1. Introduction
Play can be classified into a number of categories (which are not mutually exclusive), such as social play, individual play, object play and locomotor play. Social play is defined as “a play directed at a conspecific [
1,
2] and in Canidae includes behaviours such as chasing and play-fighting games, mounting behaviour (i.e., mimicking copulatory behaviour) and inhibited biting [
3,
4]” [
5]. In domestic dogs, social play is represented by actions exhibited that are also used during hunting, sexual interactions or agonistic behaviours [
4,
6,
7], including actions unique for play (rearing up and jaw sparring). This category of play requires cooperation with play partners and therefore, we can assume that prosocial mechanisms are important for sustaining it [
8]. In comparison with either object or solitary play, social play is usually performed more frequently and for a longer period in many species [
3]. Hence, it can be understood as a part of the “prosocial toolkit” that needs to be rehearsed and developed in order to facilitate the establishment of longer-term social relationships [
9].
A repertoire of social play in dogs expanded during domestication from unique dog–dog play into dog–human play, and studies show that social play in dogs is a marker of healthy development with a positive effect on social bonds [
10,
11]. However, we must remember that dog–dog play and dog–human play appear to be motivationally distinct [
12] so we focused our study on the intraspecies social play.
The first evidence of social and sexual play in Canids is from 3 to 12 weeks of age, during the period of socialization [
13]. Following the socialization period, the juvenile period lasts from 12 weeks to 6 months or until sexual maturity [
13,
14]. During this time, play continues to be common and play-partner preferences begin to form, so at the age of 6 months, there is a strong unidirectional preference for a play partner [
5]. Dogs and wolves exhibit high frequencies of play behaviours even as adults [
15].
The role of play in Canidae has been questioned for many decades and, currently, many hypotheses about its function exist. Play is likely to be complex and multifunctional, and therefore what appears to be an explanation of function in one species may not be applicable in another [
3,
16]. Firstly, the function of social play may be a reorganization of the youngsters into the adult pattern [
17] and secondly, there is an adaptive function. Play has an impact both on the mental and physical traits of the individual; it can enhance the learning of social skills [
18] and improve motor skills [
19]. Play is also important for testing and strengthening social bonds [
20,
21,
22,
23], the development of emotional flexibility and ability to cope with unexpected situations [
8]. It may be viewed as a kind of “cognitive training” [
8,
24] as it is a way for individuals to understand what specific communication signals mean, e.g., that biting and a specific vocalization is associated with pain [
25,
26]. It is also speculated that social play helps dogs develop the skills necessary for cooperation—for example, reciprocity, development of negotiation and an understanding of which individuals to trust as social partners [
22]. Some studies [
13,
27,
28] support the theory that in Canidae (both domestic and wild ones), social play has an importance for the formation of dominance relationships within litters. When puppies are raised together, they can establish their dominance hierarchy over food (competition for source), or during play without serious injuries [
29].
Play has a definitive direct effect on other social interactions within the group, even if the group consists of members from different species. Some authors state that “winning or losing” games can affect the hierarchy among humans and their dogs [
30]. Others contradict this by stating that play pursues the goal of “equality” [
22]. It means for play to occur, both partners must win an equal proportion of play encounters [
16,
31,
32]. This is called the 50:50 rule [
32] or the egalitarian style of play. For the purpose of equal play, partners use self-handicapping (muzzle lick or voluntary laying down) or role-reversal behaviour. It is known from a previous study by Bekoff [
33] that individuals who do not play symmetrically with partners may be excluded from play, therefore individuals with some advantages should express more self-handicapping or role-reversal behaviour within the play to make it more symmetrical [
21,
32] and to continue playing.
Contrary studies indicate that mammals initiate and play more often with individuals they can dominate [
18,
34]. Bauer and Smuts [
35] tested the 50:50 hypothesis in adult dogs and they found, contrary to prediction, that only 5.4% of the 55 dyads tested displayed 50:50 symmetry during play, whereas 21.8% displayed complete asymmetry. Instead of egalitarian play, older animals showed significantly fewer self-handicapping behaviours than the younger animals did, and smaller dogs self-handicapped more often than larger dogs. It seems that older and larger dogs tend to be more dominant. Ward et al. [
5] also tested this hypothesis within 39 dyads of dog puppies (all of whom were littermates) and they found no evidence supporting the 50:50 rule, actually finding that asymmetry increased as puppies grew older. It seems that dogs use their size, age or higher rank as an advantage over the partner. Therefore, more research is still needed to find a definitive answer for the question of how the 50:50 rule applies to Canid relationships.
