Animal welfare: Could adult play be a false friend

- There is no consensus regarding the functions of play. As play behavior is a characteristic of young stages of development, it has been suggested that the higher prevalence of adult play observed in domestic animals could be the result of their “neotenic retardation.” Functional hypotheses have dealt with the long term benefits, such as “rehearsal,” “motor training” for future adult competencies or “training for the unexpected.” However, there is little consistent experimental evidence favoring a particular hypothesis. The present study aimed to test the functional significance of adult play as a potential reliable indicator of good welfare, a by-product of domestication or a tool for social cohesion. Observations of both a domestic species (the horse) and wild/captive animals (cercopithecids) confirm the literature data that show the greater prevalence of adult play in the domestic/captive situations. This convergence between a domestic and a wild species argue against the idea that adult play may be a mere product of domestication. Moreover, animals living in naturalistic situations had the same low level of adult play as observed in wild animals suggesting that captive/domestic animals do not play only because they are stress free or well fed. Play is not a reliable indicator of welfare: Horses and adult macaques that played the most were also those that exhibited the greatest signals of poor welfare as stereotypic behaviors. Furthermore, adult play was more frequent at times of social disturbances and instability. Adult play is a sign showing that the adult organism needs to evacuate stress. The subjects lived in a large multimale multi-female group of ten males, 40 females and 30 sub-adults and juveniles, on a 2 ha area planted with trees and covered with grass. The group had been stable for several years at the time of the observations and no weaning was done until two years. The animals were fed with fruit, vegetables, and barley or bred once daily and had water ad libitum in a pond. The observations were done during one month in February 2011. We used individual scan-sampling (Altman, 1974) occurring every two minutes during the focal observations (i.e., 60 scans/individual). We assessed the individual behavior, posture, body orientation, location and distance to the nearest congener. Data were collected in a randomized order twice a day (morning and afternoon), outside of feeding times. play the development

may favor "emotional resilience" and more and more authors wonder whether stressful situations may diminish play or, on the contrary, enhance it (Pellis & Pellis, 2009).
Of course this contradicts the common view that play is suppressed by stress (Palagi et al., 2004) and is indicative of good welfare. Since the early hypotheses of Newberry and Wood-Gush (1988) and others, and in agreement with the statement that play is indicative of "having fun" (Spinka et al.,p. 144), play behavior has been proposed as one of the three reliable indicators of good welfare (Boissy et al., 2007). According to Held and Spinka (2011), the occasional occurrences of adult play would indicate the well-being of the animals displaying it. However, one must pay attention to two aspects: a) causality and function have to be distinguished because they are often mixed up: a happy and stress free state may trigger play (causality) and/or play offers a beneficial outcome (function) (Pellis, Pellis, & Reinhart, 2010) b) acute and chronic stress have to be distinguished: play diminishes in rats under acute restraint stress, but not under a repeated chronic stress (that nevertheless diminishes other social interactions) (Klein, Padow, & Romeo, 2010). This may be especially true when rebound effects are considered: Rats and horses play more when they are placed with peers after a social isolation period (Christensen et al., 2002;Loranca, Terrero, & Salas, 1999), calves and piglets play more when they are placed in larger pens (Jensen & Kyhn, 2000;Mintline, Wood, de Passillé, Rushen, & Tucker, 2012;Wood-Gush et al., 1990). While the individual is certainly excited and potentially enjoying an increased immediate "well-being," this does not mean that it is overall displaying a good welfare, what we mean a durable state of wellbeing. An alternative view is that, precisely because they are in poor welfare conditions, the individuals may require play to "eliminate toxic substances" (Carr, 1902) and the surplus of energy (Spencer-Schiller theory) and to perform a "self-medication" thanks to the endogenous opioids that are triggered by play (Pellis & Pellis, 2009).
In the present study, we hypothesized that adult play may, as opposed to the dominant view, be a signal of potential altered welfare that animals try to compensate when given the opportunity. Two sets of facts support this statement:  It is intriguing that adult play is present/more frequent in domestic or captive situations compared to wild/naturalistic conditions. This is something that reminds us of abnormal behaviors, such as stereotypies. If the increased adult play observed in domestic animals is due to selection for neotenic features (Coppinger & Smith, 1989), there should be more differences between wild/feral animals and their domestic counterparts than between wild species under natural versus captive conditions. However, few studies have precisely compared (using the same sampling) the prevalence of play in the same species under such opposite conditions (natural/domestic; natural/captive). Therefore, in the present study, we will a) test the domestication hypothesis by comparing data obtained in a domestic species, the horse, between animals living in naturalistic conditions versus in the usual domestic conditions on one hand, to those obtained in a wild species (rhesus macaque) living in naturalistic conditions versus in intensive husbandry. Horses are interesting also because data are available from feral populations and the close wild Przewalski species, which all confirm the rarity of adult play in natural conditions. The choice of these two species is both opportunistic (data are available in the literature) and related to the fact that both express low levels of adult play in natural conditions while anecdotal reports indicate that it does occur in the domestic/captive situations. Although not close phylogenetically, if the pattern is the same in both species (more adult play in captive/domestic situation), then the effect of domestication through neoteny is not the primary explanation to the increased level of play observed in domestic animals. Thus, if the domestication hypothesis is not the sole explanation (e.g., both captive rhesus macaques and domestic horses show increased levels of adult play), then the alternative/complementary hypothesis would be that captive/domestic conditions may alter welfare and induce increased adult play. Thus, we will also b) test whether there is cooccurrence between indicators of poor welfare (e.g., stereotypic behaviors considered as reliable indicators of poor psychological wellbeing in primates by the International community of primatologist) and the prevalence of adult play in both horses and macaque species. If this hypothesis is confirmed, we will c) examine some of the possible factors that may be involved.  Adult play seems to occur mainly in fluid tolerant social organizations (Mancini & Palagi, 2009) and to be involved in tension reduction and social cohesion. We will study primates with different types of social organizations to investigate the effects of social instability and disturbances.

Prevalence of Adult Play in Different Conditions
The Insight from a Domestic Species: the Horse. Play behavior, locomotor, social or with objects, starts in foals a few hours after birth and peaks when they are about two weeks old (play occurs then 1 to 4 times per hour and represents 6% of the time budget: Crowell-Davis, Houpt, & Kane, 1987;Sigurjonsdottir, van Dierendonck, Snorrason, & Thorhallsdottir, 2003). Young animals and two year old mares can spend 1.5% of their time playing (Sigurjonsdottir et al., 2003;Wells & von Goldsmith-Rotschild, 1979), but play behavior is totally absent in the repertoire of adult mares, generally from their first foaling on three years of age (Feh, 2005;Feist & McCullough, 1976). The young males, will generally join bachelor groups where social play, especially play fighting may reach 10% of their time budget (Bourjade, de Boyer des Roches, & Hausberger, 2009;Hoffmann, 1985). However, play behavior decreases from the age of four and becomes a rare behavior mostly observed in winter and only in bachelor males (Hoffmann, 1985). Harem stallions do not play, apart from a few responses to solicitations from juveniles.
Play is frequently stimulated by novel conditions in foals as in young adult males (changes in weather, novel site) and is often interspersed with exploration. All studies agree to consider that play is absent in the repertoire of most adult horses living in natural conditions. While all field studies agree on the quasi absence of play behavior in adult horses, reports of adult play are frequent in domestic or captive (e.g., Przewalski in zoo: Budiansky, 1997) situations.

