Due to its multifunctional and complex nature, play is one of the most difficult behaviors to study [1]–[3]. Play apparently is difficult to define and definitions vary widely among researchers. For instance, in humans, there is considerable confusion surrounding the definition of play. In fact, in the child development literature, a variety of children's social and non-social behaviors are grouped under the term “play” [2]. For clarity, in this paper we will use ethological and sociobiological definition of play [2], [4], [5].

Ethologists and sociobiologists often define play as all activity, which has no clear, immediate benefits. Frequently, play seems to involve an array of motor patterns, both typical of serious functional contexts (e.g. agonistic, antipredatory, and mating behavior) [6]–[8] as well as playful actions [9]. However, the main difference between playful and serious contexts is not in the actual behavioral patterns, but how they are performed [10], [11]. Burghardt [1] listed five criteria that a behavior must follow to be considered play; a playful behavior must be incompletely functional, rewarding/voluntary, structurally or temporally modified, performed in a repeated manner, and initiated in a relaxed context [1].

Physical activity play is one of the most common forms of play. It typically implies locomotor-rotational/acrobatic (LR-play) patterns that can be carried out both solitarily and socially. Rough and tumble play (R&T), is characterized by fighting elements (performed in non-serious way), which involve more than one player [12]–[16]. Here, we focus on social, physical activity involving both LR-play and R&T.

Social play is widespread in mammals [1]. This behavior has different functions according to species, sex, age, relationship quality between playmates, group membership, context, and habitat [17]–[22]. Clearly then, play joins and cuts across a variety of disciplines. It leads directly to inquiries connecting individual development with species adaptation. It is not surprising that comparative studies of play behavior can make contributions to a wide variety of fields [23].

Social play is a fundamental component of the behavioral repertoire of the youngsters of many species of mammals, including humans, and its developmental trajectories (onset, peak, and offset) have evolved in concert with the extension of the immaturity period [20], [24]–[27]. Social play is first experienced between mother and offspring. Many good examples come from human and non-human primates [28], [29]. Peek-a-boo, a typical mother-child game, also occurs in the great apes as do other locomotor activities that involve bouncing, throwing, and swinging infants [4]. Interactions with their mothers represent for infants a good starting-point to learn how to manage play sessions (fine-tuning) that will later be transferred and fully developed in peer-peer interactions [4], [30], [31]. The quality and quantity of mother-infant play seem to predict the quality and frequency of infants' play with peers [2], [32]. However, peer-peer play can remediate deficits in mother-infant play and contribute sufficiently to normal social development in monkeys [33], great apes [34], and children [35]–[37]. This suggests that infant-infant play is important for acquiring social competence and developing affinitive bonds.

Primate social play in older immature subjects also functions to establish a dominance order among individuals. Individuals acquire information on the strength and weakness of group members by engaging in R&T (humans [38], [39]; chimpanzees, [40]).

In chimpanzees social play serves different functions at different ages (Hypothesis 1) [11], [14], [19], [41]. According to Hypothesis 1, chimpanzee play should vary in modality (e.g. asymmetry in R&T, competition), complexity (variability in the play patterns used), selectivity of playmates (e.g. age selectivity), and frequency as a function of individual age classes (Prediction 1).

One of the most important reasons for studying play in chimpanzees is to shed light on the biological roots of human behavior. Due to their phylogenetic closeness and prolonged immaturity phase [1], [3], [20], [23], chimpanzees and humans have similar developmental pathways for play (Hypothesis 2). Accordingly, both species should show similar play parameters (e.g., modality, complexity, selectivity of playmates, and frequency) across age-classes (Prediction 2).

Specific facial displays (the relaxed open-mouth display or play face is usually associated with pant-like vocalizations) often accompany play sessions [11], [13], [42] In primate species, facial displays are a fundamental key for successfully managing play bouts [43]. Playful facial signals have to be considered as an integral part of play behavior development [44]. Great apes perform playful facial displays via two different configurations; play face (PF), where the mouth is open with only the lower teeth exposed, and full play face (FPF), where the mouth is opened with upper and lower teeth exposed. Some authors contend that the two expressions are used differently in relation to the intensity of play [11], [45], [46]. In humans, laughter, which is a universal expression [47], [48], seems to derive from non-human primate play faces and pant-like vocalization [49]. Recently, an affect-induction hypothesis has been proposed to elucidate the function of human laughter: this expression does not give simple information, but induces a positive influence on the receiver behavior [49]–[54]. Humans and apes smile spontaneously during pleasurable experiences, including visual, auditory, and tactile stimulation. Smiles and play faces, being the expression of positive emotional states, reinforce the behaviors that elicited them in the first place. Such kind of visual reinforcement is essential for learning in infants, when mothers smile at babies to encourage desired behaviors (see [55]–[57] for an extensive review).

