Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Mammalian Play: Training for the Unexpected Author(s): Marek Spinka, Ruth C. Newberry and Marc Bekoff Source: The Quarterly Review of Biology, Vol. 76, No. 2 (Jun., 2001), pp. 141-168 Published by: The University of Chicago Press Stable URL: http://www.jstor.org/stable/2664002 Accessed: 14-09-2015 20:39 UTC Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/ info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access to The Quarterly Review of Biology. http://www.jstor.org This content downloaded from 169.237.90.178 on Mon, 14 Sep 2015 20:39:51 UTC All use subject to JSTOR Terms and Conditions VOLUME 76, No. 2 JUNE THE QUARTERLY REVIEW of BIJOLOGY MAMMALIAN PLAY: TRAINING FOR THE UNEXPECTED MAREK SPINKA EthologyGroup, Research InstituteofAnimal Production CZ-104 01 Prague-Uhffneves,Czech Republic E-MAIL: [email protected] RUTH C. NEWBERRY CenterfortheStudy ofAnimal Welfare,WashingtonState University Pullman, Washington 99164-6520 USA E-MAIL: [email protected] MARC BEKOFF DepartmentofEnvironmental,Population, and Organismic Biology, Universityof Colorado Boulder, Colorado 80309-0334 USA E-MAIL: [email protected] ABSTRACT In thisreview, wepresenta newconceptualframeworkfor thestudyofplaybehavior, a hitherto puzzlingarrayofseemingly and unrelatedbehavioralelements thatarerecognizable purposeless as play throughout themammalianlineage.Our majornewfunctionalhypothesis is thatplay enablesanimals to developflexiblekinematicand emotionalresponses to unexpected eventsin a suddenlossofcontrol. whichthey weproposethatplayfunctions toincrease experience Specifically, theversatility usedtorecoverfrom suddenshockssuchas lossofbalanceandfalling ofmovements withunexpected situations. stressful over,and toenhancetheabilityofanimalstocopeemotionally " we suggestthatanimals activelyseekand create To obtainthis "training for theunexpected, situationsin play through unexpected thatis, deliberately self-handicapping; relaxingcontrol overtheirmovements oractively intodisadvantageous puttingthemselves positionsand situations. in whichtheplayers switch between moveThus,playis comprised ofsequences well-controlled rapidly The Quarterly ReviewofBiology, June 2001, Vol. 76, No. 2 of Chicago. All rightsreserved. Copyright? 2001 by The University 0033-5770/2001/7602-0001$02.00 141 This content downloaded from 169.237.90.178 on Mon, 14 Sep 2015 20:39:51 UTC All use subject to JSTOR Terms and Conditions 2001 142 THE QUARTERLY REVIEW OF BIOLOGY VOLUME 76 movements thatresult mentssimilarto thoseused in "serious"behaviorand self-handicapping between in-control and outin temporary lossofcontrol.Weproposethatthisplayfulswitching elements is cognitively and ontogenetic constraints on demanding,setting phylogenetic of-control emotional state responses thatproducea complex play,and is underlainbyneuroendocrinological weproposethatplay is often prompted byrelatively novel knownas "havingfun."Furthermore, Wepresent from,exploration. orunpredictable stimuli,and is thusrelatedto,althoughdistinct examiningtheextenttowhichthey 24 predictions thatarisefromournewtheoreticalframework, evidenceand contrasting themwiththepredictions offour empirical aresupported bytheexisting " hypotheaboutplay. Wearguethatour "trainingfor theunexpected majoralternative hypotheses structural,motivational, emotional, sis can accountfor somepreviously puzzlingkinematic, a diversity and phylogenetic aspectsofplay.It mayalso accountfor cognitive, social,ontogenetic, misfortunes. ofindividualmethods forcopingwithunexpected A forthosewho ethologistsagree in their classificationof beMAJORAREAofinterest study "play behavior" centers on the questions: Why and how has play behavior evolved, and how has it been maintained in populations of differentspecies bynatural selection (Bekoffand Byers1998; Power 2000)? There has been little success in achieving a generallyvalid and empiricallysupported answerto thesequestions.The extensivevariabilityof play patternssuggeststo some researchers that play may serve differentfunctionsin animals of different species,and in individuals of differentages and sex withinthe same species (Gomendio 1988; Bekoffand Byers1998; Pellegriniand Smith1998). Some authorseven think that attemptsto find a single function for all play behavior are doomed to failure. For instance,Burghardt(1998a) regardsplay as almost certainlya "heterogeneous category linked togetherby characteristicsthatmaybe superficiallysimilar,but have separate origins, causes, functions,and ontogenies" (p 22). While a pluralistic approach may be the mostappropriatewayto tacklequestionsabout the specific functions of differentmodes of play, the implication that no comprehensive understandingof playis