Social Play and Play-Soliciting by Infant Canids. Department of Biology, University of Missouri, St. Louis, Missouri 63121

Transcription

1 AMER. ZOOL., 14: (1974). Social Play and Play-Soliciting by Infant Canids MARC BEKOFF Department of Biology, University of Missouri, St. Louis, Missouri SYNOPSIS. The development of social interaction was studied in infant coyotes, beagles, and wolves. In this paper, social play behavior is discussed. Social play may be characterized in a number of ways: (i) actions from various contexts are incorporated into labile (unpredictable) temporal sequences; (ii) the "play bout" is typically preceded by a metacommunicative signal which indicates "what follows is play"; these signals are also observed during the bout; (iii) certain actions may be repeated and performed in an exaggerated manner; (iv) the activity appears "pleasurable" to the players. By comparing these three species, some insight into the dynamics of social play may be gained. Coyotes were the least successful in soliciting play. When they did play, 90% of all bouts had been previously solicited. Coyotes also tended to use the most successful signals most frequently. This trend was not observed in the beagles or the wolves. By taking into account the fact that infant coyotes are significantly more aggressive than either infant wolves or beagles, the differential ontogeny of social play can be explained. Some of the functions of social play in canids are discussed, and it is concluded that social play is a valid class of social behavior and lends itself nicely to quantitative study. "By the very clarity and simplicity of its expression, the play of children is often an excellent window through which an observer may look upon a culture otherwise concealed by complexities, suspicion, and the barriers of language." (Gardner and Heider, 1968, p. 63). Although there have been many recent reviews of social play behavior in various mammals (e.g., Loizos, 1966, 1967; Millar, 1968; Berlyne, 1969; Dolhinow and Bishop, 1970; Muller-Schwarze, 1971; Bekoff, 1972a, 1974), few quantitative data have been amassed concerning this important category of social behavior. Many workers have concluded that "play" is not a valid class of behavior (e.g., Schlosberg, 1947; Berlyne, 1969; Welker, 1971), because they lump social play behavior together with inanimate object play and/or exploration. These are not the same behaviors (Hutt, 1970; I would like to thank Suzanne King for helping me with observations. I am also grateful to Ms. King and Professor Viktor Hamburger for help in translating some papers. Thanks also to Ms. Carolyn Reynolds for typing the manuscript. At various times our facility had support from PHS grants GM and ES (the latter through the Center for the Biology of Natural Systems, Washington University). 323 Muller-Schwarze, 1971). Social play must be studied in its own right, as involving two (or more) living organisms, capable of exchanging information about an on-going interaction. The level of study may involve the observation of gross movements (postures, gestures, facial expressions), the analysis of vocalizations, or the monitoring of neuromuscular processes (e.g., Welker, 1971). Another source of confusion stems from the fact that definitions of social play range from being totally uninformative (e.g., play is "any socially directed play activity, including rough-and-tumble, approach-withdraw, and noncontact play"; Suomi, 1973, p. 73) to being fairly comprehensive (e.g., Bertrand, 1967, p. 58ff; Muller-Schwarze, 1971; Bekoff, 1972a). Workers should make every effort to indicate what they mean by "social play behavior" in the organisms that they are observing. In canids, there have been few in-depth studies of social play behavior (e.g., Tembrock, 1958; Ludwig, 1965; Zimen, 1971; Bekoff, 1972a,&; Bekoff and Jamieson, unpublished). Social play in canids (and other mammals) may be characterized as follows: (i) actions from various contexts are incor-

2 324 MARC BEKOFF porated into labile (unpredictable) temporal sequences; (ii) the "play bout" is typically preceded by a metacommunicative signal which indicates "what follows is play" (Bateson, 1955); these signals are also observed during the bout; (iii) certain actions may be repeated and performed in an exaggerated manner; (iv) the activity appears "pleasurable" to the participants (Bertrand, 1969; Csikszentmihalyi and Bennett, 1971; Csikszentmihalyi, personal communication; 1 Bekoff, 1974; Reynolds, unpublished). There is usually an initial decrease in distance between the interactants (Bekoff, 1972a), however, this distance may decrease and then increase, such as during chase and running play. In this paper I would like to present data which have been collected on pairs of wolves, coyotes, and beagles, observed under controlled conditions from 3 to 7 weeks of age. This period of life is characterized by important changes in the central nervous system (Fuller and Fox, 1969), and it is also during this period of time that the animals form intraspecific (and interspecific) social relations (Mills, 1898; Scott and Fuller, 1965). These three species of canids show very different courses of social development during early life (Fox, 1969; Bekoff, 1972&), and by following the ontogeny of social play under standardized conditions, a comparative analysis is possible. METHODS The animals observed in this study included four coyotes (Canis latrans), four wolves (C. lupus), and four beagles (C. familiaris). The animals were hand-reared from approximately 10 days of age and were all treated similarly. Until they were able to chew semi-solid food, all animals were tube-fed liquified Esbilac (Borden's bitch's milk supplement). In order to regulate social interaction, the animals were housed separately, and allowed to interact x The interested reader may contact M. Csikszentmihalyi at the University of Chicago for a copy of his paper entitled "Flowing: A general model of intrinsically rewarding experiences." for 15 min per day (at the same time each day, 1 hr after feeding), from 21 to 50 days of age with a similarly reared littermate in a 1.5-mX 1-5-m four-wall arena. The animals were housed together in pairs (with the same littermate with whom they had had previous interaction) after dominance relations were formed or on day 42 (whichever came first), in order to determine if any changes in behavior would result due to cohabitation (e.g., food competition, increased proximity). Thirty-five discrete motor action patterns (body postures, facial expressions, gestures) were catalogued and coded (Bekoff, 1972a, b). Coding the actions facilitated recording the data. Detailed daily notes were also taken. During all observation periods I and the same co-observer were present. DESCRIPTIONS OF ACTIONS USED TO INITIATE SOCIAL PLAY Bow (B): The soliciting animal crouches on its forelegs and elevates its hind-end (Fig. 1); from this position the animal is able to perform a wide range of other movements such as leaping, dodging, and springing back-and-forth. Exaggerated approach (EA): The soliciting animal approaches its prospective playmate in a "loose, bouncy" gait, at a speed greater than that observed during normal walking; this has also been called a "play rush" and "gamboling" in the non-human primate literature; during the approach, the shoulders and head are frequently moved from side-to-side in an exaggerated fashion. Approach /withdrawal (A/W): The soliciting animal approaches its prospective playmate and then withdraws; withdrawal may involve stepping away slowly, or running away for a few meters and then approaching (and withdrawing) once again; it is not uncommon to observe the soliciting animal approach, stop, and then rock backand-forth in one spot, making intention movements of running away. General movements (GM): These include movements of the head and eyes, such as head-tossing and eye-rolling, and also body

