Adult–adult play in captive lowland gorillas (Gorilla gorilla gorilla)

Among African great apes, play is virtually absent between adult lowland gorillas (Gorilla gorilla gorilla). Here, we report an extremely rare case of adult–adult play observed in the lowland gorilla group housed at La Vallée de Singes (France). We recorded three playful interactions between the silverback and an adult lactating female. Given the diverse causal and functional nature of play, different factors may join in promoting this behaviour. In our group, contrary to what has been shown by previous studies in wild and captive gorillas, adult females spent more time in spatial proximity with the silverback than with other females. Hence, the probability of social interaction (including play) between silverback and adult females was enhanced. Moreover, the motivation of the lactating female to play might be an effect of oxytocin, a hormone that reaches high concentration levels during lactation and that promotes social affiliation. The fact that play invitations were always performed by the female might support this hypothesis. Certainly, we cannot exclude the possibility that the play event is a group idiosyncrasy or an artefact of captivity, even though the subjects never showed abnormal behaviour. Structurally, play sessions showed a suitable degree of pattern variability and switching frequency from one pattern to another. The proportion of offensive patterns was higher in the female during play and in the male during aggression, which conforms to the role reversal play phenomenon. In conclusion, this report confirms that the absence of evidence is not the evidence of absence. It is likely that under particular physiological or socio-ecological conditions, adult–adult play may be manifested as an “unconventional” part of gorilla social behaviour. Supplementary Information The online version contains supplementary material available at 10.1007/s10329-022-00973-7.


Introduction
In human and non-human primates, social play is widespread among immature individuals and is considered the main social interaction characterizing the juvenile developmental phase (Fairbanks 2000;Cordoni and Palagi 2011). Conversely, adult-adult play has been described in only about 50% of primate species Iwaniuk 1999, 2000). Following a phylogenetic logistic regression analysis, O'Meara and colleagues (2015) could not single out any correlations between the maintenance of play in adulthood and metabolic (e.g. basal metabolic rate), socioecological (e.g. group size) or life-history (e.g. age at sexual maturity) variables. This finding supports the diverse causal and functional nature of adult social play behaviour in primates.
African great apes show frequent and quite stable levels of social play in the immature phase (Fagen 1981;Cordoni and Palagi 2011;Palagi and Cordoni 2012;Cordoni et al. 2018), but the level of play during adulthood varies greatly depending on the species. In bonobos (Pan paniscus), adult-adult play remains relatively frequent across different sex/age-class combinations and contexts Palagi and Paoli 2007;Palagi and Cordoni 2012). In chimpanzees (Pan troglodytes), adult-adult playful interactions are less frequent compared to bonobos, although they can still be observed as a means for reducing social tension and strengthening affiliative bonds (Palagi et al. 2004;Cordoni and Palagi 2011;Yamanashi et al. 2018). Although infrequent, social play in adult mountain 1 3 gorillas (Gorilla beringei beringei) has mainly been reported between mature males living in multi-male or bachelor groups (Yamagiwa 1992;Watts and Pusey 1993;Grueter et al. 2016). In lowland gorillas (Gorilla gorilla gorilla), social play plummets when approaching sexual maturity and is virtually absent between adults (Stewart and Harcourt 1987;Palagi et al. , 2019Masi et al. 2009;Cordoni et al. 2018).
Here, we report and describe a rare case of adult-adult play in captive lowland gorillas that was observed between the silverback and an adult female. In the wild, lowland gorillas mainly constitute one-male units comprising an adult male (the silverback), several adult females and their offspring (Fleagle 2013). Mature males can temporarily associate in the so-called bachelor groups during the period of dispersion from their natal groups (Robbins and Robbins 2018;Forcina et al. 2019;Hagemann et al. 2019). Before their first reproductive event, females also disperse from their original groups, and they preferentially join smaller groups led by younger and stronger silverbacks (Stokes 2004;Manguette et al. 2020). Even if affinitive interactions may be present at variable levels in lowland gorillas (Forcina et al. 2019;Cooksey et al. 2020), social contacts (e.g. allo-grooming) between adults are less frequent compared to other great apes (Stokes 2004;Cordoni and Palagi 2007;Masi et al. 2009;Cordoni et al. 2018). In the wild, Stokes (2004) reported only eight affiliative events (i.e. sexual interactions and contact sitting) between adults during 802 hours of observation; intriguingly, six out of eight contacts occurred between the silverback and the reproductive females. In both captive and wild gorillas, first postconflict affinitive contacts between former opponents (reconciliation sensu de Waal and Roosmalen 1979) occurred more frequently between the alpha male and adult females compared to other dyads, possibly aimed at restoring a peaceful relationship between these subjects (Watts 1995(Watts , 2000Cords and Aureli 2000;Cordoni et al. 2006). Different results have been obtained for spatial proximity (Watts 1994;Stokes 2004;Lemasson et al. 2018). Indeed, in captive lowland gorillas, it has been observed that adult females (particularly females with newborns) did not maintain close proximity with the silverback, and they even preferred staying in spatial closeness with other group members (Fischer 1983;Nakamichi and Kato 2001;Stoinski et al. 2003). As concerns play, in two captive groups housed at the Zoo-Parc de Beauval (France), Cordoni and co-authors (2018) reported extremely low frequencies of playful interactions between adults and immature subjects (mean individual hourly frequency = 0.023 ± 0.015SE), and social play was never observed between adults. In the wild (Odzala-Kokoua National Park; Congo), Forcina and colleagues (2019) qualitatively reported playful interactions between adult females of different groups, but never between adult females and the silverback.
Given the extremely rare occurrence of adult-adult play in lowland gorillas, very little is known about how adults engage in and manage playful interactions. In this report, we structurally describe the social play observed between the silverback and a lactating female, as a starting point for future research on modality, complexity and possible function/s of play in adult gorillas. Due to the complex nature of play, different factors, not mutually exclusive, may join in promoting adult-adult play, such as inter-individual closeness (e.g. spatial proximity), physiological conditions (e.g. lactating period) and individual characteristics (e.g. play propensity). We discuss different scenarios in the possible hypotheses explaining the occurrence of this playful event.