possible is discouraging. If play is a heterogeneous categorywith neither a traceable phylogenetic root nor a common function,then whystudyit as a general phenomenon? Would it not be better to tryto splitit into smaller homogeneous categories? Should we, in the end, treatplayjust on a case-by-casebasis? There are tworeasons to rejectthisposition. First,playbehavior patterns,as diverseas they mightbe, do have featuresin common (Fagen 1981; Burghardt1984; Bekoffand Byers1998). In most cases, both novice and professional havior as either play or nonplay (e.g., Rasa 1984). It is easier to see the unityof play than to find any clear natural borders withinthe richness of play that could be used to split it into more manageable categories. Second, playis nearlyubiquitous in all mammalian orders. This suggestseithera common ancestry or a similarsetofselectionpressuresthatacted in all divergingmammalian orders in the Cenozoic (Byers 1984). In our view, it is more parsimonious to assume that there is a basic phylogeneticand functionalunityunderlying mammalian play than to assume thatthe "superficiallysimilar"playpatternswidelydistributed in mammalsevolvedindependendymany times during mammalian phylogeny.Hence the search fora major functionof play seems to be justifiedand important,as it is forother widelydistributedmammalian characteristics such as sleep (Tobler 1995). In this paper, we propose a major new hypothesis about the adaptive value of play behavior,withrelated structural,psychological, and cognitiveconstituents.We followwith a set of testablepredictions thatarise fromthis theoreticalframework. We then examine some previous hypothesesabout the function(s) of play, contrastingtheir performance against thatof our new hypothesis,based on how parsimoniouslyand consistentlyeach hypothesis accounts for the available empirical information on mammalian play. We argue that our hypothesiscan account for some well-documented kinematic, structural,motivational, emotional, cognitive,social, ontogenetic,and phylogeneticaspects of play. This content downloaded from 169.237.90.178 on Mon, 14 Sep 2015 20:39:51 UTC All use subject to JSTOR Terms and Conditions JUNE MAMMALIAN PLAY 2001 THE "TRAINING FOR THE UNEXPECTED" HYPOTHESIS OF PLAY THE ADAPTIVE VALUE OF PLAY We hypothesizethata major ancestralfunctionofplayis to rehearsebehavioralsequences in which animals lose full control over their locomotion,position,or sensory/spatialinput and need to regain thesefacultiesquickly.Animals learn how to improvisetheirbehavior by chaining conventional movementswithatypical movementsto get themselvesback into a standard position. Sequences thatlink highly efficientspecies-typicalmotor patterns and standard body positions with atypical movements necessary for recoveryfrom awkward positions often occur in biologically significant situations.For example, when fleeing a predator,an animal triesto use the most efficient patternof flight,but maybe disoriented or interruptedunpredictablyby rapid changes in visualinput,actions of the predator,or collisionswithother herd membersor inanimate obstacles. The abilityof the animal to recover rapidlyusing atypicalmovementscould mean the differencebetween life and death in a predator attack. Similar mishaps may occur during intraspecificinteractionsand during pursuit of prey.The opponent (or the prey) adds significantunpredictabilityto the environment.Skilled movementsoftencannot be completed or properlysequenced because of interruptionby the other animal. Besides the development of locomotor versatilityin unanticipated situations,we hypothesize that animals in play learn how to deal with the emotional aspect of being surprised or temporarilydisoriented or disabled. Loss of controlin a serious situation,despite active attemptsto cope, willnormallyresultin activationofbothsympathetico-adrenomedullary and pituitary-adrenocorticalsystems (von Holst 1998). These systemsprepare the animal for immediate action, but can have long-term costs, especially in suppressed immunocompetence (Apanius 1998). In adversesocial situations,emotional overreactionmaylead to undue escalation of conflicts.In the presence of a predator,emotional overreactionleading to aimlesspanic willdecrease an animal's chance of survival.In general, adaptive responses in serious situationsdepend upon the animal's 143 ability to avoid incapacitation via negative emotions. We propose thatthe experience of "self-induced"mishaps during play helps animals to avoid emotional overreactionduring unexpected stressfulsituations. The ultimate benefits obtained from play are probably low,judging from the fact that playis dropped fromthe behavioral timebudget under harsh conditions (Baldwin and Baldwin 1974; Berger 1980; Barrettet