3 SOCIAL PLAY IN CANIDS FIG. 1. A "bow" performed by the dog on the right. (After Bekoff 1972a, with kind permission o the movements such as shoulder swaying; stalking of the prospective playmate is also included one animal slowly circles its partner, and then slowly, stealthily approaches; the approach after a stalk may be Quarterly Review of Biology.) exaggerated, and may also involve rapid approach/withdrawals. Face-pawing (FP): This action involves extension of one of the forelimbs toward the face of the other animal (Fig. 2); it is FIG. 2. Face-oriented pawing by the beagle on the right.

4 326 MARC BEKOFF FIG. 3. High leaping during play soliciting by a red fox. not uncommon to observe rapid extension and flexion of the forelimb when the animals are at a few meters from one another (in this case the action is called a "paw intention"). Leap-heap (LL): This action involves two high-amplitude leaps in which the forelimbs are lifted off the ground, and hit the ground, simultaneously. Barking (Bk): See below. We have also had the opportunity to observe some red foxes {Vulpes vulpes) (Bekoff and Fox, personal observation). Some actions which they perform to initiate social play, that are not observed in other Canidae include high-leaping (Fig. 3), flattening out (the animal lies flat on the ground and moves its head from side-to-side), and "high-stepping" with the hind-legs. RESULTS Interobserver agreement was consistently over 95%. The observed animals demonstrated large differences in the development of social behavior (Fig. 4). Note the high frequency of agonistic behavior observed in the coyotes from days 21-28, and the low frequency of play-soliciting during this same time period. Coyotes typically form their dominance relationships between 25 to 30 days of age (Fox and Clark, 1971; Bekoff, 19726; Bekoff and Jamieson, unpublished), and with few exceptions, these social relations are the outcome of severe, unritualized fights. The beagles show an early ontogeny of play and play-soliciting and very low

5 SOCIAL PLAY IN CANIDS 327 levels of agonistic behavior. In fact, the levels indicated on the graph represent very mild threat behavior and absolutely no fighting at all. Likewise, the wolves initially show low levels of agonistic behavior and a higher percentage of play-soliciting. As the frequency of agonistic behavior by the wolves decreased from days (the animals were housed together in pairs on day 42), the incidence of play-soliciting increased. All agonistic behavior consisted of threat, and there was no fighting at all. Table 1 presents data concerning the relative success (P{P/S}) of each of the actions used to initiate social play. The beagles were successful greater than 56% of the S-PLAY-SOLICITING COYOTE AGONISTIC BEHAVIOR t t t DAYS FIG. 4. The median frequency of occurrence (%) of action patterns observed during both play soliciting and agonistic interactions, in relation to the total number of actions performed during the OF AGE stated time periods, t = animals housed together in pairs at the beginning of this time period; vertical bars z=. range.

6 328 MARC BEKOFF Action B EA A/W GM FP LL BK TABLE 1. Play-soliciting by beagles, wolves, and coyotes (days 21-50). Beagles Wolves S(#) a U(#) a P(P/S) a S(#) U(#) P(P/S) x S.D. = rho _77b.c.12 _.48 d Coyotes S(#) U(#) P(P/S) * S = frequency of occurrence of successful play solicits. U = frequency of occurrence of unsuccessful play solicits. P(P/S) = given that there was a play-signal (S), what was the probability that social play (P) resulted. In other words, P(P/S) is equal to the success of soliciting social play. "Based on first 5 actions; X for beagles based on all 7 actions =.75, S.D. =.13. c See figure 5 for "P" values (levels of significance). a Correlation between success of an action (P(P/S)) and the frequency of occurrence of that action. time, using any of the actions. In the wolves and coyotes, general movements were most successful, and bows, second-most. Figure 5 shows the total number of play bouts observed, the overall success (P{P/S}) rates, the levels of significance for the differences P (play/s oliciting) r P(soliciting/play) bouts: FIG. 5. The success with which beagles (B), wolves (W), and coyotes (C) initiated social play. P(play/ soliciting) see Table 1 for explanation; P (soliciting/play) see Table 6 for explanation. The small numbers on the left-hand graph (e.g.,.01) indicate levels of significance (see text). The small numbers on the right-hand graph indicate the individual B w scores for each member of each pair of animals. (1 applies to the first pair of animals and 2 to the second pair see Table 6). Rho = the correlation between the success of a particular action and the frequency with which it was used. Bouts the total number of play bouts observed.

