. Such pleasurable experiences may decrease the threshold of social interactions even among dyads with a low relationship quality.

Social play increased before feeding, also in all-male groups. Both situations can be related to heightened social tension as we observed increased aggressive interactions in both situations. In our previous study, we found that the hair cortisol level was lower in males from the all-male groups than in those from mixed-sex groups. Combined with the present findings, males in the all-male groups may use social play as a means to cope with social tensions, especially among conspecifics with a low-quality relationship. Social play may not necessarily facilitate close bonding because it has a negative correlation with mutual social grooming in the already established groups. However, individuals with a low-quality relationship might assess and tolerate each other by engaging in social play. Therefore, social play is important for the coexistence of several males who do not always get along well.

Interestingly, bonobos are more playful than chimpanzees, and they are also less aggressive and more cohesive than chimpanzees (Hare and Yamamoto, 2015). Hare et al. (2012) proposed the hypothesis that self-domestication occurs in bonobos, because the above-mentioned traits of bonobos are similar to those of animals selected by means of artificial selection. Adult chimpanzees were considered to play less, but our study revealed that this also depends on their environment. We observed the increase of social play in the captive environment lacking fitness threat, but socially closed environment with several unrelated males. Therefore, understanding the social play of adult chimpanzees living in various captive social environments may provide more clues for how certain traits of humans and bonobos evolved.

Our study’s findings question the view that social play can be used as an indicator of positive animal welfare. Although we need more discussion on what is “positive” animal welfare, basically positive states can be related to relaxed states without fitness threat. Additionally, it should accompany good social relationships, as our previous studies reported the importance of social relationships for the long-term stress levels of captive chimpanzees (Yamanashi et al., 2013, Yamanashi et al., 2016a). The fact that social play increased with increasing age in all-male groups, sex and group differences and time of social play, correlation with other social behaviours, and absence of correlation with self-directed and abnormal behaviours, indicate the complexity of social play. Because social play is important for maintaining social groups among adult male chimpanzees, it is essential to consider this behaviour in the context of captive group management. However, it is difficult to assert that individuals maintain good relationships with each other. Therefore, we did not ascertain that social play is a negative indicator; instead, the interpretation of the behaviour as animal welfare indicators should be determined by combining it with other welfare indicators and treating it in a similar manner.

There are some limitations of our study regarding understanding the social play among adult chimpanzees. Although we found that males in the all-male groups played and engaged in aggressive behaviours more than those in the mixed-sex groups, we could not differentiate whether the number of males or other factors in the all-male groups (e.g. frequent social changes) were responsible for the differences. We did not observe the same behaviour in the social group, which included several males and females, because it is rare in a captive environment (Bloomsmith and Baker, 2001). Another point is that although this study treated social play as one category of behaviour, social play includes behaviours with different modalities and intensity (Palagi and Paoli, 2007). There might be a difference between underlying motivation and emotion, depending on such modality differences. Furthermore, although we did not find evidence that social play deepens the social relationship among individuals, there is still such a possibility in the case of introduction of new individuals because social play can have a function of social assessment (Graham and Burghardt, 2010, Pellis and Iwaniuk, 2000). Future studies should investigate social play and its relevance to the internal states in more diverse types of social groups and in more diverse social situations in depth.

To sum up, social play and unilateral social grooming are similar in terms of the occurring behaviours’ timing. However, social play and mutual grooming were not similar in terms of the occurring behaviours’ timing and also did not imply a similar social relationship, as social play does not necessarily indicate an affiliative social relationship in the same way that mutual social grooming does. Individuals with low relationship quality also engaged in social play to reduce social tension. Combined with the finding that social play increased in all-male group chimpanzees, social play is important for the coexistence of multiple adult males who do not always get along well. Although this is an important behaviour in the context of the social management and welfare assessment of captive chimpanzees, we have to be cautious when we assess animal welfare using social play. The notion that all social play implies ‘positive’ welfare is oversimplified, since individuals do not always have affiliative relationships with others, which is an important point when considering animal welfare. Therefore, as we treated other welfare indicators, we need a more holistic view to interpret social play by combining multiple welfare indicators.

The care of the chimpanzees and the present study were supported financially by grants from the Japan Society for the Promotion of Science (#13J04636 and #17K17828 to YY, #25119008 to SH, #15H0579, #16H06301, #16H06283, JSPS-LGP-U04, JSPS core-to-core CCSN) and also by Future Development Funding Program of Kyoto University Research Coordination Alliance and Great Ape Information Network. We are grateful to the following people and institutions for their support of our study: Yusuke Mori, Toshifumi Udono and the staff at the Kumamoto Sanctuary. We also thank Masaki Shimada for invaluable discussion regarding this study and Yusuke Hori and Daisuke Muramatsu for statistical advice.

* p < 0.05, + 0.05 < p < 0.10.