al. 1992). However,itmaybe thatindividualdifferences in retainingplayduringharsh timeswere beneficial during evolution. All in all, play could probablybe counted among "opportunitybehaviors"(a termcoined byFraserand Duncan 1998); that is, those behaviors thatbring low ultimatebenefitand are thereforeactuated at moments when the cost of performingthem drops to an even lower level. In short,we propose that play: (i) results in increased versatility of movementsused to recover from sudden "gravitational,""kinematic," or "positional" shocks such as losing ground underfoot,fallingover,being knocked over,being pinned down,or being shaken vigorously; and (ii) enhances the abilityof animals to cope emotionallywithunexpected situations. These mayinclude both "locomotor" shocks as described above, and "psychological" shocks such as suddenlybeing faced with frighteningor dangerous stimuli,unexpectedly meeting a stranger,or experiencing a sudden reversalin dominance. SELF-HANDICAPPING-SEEKING AND CREATING THE UNEXPECTED IN PLAY If play has the functionof trainingfor the unexpected,thenunforeseensituationsshould occur frequendyin play.We suggestthatmammalsactivelyseek and createunexpected situations in play. Specifically,we propose that mammalian playis a sequential mixtureof: (a) well-controlled vigorous locomotor movementssimilarto those used in "serious"behavior that load heavilyon fitnesstraitssuch as escape frompredators,intraspecificagonism, or hunting fast or dangerous prey; and (b) movements during which postural control is compromised, or the chance forrandom factors to influence movement is increased so that the animal is more likelyto be knocked This content downloaded from 169.237.90.178 on Mon, 14 Sep 2015 20:39:51 UTC All use subject to JSTOR Terms and Conditions THE QUARTERLY REVIEW OF BIOLOGY 144 offbalance, fallover,lose controlof a play ob ject, or fail to counter the actions of another animal. Animals can activelyseek and create unexpected events in play through self-handicapping; thatis,deliberatelyrelaxingcontrolover their movements or activelyputting themselvesinto disadvantageouspositionsand situations. For example, animals may self-handicap by moving in a way that is less than fully or byperformingobject mastable or efficient, nipulation while positioned in a waynot best suited for full control over the object. They may also impair theirsensoryand spatial orientation throughhigh-speed angular and rotatorymovements of the head, puttingtheir head into unusual positionsin relationto gravityor horizon, or twistingtheirbody in an unusual way.By using physicalpropertiesof the environmentsuch as deep softsnow,a slippery water, animals slope, or gravity-attenuating can enhance the probabilitythat theywill be thrown off balance into unusual positions. They can also increase the probabilityofexperiencingunexpected eventsbyplayingwithor among relativelynovel environmentalfeatures or among featuresthat are moving in unpredictable ways (e.g., due to wind). In social interactions,animals can self-handicapbyusing positions and movements that impair their competitiveabilityand enable theirplaymates to gain the "attack" position. For example, theymay inhibit the force of their bites and pushes, and allow themselves to be pushed overand chased, even when theyhave the abilityto harm or dominate a playmate.They may also put themselvesat a self-induceddisadvantage byplayingwithlarger,stronger,or more experienced play partners,or even with animals ofa different species. Because playis only performed when its costs remain low, however,there is an upper limitof unpredictabilityand loss ofcontrolabove whichanimalswill not play. THE RELATIONSHIP EXPLORATION BETWEEN AND PLAY Explorationcan be viewedas a seriouscounterpartto play. During an initial encounter with a novel environmentalfeature,animals firstinvestigateit through "serious" exploration, examining whetherit is dangerous and VOLUME 76 whetherit has any resource value to them. If theyfindthenovel featureto be relativelynonthreatening,play may follow. Through their playfulbehavior the animals seem to address the question, "What if it reallywere dangerous?" Although exploration is often temporallyassociated withplay, it differsfromplay in three importantaspects. First,a function of exploration may be to learn how to avoid getting into trouble (by gathering information), whereas we propose that an important aim ofplayis to learn how to get out of trouble (throughenhanced improvisationskills).Second, there is no deliberate self-handicapping in exploration. Third, whereas play is associated witha relativelyrelaxed and secure state, exploration is more closely associated with fear and perceived danger. If an animal loses recontrolin