7 SOCIAL PLAY IN CANIDS 329 TABLE 2. Frequency of occurrence of mounting (M), clasping (CL), and pelvic thrusting (PT) by beagles during play.' Days M Males CL PT M o oo Females CL 1 3 T PT ol o o " No mounts, clasps, or thrusts were observed in the coyotes; on day 45, one female wol mounted her brother. observed, the Spearman rank-order correlation (rho) for the relationship between the success of a particular action and the frequency of its occurrence, and the frequency with which bouts were solicited (P{S/P})> The beagles were significantly more successful than either the wolves (df = 8, t = 3.42, P <.01) or the coyotes (df = 8, t = 6.7, P <.001), and the wolves more successful than the coyotes (df = 8, t = 2.20, P <.05). Exaggerated approaches and face-pawing were least successful in the coyotes. The low success of approach/withdrawals in the wolves was due to the inactivity and lack of "playfulness" on the part of one of the animals (see below). It is also interesting that the correlations between success of a particular action and the frequency of occurrence of that action was negative in both the beagles and the wolves, and highly positive in the coyotes. That is, coyotes tended to use the most successful actions most frequently. Table 2 shows the frequency of occurrence of "sexual" behaviors during play bouts. The male beagles performed more of these actions than did the females; mounting and thrusting were first observed on day 27. No "sexual" behaviors were observed in any of the coyotes, and one female wolf mounted her brother on day 45. Table 3 shows the frequency of occurrence of head-shaking (side-to-side shaking of the head while biting), face-pawing and also the frequency of occurrence of biting directed to various parts of the body. The beagles did more face-pawing (it was very successful in soliciting play) and this action persisted throughout" the course of observations. In the coyotes and the wolves, the frequency of face-pawing began to decrease on approximately day 35. The wolves performed more head-shaking and also more scruff-biting, while the beagles bit more at the face and the body (legs, flank, and back). Individual differences TABLE 3. Frequency of occurrence of various actions and bite orientation (days 21-50). Coyote pair 1. In this pair there was a severe, unritualized fight (uninhibited biting accompanied by vigorous head-shaking) on day 25, U female emerging dominant. On this day there was no play-soliciting by either animal, and there was a large increase in self-directed play (tail-chasing) by the subordinate male (Fig. 6). There was a further increase on self-directed play on day 26. Similarly, on days 45 and 46, when there was an increase in aggression by the dominant female, there was an increase in selfdirected play by her subordinate brother. Overall, the subordinate male (C) performed significantly more self-directed play behavior than his dominant sister (135:33; P <.005, sign-test). Usually this behavior either followed an unsuccessful play-solicit (28%) or occurred in the absence of any play- Hsh«FP FB SB GB Beagles (26)- ^90) (287) i(46) Coyotes (i (59) (16) (67) i(57) Wolves ' 881 (68) (35) (130) (115) (246) * Hsh = side-to-side head-shaking accompanying a bite FP = face-pawing FB = face-biting SB z= scruff-biting GB = general biting (flank, body, tail) 'Median

8 TABLE 4. Day of first occurrence of actions used to initiate social play. Coyotes Beagles (1) (2) (1) (2) CU CU BW BW Action Male Female' Male Female* Male Female Male Female 28 b (27) (30) (32) (27) (30) (26) (30) EA (46) (25) A/W (24) (26) (31) J9 (28) (27) (40) (28) (29) (29) (26) (30) (31) GM (27) (35) (29) (27) (25) (26) FP (26) (29) (25) (42) (23) (38) LL (30) (43) B ^ 27 (23) (37) X (23) 28 (26) (33) (28) (28) (27) (34) (28) (35) * Dominant animal. b Day of first success. ( ) = day of first failure. c See text for code. Wolves (1) (2) c u c u Female Male Female Male (28) V) (39) (27) - C I(24) s(34) (26) C 26 o(26) (41) 5 (24) 40 5 (24) 24(26) 9(4 3) (42) (27) (43) (26) 30 (25) 30 (29) 30 (31) 36 (38) n W

9 SOCIAL PLAY IN CANIDS 331 T 24 FIG. 6. The relationship between self-directed play and play-soliciting by the subordinate male in coyote pair 1. The star indicates that on day 25, dominance relations were formed (see text). soliciting (in vacuo) (67%). On a number of occasions, tail-chasing occurred, and then the male briefly stopped and looked at his sister, and then commenced chasing once again. Between days there were only 17 bouts of play, and between days 45-50, there were 23 bouts. The animals did not get more successful in soliciting play as length of time with social experience increased. The subordinate male first successfully solicited social play on day 28 after having been unsuccessful on days 24, 26, and 27 (Table 4). The dominant female first successfully solicited social play on day 28. On day 32 there was an increase in social play by the female, and on day 33, the longest fight of all was observed. Overall, playsoliciting resulted in an aggressive response only four times, the remainder of the unsuccessful solicitations being responded to with "indifference" by the recipient On day 45, when there was an increase in aggressive behavior by the female, the male approached her after play soliciting and she did not respond. Ten seconds later, he approached once again without previously signalling a "play intention," and she responded aggressively. It is also important to point out that of 40 attempts to initiate chase play by the dominant female, only 1 was successful, and this was the only one which was preceded by a play signal (general movements) (Table 5). Table 6 presents the data dealing with the success (P{P/S}) of play-soliciting and also the probability that a bout had been previously solicited (P{S/P}. The dominant female of this pair was both less successful and less solicitous than her brother, and when play did occur, the probability that the subordinate male had previously solicited playful interaction was higher than that observed for his sister. A total of 90% of all bouts were signalled by either animal (see also Fig. 5). Play bouts typically consisted of wrestling, jaw-wrestling, and inhibited biting. "Role reversals" were also observed, the dominant female allowing her subordinate brother to "dominate" her. On a few occasions, play fighting did lead to real fighting. On day 40, for example, the animals were engaged in a reciprocal bout of wrestling and play-fighting, when all of a sudden the dominant female aggressively vocalized and began attacking her brother. He rolled over into a passive-submissive posture (Bekoff, 19726) and remained motionless until she Chase 5dominant y female TABLE 5. Chase sequences in coyotes. Response by subordinate male Turning-away Rolling-over Def. threat No reaction Play Total * The only instance in which chase was preceded by a play-signal.