The aim of our work was to contribute to the evaluation of symmetry within social play interactions in domestic canines. The second aim is to focus on the effect of sex and body size on social play in two puppy packs of German Shepherd dogs (14 puppies) and if these two variables can possibly affect the equality of play interactions and the frequencies of “winning” within the dyads of German Shepherd dogs (GSDs). We had a unique chance to observe one specific breed (German Shepherd dogs) when all the studied individuals were at the same age. It brings an opportunity to evaluate the impact of body size and sex as primary variables in social play and its symmetry (as the age effect is theoretically removed).
The environmental background of the subjects studied is also unique as the subjects did not need to compete for resources. This is unique to previous studies which were performed within groups of wolves or free-ranging dogs (where competition for limited resources influence the symmetry of play) or within mixed-aged groups (which may be a reason for asymmetry due to advantage of age and experiences). We first tested the hypothesis regarding the proportional distribution of interactions within the group in relation to the sex of individuals. We then tested the second hypothesis that a larger body size correlates to a higher probability of being the “winner” in play interactions. Finally, we investigated whether play in puppies of GSDs is egalitarian.
2. Materials and Methods
The study began in March and lasted until July of the same year. In April we carried out preliminary observations in order to become familiar with the study area and accustom the dogs used in the study to the presence of the observer. We also established the most visible behavioural patterns in the group of dogs and the data collection methods.
2.1. Study Area and Subjects of the Study
Our study was carried out in collaboration with the Slovak Police Headquarters, Department of Kynology and Hippology, where an integral part of the Department of Kynology and Hippology of the Slovak Police Headquarters is the centre for the breeding and training of police dogs. The centre is responsible for the selection of the most appropriate individuals from a genetic pool of working dogs from the German Shepherd breed, and for the breeding of selected individuals which are later used for training and service. This centre allowed us to record behavioural interactions within two litters of German Shepherd puppies (first at the age of 7 weeks and again at 9 weeks of age).
Puppy pack 1 consisted of eight puppies that were German Shepherd dogs (GSDs) (see
Table 1) with an even sex distribution (4 females and 4 males). Puppy pack 2 consisted of six puppies that were GSDs (see
Table 1) which also had an even sex distribution (3 females and 3 males). The total number of individuals from the combined litters (14 puppies equally divided into 7 males and 7 females) remained stable throughout the period of study. The sex of dogs was determined based on morphological characteristics. All dogs were born in the study area, so the dates of birth of each puppy was known and they were bred under the same conditions (including appropriate shelter, space for resting, quality of food and feeding regime, medical care and welfare). They were raised by professional trainers and thus spent some time in the presence of humans from birth. Details about the packs used for the purpose of our study, as well as the information about individuals, are available in
Table 1.
All the animals included in our study were marked by specific dog collars (with a unique visual pattern on the collar for each individual, e.g., horizontal lines, vertical lines, dots, etc.). The filming of interactions took place in the animals’ home enclosures (indoor shelters and fenced outdoor areas).
2.2. Data Collection, Behavioural Observations, and Analysis
All interactions in the litter were video recorded from May until June/July and used for subsequent coding. Observations of groups were recorded twice: first within the one-week period when the puppies were 7 weeks old and when they were 9 weeks old. We obtained 4 sets of continuous records altogether, each approximately 150 to 180 h in duration. After each observation period we measured the chest circumference and weight of each puppy in both litters.
Subsequently, the records were edited so only sequences of footage containing interactions between 2 puppies remained for evaluation. Therefore, any footage that we deemed unimportant to the study, for example when puppies were asleep or during periods of no activity, were cut out. Finally, we randomly selected 10 h of footage from each of one-week period and evaluated them in terms of separate litters and ages.
The behaviour of the puppies was observed when there was no source of competition (no food, treats, toys etc.).