Material and Method
We performed two sets of observations, to test the differences between horses living in different conditions:  Five domestic horses (2 stallions and 3 geldings of mixed breeds: Haflinger, Fjord, Cross Haflinger, Irish cob type, Appaloosa; aged 7 to 17 years old) living in a stable group in naturalistic conditions (large pastures). This group was observed twice a month for 13 months during 2 hours. Play behavior was recorded ad libitum (Altmann, 1974). A total of 52 hrs observations were recorded (i.e., 2 hrs every 15 days).  Twenty-nine domestic horses (23 geldings and 6 mares; 26 of which were French saddlebreds; aged 7 to 17 years old) living in a riding center. They spent most of their non-working time in single stalls (3x3m) where they received three meals per day and hay once a day (see also Hausberger et al., 2012). They were used in classical riding school lessons for 4 to 6 hrs per week and they were released in bare (soil) paddocks in (always the same) groups of about seven horses (always the same) on weekends for two 5 hrs session in the morning. Four groups were observed (3 groups comprised of 7 horses and 1 group of 8 horses). Each group was observed twice, in sessions of one hour. Play behavior was recorded ad libitum (Altmann, 1974). In both cases, only social play was observed.
Play can easily be distinguished from "real" fights as the ears are never pinned back and the interaction lacks the vocalizations that are associated with real fights (Bourjade et al., 2009;McDonnell & Poulin, 2002).

Results (Figure 1A)
The results on the domestic horses living in naturalistic conditions confirm data from feral animals and other natural conditions: Only 10 play bouts (social play) were observed over the 52 hrs observations, in three of the five individuals. Within these 10 bouts, 6 were observed in December, 2 in April and 1 in June and July respectively. There was no relation with age nor sexual status (1 out of 2 stallions, 2 out of 3 geldings involved).
The results were very different for the riding center horses: 22 of the 29 horses (only two mares out of 6 but 20 geldings) have played at least once (3+ 5 SE in 2 hrs, 1 to 20 times). One horse played 20 times in 2 hrs, which is more than the play rate observed in foals in natural conditions. There was no correlation with age (Spearman correlation rs = 0.02, N = 29, p > 0.05).

Conclusion
These data clearly confirm previous data on feral and domesticated but free living horses: adult domestic horses living in naturalistic conditions rarely play compared to conspecifics living in a restricted domestic situation. These differences cannot be explained solely by domestication-induced neoteny although genetic differences could occur (Wolff & Hausberger, 1994).

A Wild Species in Captive Conditions: An Impact on Adult Play?
Rhesus macaques are known for their "despotic" social system, which would explain the absence of adult play in natural conditions (Mancini & Palagi, 2009;Thierry, 1985). We compared one large group of rhesus macaques living in naturalistic conditions with conspecifics living in an intensive system of breeding, using similar protocol of observation.

Material and Method
Naturalistic conditions. Fifteen adult females and five adult males (mean age unknown) were observed among 70 individuals in a zoo (Planète Sauvage, Nantes). The subjects lived in a large multimale multi-female group of ten males, 40 females and 30 sub-adults and juveniles, on a 2 ha area planted with trees and covered with grass. The group had been stable for several years at the time of the observations and no weaning was done until two years. The animals were fed with fruit, vegetables, and barley or bred once daily and had water ad libitum in a pond. The observations were done during one month in February 2011. We used individual scan-sampling (Altman, 1974) occurring every two minutes during the focal observations (i.e., 60 scans/individual). We assessed the individual behavior, posture, body orientation, location and distance to the nearest congener. Data were collected in a randomized order twice a day (morning and afternoon), outside of feeding times.
Breeding facility (Institute of Beijing Xierxin Biology Resource, China) (Camus et al., 2014). This facility produces macaques for pharmacological and medical laboratories. Weaning was done early at 6 months and the density of macaques was high. The animals lived in stable (> 1 year) single male groups (one male with eight or nine females) in an indoor/outdoor enclosure (14m 2 ). They were fed with monkeys pellets twice a day and fruit once a day. Thirty five females (9.06 years old ± 0.6) from 8 different stable groups were observed in both conditions and their behaviors were recorded using a scansampling method, in which behavioral parameters were assessed every 6 min, during 2 hr sessions, resulting in 240 scans per female. For each group, data were collected in a randomized order twice a day (morning and afternoon) on 6 consecutive days. Following a 4-hr habituation phase performed one day before the beginning of the observations, macaque behavior was voice-recorded and scored live, by one trained observer, outside of feeding and cleaning times. The observations were performed during 2 months: September and October 2010.
For both populations, the observation methods were the same: scan sampling in the morning and in the afternoon, outside of feeding time and with the same observer (SC) in China and Nantes, but with another additional observer in Nantes (CR) (inter-observer reliability: Spearman rank order correlation rs = 0.86, p < 0.05). Four categories of play were distinguished: social play, object play, water play and movement play.

Results (Figure 1B)
The results show that adult play occurred only in the intensive breeding situation, with 6 adult females exhibiting play behavior, at a rate from 0.41% to 2.31% (0.17% ± 0.08 SE), while none of the females played in the naturalistic setting.

A) B) Conclusion
This result clearly shows the impact of captivity on the prevalence of adult play in a wild species.

Is Adult Play Triggered by Intensive Captive Conditions?
Long-tailed macaques have a more tolerant social organization than rhesus macaques, so adult play should be more frequent in long-tailed macaques. However, the studies show possible but rare occurrences of adult play in natural conditions. We had no access to a "reference" population of long-tailed macaques, so we studied the prevalence of adult play in intensive breeding facilities, and compared with data from the literature (Leon & Taylor, 1993cited in Crockett et al., 1995Son, 2004). According to Son (2004), long-tailed macaques spend around 2.5 % of their time in "other activities," including play behavior. Xu et al. (2012) reported that play behaviors occurred at a lower average daily count per subject (1.22, i.e., 0.20% of the total behaviors expressed by macaque) compared to other behaviors. In addition, in rhesus macaques, play is also a minor activity, representing around 5% of the group activities (Bernstein & Mason, 1963;Post & Baulu, 1978), or less (1.13%, Fisler, 1967). All these authors underlined that play is typically an activity of juveniles, even though adult males (lower ranking) sometimes play with juveniles and higher ranking males and females may play with infants (Post & Baulu, 1978).