Study site and subjects
The study was carried out on the family group of lowland gorillas (Gorilla gorilla gorilla) housed at La Vallée des Singes (Romagne, France). The group comprised 10 individuals (see Table 1). The silverback sired all immature subjects. Only one adult female (Virunga) in the group had no offspring. During the observations, the primiparous female Mahmah was lactating. The gorilla enclosure included both an indoor and an outdoor facility of about 150 and 3400 m 2 , respectively. The outdoor space was a natural wooded island surrounded by a water canal, and the indoor enclosure was enriched with lianas, trunks, straw and platforms. Gorillas were fed outdoors with fruit and vegetables five times per day during spring and summer (May-Aug: 11:15, 14:00, 15:30, 17:00 and 18:00) and twice per day starting from September (Sep-Apr: 11:15 and 15:30). Water was provided ad libitum. Gorillas were free to move between the indoor and outdoor enclosures and to socially interact. No aberrant or stereotypic behaviour was observed.

Data collection and operational definitions
We collected video data during August-October 2020 and March-June 2021 on a daily basis spanning morning and afternoon and including all feeding times. The videos were recorded by one author (SE) using a Panasonic HDC-SD9 video camera. In total, about 72.5 hours of videos were collected. The videos were analysed-also through a frame-byframe analysis when necessary-via the freeware Avidemux 2.7.8. Before starting the video analysis, GC supervised LP in a training period of 15 hours to reach suitable agreement in individual gorilla recognition and in the classification of behavioural patterns and facial expressions. The 1 3 interobserver reliability reached Cohen's κ value of 0.78 for individual recognition and 0.74 for behaviour classification (considered behaviours: play, playful facial expressions, aggression, affiliation).
Following Altmann (1974), we employed scan sampling for evaluating baseline activities and spatial distance between subjects (i.e. resting, moving, feeding, spatial proximity, body contact, grooming, play, aggression). Spatial proximity was defined as an arm-length distance between two gorillas. We carried out group scans at 10-minute intervals by collecting a total of 132 scans (22 hours of observation). We also employed all occurrences sampling for recording playful/aggressive events (50.5 hours of observation).
For each playful/aggressive event we reported (i) the identities of the interacting subjects (age and sex) and which of them initiated the session, (ii) behavioural patterns performed in their chronological order (see Table 2 for definition of behavioural patterns), and only for play, (iii) facial expressions (play face, PF; full play face, FPF; see Table 2) and their duration (centiseconds) and (iv) session duration (seconds).
We considered that a play session started when an individual performed any playful pattern (see Table 2) towards a companion and finished when both subjects stopped the interaction (Cordoni et al. 2021). As suggested in the play literature, two consecutive sessions were considered as different sessions if the play interaction stopped for more than 10 seconds (Mancini et al., 2013;Davila-Ross et al., 2015;Cordoni et al. 2016Cordoni et al. , 2018. In our case, between the first (duration = 35 seconds) and the second (duration = 30 seconds) play session there was a time gap of 35 seconds and between the second and the third session (duration = 270 seconds) the time gap was 20 seconds. Hence, we treated the sequences as three separate playful events. During the play-pause, the silverback and the adult female maintained their spatial proximity and performed self-grooming.
We considered that an aggressive session started when a gorilla directed any aggressive pattern (see Table 2) towards a group mate and usually ended with one of the opponents moving or fleeing.
We classified playful patterns into three categories (see Table 2): (1) offensive (i.e. attack/pursuit playful patterns giving one of the playmates a clear physical advantage over the partner); (2) defensive (i.e. playful patterns by which the player tries to shelter from the playful attack by the partner); (3) neutral (locomotor/acrobatic patterns, possibly involving an object).
To evaluate the asymmetry of each interaction, for each involved party we defined a play asymmetry index (PAI) as "the proportion of offensive patterns performed by A towards B plus the defensive patterns performed by B towards A" subtracted from "the proportion of offensive patterns performed by B towards A plus the defensive patterns performed by A towards B" divided by "the total number of patterns performed by both playmates" (Cordoni et al. 2016(Cordoni et al. , 2018. The PAI ranges from −1 to +1, with main values indicating (i) complete symmetry of the session (zero), (ii) complete asymmetry of the session in favour of A (+1) and (iii) complete asymmetry of the session in favour of B (−1).