2.3. Play Observations
For the purpose of the study, selected behaviours were assessed. Behavioural patterns during dog–dog interactions were defined and characterized using indicators described by recent studies to be the indicators of play in canines [
5,
35]. We observed interactions defined as offensive behavioural patterns (see
Table 2). Mutual interaction began after an “invitation” from one puppy to another and ended right after one of the playmates moved away or stopped the action. We did not include interactions of unsuccessful attempts ignored by the opposite partner. Only play interactions lasting for at least 60 s were included in the study.
Play interactions were coded as a second play if one of the playmates stopped playing for 20 s or more. Video recordings of the sessions were coded using Observer XT 9.0 software (Noldus Information Technologies). All videos were coded by three independent observers, J.K., T.J., and A.D. The interobserver reliability (three coders) for the play observations was 91.5%.
2.4. Testing of the 50:50 Ratio
According to the 50:50 rule hypothesis, the win ratio (or winning index) of most puppy dyads should be around 50:50 in the context of social play [
20,
22]. To test the equality in the context of social play, a relative winning index was calculated for each individual within each dyad.
In the context of play, offensive behaviours were considered “winning” behaviours, while submissive or self-handicapping behaviours were considered “losing” behaviours (the same terms were used also in other studies on play in dogs or wolves [
5,
35,
36]). In this way, it was possible to calculate the ratio for each dyad. The proportion of winning interactions ranged from 50:50 (complete symmetry of a play) to 100:0 (complete asymmetry).
Subsequently, we analysed how size or sex of the partners within the dyads influenced play behaviour.
3. Results
3.1. Play Interactions within the Litter in Relation to the Sex of Puppies
Data obtained from our observations are presented for each litter (puppy pack) separately at the age of 7 weeks and the age of 9 weeks. We observed 28 dyads in puppy pack 1 and 15 dyads in puppy pack 2. Therefore, 43 dyadic interactions were observed altogether throughout the study. The total number of interactions observed was 2542 (see
Table 3). At 7 weeks of age we observed 1287 dyadic interactions within both puppy packs and at 9 weeks of age, we observed 1255 dyadic interactions in both puppy packs.
We tested the hypotheses discussing whether the distribution of interactions between sexes was proportional within the puppy packs. In both puppy packs, the proportion of sexes was equal (50:50) (see
Table 1). In puppy pack 1, any individual of any sex had an opportunity to interact with the individual of same sex in three out of seven cases (43%) or with the individual of opposite sex in four out of seven cases (57%). In puppy pack 2, any of the puppies has an opportunity to interact with the individual of same sex in two out of five cases (40%) or with the individual of opposite sex in three out of five (60%) of cases. The total numbers of interactions observed within each puppy pack are presented in
Table 3. Results (presented in
Table 3) were statistically analysed with a χ
2 test for goodness of fit at the significance level of alpha < 0.05 (*) (
Table 4). We found that both males and females preferred a partner of the opposite sex slightly more than expected. These results were the same in both puppy packs and at the age of 7 weeks and the age of 9 weeks (
Table 4).
3.2. Dyadic Play Interactions within the Litter in Relation to Size (Weight vs. Chest Circumference) of Puppies
Data on the individuals in both packs are in
Table 1. To evaluate the effect of the body size of puppies on dyadic interactions, we measured the chest circumference and weight of the puppies. To decide which of the two given variables were going to be used for the purpose of evaluation, we analysed the correlation between the weight and chest circumference of the puppies in puppy pack 1 and puppy pack 2. The results were analysed statistically and Pearson’s correlation coefficient was calculated. A moderate to strong positive correlation was found between the two variables (weight and chest circumference) at the level of significance
p < 0.05 (*).
Based on the previous results (correlation of the two variables), we decided to use only weight data to test the hypothesis that puppy size and winning during interactions are positively correlated. Winning indices were calculated for each dyad and both individuals interacting in these dyads (this was calculated separately for pack 1 (
Table 5) and pack 2 (
Table 6)), as was the mean value, median and standard deviation.
A winning index of 0 would represent complete symmetry, with each partner being in the winning position an equal amount of times. A winning index of −1 or 1 represents complete asymmetry, with one individual being in the winning position the entire time the dyad was seen playing [
36].