Material and Method
A total of 80 females (mean age: 5.9 ± 0.1 yrs old) from eight "harems" were observed during 2hr sessions, using a 6 min interval scan-sampling resulting in 240 scans per individual, at the Fang Cheng Gang facility (the Fangcheng Gang Spring Biological Technology Development Corporation, Guangxi province, China). The observations were performed in March and April 2010 (see also Camus, Rochais et al., 2013;Camus et al., 2014).

Results
The results show that most adult females played at least once (68/80) with variable rates, from 0.42 to 13.87. The mean rate of adult play (2.2 % ± 0.32 SE) was equivalent to that observed in the wild for mixed age groups for behaviors including play (Son, 2004). Therefore, captive conditions, such as those observed in breeding facilities, seem to trigger adult play.

Conclusion
Adult long-tailed macaques played more in intensive breeding captive conditions compared to natural conditions. More than half of the captive animals exhibited play behaviors, which is much more than reported in field studies. Moreover the captive animals performed play at a quite high rate for adults.

Conclusion: Domestication or Living Conditions?
Both domestic species (the horse) and wild species (rhesus and long-tailed macaques) showed a different prevalence of adult play according to life conditions. This convergence between a domestic and a captive species is very interesting and suggests that the "neotenic hypotheses" (Coppinger et al., 1987) can not entirely explain why adult animals play more in restricted conditions. This hypothesis may apply better to pet species, such as dogs or cats, that may have been selected in this direction, in order to exhibit attractive features (amongst which play) and less fear of humans, but it does not explain why such an increase in play is observed in wild species that were not selected for such traits.
It is generally admitted that adult captive animals play more because they are stress free, having no predation pressure nor need to search for food (Burghardt, 2005;Graham, 2010). This is possible because experiencing less fear and having more time and energy for non-vital activities could lead to an increase of "luxury" behaviors, such as play. Nevertheless, we have shown that even domestic animals, when they are maintained in naturalistic situations, i.e., predator free and with food available, exhibit low level of adult play, like in the field situation. Thus, this is not just a matter of being careless and full of energy. In fact, the conditions where increased adult play is observed seem to correspond to situations where food, space and social contacts are restricted and where life experiences (e.g., early weaning) are often stressful (Table 1).
Many factors concur to alter the animal welfare in these conditions and this higher prevalence of adult play, which reminds us of the higher prevalence of abnormal behaviors, such as stereotypies, in these same situations, questions the possible relation of adult play with welfare states.

Co-occurrence of Adult Play and Welfare Indicators
We asked whether or not adult play may be a way of coping with stress. The first unsolved question is whether play may be triggered by stressful situations, or whether it may be, as generally admitted, an indicator of a stress free situation. The separation between acute stress, that indeed may indeed suppress play in all cases, and chronic stress, that may lead to altered welfare states, is particularly important here. When playing we expect animals to experience positive emotions, triggered by the underlying neurotransmitters (Trezza et al., 2010), but this may not reflect their chronic welfare state that results from their daily experiences. Both stereotypic behaviors and play are associated with endogenous opioids (Burghardt, 2005), but it is generally admitted that stereotypic behaviors reflect inappropriate living conditions (Mason & Latham, 2004;Mason & Rushen, 2006).
Testing the hypothesis that adult play may reflect altered welfare is difficult because play is a highly pleasurable behavior at the moment of its performance. Thus, we chose to look at possible cooccurrence between adult play prevalence or rate and reliable indicators of poor welfare.
Adult Play and Welfare in Riding School Horses Domestic horses face a set of challenges in the usual riding school settings (see Table 1), which results in physical and psychological disturbances (see also Hausberger et al., 2012). Vertebral problems are often underestimated, while they are a major and highly common problem (Fonseca et al., 2006;Jeffcott, 1980;Lesimple, Fureix, Biquand, & Hausberger, 2013). Recent studies, as well as veterinarian reports, showed that horses with more vertebral problems, hence chronic discomfort, are more aggressive towards humans (Fureix, Menguy, & Hausberger, 2010). Aggressiveness reflects altered well-being in humans (Anderson & Bushman, 2002) as well as pigs (Day, de Weerd, & Edwards, 2008;Turner, Ewen, Rooke, & Edwards, 2000). We used it here as one indicator of altered welfare. The horses were submitted to standardized human-animal relationship tests (e.g., Fureix, Jégo, Sankey, & Hausberger, 2009;Hausberger, Roche, Henry, & Visser, 2008) and the occurrence of aggressive behaviors (threats) was recorded.On the other hand, likely because they repeatedly face unavoidable stress, horses may develop a "depressive-like syndrome," becoming withdrawn and unresponsive to environmental stimuli (Fureix, Jégo, Henry, Lansade, & Hausberger, 2012; but see also Burn, Dennison & Whay, 2010;Hall, Goodwin, Heleski, Randle, & Waran, 2008). The horses tend to face the wall. Their unresponsiveness can be demonstrated through tactile stimulation using von Frey filaments .
Horses, like many other species, also exhibit abnormal repetitive behaviors (ARB) in inappropriate settings. The prevalence of stereotypies and other ARB in horses has been shown to relate to frustration related to restricted roughage availability, space and social restriction, in addition to early stresses (such as early weaning) or to working conditions (e.g., Hausberger, Gautier, Biquand, Lunel, & Jégo, 2009;McGreevy, Cripps, French, Green, & Nicol, 1995;Mills, 2005). Thus, we performed observations of the horse's behavior in their stall.
Possible physiological correlates of altered welfare are more difficult to assess. "Classical" measures of plasma or fecal cortisol do not reliably account for the expression of stereotypic behaviors in horses . Nevertheless, this remains a common way to estimate welfare in farm animals (Mormède et al., 2007), so we measured it here.
Finally, there is a novel approach of profound alterations of welfare: the measure of oxidative stress that occurs when antioxidant defense is overwhelmed, in relation with muscular tiredness and disorders, immune disorders, inflammatory and degenerative pathologies and stress (De Moffarts, Kirschvink, Pincemail, & Lekeux, 2005).
We tested the potential relationship between adult play and this array of indicators on the 29 riding school horses that lived in restricted conditions but were released in groups twice a week.