PAI =
(of fensive A→B + defensive B→A ) -(of fensive B→A + defensive A→B ) (of fensive A→B + defensive B→A ) + (of fensive B→A + defensive A→B ) + neutral A+B Table 1 The composition of the lowland gorilla family group hosted at La Vallée des Singes (Romagne, France) M male, F female, Ad adult (female > 8 years; male > 12 years), SubAd sub-adult (female 6-8 years; male 6-12 years), Inf infant (0-4 years). The play sessions that are the focus of this study were observed between the two individuals marked with asterisks  For comparative purposes, outside of the play context we also calculated an aggression asymmetry index (AAI). We considered as (i) "offensive" all patterns performed to threaten/attack/pursue an opponent, (ii) "defensive" all patterns performed to avoid/shelter from an opponent's attack and (iii) "neutral" all patterns not classified as offensive or defensive (see Table 2). AAI was defined as follows: "the proportion of offensive patterns performed by A towards B plus the defensive patterns performed by B towards A" subtracted from "the proportion of offensive patterns performed by B towards A plus the defensive patterns performed by A towards B" divided by "the total number of patterns performed by both opponents". Like the PAI, the AAI ranges from −1 to +1, with main values indicating (i) complete symmetry of the session (zero), (ii) complete asymmetry of the session in favour of A (+1) and (iii) complete asymmetry of the session in favour of B (−1).

AAI =
(of fensive A→B + defensive B→A ) -(of fensive B→A + defensive A→B ) (total pattern performed) We carried out a sequential analysis to evaluate the temporal association between different behavioural patterns. We created a string for each play session by reporting the patterns separated by a break symbol (i.e. |). The resulting string represented the ordered concatenation of patterns as they occurred during each playful interaction. Then, we employed the free open-source software Behatrix 0.9.11 (http:// www. boris. unito. it/ pages/ behat rix; Friard and Gamba, 2020) to analyse the sets of behavioural sequences and organize data into contingency tables. The program generates the code for a flow diagram (Graphviz script) of behaviour-to-behaviour transitions. Via Behatrix, we also calculated the Levenshtein distance, a string metric for measuring the difference between two sequences (Levenshtein 1966;Kruskal, 1999).

General and contextual data
In the gorilla colony under study, we collected a total of 590 play sessions, 83 of which involved at least one adult subject. In particular, we recorded three sessions between the silverback (Yaoundé) and the lactating female (Mahmah) and 80 sessions between an adult and an immature subject (number of sessions: 62 Mahmah, 13 Hakuna and 5 Yaoundé). Moreover, we collected 226 aggressive interactions, three of which involved adults. The three aggressions were directed by the silverback towards two adult females (Virunga and Hakuna). No aggression was recorded between the silverback and the lactating female (Mahmah). The spatial proximity between the silverback and adult females accounted for 26% of the proximity bouts (mean dyadic value 1.6 ± 2.1 SD), whereas adult female dyads only accounted for 9.3% of the proximity bouts (mean dyadic value 0.8 ± 3.5 SD).