When testing the hypothesis regarding the correlation between the winning index and puppy’s body size and when calculating the Pearson’s correlation coefficient, we found that although there is technically a negative correlation between the two variables, the relationship between variables was only weak and the results were not significant at
p < 0.05 (
Figure 1 and
Figure 2).
3.3. Testing the 50:50 Rule
Firstly, we identified individuals winning and losing in each interaction within the dyad. Next, we calculated the winning indexes for each individual in a dyad. We also took into account the number of wins of the opposite partner, but did not consider the occurrence of self-handicapping behaviours or role-reversal behaviours.
The 50:50 rule means that the number of wins within the dyad should be the same for both partners. Then the style of play within the dyad could be considered as symmetric or equalitarian. A 50:50 distribution is represented when the value of winning index = 0 and a 40:60 distribution of wins is represented with the values 0.20 (or −0.20).
We tested the 50:50 rule in relation to the age of puppies (7 weeks vs. 9 weeks) and in relation to the combination of play partners’ sexes within the dyads (female–female dyads vs. male–male dyads vs. mixed-sex dyads). The total number of dyads tested was 43.
In relation to the age of puppies, at the age of 7 weeks a distribution of 50:50 (complete symmetry) was observed in 6.98% of dyads and a distribution from 50:50 to 60:40 was observed in 58.14% of dyads. At the age of 9 weeks, a distribution of 50:50 was observed within 11.63% of dyads and the distribution of 50:50–60:40 was observed within 48.84%. Winning indexes represented the degree of asymmetry (or symmetry) within each dyad and were used to test the distribution of that data against the normal distribution. Overall data in our study did not differ from the normal distribution (Kolmogorov–Smirnov test of normality: D = 0.10968, n = 86, p = 0.23441, SD = 0.333617) as well as data for male–male dyads or female–female dyads (not different from the normal distribution).
To test symmetry in play, the Wilcoxon signed rank test was used (WI = 0) for each age group separately. Different results were confirmed for 7- and 9-week old groups. In the younger group, symmetry was confirmed (sum of signed ranks W = 93, 9 = 0.5470, n = 43), but in the older group it was not (W = 370, 9 = 0.0063, n = 43). When testing the symmetry of each litter at different ages (4 groups), symmetry was confirmed for litter 1 at the age of 7 weeks (W = 22, p = 0.7762, n = 28) and at the age of 9 weeks (W = 80, p = 0.2310, n = 28). It was also confirmed for litter 2 at the age of 7 weeks (W = 22, p = 0.5614, n = 15). The symmetry of play was not confirmed for puppies in litter 2 at the age of 9 weeks (W = 98, p = 0.0034, n = 15).
Taking into account the fact that these two age groups must be treated as repeated measures, a Wilcoxon matched-pairs signed rank test was used to test the difference between them (W = 365, p = 0.0267, n = 43). A significant difference was confirmed with the median of differences 0.06 (95% CI: (−0.02 to 0.16)). The Spearman coefficient confirmed the effect of pairing as it was significant and reached a value of 0.5735 with p < 0.0001. Statistical analyses were conducted for the first litter with the following results: (W = 82, p = 0.3588, n = 28), median of differences was 0.06 (−0.02 to 0.16), Spearman coefficient was significant and reached a value of 0.5735 with p < 0.0001. Significant differences between age groups were not confirmed. Results of analyses for the second litter: (W = 83, p = 0.0157, n = 15), median of differences was 0.18 (−0.13 to 0.32), Spearman coefficient did not confirm effecting of pairing, it was not significant −0.1321 with p = 0.3195. Significant differences between age groups were confirmed, but the nonsignificant Spearman coefficient means that the effect of pairing was not significant.
In relation to combination of play partners’ sexes within the dyads, the testing with the Kruskal–Wallis ANOVA was used. The box-plot graph of winning indexes (WIs) for sexually different (F, M, FM) groups of dyads in both age pseudoreplicates shows that the WIs oscillate more or less about 0 (
Figure 3).
Significant differences were not confirmed for the younger group (Kruskal–Wallis test: H (2, n = 43) = 1.2046 p = 0.5475) nor for the older group (H (2, n = 43) = 1.9808 p = 0.3714). Significant differences for symmetry in relation to the sex composition of dyads were not confirmed for litter 1 nor litter 2 (tested separately), but we assume that nonsignificance is likely to be due to the low sample size.