10 332 MARC BEKOFF Coyote 1: Coyote 2: Beagle 1: Beagle 2: Wolf 1: Wolf 2: C Male U Female" C Female U Female" B Male W Female B Male W Female C Female U Male C Female U Male TABLE 6. Play-soliciting by individual animals observed in pairs." S(#) b U(#) b P(P/S) b # Bouts (P(S/P) b Based on the total frequency of play-soliciting actions b See Table 1 for legend concerning S, U, and P(P/S); P(S/P) = given that there was play (P), what was the probability that it was preceded by a play-signal(s). In other words, P(S/P) is equal to the % of solicited play bouts. c Dominant animal stopped attacking. Eye-contact by the female was sufficient to induce flight and turning away by the male (Fig. 7). Coyote pair 2. In this pair of coyotes, C appeared to be dominant over her sister (U) between 26 and 30 days of age. There was no fight and the relationship was not as clear-cut as had been observed in coyote pair 1. On day 30, U began asserting dominance over C and controlling their interactions from then on. Of the 24 periods in which there appeared to be an assertion of dominance by one of the animals, U was clearly dominant during 20 (83%). From 3 to 4 weeks of age, there was little difference in their success of soliciting play and their "playfulness," C attempting to solicit play eight times, and U attempting to solicit play six times. After U began asserting dominance on day 30, C became more successful in soliciting social play. The animals were housed together on day 35, and between days 36-42, there were no bouts of play, as U became more aggressive. On day 42 there was a dominance fight, U clearly dominant over C. Between days 43-50, C was clearly more solicitous, attempting to initiate social play 21 times to her sister's 8. U was less successful overall, and more bouts were preceded by a play signal sent by her subordinate sister (Table 6). A total of 90% of all play bouts were preceded by a play signal sent by either animal (Table 6). Beagle pairs 1 and 2. It is not misleading to clump the data for the beagles, since the individual animals did not display individual differences as large as those observed in the coyotes or the wolves. In both dyads, Typical BEAGL ES: day 29- play bow. COYOTES: day 32: AIW wrestling eye-con ta c t Sequences pelvicthrusts FIG. 7. Some typical sequences of social play in the beagles and coyotes. EA = exaggerated approach; TA = turning away; CH chase; GB = general body biting; hsh = side-to-side head shaking accompanying the biting; RO = rolling over; A = approach; CR = chin rest; NR = no response; ISO = incomplete stand-over; A/W = approach/ withdrawal; PS == play solicitations. For pictoral representations and full descriptions of these actions see Bekoff, )

11 SOCIAL PLAY IN CANIDS 3S3 FIG. 8. This sequence of photos shows a play sequence between two 30-day old beagles. The beagle on the right performs a bow (A) and his littermate approaches (B). As she approaches he makes direct eye-contact and she turns her head away for a brief moment (C). D, the female (on the left) rears and leaps off the ground; E, the male responds by moving closer. A play bout consisting of wrestling, rolling, and inhibited face-biting occurred (f).