Adult playful interactions
On 27 September 2020, we video-recorded three playful sessions between the silverback (Yaoundé) and a lactating female (Mahmah). It was a rainy day and the gorillas spent the whole day indoors, although they had free access to the outdoor enclosure. The group did not receive any energy-rich or special food. No changes were made in the usual routine management. No particular social or environmental stressful event or aggression among groupmembers occurred during the day. Play sessions occurred during resting time (session #1 occurred at 16:11; session #2 at 16:12; session #3 at 16:15; see video clip in Supporting material), and the two players were in strict proximity or body contact with the immature individuals of the group, including Mahmah's 3-month old son (Basoko; see Table 1). We did not observe any playful interactions among immature subjects immediately before the Mahmah-Yaoundé play sessions. In all cases, the adult female invited the silverback to play by directing playful contact patterns towards him (i.e. play bite and play pat; see Table 2). It is worth noting that throughout the three adult-adult play sessions, the immature individuals (excluding the newborn, Basoko) tried unsuccessfully to interrupt or join the interaction.

Play behavioural sequence
We recorded 13 different types of patterns constituting the three play sessions. A total of 37 transitions occurred between behavioural patterns (see Table 3). The Levenshtein distance values ranged from 6 to 20, thus indicating a difference in the composition of the three sessions, particularly between sessions #1 and #2 and #2 and #3 (Table 3).

Play and symmetry
As concerns the symmetry of the session, the medians of PAI (−0.83) and AAI (1.00) indicated two opposite directions (Fig. 1). While in the aggressive context the interaction was completely asymmetric in favour of the silverback, during play the roles of playmates were reversed, with the adult female "dominating" the interaction with the silverback. Fig. 2 Flow diagram representing behaviour-to-behaviour transitions in the three playful sessions recorded between the silverback and lactating adult female. Legend (see Table 2 for pattern definitions): abit attempt play bite, fpf full play face, pat play pat, pbit play bite, pf play face, pinv play invitation, plgr play grab, pps play push, ppu play pull, pre play retrieve, psh play push, psl play slap

Playful facial expressions
Among the recorded playful patterns, 17.5% were facial expressions mainly performed by the silverback (two play faces, PF, and four full play faces, FPF). The sequential analysis (Fig. 2) showed that PF and FPF performed by both the silverback and the adult female were immediately followed or preceded by a typical playful offensive pattern, that is, play bite (transition occurrence % FPF ↔ play bite = 76.7%; transition occurrence % PF ↔ play bite = 73.4%). In one of the three sessions, namely session #2, the silverback directed a play bite towards the female, and immediately after he performed a FPF while engaging in a face-to-face interaction with Mahmah. Mahmah responded to Yaoundé with a PF and concomitantly directed a play bite towards him (see supporting video material).

Play and affiliation
During session #2, between two play-biting events the silverback groomed the female for less than 10 seconds (grooming was not part of play; transition occurrence % play bite → grooming 6.67% and grooming → play bite 100%). Generally, during grooming, a gorilla manipulates the fur, extremity or orifice of a companion; grooming may include both manual and oral components . No other grooming bout involving the adults was recorded throughout the duration of our study.