play and has too much difficulty gaining it, or the situation becomes dangerous, the animal should withdrawand reassess the situationthroughfurtherexploration. HAVING FUN-THE EMOTION UNDERLYING IN PLAY playenables aniAccordingtoour hypothesis, byrehearsmals to develop emotionalflexibility ing the emotional aspect ofbeing surprisedor temporarily disorientatedor disabled.Although unexpected events thatoccur in a dangerous in inexperisituationwould likelymagnifyfear enced animals, we suggest that fear is modulated in play by the relativelysafe context in which play occurs and the improbabilitythat losing controlwillhave serious consequences. In addition, regaining control following an unexpected challengeis likelyto be rewarding, and the positivenature of thisexperience may be intensifiedby the rapid repetition of incontrol and out-of-controlelements that occur in play. Thus, we hypothesizethat play is emotionally exciting(perhaps even thrilling, thoughnot intenselyfrightening)and rewardwhile at the same ing,maybe even pleasurable, timebeing relaxed. We suggestthatthiscombination of affectiveattributesis unique to play, producing the complex emotional state that is referredto as "havingfun" in human folk psychology.We propose that the three phenomenological aspects of this "having fun" feelingare directlyreflectedin the kinematic, structural,and motivationalcharacterof play This content downloaded from 169.237.90.178 on Mon, 14 Sep 2015 20:39:51 UTC All use subject to JSTOR Terms and Conditions JUNE MAMMALIAN PLAY 2001 behavior. The excitement is revealed in the vigor and speed of play movements and sequences, thepleasurable aspect in thefactthat animals activelyseek out and workforopportunitiesto play,and the relaxationby the willingnessofanimals to self-handicapin playand play only when theyare relativelysafe or unstressed. It may also be that play behavior is supported by a unique pattern of neurobiological response in the brain centers associated with complex somatosensation, motor patterncontrol,emotionality,and reward. RICH COGNITIVE CONTENT OF PLAY 145 locomotion, increased speed ofbalance recoveryand heightened abilityto regain fullorientation aftera collision or fall; in agonistic behavior,knowledge to avoid undue escalation, fasterconflictresolution, and fewerinjuries sustained;in predator-preyinteractions,more effectiveevasive movements in response to movementsofthepredator,decreased chance of panic and collision withobstacles, and increased chances of escape; in response to novel physicalfeaturesin the environment,a shorterlatencyto shiftfromfreezingto exploration and then to play (exploiting or ignoring the feature); and when encounteringunexpected situationsin a serious context,a less pronounced physiological stress responseespeciallyin relation to the chronic elevation of adrenal glucocorticoidsand compromised immunocompetence. Prediction 2: The benefitfromplaytowards increased locomotorversatility when handling unexpected eventsis mostpronounced within the currentphase of ontogeny,whereasthe improved abilityto cope emotionallywithunexcan be both immediateand pected misfortunes long lasting.Withongoing development,body proportions and locomotor abilities change and the body movement strategiesthat have been rehearsed earlier may no longer be applicable. By contrast,there is no obvious reason whyincreasedemotionalflexibility resulting fromplaywould disappearwithincreasingage. During play,rapid alternationbetweencontrolledand uncontrolled actions requires frequent and rapid assessmentand reassessment of qualitativelydifferentsituationsthatdo not normallyfolloweach other. This implies that playis,even when a solitarylocomotor activity, a cognitively demandingactivity(Allen and Bekoff1997; Bekoffand Allen 1998). When (unpredictable) social interactionis involved,the movements,their sequencing, and, presumably,the accompanyingmental statesbecome even more complicated because theplaymates change the situation with their everyaction. Moreover, the abilityto experience the complex feeling of "having fun" may require a richlydeveloped cognitivesystem.Hence, our functional hypothesisabout play as training forthe unexpected implies that,though such trainingmightbe advantageous formanyanimals,playevolvedonlyin thosewiththe requiKINEMATICS AND STRUCTURE OF PLAY site cognitivecapacity.Also, the level of cogniPrediction 3: Self-handicappingactions are tivefunctionneeded forplay may set a lower ubiquitous elementsofplaythatfrequentlyrelimiton the stageofontogeneticdevelopment sult in temporaryloss of control. That is, play at which play firstappears. sequences in which an apparentlyself-handicapping action is followed immediately by PREDICTIONS being knocked