12 334 MARC BEKOFF no dominance relations were formed, and there was a lot of reciprocal interaction. The beagles were the most solicitous and playful of the animals (engaging in 483 bouts of play) and successfully solicited play earlier than their congeners (Table 4). The frequency of occurrence of play bouts for any seven-day period (e.g., 21 to 28, 29 to 35 days of age) ranged from 45 to 69. In the first pair, when W female was inactive between days 29-35, B male engaged in selfdirected play 112 times. Typical sequences of beagle play are illustrated in Figures 7 and 8. During play bouts, more "sexual" actions were noted in the beagles. Also, the beagles were the only animals which barked during the solicitation of social play. (We also observed barking at inanimate objects such as the food bowl and water bucket.) None of the beagles were as successful as the subordinate coyotes (Table 6). Play consisted of reciprocal chase, wrestling, inhibited biting, and rearing and pushing with the forepaws. A breed-typical action pattern, the leapleap, was performed in its highest frequency by the beagles (mean = 165), and it was highly successful in the initiation of social play (Table 1). It was never observed in the coyotes and very rarely in the wolves (mean = 9). When it occurred in the wolves, the leaps were of lesser amplitude. Wolf dyad 1. The wolves in this pair behaved very much alike and there was a good deal of reciprocal play interaction. No dominance relations were formed. Between days and days 29-35, there were 33 bouts of play during each period. The frequency of occurrence of social play precipitously decreased between days to 9, and between days to 10. Play consisted of chase and wrestling. Wolf dyad 2. As in pair 1, no dominance relations were formed during the course of observation. Although both animals were approximately equally successful in soliciting play (P{P/S}) (Table 6), there were large individual differences in "playfulness," U male attempting to solicit social play only 13 times. Unlike pair 1, there was a great deal of play from days 35-50, there being 28 bouts from days and 29 bouts from days The large difference in P(S/P) was due to U's inactivity. C female performed significantly more self-directed play when U was inactive and unresponsive (38:8; P <.05, sign-test). DISCUSSION In canids, play may be characterized as indicated in the introduction to this paper (see Tembrock, 1958; Ludwig, 1965; Eisfeld, 1966; Altmann and Recker, 1971; Bekoff, 1972a,6). To an observer familiar with the species or individuals under study, it is usually not difficult to differentiate social play from other categories of social behavior. Miller (1973) has correctly written that an "... observer who 'knows' he sees play, even if he doesn't know what it 'is', is not necessarily talking off the top of his head" (p. 89). By observing pairs of animals, much may be learned about the dynamics of social interaction in a particular species (Kalmus, 1969; Poole, 1972). In order to gain insight into social ontogeny, conditions of captivity provide a good setting, since individuals may be identified at an early age. In canids in the wild, much behavior occurs inside the den during the first weeks of life, and in coyotes in particular, many valuable data would be lost. Recent observations of an intact litter of coyotes, reared in captivity with their mother, indicated that between days 18 and 35, 85-90% of all social activity took place within the nest box, or directly in front of it (Bekoff and Jamieson, unpublished). A most difficult aspect of the analysis of social play is the study of the temporal sequencing of the actions which are incorporated into a play bout. Most of these actions originate in other contexts. Preliminary analysis (Bekoff, in progress) indicates that the probability that an action will occur during a bout is very little dependent on the nature of the preceding event. The sequences appear more random and less predictable than those observed during actual fighting, courtship, or prey-killing. This is especially true for the beagles and wolves. In mathematical terminology, it appears that most

13 SOCIAL PLAY IN CANIDS 335 play sequences would be categorized as zeroth-order Markov chains (Slater, 1973). That is, after one animal delivers a bite to the rear leg of another animal, side-to-side head-shaking occurs as frequently as biting the other leg, scruff-biting, or rapid withdrawal. This lability of temporal sequencing has also been observed in free-roaming dogs (Ludwig, 1965; Bekoff et al., in progress), and in wolves (Zimen, 1971). Slater (personal communication) has suggested that perhaps different rules apply at different points in the play bout, and that the analysis of temporal sequences is complicated by the fact that two interacting organisms are involved. I do not agree with Lazar and Beckhorn (1974) that "... play activities are seen to be exaggerated, out of sequence, incomplete... only when viewed with respect to adult behavior patterns." Fox (1969) and this author have observed infant coyotes perform normal species-typical predatory sequences as well as engage in true agonistic encounters. When these actions are performed by the same animals during social play, there is exaggeration, lack of sequencing, and incompleteness of action. Thus, play need not be characterized only by reference to adult behavior. Furthermore, I do not agree with Tobach and Schneirla (1968) that "play is a term for immature undifferentiated behavior patterns in the process of organization..." for the above reasons, and for the more obvious reason that adult animals do engage in social play. Species and individual differences The early ontogeny of agonistic behavior in coyotes which results in clear-cut dominance relations (Fox and Clark, 1971; Bekoff, 19726; Bekoff and Jamieson, unpublished) clearly differentiates coyotes from beagles and wolves. Overall, the beagles were seven times as playful as the coyotes and three times as playful as the wolves. (In the domestic dog, breed differences are also evident [Ludwig, 1965; Scott and Fuller, 1965] and whether or not these breed differences are reflected in early ontogeny would be interesting to study [Ewer, 1968].) There were virtually no instances in which we were unable to differentiate play-fighting from real fighting, and I feel that rather than labeling play-fighting as "quasi-agonistic" (McGrew, 1972), I would retain the label "play-fighting." This is because, at least in canids, both the sequencing of the actions and the outcome of the interaction differ from that observed during true fighting. Poole (1973) has recently differentiated true fighting from play-fighting in polecats. Recent data collected on a litter of coyotes (Bekoff and Jamieson, unpublished) suggest that the differences in ontogeny which are evident in the early life of canids, may be useful in understanding later adult social organization. And, as has been suggested for various mammals (Etkin, 1963; Baldwin and Baldwin, 1974), it appears that "animals which play together, tend to stay together." Animals which play less appear to have weaker social ties to their group. Play-soliciting Animals have to be able to differentiate play from "not play." Before the play bout commences, one organism may "metacommunicate" (Bateson, 1955) its intention to play by using various signals. These signals are usually clear and unambiguous, and some appear to be specific to the context of social play (Bekoff, 1972a; Sade, 1973). Furthermore, when two seemingly contradictory signals are given, one of which is a playsoliciting action, priority is most usually given to the play signal (Loizos, 1966; Bekoff, 1972a). In a previous paper I have discussed metacommunication in more depth (Bekoff, 1972a), and I also suggested that some of the play-signals in canids were specifically used to initiate social play. I was referring to the "bow." It still appears that although the bow and other movements are also observed in other contexts (during courtship for example) "play intention" cannot be ignored. Christie and Bell (1972) present quantitative data showing that there is an increase in male-female play during both pre-estrous and estrous in domestic dogs. We have observed courtship between a male malemute and a female