Presence of adult-adult play
In the current captive study on lowland gorillas, we structurally described three social play interactions between a silverback and an adult female. The novelty of our investigation relies on the report that adult-adult play-although extremely rare-is part of the behaviour of lowland gorillas and can be shown at least under certain circumstances. To our knowledge, no report on play modalities between the silverback and adult females is available in the literature for the study species. Considering that the two subjects involved in the play sessions were born and mother-reared in captivity and thus had no particular history of exploitation in the entertainment industry, we can reasonably exclude that the observed patterns have to be considered as abnormal behaviours. It is possible that the occurrence of such behaviour can be facilitated by the safe social and ecological environment in which the gorillas lived.
In this study, contrary to what is reported in the literature on wild and captive lowland gorillas (Fischer 1983;Stoinski et al. 2003;Stokes 2004;Nakamichi et al. 2014;Klailova and Lee 2014), adult females spent more time in spatial proximity to the silverback than to other females. In this view, the probability of social interaction (including play) between the silverback and adult females was enhanced. However, we observed social play between the silverback and one specific adult female, but not between the silverback and any female. Play specifically involved the only lactating female.
Greater concentrations of oxytocin-a neuropeptide hormone-are associated with lactation and milk ejection (Carter et al. 2007;White-Traut et al. 2009). Moreover, even though the relation between oxytocin and social behaviour is highly complex, this hormone could have an effect in creating mother-infant bond, increasing affinitive interactions, forming inter-individual relationships and, in some cases, promoting play (Wallen and Hassett 2009;Insel 2010;Carter 2014;Vanderschuren and Trezza 2014;Numan and Young 2016;Vanderschuren et al. 2016). Hence, in our study the motivation of the female to play might have been increased during lactation. This hypothesis is also supported by the fact that play invitations were always performed by the lactating female towards the silverback. Of course, a much larger sample is necessary to verify this hypothesis. In addition, general levels of play performed by adult females with both other adults and immature subjects should be evaluated before and after the lactation period and compared with levels during lactation. Higher general levels of play during lactation than during non-lactating periods would further support our hypothesis.
Play is considered a by-product of the interaction and rearrangement of different behavioural systems (e.g. aggressive and sexual domains), and its motor patterns resemble those used in these "serious" contexts (Pellis et al. 2019). One of the major risks of play is the misinterpretation of the pattern performed/received by playmates and the consequent escalation of the session into aggression (Pellis and Pellis 1996;Palagi et al. 2016a). In the wild, both mountain and lowland lactating female gorillas receive less aggression by the silverback compared to cycling and pregnant females (Robbins 2009;Breuer et al. 2016). Hence, in our case the risk of the lactating female of receiving an aggressive response by the silverback could be lowered, thus facilitating the start and maintenance of the playful interaction.
In wild mountain gorillas, Grueter and colleagues (2016) recorded two playful interactions between the silverback and adult females in one out of the three groups under observation. The authors did not exclude the possibility that social play could be an effect of food intoxication. In this respect, social play was considered an abnormal, toxin-induced behaviour. Intoxication due to plant consumption was never observed in the study group, and no toxic plants seem to be present in the outdoor enclosure. Therefore, it seems 1 3 unlikely that social play in our study group was triggered by toxins.
It is also worth noting that during the day in which adult-adult play was observed, gorillas remained mainly inside, and they probably had been expending less energy compared to other ordinary days. The surplus energy theory of play (Barber 1991) proposed that animals can consume the energy in "excess of need" by playing. In this view, adult-adult play could be favoured by an excess of energy subjects have to expend. Nevertheless, in our case the gorilla keepers who have been working with this group for over 20 years told us that they had never observed Yaoundé playing with adult females during "less energy-expending days".
We can hypothesize that under particular physiological and/or socio-ecological conditions-in our case the possible effects of oxytocin and more frequent spatial proximityadult-adult play may be present as a rare and "unconventional" part of gorilla social behaviour. Certainly, we cannot exclude the possibility that what we observed is linked to idiosyncratic factors deriving from particular individual life history events that we ignore (e.g. behavioural acquisition caused by prolonged interactions with humans) or to more extroverted personality traits (Racevska and Hill 2017) favouring the play propensity of the lactating female. Moreover, we cannot rule out the possibility that play behaviour in adult gorillas might be an artefact of captivity, although we believe that if this were the case, play behaviour should be much more commonly observed in the captive groups of this species.

Structure of the play sessions
From a structural point of view, the recorded play sessions showed a degree of variability in the motor patterns performed (13 different types of play patterns; see Table 3) and in the frequency of switching from one pattern to another (37 transitions for a total of 40 patterns performed; see Table 3). In 2011, Burghardt (2011) proposed five criteria for a behaviour to be considered play. In particular, play behaviour (i) does not appear completely functional, (ii) is spontaneous, voluntary and rewarding, (iii) includes patterns exaggerated and modified in their sequences, (iv) comprises repeated but not stereotyped patterns and (iv) is performed in a relatively relaxed context. Our observations are in line with the third and the fourth criteria, which are mainly centred on the behavioural structure of play in terms of pattern repetition and exaggeration.
In our case, we observed that play facial expressions (PF and FPF) mainly occurred immediately before or after an offensive playful pattern. From a classical ethological point of view, PF and FPF are ritualized displays that drive information to the receiver on possible future actions of the play companion. By doing so, such displays reduce the uncertainty concerning the nature of the interaction (Bekoff and Allen 1998;Pellis and Pellis, 1996;Bekoff 2001). Play faces also convey a positive emotional state of playmates (Palagi et al. 2016b(Palagi et al. , 2019(Palagi et al. , 2020; but see: Bliss-Moreau and Moadab 2017). It is likely that PF and FPF not only function in communicating the benign intent of the playmate (particularly the silverback), but also allow individuals to potentially share their playful mood.
Finally, the partner roles during play were reversed; indeed, the adult female directed more offensive patterns towards the silverback than vice versa. This strategy is often used by strong, old and dominant individuals for maintaining play with partners having fewer physical abilities and lower hierarchical status (self-handicapping and role reversal ;Power 2000;Petru et al. 2009).
In conclusion, this case report confirms that "the absence of evidence does not indicate the evidence of absence". Although extremely rare, it can be stated that social play between adults is present in lowland gorillas, although its expression may occur in particular individuals living in particular social groups and under particular (if not exceptional) circumstances.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will 1 3 need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.