Although playful facial configurations can differ across species [8], they may have a common role in signaling non-agonistic intent to a playmate and/or in expressing emotion even when alone [42], [44] (Hypothesis 3). If in chimpanzees play faces are used in a strategic way and function to signal benign intent, they should vary in frequency, timing, and type (PF & FPF) according to the age of playmates (Prediction 3a). If play faces mainly signal internal emotional states of the player [58], they should not necessarily vary across the playmates' ages (Prediction 3b).

The present study was carried out on two captive groups of Pan troglodytes hosted at the ZooParc de Beauval (S. Aignan sur Cher, France) and the Dierenpark Amersfoort (Amersfoort, The Netherlands). The Beauval colony was composed of 2 adult males, 8 adult females, 3 juvenile males, 2 juvenile females, 2 infant males and 2 infant females; the Amersfoort colony was made up of 2 adult males, 9 adult females, 2 juvenile females, 2 infant males and 2 infant females (Table 1).

Via focal animal sampling [59] we collected data on both social and solitary play behavior: Locomotor/Rotational-play (LR-play), Rough&Tumble (R&T), Object play (O-play). See the Methods and Table 2 for the definitions.

Predictions 1 & 2: There should be age-related changes in the frequency and content of chimpanzee play (P1) and strong similarities between chimpanzee and human play development (P2)

As it occurs in humans, play in chimpanzees varied according to the age of the players in relation to different parameters such as modality, complexity, selectivity of playmates, and frequency (Prediction 1 and 2 supported, Table 3). Play signals did not vary in terms of frequency, timing and type across the ages of playmates, even though we found interesting results indicating that both infants and juveniles perform playful facial displays in a strategic way (Prediction 3a partially supported). Moreover, we found that playful expressions were performed even when chimpanzees played alone, thus suggesting that PF and FPF may be directly linked to the emotional state of the sender (Prediction 3b supported).

In humans, the three types of physical activity play (rhythmic stereotypies, exercise play, R&T) show three successive peaks, thus reflecting different functions for these different forms of play [16] (Table 3). Our finding on solitary play is in agreement with data coming from children. Even though parallels across ages of different species have to be taken cautiously, there is a marked overlapping between the percentage of solitary play in infant (29.27%±5.9SE) and juvenile chimpanzees (14.45%±5.2SE) with those of preschoolers (0–3 yrs; 17%–23%) and kindergarten-aged children (3–6 yrs; 17%), respectively [60]–[62] (Prediction 2 supported). Such overlapping seems to disappear when we consider social play. In humans, the transition from solitary to social play occurs during the preschool period [2]; whereas, in chimpanzees social play constantly covers a wide time-window, from infancy (0–3 years) to juvenility (4–7 years). However, if within social play we consider R&T, there are striking similarities between humans and chimpanzees (Prediction 2 supported). Consistent with our findings, Scott and Panksepp [63] demonstrated that R&T play is at a plateau until children are at least 7-years old.

Play effectiveness in the development of social skills often means a choice of an appropriate playmate [13]. Even though studies of children's social play rarely focused on the effects of age on partner choice, it seems that human and non-human primates show selectivity for peers, especially in terms of strength/size matching (Prediction 2 supported, see Table 3) [3], [31]. In humans, for example, this is true also for 9-month-old babies, who show peer preference when they watch movies of same-age infants [64]. In chimpanzees, the selectivity for play partner choice may be evaluated by analyzing the directionality of play invitations. Our data showed that chimpanzees invited peers to play more often than non-peers, thus suggesting a preference to engage in play with matched individuals (Prediction 1 supported). Accordingly, other studies, carried out both in captivity and in the wild, demonstrated that immature chimpanzees tend to play with partners who are closest to themselves in age; for juveniles playing with infants might not have been challenging enough because of their limited motor skills and, on the other hand, for infants playing with juveniles might be too dangerous [41], [65].

Infant-infant dyads performed R&T more frequently than infant-juvenile and juvenile-juvenile dyads did (Prediction 1 supported); on the other hand, no difference was found for LR-play. Although, it was difficult to make direct comparisons between chimpanzee and human R&T due to the lack of quantitative data in children across different age stages [3], [16], some parallels seem to emerge. As in humans, the functions of R&T in chimpanzees shifted through the different developmental stages (Prediction 2 supported, see Table 3). In infancy, R&T seemed to have a role in socialization and in developing motor and psychological skills [3], [4]. In juveniles, R&T begins to include competitive elements that will be used by animals to establish social dominance relationships [16], [40], [66]. In humans, up to around 11 years, most evidence suggests that the great majority of R&T is purely playful, and that when play fighting does turn into real fighting, this is due to a lack of social skills and not the conscious manipulation that characterizes adolescents' R&T [67], [68]. Accordingly, Pellegrini [38] found that adolescent R&T was positively correlated with aggression and negatively correlated with social preference, thus suggesting that R&T could be a sort of training to acquire information on partner's skills; this information will be useful in the future to gain an advantage during real fights.