Material and Method
As mentioned above, the twenty nine horses belonged to a same riding school, were living daily in single stalls but were released on the week-ends during two 5 hr sessions, in a bare paddock, where we performed the observations of play behavior (for more details see Hausberger et al., 2012).
In their usual stall, and in periods of quietness in the school, the horses were submitted to:  5 human-horse relation tests: motionless person, approach, sudden approach, saddle and halter fitting and all the behaviors directed towards humans were recorded, with a special emphasis on aggressiveness  behavioral observations: 6 sessions of 5 min continuous observations, as well as 3 sessions of 1 hr instantaneous scan sampling (scan every 2 min) were performed for each horse (with special emphasis on abnormal repetitive behaviors and orientations)  tactile reactivity tests using von Frey filaments that induce trembling when applied (Lansade, Pichard, & Leconte, 2008)  blood sampling that allowed cortisol measurements, as well as measures of oxidative stress using spectrophotometry (Corino, Rossi, Musella, Cannata, & Pastorelli, 2007) To obtain a more integrative evaluation of horse welfare, we calculated a score based on the behavioral indicators (physiological measures could have a non-linear relation with chronic stress). A total Chronic Stress Score (TCSS) was calculated, based on the ranking of the horses for the behavioral, sanitary (vertebrae) and reactivity (tactile) measures.

Results
The frequency of play was correlated with the total chronic stress score, the aggressiveness towards humans but also with the oxidative stress value. Players (P -in paddock) differed from nonplayers (NP) on a variety of welfare indicators measured in the stall, including back disorders, aggressiveness towards humans, orientation towards the wall, tactile reactivity, as well as oxidative stress measures ( Figure 2). There was no difference in the cortisol levels, which is not surprising given the complex relation between cortisol evaluations and chronic stress in horses. There was also no clear link between adult play and stereotypic behaviors, although these were included in the TCSS that differed between players and non-players. Positive correlations between play frequency and aggressive behavior, oxidative stress and total chronic stress score were found ( Figure 3). Relations between welfare and play indicators in horses. Welfare indicators measured in box for "players" horses (i.e., at least one occurrence in 2 hours; P) and "non-players" horses (NP). Vertebral disorder: number of reached vertebral site, mean ± SE), aggressiveness toward humans during standardized tests: % of horses threatening the experimenter at least one, Horses orientation in their box: mean% of scan spent facing a wall ± SE; reactivity to touching chest: % of horses reacting at least one; measure of oxidative stress: mean ± SE, Carrelli unit. Chi-square and Mann-Whitney tests: *p < 0.05, **p < 0.01, ***p < 0.001. From Hausberger et al. (2012). Relations between welfare and play indicators in horses. Spearman correlations between play frequency and aggressive behavior, measure of oxidative stress and total chronic stress score (based on the ranking of the horses for the behavioral, sanitary and reactivity measures, the higher the score is, the more stressed the horse is, Hausberger et al., 2012). Note that a similar pattern of results appeared when similar analyses were performed but excluding the two extreme outliers (rs*).

Conclusion
Both rates and play, but also the presence/absence of this behavior in the repertoire of the adult horses could be related to measures of altered welfare. It seems that adult players are the individuals that suffer the most from the restricted conditions of the current riding school living conditions. Social play in paddocks may be a mere rebound effect (which would also indicate that the usual situation does not satisfy the primary needs), but this would not explain the differences observed between players and nonplayers in terms of welfare. One explanation could be that play may help coping with the poor welfare state (e.g., Oliveira, Rossi, Silva, Lau, & Barreto, 2010) and enable temporary emotional resilience (Spinka et al., 2001). Hausberger et al. (2012) suggested that the Carr's "recuperation theory" could make sense here, given the strong relationship with oxidative stress: the most affected horses could use play, when given the chance, to momentarily "escape" from stress, by using "muscular and circular actions to remove toxic substances build up in response to physical and mental work" (Burghardt, 2005, p. 36). Indeed, play could increase oxygenation through muscular exercise (Byers, 1998). While it is certainly pleasurable at the precise time of play, adult play in horses (especially when occurring at high rates) seems to indicate that the animal is "evacuating" excess stress rather than being a reliable indicator of welfare.

Stereotypic Behavior and Adult Play: A co-occurrence in Primates?
While in farm animals the status of stereotypic behaviors (STB) in terms of welfare assessment is regularly debated, stereotypic behaviors have been recognized as being a reliable indicator of poor welfare in captive primates by the International Primatological Society in 1993 (Crockett et al., 1995). Earlier studies have tried to relate derived activities with STB in primates (Reamer, Tooze, Coulson, & Semple, 2010) or in other mammals (Fureix, Gorecka-Bruzda, Gautier, & Hausberger, 2011) or have based the assessment of stress on activities such as self-scratching (Norscia & Palagi, 2011). Mancini and Palagi (2009) found that play behavior and allogrooming were more frequent in the pre-feeding periods and concluded that play may be used to release tensions in such situations that are characterized by conflicts of interest. The increase of excitement due to the anticipation of food or reward arrival is well known to increase the rate of STB in horses  but also locomotor play in piglets (Dudink, Simonse, Marks, de Jonge & Spruijt, 2006). In a study on horses, Fureix et al. (2011) showed a co-occurrence between yawning and STB: The rates were correlated in some populations of horses, but could also just occur mostly in the same circumstances such as the pre-feeding period or periods without roughage. These findings do not lead to causal explanations, but provide hints about common factors of causation (e.g., frustration in this case). Therefore, we investigated the possible co-occurrence of adult play and stereotypic behaviors in captive rhesus and long-tailed macaques living under intensive husbandry conditions (see above). There are reasons to believe that the two categories of behaviors may be related: they are both quasi absent of the repertoire of these species in natural conditions and they are both related to the release of endogenous opioids and to basal ganglia and striatum brain structures (Graham, 2011;Normansell & Panksepp, 1990;Pellis & McKenna, 1992;Trezza et al., 2010;Trezza, Damsteegt, Achterberg, & Vanderschuren, 2011).

Material and Method
Thirty-five female rhesus macaques and 118 female long-tailed macaques were observed in three facilities that breed monkeys for laboratories (see also Camus et al., 2014). The conditions of life have been described earlier. All the rhesus macaques were at Xierxin (China), belonged to 8 groups composed of one male with 8-9 females. Eighty female long-tailed were at Fang Cheng Gang (China), belonged to 8 groups composed of one male with 17-27 females, and 38 female long-tailed, belonged to 10 groups composed of one male with 10-20 females, were at Haikou (Hannan Jingang Laboratory Animal Corporation, Hannan Province, China). All groups have been stable for at least 9 months, all the females were adult (7 ± 0.4 yrs old; from 4 to 21 yrs old). The facilities differed in some aspects of management (Table 2), which led us to compare the female behavior in the two facilities for long-tailed macaques. Socially-housed female macaque behavior was scored following the same protocol, whatever the facility, and recorded live by one (in Xierxin and Haikou; SC) or two (in Fangcheng Gang SC and CR) trained observers (inter-observer reliability: Spearman correlation rs = 0.86). Colored medals were added to the monkeys' collar to facilitate the identification of the females using gentle restraint. The observations occurred outside of the feeding and cleaning times. For each group, the data were collected using exactly the same protocol as the one described earlier resulting in 216 scans per female in Xierxin and 240 scans per female in Fang Chen Gang and Haikou. Different types of stereotypies were observed, such as motor stereotypies, oral stereotypies and manual stereotypies (Table 3).   (Figure 4). Rhesus macaques. The two categories of behaviors were observed in all the facilities. As mentioned earlier, 6 rhesus macaques in Xierxin performed play, with a rate of 1% ± 0.30 SE (from 0.46 to 2.31% of scans). Thirty-one individuals (88.6%) performed a stereotypic behavior with rates of 3.16 ± 0.64% of time budget (from 0.46 to 18.52%). All the players performed stereotypies as well. The players performed more STB than the nonplayers (Mann-Whitney-U test N = 6, 25, Z = -2.22, p = 0.02).The rates of adult play and stereotypic behaviors were correlated (Spearman correlation: N = 35, rs = 0.43, p < 0.05), and both were respectively negatively correlated with resting behavior (Spearman correlation: N = 35, rs STB = -0.35; rs PLAY = -0.37, p < 0.05).