14 336 MARC BEKOFF wolf (Lockwood, Shideler, Bekoff, and Fox, unpublished) and observed numerous bows, along with other actions used to initiate play. Perhaps play serves to strengthen the bond between the courting animals and reduces flight tendencies. After the bout is underway, there is continual feedback between the interactants, and it appears possible that the animals "know" that they are playing because of the changes in the temporal sequencing (Chevalier-Skolnikoff, 1973). Therefore, when an observer does not perceive a play signal, he does not have to feel that he has necessarily missed something, for the "signal" may be in the sequencing itself. On the other hand, a signal may be given which is so subtle as to be almost undetectable to the human observer. By doing frame-by-frame analysis of movie film shot at the "eye level" of the interacting animals, we have observed very slight eye and head movements directed toward the prospective playmate, and at the moment, can only assume that they are of some import to the animals (Bekoff and Jamieson, in progress). The coyotes were significantly less successful than the beagles or the wolves in initiating play. However, when play did occur, 90% of the time it was preceded by a play invitation signal. The subordinate animals were more playful than their dominant partners. Furthermore, in the coyotes, there was a high positive correlation (rho) between the success of a particular action and the frequency with which it was used. That is, coyotes showed a strong tendency to use the more successful signals more frequently, while in the beagles and the wolves there were negative correlations. (In an ongoing study of free-roaming domestic dogs, we have found a correlation of rho =.04.) These facts fit in nicely with what has already been discussed concerning the differences in social development between these three species. Since in the coyote there is an "aggressive atmosphere" prevailing, and during social play the actions observed are usually those otherwise seen in true aggression, it would be important for an animal to communicate its intention to engage in social play and that this intention be perceived and shared by the recipient. In the coyotes, the least ambiguous signals (at least to the observers) were the most successful. The importance of the signal in coyotes is also indicated by the fact that if play did occur after a signal was sent, it was most usually the subordinate who had solicited the interaction, in a sense, assuring the most dominant sib that "what follows is play, and not an attempt to overthrow you!" Similarly, when the dominant female in pair 1 a'ttempted to initiate chase play without previously communicating play intention, she failed 39 out of 39 times. The only successful initiation of chase play occurred after she had sent a play signal. Loizos (1969), in her study of chimpanzee social play, found that in order for chase to result in play, and not flight, the subordinate had to do the chasing. In the wolves, exaggerated head and body movements (e.g., head-tossing and side-toside swaying of the shoulders) were most successfully used to initiate play. In the beagles, all actions were successful mo're than 56% of the time. High-pitched barking by the beagles was also successful in initiating an approach by the partner and subsequent play 56% of the time, and it was often combined with another play-soliciting action. Barking may be a way in which an animal calls attention to itself and play signals may also serve as attention-getting devices (Steiner, 1971), in addition to intention movements. Beer (1973) has written that "an effective signal is one that gets attention and is unambiguous" (p. 69). Sex differences In the present study, no sex differences in social play were observed with respect to the "roughness" of the playing. It is possible that this was due to the fact that the animals were not reared with their mothers, however, observations of other canids reared with their mothers (Bekoff and Jamieson, unpublished, Bekoff, personal observation) indicate no sex differences during 3 to 8 weeks of age. The only sex differences involved the frequency of occurrence of "precocious" sexual behavior by the male bea-

15 gles. Rheingold (1963) points out that only male dogs were observed to mount other animals during the course of her observations. Self-directed play Qualitative observations of the development of social behavior in various canids led Fox (1971, p. 52) to suggest that self directed play might serve as a substitute for social play when the possibility for social play was blocked. Rather than block social play using experimental techniques (e.g., Muller-Schwarze, 1968; Chepko, 1971; Reynolds, Oakley, and Noble, personal communication) we found that this occurs naturally. It appeared that an animal could be "play deprived" even in the presence of a partner(s), if that partner was unresponsive to play solicitations and/or generally inactive. Being unable to successfully solicit play, an organism might then redirect its play towards its own body. When the possibility for social play was blocked, due either to the intolerance of one animal for the proximity of its partner (coyotes) or to one animal's inactivity (beagles and wolves), there were substantial increases in self-directed play behavior by the more active animal. Kruijt (1971) wrote that the most important influence of social experience appears to be that it guides the responses of the organism toward social companions, and in the absence of such partners, the animal forms relationships with its own body. Mason (1965) suggested that in the absence of social companions, an animal disposed toward social play would engage in strenuous motor activities. As mentioned above, absence does not necessarily imply physical absence, but can also refer to unresponsiveness. While the animals we studied did not have the opportunity to play with inanimate objects during observation sessions, observations of free-roaming dogs has indicated that it is not uncommon to observe an unsuccessful solicitor immediately begin playing with an inanimate object. Finally, in-cage observations of these and other canids that have been housed individually for methodological purposes have shown that these animals SOCIAL PLAY IN CANIDS 337 perform very little self-directed play during this limited social isolation at this age (Bekoff and Fox, personal observation). The presence of another animal appears to serve as a stimulus. Self-directed play does not always follow an unsuccessful play solicitation. After a period of time, self-directed play occurred without a prior play soliciting. Since animals adjust their actions to the "expected" actions of others (McFarland, 1966; Bindra, 1969), an animal may learn not to expect another animal to play. After repeated "play deprivation," it may forego soliciting play and engage in self-directed play without prior play soliciting. Bindra (1969) has discussed the above idea in terms of the central motive state (CMS). The CMS occurs through an interaction between physiological states and incentive stimuli. Indeed, more studies are needed concerning the physiological state of playing animals (Muller-Schwarze, 1971). Why play? Space does not allow me to go into a full discussion of why animals play. Various authors have considered the functional aspects of social play (Beach, 1945; Millar, 1968: Dolhinow and Bishop, 1970; Muller- Schwarze, 1971; Bekoff, 1972a), and no one theory of play has been found to be applicable to all animals (Lowenfeld, 1936; Beach, 1945). Indeed, ecological factors are important to consider when discussing even a single species (e.g., Barash, 1973). Animals obviously get physical exercise while playing (Brownlee, 1954), and they may also acquire skills which will increase the facility with which they can move through their environment (Ripley, 1967). There have also been repeated assertions that social play experience is necessary for animals to become "socialized" to conspecifics (Harlow, 1969; Jolly, 1972). In an earlier paper, I overenthusiastically (and perhaps naively) accepted this point of view. There is some evidence that it is not mandatory for animals to have had social play experience in order to acquire species-typical social communicatory skills (Baldwin