The asymmetry index of social play sessions seems not to be affected by the size and age of the players; however, when focusing on the matched dyads we found an interesting result. Social play between infants seems to be characterized by a higher degree of asymmetry compared to social play between juveniles (Prediction 1 supported). As play in juveniles is more competitive than in infants, the former have to restrain themselves in order to maintain as much as possible a symmetrical session and to limit the risk of escalation into serious fighting [8], [69]. Such an interpretation agrees with data on human youngsters, whose play bouts can turn into overt aggression [35], [38] (Prediction 2 supported). The unbalanced play sessions between infants may reflect different degrees of maturity in motor skills and/or the lack of capacity to fine-tune sessions due to difficulties in performing self-handicapping (a playmate, independently from his/her age, puts him/herself into unnecessary disadvantageous positions or situations) [3], [9] and motor inhibition [70]. This conclusion is supported by the lower complexity of infant play sessions characterized by few motor patterns which are highly repeated (see Fig. 4). In this view, the simpler R&T performed by infants could be a sort of training to assemble a more complex and sophisticated form of R&T typical of juvenility (play for play itself) (Prediction 1 supported). When playing with infants, juveniles have to self-handicap, promote reciprocal role-taking, and limit the number of types of patterns used. This capacity, also known as high-level response inhibition, relies on control mechanisms in anterior (in particular pre­frontal) cortical areas [70], [71]. This explains the lack of difference in play complexity recorded between I-I and I-J dyads. The higher complexity level recorded in juvenile dyads suggests that juvenile play is characterized by a greater number of innovative and unpredictable elements, thus indicating the development of novel control mechanisms in the cerebral cortex [70], [72]. The experience gained by switching between different play patterns may improve behavioral flexibility in chimpanzees in order to cope with unexpected situations (the training for the unexpected theory, [73]). Some studies on humans correspond to our data; the ability of children to switch activities during R&T was correlated with their capacity to face new social challenges [38], [74] (Prediction 2 supported).

Considering playful signals, infants and juveniles performed facial expressions with comparable frequency to maintain a playful interaction, thus indicating that there is no quantitative variation in the use of the two facial signals relating to the age of the performer. Moreover, no difference was found in the use of the two variants (PF and FPF) of play signals according to the age-phase. If we ended our analysis here, we should have to affirm that the Prediction 3a is not supported. However, the use of playful expressions varied according to the asymmetry of the session in infants, and to the receiver identity in juveniles (Prediction 3a partially supported). In infants, whose play sessions were the most unbalanced, we found a positive correlation between the frequency of playful facial displays and the degree of asymmetry characterizing each single session. In juveniles, we found that most of the facial signals were directed towards other juveniles. This result is not surprising if we consider the high complexity and competition levels characterizing chimpanzee juvenile play. Probably, when play becomes more competitive, as occurs in juvenile chimpanzees and humans [38], [67], [68], there needs to be clearer signaling to maintain the session and to avoid it turning into overt aggression [75]. Therefore, like laughter in humans (Duchenne laughter: [53], [55], [57], [76]), the playful expressions in chimpanzees (at the different age phases) seem to have a role in advertising cooperative dispositions and intentions thus increasing the likelihood of engaging in solid social relationships (Prediction 3a partially supported, see Table 3).

The presence of play faces during solitary play both in infant and juvenile chimpanzees indicates that, like in humans, playful facial displays can be an expression of an emotional state [57] (Prediction 3b supported), thus suggesting that infants of both species have the capacity for self-reflection or self-awareness (the precursors to more complex forms of social cognition, [44]). Recently, Pellis and Pellis [77] demonstrated that the role of play signals in self-regulating emotional state is also present in spider monkeys. Yet, there are some greater cognitive overlap and developmental similarities between Pan and Homo than between either and other primate species. This makes clear why the Pan genus offers special insights into the Homo genus in relation to some general aspects of play that are true for all species that play irrespective of their phylogenetic relationships [78].

In conclusion, like in humans, play in immature chimpanzees shows a number of changes, both quantitative and qualitative, across the ontogenetic pathway from infancy to juvenility, thus suggesting that chimpanzee play can have different functions according to the developmental stages of animals. This appears to be valid also for playful facial displays which, both in humans and chimpanzees, seem to function in modulating/enhancing social interactions and in expressing private emotions.