Prevalence of STB and Adult Play
Long-tailed macaques. Similar findings were found in long-tailed macaques in Fang Cheng Gang, where 67 (83.75 %) females were observed playing, with a rate of 2.62 % ± 0.37 SE (from 0.42 to 13.87%) and 57 females (71.25%) expressed stereotypic behaviors, which is more than half of the females observed, with a rate of 5.40% ± 0.93 SE (from 0.42 to 33.61 %). Play and STB rates were correlated (Spearman correlation, N = 80, rs = 0.41, p < 0.05) and there were more stereotypic individuals amongst the players than amongst the non-players (Chi-square test, df = 1, Chi-square = 32.44, p < 0.0001).
The co-occurrence between adult play and STB is striking: there were clear correlations between the two categories of behavior, for the two species and in two facilities (Xierxin and Fang Cheng Gang). Adult players had greater chances to also be stereotypic. Moreover, one facility appeared different in terms of the welfare state of the animals (as shown by the much lower prevalence of STB) and also appeared to have a lower prevalence of adult play.

Figure 4. Effect of monkey species and facility on the prevalence and rate of play and stereotypic behaviors (STB) (means ± SE).
A) Comparison of rate (% time budget) revealing higher rates of both play and STB in Fang Cheng Gang facility (FCG) longtailed macaques. B) Prevalence (% of individuals that performed the behavior at least once) showing a species difference (less players and more stereotypic individuals in Xierxin rhesus macaques), but also a facility effect (FCG long-tailed macaques showed a higher prevalence of both adult play and STB than Haikou macaques). C) Correlations between adult play and STB in rhesus and long-tailed (in FCG) macaques. (Camus, Rochais et al., 2013). Some reports suggest that a "normal ontogenesis" may protect from later challenging situations, which would explain the lower prevalence of stereotypies in wild born animals (chimpanzees: Davenport & Menzel, 1963;macaques: Mason & Green, 1962). Social deprivation and maternal separation are considered as major causes of emergence of abnormal behaviors in primates (Khan, 2013;Martin, 2002). According to Novak, Meyer, Lutz, and Tiefenbacher (2006), all socially raised monkeys that show self-mutilation have been separated from their mother before the age of 2 years. However, the relation between early care and play at adolescence is complex: while in most species, young animals are more likely to play when well fed and securely attached (e.g., Henry et al., 2009); in rats, low maternal licking and grooming (an indicator of maternal care) is associated with an increased rate of play at adolescence (Van Hasselt et al., 2012) and in feral horses, foals of low condition mares play more than those of high condition mares (Cameron et al., 2008).

Effects of early experience
Here we observed (see above) differences in the prevalence of play in long tailed macaques between facilities. Interestingly, the only major difference between facilities was early experience as the timing and type of weaning differed (Table 4). Therefore, we investigated the potential effect of early experience by comparing long-tailed macaques in Fang Cheng Gang according to their origins: in this facility individual could be either captive-born and, as mentioned earlier submitted to early weaning, hence to maternal or successive peer separations, or wild-caught as adults and hence had experienced a normal ontogeny. Both categories were kept in separate social groups. We hypothesized that wild caught animals could have been "protected" when facing challenging situations, which is the case of the intensive husbandry system in which they were housed at the time of the observations. Hence, indicators of poor welfare, such as STB, should be more present in captive-born animals. We investigated whether adult play would follow the same pattern.
All the animals had been living in exactly the same conditions and on the same site for years, at the time of the beginning of the study and therefore, the animals differed only regarding their early experience.
Forty long-tailed macaques (N = 40 of each type) of FCG facility were observed according to the protocols described earlier. The results reveal a clear co-occurrence between play and STB. In captiveborn individuals (CB), both play and STB occurred with a high prevalence, respectively 99% and 82.5%) and co-occurrence (33/40 females performed both). All stereotypic females performed play. On the contrary, wild-caught animals (WC) showed a slightly lower prevalence of STB (60%) but still more so play (70%). Moreover, there was no systematic association between STB and play (Table 5). Note. Number of players (performed at least once play behavior) and non-players according to early experience being (STB) or not stereotype (NSTB) (stereotypic behavior observed at least once).
The rates of STB and adult play also differed strikingly between WC and CB individuals, with CB females performing three times more play (Mann-Whitney-U test, N = 40, Z = 3.70, p = 0.0002) and 7 times more STB than the WC females (Mann-Whitney-U test, N = 40, Z = 4.15, p = 0.00003). This was observed for all types of STB (motor: p = 0.002, oral: p = 0.0001).

Conclusion
The comparison of adult female macaques that had experienced different ontogenies before living in the same unfavorable conditions as adults again reveals that adult play and STB covary. Since, STB are more frequent in the CB animals, one can consider that the early management of the young monkeys (early separation, peer rearing) is a source of stress that makes them sensitive to future challenges, as reported in the literature for rodents or horses. This adult play covariation suggests that play may be a way of escaping from chronic stress and cannot be considered as a reliable indicator of positive welfare, in any case. Interestingly, while STB were also highly present in wild caught animals, this was less the case for play. This suggests that play, was still more dependent on early experience than STB, which deserves further investigation.

Adult Play and Social Organization: Is There a Relationship?
The Importance of the Type of Social Organization in Cercopithecids Mancini and Palagi (2009) mentioned that in many species of cercopithecids, characterized by large groups including cohesive and structured relationships, play has not been reported. We took the opportunity to be able to work on different species of cercopithecids to compare the rate of adult play in naturalistic situations, between Campbell monkeys, a highly tolerant and fluid social species, mangabeys (white cheeked and red-capped mangabeys), with a more hierarchical type of society, and rhesus macaques, considered as "despotic" (Thierry, 1985).