16 338 MARC BEKOFF and Baldwin, 1974) and that methods which have been used to experimentally deprive animals of social play behavior have not specifically affected only social play experience. However, social play experience does appear to lead to an increase in flexibility of an individual's behavior repertoire (Miller, 1973) and in the subtlety of social cues to which it can appropriately respond (Baldwin and Baldwin, 1974). It has also been suggested that animals play in order to pre-exercise "instincts" which will be needed in later life (Groos, 1898). Suffice it to say, this theory has not been supported in various mammalian groups (Poole, 1966; Fox, 1969; Welker, 1971; Bekoff, 1972a). What appears to be the case is that play experience increases the "smoothness" of carrying out certain sequences of behavior, but that it is not necessary for achieving the consummatory phase of the sequence (e.g., the killing of prey) (see Beach, 1968). In canids, play appears to be important in learning to control the intensity of the bite and in facilitating the formation and continued maintenance of social relationships within a group. In coyotes (and red foxes) play does not usually occur until after dominance relations are formed, while in wolves, play appears before dominance fights. That the more social canids play more earlier in life is interesting, and there appears to be some relationship between social ontogeny and later social organization in Canidae, namely, that "animals that play together, tend to stay together" (Bekoff and Jamieson, unpublished). Tembrock (1958) wrote that the frequency of play between two individuals was a measure of "fondness." Finally, animals may play because it is a "pleasurable" experience (Bertrand, 1969; Bekoff, 1974). Young animals devote a lot of time and energy to social play, and perhaps this good "feeling" is indicated by the looseness of their gait, the bouncy movements observed during play, and the "smilelike" facial expression. External cues are frequently used by ethologists and other behavioral scientists to infer mood (e.g., Darwin, 1872; Hebb, 1946; Vine, 1970; Dittmann, 1972; Ekman, 1973), and the overt behavior associated with social play indicates a "pleasurable" experience. Neurophysiological studies may provide some further support for this contention (e.g., Lindsley, 1951; Clynes, 1973). In conclusion, social play appears to be an important category of social behavior and should not be dispensed with as being a "wastebasket" into which unmanageable concepts are deposited. Data which have been and are currently being collected on a wide range of animals are providing evidence that social play may be studied in its own right. REFERENCES Altmann, D., and W. Recker Verhaltensanalyse der ontogenese von Steppenfuchsen, Vulpes corsac L. Zool. Gart. Leipzig. 41:1-6. Baldwin, J. D., and J. I. Baldwin Exploration and play in squirrel monkeys (Saimiri). Araer. Zool. 14: Barash, D. P Social variety in the yellowbellied marmot (Martnota flaviventris). Anim. Behav. 21: Bateson, G A theory of play and fantasy. Psychiat. Res. Rep. A 2: Beach, F. A Current concepts of play. Amer. Natur. 79: Beach, F. A Coital behavior in dogs: III. Effects of early isolation on mating in males. Behaviour 30: Beer, C Species-typical behavior and ethology, p In D. A. Dewsbury and D. A. Rethlingshafer [ed.], Comparative psychology: a modern survey. McGraw-Hill, New York. Bekoff, M. 1972a. The development of social interaction, play, and metacommunication in mammals: An ethological perspective. Quart. Rev. Biol. 47: Bekoff, M An ethological study of the development of social interaction in the genus Canis: a dyadic analysis. Ph.D. Diss., Washington Univ., St. Louis. 164 p. Bekoff, M Nonhuman mammalian social play and social development: some perspectives and speculations. Perspect. Ethol. (In press). Berlyne, D. E Laughter, humor, and play, p In G. Lindzey and E. Aronson [ed.], The handbook of social psychology. Vol. 3. Addison-Wesley, Massachusetts. Bertrand, M The behavioral repertoire of the stumptail macaque. Bibl. Primatol. No. 11. Bindra, D A unified interpretation of emotion and motivation. Ann. N.Y. Acad. Sci. 159: Brownlee, A Play in domestic cattle in Brit-