offbalance, slipping, falling, FUNCTIONAL CONSEQUENCES OF PLAY sliding, rolling over, losing control of a play Prediction 1: The amount of play experi- object, or losing the "attack" orientation toence obtained affectsthe abilityto physically ward another animal are frequent. This preand emotionally handle unexpected events diction followsfromour hypothesisthat aniand temporaryhandicaps. Animals deprived mals benefitfromplaybylearninghow to cope ofplay,be itbynaturalcircumstancesor artifi- withsudden losses of control. Prediction 4: Play is sequentially variable. cial manipulation,willbe less able to deal with such situations. Conversely, animals stimu- Play movements that closely resemble those lated into higher-than-usualamounts of play used in "serious" behavior such as predation, willsurpass othersin theirabilityto cope with aggression, or escape are interspersed with such situations.Evidence of improvedcoping other movementsthatare seeminglycountermay include one or more of the following:in productive towards achieving the proximate This content downloaded from 169.237.90.178 on Mon, 14 Sep 2015 20:39:51 UTC All use subject to JSTOR Terms and Conditions 146 THE QUARTERLY REVIEW OF BIOLOGY goal ofthe "serious"behavior.In addition,elements usuallyperformedin differentmotivational contextsmayfolloweach other in rapid succession during a play bout. This predicted variabilityis derived fromour hypothesisthat playenables animals to rehearse how to regain controlafterunexpected interruptionsofnormal behavioral sequences. Prediction 5: Under favorable conditions, mammals of differentspecies are able to play with each other, or at least stimulate each other to play. This prediction arises because certain common features of play, like performingvigorousrotationalmovements,moving body parts in the frontalplane, assuming asymmetrical positions, and switching betweenhighlyskilledand deliberatelyawkward movements,are universaland should be recognized by individualsof differentspecies. SOCIAL ASPECTS OF PLAY Prediction 6: Animalspreferto playwithfamiliarconspecificsover unfamiliarconspecifics, even if the latter are better matched in terms of body size. This is because play involves self-handicapping,which is vulnerable to cheating in animals thatdo not interactrepeatedlywitheach other. Prediction 7: Among pairs of familiarconin bodysize or expespecificsthatdifferwidely rience, play is more often initiated by the smalleror less experienced partner.This predictionis derivedfromthe hypothesisthatthe much smaller or less experienced individual willgain a benefitfromthe experience of copingwithbeing at a disadvantagein the interaction. However,ifthe much largeror more experienced animal is unable to self-handicap sufficiently-tolose some controland experience something unexpected from the movements of the playmate-it should terminate the interactionor fail to respond to attempts by the other individual to initiateplay. Prediction 8: When the strengthor dominance statusof twoplay partnerschanges, the relative degree of self-handicappingby the twoindividualschanges. The strongerindividual will usually performthe most self-handicapping. This is because the main functionof play is to train for the unexpected, not to achieve dominance over,or predate, the playmate. VOLUME 76 Prediction 9: Play is contagious and more than two animals often interact simultaneously with each other in play. There are two reasons forthisprediction. First,conspecifics at play bring much unpredictabilityinto the environment,increasing the number of opportunitiesto respond to unexpected situations.Second, an animal's assessmentthatthe safe to encurrentenvironmentis sufficiently gage in play is strengthenedby the fact that other animals have already startedto play. Prediction 10: Play signals-elements that promote the initiationor continuationof play interaction between two.or more animalsare often derived fromself-handicappingactionsused in play.This predictionarisesfrom our functionalhypothesisthat play involving conspecifics originallyserved as a lesson in how to regain control in social and predatorprey interactions; thus play necessarily included vigorous species-typicalactions that, when performedin a serious manner, would pose a serious threator even immediate danger to the recipient. Hence, for play to be reciprocated, it would have been necessary to informthe recipient thatthese actions would be performedin a nonharmfulmanner. The self-handicappingmovements and positions used to practice for awkwardsituationswere good precursorsforsignals of an individual's playfulintention.When an animal perceives another animal in a compromised position or performinga self-handicappingmovement, the informationthatthisanimal cannot attack