Material and Method
The observations were performed using exactly similar protocol (for more detail see paragraph 2.1.b). They took place at the biological station of Paimpont for Campbell (ten adults), red-capped mangabeys (nine adults). Rhesus macaques were observed in a zoo (Planète Sauvage, Nantes, France). The behavior of 20 macaques (Nmale = 5; Nfemale = 15) was scored and recorded live by two trained observers, as in the Xierxin facility. In all situations, food was delivered in an equal manner in different parts of the enclosure.

Results
For all three species there were very few individuals exhibiting adult play behavior, in this situation of undisturbed stable groups, despite the many hours of observations: None of the Rhesus macaques adults and Campbell monkeys showed any play and 3 mangabeys showed some play (at a very low rate: from none to 2).
These results show that spontaneous adult play does not arise in stable conditions, maybe because there is no need for social assessment, in a stable group, or there is not much social tension when food is delivered in an equal manner in different parts of the enclosure. Nevertheless in more challenging situations, long-tailed macaques exhibited more adult play than rhesus (see above), confirming interspecies differences.
Adult play behavior may need some kind of environmental or social challenge to appear. The social organization of the species is not sufficient to explain the prevalence, sometimes high, of adult play in some studies.

In Horses Material and Method
We performed a study on 52 horses aged 3 -17 years old, of varied breeds and sex (2 stallions, 29 geldings and 21 mares), in Fall when they were in pastures in groups of five to fifteen individuals. Some of these groups were highly stable (G1,2,6,7) and the horses had been going out together every day for at least 6 months, while others changed partners regularly when going out, according to the riding center's needs. The latter were then in a highly changing group composition. Each group was observed twice for two hrs and social interactions were noted ad libitum, while the identity and the distance of nearest neighbor were recorded using instantaneous scan sampling.

Results
Adult social play was observed at least once in 7 groups, but most play bouts were observed in G5, G7 and G8 (Fig. 5). This is interesting as these three groups corresponded to the most often modified groups that also have more work hours. With G6, they also corresponded to the groups that had larger distances between neighbors but also less social cohesion, as half of them had no preferred partner, a well-known feature of social cohesion in horses (Bourjade, Moulinot, Richard-Yris, & Hausberger, 2008;Bourjade et al., 2009;Waring, 2003).

Material and Method
Behavioral observations were made on a captive group of 14 Campbell's monkeys that included seven adults (one male and 6 females), from January 1999 to June 2001. These observations were made during 8 different periods (Figure 6). Two types of periods can be distinguished: Periods without (A, C, and P5) and with (B, P1, P2, P3, and P4) social disturbances (removal of females, replacement of the male or birth of 2 babies). Each period lasted 10 to 12 days. Three observation sessions were performed every day (09:30-10:00 hr, 13:30-14:30 hr, and 16:30-17:00 hr).

Results
After each social change, the rate of aggressive behavior increased. In addition, aggression decreased strikingly from P4 to P5, after the group composition remained unchanged for a longer time. There was more play during the periods with disturbances (3.41 ± 1.21 SE per day) compared to periods without disturbances (1.19 ± 0.56 SE per day). Campell's monkeys, over 2 years. Two types of periods can be distinguished: periods without social perturbation (A, C and P5) and periods with (B, P1, P2, P3, and P4) social disturbances (removal of females, replacement of the male or birth of 2 babies). Each period lasted 10 or to 12 days. From Lemasson et al. (2006).

Conclusion
These data suggest that adult play may be triggered by social instability, which is reflected by the longer distances and the lower level of social bonding observed in horses and by the increased aggressive behaviors in monkeys. In horses, our results go along observations in the wild showing that the rare occurrences of adult play are mostly in stallions' bachelor groups, which are the less stable groups. One can hypothesize that play may help to evacuate stress or may enable to reduce tensions between the individuals (as in Mancini & Palagi, 2009).

Discussion
Our study of adult play in horses and cercopithecids species aimed at investigating whether adult play may be a reliable indicator of good welfare (Boissy et al., 2007;Spinka et al., 2011), a by-product of domestication or a tool for social cohesion in tolerant type societies of monkeys. Data from a domestic species (the horse) and wild/captive animals enabled us to confirm the literature data showing a greater prevalence (or even presence/absence) of adult play in both domestic and captive wild species. These findings argue against the idea that adult play may be a mere product of domestication, related to a "neotenic" origin of adult play. On the contrary, comparison of different life conditions revealed that in both horses and monkeys, adult play is more likely to arise in conditions that imply social/spatial/feeding restrictions. We demonstrated that animals living in naturalistic situations, without predators and with continuous access to (food) resources expressed the same low level as wild animals suggesting that captive/domestic animals do not play only because they are stress free (no predators) or well fed. Domestic/captive animals in naturalistic settings just spend their time foraging, resting and/or interacting. These observations led us to question the relation between adult play and welfare. Data on horses clearly indicate that play is not a reliable indicator of welfare: 1) Adult domestic horses in restricted domestic conditions play more than conspecifics that live in natural conditions; 2) within a population of riding school horses, the horses that played the most were also those that exhibited the greatest signals of poor welfare.
Interestingly, similar conclusions were drawn from our macaque studies. Not only did the adult macaques living in intensive husbandry play more, but a clear co-occurrence was observed between play and stereotypic behaviors, which are indicators of inappropriate living conditions in primates. These findings could be related to 1) the restricted access to time-limited resources and 2) the early life experiences. Captive born animals that had experienced early weaning (and often successive peer separations) were more likely to exhibit abnormal behaviors and play than wild born animals. Other aspects, such as physical obstacles were shown to increase both aggressiveness and adult play in cercopithecid species, indicating that adult play may be triggered by tensions and stress, as suggested by several authors (Fagen & Fagen, 2004;Norscia & Palagi, 2011;Palagi, 2007). We tested whether the prevalence of adult play in monkeys may reflect social organization and social bonding. Actually, none of the two species observed (rhesus macaques, Campbells' monkeys) and few individuals of the two species of mangabeys produced adult play when in naturalistic conditions and in stable groups. Increases of adult play did however occur at times of disturbances: Disappearance of familiar animals, arrivals of new animals. Thus, adult play was more frequent at times of social disturbances and instability, in horses and monkeys, suggesting it could be a general trend.