17 SOCIAL PLAY IN CANIDS 339 ain: an analysis o its nature. Brit. Vet. J. 110: Chepko, B. D A preliminary study of the effects of play deprivation on young goats. Z. Tierpsychol. 28: Chevalier-Skolnikoff, S The primate playface: a possible key to the determinants and evolution of play. Paper presented at the American Anthropological Association meetings, New Orleans. Christie, D. W., and E. Bell Studies on canine reproductive behaviour during the normal estrous cycle. Anim. Behav. 20: Clynes, M Sentics: biocybernetics of emotion communication. Ann. N.Y. Acad. Sci. 220 (art. 3): Csikszentmihalyi, M., and S. Bennett An exploratory model of play. Amer. Anthropol. 73: Darwin, C The expression of the emotions in man and animals. D. Appleton and Co., New York (1965, Univ. of Chicago Press). Dittman, A. T Interpersonal messages of emotion. Springer Publ. Co., New York. Dolhinow, P. J., and N. Bishop The development of motor skills and social relationships among primates through play. Minn. Symp. Child Psychol. 4: Ekman, P. [ed.] Darwin and facial expression: a century of research in review. Academic Press, New York. Eisfeld, D Verhaltensbeobactungen an einigen Wildcaniden. Z. Wiss. Zool. 174: Etkin, W Communication among animals, p In J. Eisenson, J. J. Auer, and J. V. Irwin [ed.], The psychology of communication. Appleton-Century-Crofts, New York. Ewer, R. F Ethology of mammals. Plenus Press, New York. Fox, M. W Ontogeny of prey-killing behavior in Canidae. Behaviour 35: Fox, M. W Integrative development of brain and behavior in the dog. Univ. Chicago Press, Chicago. Fox, M. W., and A. Clark The development and temporal sequencing of agonistic behavior in the coyote (Canis latrans). Z. Tierpsychol. 28: Fuller, J. L., and M. W. Fox The behavior of dogs, p In E. S. E. Hafez [ed.], The behaviour of domestic animals. Williams and Wilkins, Baltimore. Gardner, R., and K. G. Heider Gardens of war. Random House, New York. Groos, K The play of animals. D. Appleton, New York. Harlow, H. F Age-mate or peer affectional systems. Advan. Stud. Behav. 2: Hebb, D. O Emotion in man and animals: an analysis of the intuitive processes of recognition. Psychol. Rev. 53: Hutt, C Specific and diversive exploration. Advan. Child Dev. Behav. 5: Jolly, A The evolution of primate behavior. Macmillan, New York. Kalmus, H Animal behavior and theories of games and of language. Anim. Behav. 17: Kruijt, J. P Early experience and the development of social behavior in junglefowl. Psychiat. Neurol. Neurochir. 74:7-20. Lazar, J. W., and G. D. Beckhorn The concept of play or the development of social behavior in ferrets (Mustela putorious)? Amer. Zooi. 14: Lindsley, D Emotion, p In S. S. Stevens [ed.], Handbook of experimental psychology. John Wiley and Sons, New York. Loizos, C Play in mammals. Symp. Zool. Soc. London 18:1-9. Loizos, C Play behavior in higher primates: a review, p In D. Morris [ed.], Primate ethology. Aldine, Chicago. Loizos, C An ethological study of chimpanzee play. Proc. 2nd Int. Congr. Primatol. 1:87-93 Lowenfeld, M Play in childhood. J. Wiley and Sons, New York. Ludwig, J Beobachtungen uber das Spiel bei Boxern. Z. Tierpsychol. 22: Mason, W. A Determinants of social behavior in young chimpanzees. Behav. Nonhuman Prim. 2: McFarland, D. J On the causal and functional significance of displacement activities. Z. Tierpsychol. 23: McGrew, W. C An ethological study of children's behavior. Academic Press, New York. Millar, S The psychology of play. Penguin Books, Baltimore. Miller, S Ends, means, and galumphing: some leitmotifs of play. Amer. Anthropol. 75: Mills, W The nature and development of animal intelligence. Macmillan, New York. Miiller-Schwarze, D Play deprivation in deer. Behaviour 31: Miiller-Schwarze, D Ludic behavior in young mammals, p In M. B. Sterman, D. J. McGinty, and A. M. Adinolfi [ed.], Brain development and behavior. Academic Press, New York. Poole, T. B Aggressive play in polecats. Symp. Zool. Soc. London 18: Poole, T. B Diadic interactions between pairs of male polecats (Mustela furo and Mustela furo X M. putorious hybrids) under standardized environmental conditions during the breeding season. Z. Tierpsychol. 30: Poole, T. B The aggressive behavior of individual male polecats (Mustela putorius, M. furo and hybrids) towards familiar and unfamiliar opponents. J. Zool. (London) 170: Rheingold, H. L Maternal behavior in the dog, p In H. L. Rheingold [ed.], Maternal behavior in mammals. John Wiley and Sons, New York. Ripley, S The leaping of langurs: a problem in the study of locomotor adaptation. Amer. J. Phys. Anthropol. 26:

18 340 MARC BEKOFF Sadc, D. S An ethograra for rhesus monkeys. I. Antithetical contrasts in posture and movement. Amer. J. Phys. Anthropol. 38: Schlosberg, H The concept of play. Psychol. Rev. 54: Scott, J. P., and J. L. Fuller Genetics and the social behavior of the dog. Univ. Chicago Press, Chicago. Slater, P Describing sequences of behavior. Perspect. Ethol. 1: Steiner, A. L Play activity of Columbian ground squirrels. Z. Tierpsychol. 28: Suomi, S. J Surrogate rehabilitation of monkeys reared in total social isolation. J. Child Psychol. Psychiat. 14: Tembrock, G Spielverhalten beim Rolfuchs. Zool. Beitr. Berlin 3: (English translation by B. Piddack) Tobach, E., and T. C. Schneirla The biopsychology of social behavior in animals, p In R. E. Cooke and S. Levin fed.], The biologic basis of pediatric practice; infancy childhood and adolescence. McGraw-Hill, New York. Vine, I Communication by facial-visual signals, p In J. H. Crook fed.], Social behaviour in birds and mammals. Academic Press, New York. Welker, W. I Ontogeny of play and exploratory behaviors: a definition of problems and a search for new conceptual solutions, p In H. MoHz [ed.], The ontogeny of vertebrate behavior. Academic Press, New York. Zimen, E Vergleichende Verhaltensbeobactungen an VVolfen und Konigspudeln. Piper, Afunich.