available.An animal immediatelyis intiinsically on itsback or side cannot charge immediately, and an animal shakingitshead vigorouslycannot leap precisely,because itsjudgmentof distance and direction is momentarilyblurred. Play signals may also be derived fromexpressions, sounds, and odors associated with"having fun." NEUROPSYCHOLOGICAL INVOLVED MECHANISMS IN PLAY Prediction 11: Severe or prolonged negative emotions, such as fear,pain, anger, hunger, frustration,or malaise, suppress play behavior. This predictionarisesfromour hypothesis that play allows animals to develop adaptive responses to unexpected eventswithoutbeing overwhelmed by negative emotions. Clearly, This content downloaded from 169.237.90.178 on Mon, 14 Sep 2015 20:39:51 UTC All use subject to JSTOR Terms and Conditions JUNE MAMMALIAN PLAY 2001 the preexistence of overwhelmingnegative emotions,or theirarousal during play,would be counterproductiveto this process. Negative emotions are also incompatible with the "havingfun" affectthatis integralto play. Prediction 12: Vigilance decreases during play. This relationship is predicted because some playmovementsimpairthe acuityofsensoryinput. Also, because the rapid switchbetween in-control and out-of-controlmovements and positions makes play a cognitively thereis,presumably,little demanding activity, cognitive capacity leftfor other mental processes to occur simultaneously. Prediction 13: There is a similarityin the neuroendocrinological basis that underlies playbehavior acrossmammalian taxa. We predict thisbased on the hypothesisthatthe specific neurobiological processes that produce the "havingfun" emotion are constitutiveancestralelements of play. Prediction 14: Play results in measurable changes and enhancementsin the brain structures that receive and integratecomplex somatic sensations,control complex motor patterns, and modulate emotional reactions, whereas exploratorybehavior leads, primarily, to changes in associative centers of the brain. We suggest this because, in play, the brain mustdeal withsituationsthatcannot be subsumed under reliable rules but must be solved by kinematic improvisationand emotional flexibility.By contrast,exploration enables animals to gatherinformationabout regularlyoccurringcontingenciesbetweenevents. ENVIRONMENTAL CONDITIONS FOR PLAY Prediction 15: Members of species thattypically live in stable physical habitats perform less locomotoryplay than membersof species thattypicallylivein more changeable physical habitatsor those thatmove between different habitattypes.This is predictedbecause the latter are exposed to greaterenvironmentalunpredictability,and should therefore gain a greater benefit from experience in coping with sudden environmentalchanges that affect locomotion. By contrast, the former should benefitmore fromrehearsingthe specificlocomotor skillsneeded to negotiate establishedroutesthroughtheirterritory byperforming them repeatedly in a nonplayful 147 manner (i.e., throughmotor learning, as described by Stamps 1995). Prediction 16: Play increases in frequency afteranimals move between habitats,experience substantialchanges in habitatthataffect locomotion, or encounter mildlyfrightening or novel stimuli.Through play, animals gain experience in coping withthe novel situation (e.g., snowfallin temperateand arcticregions). Prediction 17: Play occurs only under relativelysafe environmental conditions and is stronglysuppressed when predators or other dangers are detected. Because playappears to produce only modest benefits,it should only occur when the costs remain low. Self-handicapping duringplayincreasestheriskofinjury and predation; however,when the riskof predation is reduced by the close presence of a vigilantparent or caregiver,a large group of similarconspecifics,or a saferefuge,the probabilityof play should increase. ONTOGENETIC DISTRIBUTION OF PLAY Prediction 18: In each species of mammal, play occurs at those ages when there is the greatestdanger of fitnessloss if the animal is caughtin an awkwardsituation.Older animals should playlesswiththe physicalenvironment because theyare morefamiliarwithit.As a result of theirpreviousexperience,the environment is more predictableand there is less need for trainingfor the unexpected. However, play thatinvolvessocial and predator-preyscenarios should continue in adultsofrelativelyintelligentspecies (i.e., withhigh encephalization quotients:Jerison1973; Eisenberg 1981) that adjust their strategiesin response to the unpredictable behavior of intelligentconspecifics, predators,or prey. Prediction 19: Play is most frequent and/ or intense during periods of rapid allometric growth.Rapid changes in body proportions make the control of movementsless precise, resultingin more