A Finding: More Adult Play in Domestic/Captive Animals
Play is a juvenile characteristic and has long been considered as such (Buytendijk, 1933cited by Thinès 1996. In most species, play is at its peak a few days or weeks after birth, but a "zero-level" is reached around 30 days for a mouse (Byers & Walker, 1995), 12 weeks for a pronghorn (Byers, 1997), a few months for a cat (West, 1974), around 5 years for a baboon (Owens, 1975). Our studied species conformed to this rule, as in natural conditions horses stop playing at around 5 years of age, at the latest, and cercopithecids show a negative correlation between age and play (as do chimpanzees) (Mendoza-Granados & Sommer, 1995). What could explain this consensual observation that domestic and captive animals do play at adult stages (Bekoff & Byers, 1981)? The privileged hypothesis was that domestication or captivity may promote a developmental retardation, the retention of neotenic characteristics (e.g., play) and diminished fear (Coppinger & Smith, 1989;Himmler et al., 2013).When wild rats and long Evans rats live in the same conditions, young wild rats play less than their domestic counterparts (Himmler et al., 2013). Strains of rats may differ in their frequency and types of play (Reinhart, McIntyre, Metz, & Pellis, 2006;Sivyi, Baliko, & Bowers, 1997), while play frequency is influenced by paternal origin in foals (Wolff & Hausberger, 1994). Carnivores tend to play more than herbivores (Burghardt, 2005). The idea that domestication, through intended or non-intended genetic selection, may have induced the persistence of juvenile features (amongst which play) at later stages is likely to be true and may explain why pet animals (dogs, cats) are particularly concerned.
However, our study, along with several previous reports (e.g., Sigursjondottir et al., 2003), shows that, when domestic animals, such as horses, live in naturalistic conditions, adult play rarely occurs. On the contrary, our macaque study showed that under intensive husbandry, wild species may develop adult play. Adult Przewalski horses also developed adult play in zoos (Budiansky, 1997). Another interpretation could be that adult play is higher in captive studies because of the removal of predation and foraging pressures (Graham, 2011). Again, this may be true, but the rhesus macaques and horses observed in naturalistic conditions, with no predation or food restriction, failed to exhibit adult play (as do Icelandic domestic horses, Sigurjonsdottir et al., 2003). Thus, this hypothesis cannot explain the observed differences between captive/domestic animals and feral/wild animals. Therefore, these observations raise two concerns: 1) To which extent can studies of play in captive/domestic animals be used for testing its adaptive significance? Most studies would just mix both situations (e.g., Graham, 2011). 2) What other alternative hypotheses could be proposed? Burghardt (2005) noticed that both adult play and stereotypic behaviors occur more frequently in the "barren/boring" environments provided by captive/domestic situations.

Adult Play as an Indicator of Welfare?
Play as an indicator of well-being has been such a consensual view that all the definitions of play include this concept as a core element (e.g., Bekoff & Byers, 1981;Burghardt, 2005). Play behavior has long been proposed as an indicator of good welfare because it is generally associated with better welfare in younger stages (Boissy et al., 2007;Held & Spinka, 2011;Newberry et al., 1988). However, even studies on juvenile stages lead to contradictory findings where social or feeding deprivation may lead to either a reduction of play (e.g., Donaldson et al., 2002) or its enhancement (e.g., Bateson et al., 1990).
As adult play is rare, it may moreover have a different functional significance from juvenile play. For Held and Spinka, "where adult play is naturally rare, its occasional occurrence most likely indicates the animal is currently faring well" (2011, p. 896). How can we explain then that captive adult animals in restricted conditions play more than conspecifics living in more "appropriate" conditions?
While more and more studies suggest a complex relationship between play behavior and stress (Biben & Champoux, 1999;Palagi et al., 2004;Pellis & Pellis, 2009) and wonder whether stressful situations invariably diminish play or whether they can actually enhance play (Pellis & Pellis, 2009), the proper investigation of this question is blocked by the assertion that the definitions of play assume its association with a positive context. However while "the commonly held view is that play is suppressed by stress," recent studies indicate that it can also regulate stress (Palagi et al., 2004, p. 960).
One major obstacle for disentangling the complex relation between play and stress may be that acute versus chronic stress may act in opposite ways on play, especially on adult play. The recent study of Klein et al. (2010) reveals that acute restraint stress diminishes play in male rats, while a repeated chronic stress diminishes all social interactions, but not plays. This finding may be a key element for understanding the contradictory findings on adult play in the literature.
The example of horses is particularly enlightening here. The rare occurrences of adult play in the wild are periods of harsh weather conditions, social disturbances in bachelor groups of young stallions or solicitations by younger animals (e.g., Bourjade et al., 2009;Sigurjonsdottir et al., 2003). This was confirmed in our semi-natural group of male horses that the almost only performed play in December, when colder weather was present and resources were scarce (see above).
Other accounts of adult play in domestic horses deserve consideration. Rebound effects are common and lead horses living in single stalls to increase their locomotors activity and exhibit excited behaviors such as trot, canter but also bucking, prancing and other types of excited behavior patterns when given the opportunity (e.g., being released in a large arena) (e.g., Lesimple, Fureix, Le Scolan, Richard-Yris, & Hausberger, 2011;Wolff, Le Scolan, & Hausberger, 1997). For scientists performing temperament tests, such behaviors are clearly indicators of excitement and emotionality (e.g., Hausberger, Muller, & Lunel, 2011). But for play specialists, these patterns of exaggerated locomotion and movements without any clear immediate function could easily be considered as locomotor play, as they are identified in the play repertoire of foals (McDonnell & Poulin, 2002). Where does play start? How can we properly measure its conditions of occurrence? Is there a common feature in situations of restricted access to space and peers, excitement and play are ways of evacuating stress when the opportunity arises? In that case, observing an adult playing tells us that it is momentarily happier, but does not tell us anything about its welfare. The problem is less crucial for social play, when horses behave clearly differently between mock aggressions and real ones (although play fighting can change into real fights) (Bourjade et al., 2009;McDonnell & Poulin, 2002), but again social deprivation leads to an increase of play fighting in young stallions when reunited with peers (Christensen et al., 2002). Such rebounds effects confirm that, in fact, play is an ambiguous indicator of welfare: In Christensen's study (2002) the young stallions that were raised along with peers did not show this increase. Should we consider that they were in a worse welfare state than the previously isolated individuals? In natural conditions (including stress free semi-natural conditions) when groups are stable, adult horses predominant activities are foraging, slow walking, resting, with very rare occurrences of 'excited' behaviors and play is absent. Does this mean that they experience poorer welfare state compared to their riding school counterparts?
There is no doubt that, at the precise time of play, animals (like humans) experience positive emotions, as play involves neurotransmitter systems that are also involved in the hedonic properties of food and drug rewards (opioids, cannabinoids and dopamine) (Trezza et al., 2011). It has been shown that opioids modulate social play by influencing its subjective positive properties (Trezza et al., 2010).These same systems are however also involved in compulsive games in humans (Frost, Steketee, & Green, 2003) and wheel running in rodents (Sherwin, 1998). Parallels with stereotypic behavior, are interesting: Absence in wild/naturalistic situations, same neurotransmitter and brain global structures involved (basal ganglia/striatum: Graham, 2010). Burghardt (2005) suggested to consider a continuum from exploration (often interspersed with play) to play to stereotypies. Our data on macaque monkeys are here compelling. They revealed a clear relationship between the prevalence of STB and adult play in unfavorable husbandry conditions: Both were more present in some facilities than in others, both were sensitive to the same temporal context and to the same early life experiences. Adult "players" had more chances to be also stereotypic. These findings strongly support Burghardt's hypothesis and open new lines for further investigations on the possible common substrates and causalities.
What these results clearly reveal is that adult play cannot be considered as a "universal indicator of good welfare." Stereotypic behaviors occur mostly in unfavorable conditions (Mason, 1991; and are considered as a reliable indicator of a poor psychological well-being in primates (Crockett et al., 1995). Our findings on adult riding school horses (see also Hausberger et al., 2012) go along with this idea, because the "players" showed more signs of a poor welfare state than the non-players. The correlation between adult play and measures of oxidative stress are especially interesting.
In conclusion, while adult play may indicate that, at the present time, the situation has just improved (e.g., rebound effects after social isolation or spatial restriction), it is clear that adult play cannot be considered as a reliable welfare indicator that could help in evaluating whether or not management of the captive/domestic animals is appropriate. It is often associated with excitement, high arousal, which are not the usual state in which animals living in naturalistic, unrestricted and socially stable groups are observed. If adult play was considered as a welfare criterion, most ungulate populations living in such conditions would be considered as in poor welfare conditions.

Adult Play and its Functional Significance
Considering that play may, in some circumstances/species, reveal an altered welfare gives a new opening to its possible functions that remain so far a mystery (Burghardt, 2005;Pellis, Pellis, & Bell, 2010).
Adult play (like stereotypies) may relieve the sensory deprivation that can prevail in captive/domestic environments (Burghardt, 2005). Play may be a way of exposing an animal to new situations. In humans, chess players have a high sensation seeking profile (Joireman, Fick, & Anderson, 2002). In foals, play is stimulated by novel conditions or stimuli and is often interspersed with exploration (McDonnell & Poulin, 2002). However, in very restricted sensory conditions, horses may on the contrary become withdrawn, unresponsive  or favor active locomotion over play when in group (Benhajali et al., 2008). Sensation seeking is certainly one possible causal factor but does not explain some of our results.
Adult play may be a way of coping with certain types of environmental stressors (Coe et al., 1989). Young rhesus monkeys play more when the troop is under stress (O'Neill-Wagner, 1994). It has also been proposed that stereotypic behaviors may be coping mechanisms. In both cases, there is still no evidence confirming this idea. It has been suggested that animals may use play as "self-medication" (Pellis & Pellis, 2009) because of its association with endogenous opioids, that it favors emotional resilience by "training earlier for the unexpected" (Spinka et al., 2001, p. 143) and that it may be used as an escape (Burghardt, 2005). Unfortunately, there is scarce evidence of such effects. It has been shown that calves exhibiting tongue play have less erosive mouth ulcerations (Burghardt, 2005). Our data do not allow us to conclude on this aspect. Adult horse players were those that showed the most altered welfare, but they were given the possibility to express play only a few hours a week, while they were in restrained conditions 6 days a week, 24 hrs per day, which may not be enough for them to compensate. In macaques, since play and stereotypic behaviors co-occurred, we could not show that being a player may improve the welfare state, but the results do not allow us to formulate clear conclusions. Regarding the hypothesis of adult/captive play being an indicator of potential welfare problems, future research should clearly investigate the animal state at different stages and in different living conditions.
Another interesting hypothesis that would also account for rebound effects after social/spatial restrictions, proposes that play, through "circular and muscular reactions would remove toxic substances that build up in response to physical or mental work and stress" ("recuperation theory, " Carr, 1902, p. 36). The finding that adult horse players showed the highest levels of oxidative stress, and that play frequency was correlated with measures of oxidative stress questions the possible role of play in oxygenation and evacuation of toxins.

Adult Play: A Part of the Behavioral Repertoire?
Our study, along with an abundant literature on a variety of primate species (Palagi et al., 2004), suggests that adult social play may be involved in tension reduction and creation of new bonds. But it most often (apart from maybe bonobos and geladas: Palagi et al., 2004) relates to stressful situations, social disturbance or instability (as shown in horses). However, many studies are performed on captive animals, and our findings show that many factors may influence the prevalence of adult play. Early experience is often disturbed in both captive and domestic animals: Pigs are weaned at 2-6 weeks (instead of 3-4 months) (Jensen & Kyhn, 2000), rats at 21 days (instead of 35 -42 days) (Mason & Latham, 2004), foals at 4-6 months (instead of 9-10 months) , elephants at 3 years (instead of 7-12 years) (Clubb & Mason, 2002), macaques are weaned at 6 months (instead of 1 year) (Camus et al., 2014) and calves are weaned at a few hours of life instead of 9-11 months (Veissier, Lamy, & Le Neindre, 1990). Given the strong impact that early experience had on the prevalence of adult play in our macaques, any study in captive/domestic situation may be biased by this (sometimes unknown) early experience. In the horses studied here, nothing was known about their early life, although most would go through routine procedures that may impact their emotionality on a long term basis (Durier, Henry, Sankey, Sizun, & Hausberger, 2012). Of course, there are differences between species. Despite differences between the social organizations of rhesus, long-tailed macaques, mangabeys and Campbell monkeys, we found the same general trend for adult play to be inexistent or limited in quiet, unrestricted situations, as was also the case in horses. Adult play has probably a different significance in different species and at different ages, but some generalities may help our understanding. Accepting that adult play is not just a sign of "having fun" (Spinka et al., 2001, p. 144) but also a sign showing that the adult organism needs to evacuate stress may help progressing in the understanding of its functions.

Towards a Challenging View of Adult Play: Could Play be a Derived Activity?
One challenging hypothesis proposes that play may be a "derived activity" (displacement). According to Tinbergen (1952), such activities are related to a "general excitement," are characterized by their unpredictability and emancipation from the context ("emancipation from normal situation"). They are generally performed in exaggerated movements and are ritualized. For Tinbergen, such displacement behaviors are a "form of defense against neurotic disorders of the CNS, enabling it to get rid of the surplus of impulses which would otherwise damage it ." (Tinbergen, 1952, p. 23).
Play bouts often include ritualized elements (e.g., canids' bow, Bekoff, 1995). They derive from "actual functional behaviors" but have more exaggerated forms, shorter durations (e.g., McDonnell & Poulin, 2002;Nunes et al., 1999) and are certainly in many cases emancipated from the context. Adult play, given the elements developed in this paper, could well (like self-scratching (Palagi et al., 2004) or yawning  be considered as a derived behavior (that potentially helps to reduce stress and to cope with otherwise unfavorable conditions) (see also Yanagi & Berman, 2014).