frequentmisjudged actions. Hence, there should be greaterbenefitsfrom playingat this time to gain practice in recovering frommisjudged actions. Prediction 20: Neonatal mammals play less than juveniles. Neonates, especially those of altricialspecies, lack the physicaland neurological development forvigorous alternation between in-control and out-of-controlposi- This content downloaded from 169.237.90.178 on Mon, 14 Sep 2015 20:39:51 UTC All use subject to JSTOR Terms and Conditions THE QUARTERLY REVIEW OF BIOLOGY 148 tions. Play should become more common as the animals become more capable of intentional self-handicapping. Prediction21: Differentmodes of playpeak at differentphases of ontogenyand, typically, a peak in locomotoryplay precedes a peak in social play.Mammalian conspecificslearn and adapt theirbehavior in response to the behavior of others;thereforethe unpredictabilityof play interactionswithconspecificsis likelyto persistlonger during ontogeny than the unpredictabilitythatoccurs in playstimulatedby elements of the physical environment.Also, the relativelyhigh cognitivedemands of play withconspecificsmay delay the peak in social playversuslocomotoryplay. Prediction 22: Withina population, play is expected in all juveniles of a particular age class. Alljuvenile mammals can gain a benefit from experience in coping with unexpected situations-not just males or the most dominant or strongestindividuals. Prediction 23: Sexual dimorphismin a particular mode of play is found in species in which the twosexes differin how importantit is to be able to cope with the typesof unexpected situations that arise in that mode of play.For instance,ifmales engage more often in forceful agonistic interactions than females, thisshould be reflectedin higherlevels of play that enable males to train for unexpected mishapsin theseinteractions.In slowly maturing species, however, dimorphism in play should be related not only to behavioral dimorphism in adult life, but also (and perhaps mainly) to dimorphismin serious behavior that occurs in the current ontogenetic behavior stage (e.g., "teasing"or showing-off injuvenile or adolescent males thatremain in natal herds, adolescent strivingfor dominance in bachelor herds). We predict that has primartrainingfor movementversatility ily immediate benefits,although benefits in termsof emotional coping can be longer lasting. It followsthat,in most species, sexual dimorphismin play increases withage. PHYLOGENETIC DISTRIBUTION OF PLAY Prediction24: The occurrence ofplayis positively correlated with the encephalization quotient (Jerison 1973; Eisenberg 1981) ratherthan withbody mass or metabolic rate VOLUME 76 withinbroad taxonomic groups. This is predicted due to the cognitivelydemanding nature of play. ALTERNATIVE HYPOTHESES ABOUT PLAY A wealth of explicit or implicithypotheses about the adaptive value of play has been proposed (Bekoff and Byers 1998; Burghardt 1998b; Power 2000). Traditionally,emphasis has been placed on the development of play, because animals are usually more playful when young than as adults (but for examples of adult play, see Breuggeman 1978; Biben 1979; Fagen 1981; Zucker et al. 1986; Pellis et al. 1993; Hall 1998). Consequently,functional explanations have usuallyfocused on waysin which play during early development could benefitanimals as adults. Hypothesized functions of play have included motor training, and social and cognitivedevelopment. However,examinationof thesehypothesesthrough play deprivation(Miller-Schwartze1968; Caro 1980), naturalsuppressionof playunder harsh environmentalconditions (Baldwin and Baldwin 1974, 1976; Bekoff 1976; Sommer and Mendoza-Granados 1995), or correlationsbetween results of play fightsand later dominance (Adams and Boice 1989; Araba and Crowell-Davis1994) has revealed surprisingly little evidence for long-termeffectsof play; attentionhas also been paid to more therefore, immediate benefitsof play in young animals. For example, play may provide opportunities for self-assessment(Thompson 1996, 1998), learningroutesthroughthe natal area (Stamps 1995), gaining dominance (Pellegrini and Smith 1998), modulatingjuvenile aggression (Drea et al. 1996), masteringimmediatefoodrelated or spatial cognition skills (Chalmers and Locke-Haydon1984; Bjorklundand Brown 1998), or achievingheat/energydispersal(Barber 1991). We brieflyoutline the four most clearly delimited and frequentlypromoted general hypothesesabout play (Bekoffand Byers 1998; Burghardt1998b). A simple explanation forthe apparent unity of mammalian play is thatplay is a side effect of other specific featuresof mammalian life historyand physiology.According to the "surplus resourcemodel"(Burghardt 1988, 1999) and itsmodificationby Coppinger and Smith (1989), young mammals, in contrastto rep- This content downloaded from 169.237.90.178 on Mon, 14 Sep 2015 20:39:51 UTC All use subject to JSTOR Terms and Conditions TUNE 2001 MAMMALIAN PLAY tiles,can affordto play because theypossess an aerobic metabolismthatenables prolonged and because theyreceive pavigorousactivity, rental care, which includes the provisioning ofmilkand a degree ofprotectionfrompredators.Play,according to thismodel, is a wayin whichjuvenilemammalian behavioris remodeled into adult behavior; this "metamorphic" character of play explains its variability.The surplus resource model is not an hypothesis about the adaptive value of play. On the contrary,it proposes that the primaryprocesses that lead to the phylogeneticappearance of play are merelyside effectsof other typically mammalian adaptations. Adaptive modifications of play evolved later and contributedto the diversificationof play in various taxa. Hence, this hypothesis does not propose a general functionofplay,but ratherdelineates theaggregateofproximate,phylogenetic,and ontogeneticfactorsthatmake up the singularityof play. The "surplusenergy as originally hypothesis," put forthby Friedrich Schiller and later by Herbert Spencer (1898), noted that "higher animals"obtain betternutritionthan "inferior species" duringearlyontogeny,and therefore have to let offthe excess energyin play.Barber (1991) made the hypothesismore specificby arguing that young mammals are often not limitedin energy.They benefitfromplay because thevigorous,3-dimensionalmovements and the associated emotional arousal activate thesympathetic nervoussystem.Resultantacute and chronic increases in metabolic rate, and especiallyin brownadipose tissuethermogenesis, may produce a defense against cold and obesity,and enhance resistanceto pathogens. According to the surplus energy hypothesis, playprovidesimmediate benefitsthatare not directlyrelated to the specificformof play. Perhaps the most intuitiveand widely accepted theoryabout the benefitsof play is the practice theoryof Groos (1898). According to thistheory,playin earlyontogenypreparesthe animal forlaterlifebyperfectingcriticalbehavioral patterns.Among numerous offshootsof this theory,the motor traininghypothesisof Byersis the best developed. Originally,Byers (1984) suggested that the ancestral form of playwas locomotoryplay and itsfunctionwas motortrainingto avoid predators.During fur- 149 therevolution,playwas modified,specialized, and diversifiedin differenttaxa according to selective pressures acting on the respective populations. For instance, in highlygregarious and polygynouslarge-bodied ungulates, social play evolved which closely mimics agonistic competition, especially among males. Later, Byers and Walker (1995) refined the hypothesis.They argued that performancedependent muscle fiber differentiationand cerebellar synaptogenesis are the probable long-termeffectsof the specific movements performed during play. Byers (1998) labels this proposal the "sensitiveperiodhypothesis." The sensitive period hypothesis propounds veryspecific and permanent (neuro) physiological effectsof playwhich lead to enhanced skill,an optimalspeed/forceratio,and greater economy of important species-typicalmovement patterns. Thompson (1998:192) recentlysuggested another general functionof play,namelythat it provides developing individualswithimmediate feedback on their physical abilities; in other words, play is an efficientmethod for "self-assessment." According to thishypothesis, young animals test themselvesby repeatedly performinga challenginglocomotor or social action, or object manipulation. If theyrepeatedlysucceed in performinga task,theyswitch to a more challenging task; if theyfail, they may switchto an easier task. The self-assessment hypothesissuggeststhatbenefitsof play are immediate and mostlyin the cognitivedomain. In Table 1, we have attempted to identify how the predictionsderivedfromour hypothesis differfrom those of the four alternate hypotheses.To do thiswe firstconsidered predictions listed explicitlyby Burghardt (1988) and Barber (1991), and some predictionssuggestedbyByers(1998) and Thompson (1998). Next we considered predictions which, according to ourjudgement, are clearlyimplied bythe hypothesisin question. For each of our predictions,we decided whethereach of the other four hypothesesleads to: (a) an identical or similar prediction (Agreement), (b) a differentbut not opposite prediction (Alternative), (c) an opposite prediction(Contrast), or (d) no prediction. In the next section, we examine our new This content downloaded from 169.237.90.178 on Mon, 14 Sep 2015 20:39:51 UTC All use subject to JSTOR Terms and Conditions THE QUARTERLY REVIEW OF BIOLOGY 150 VOLUME -C C/) CJO 00 0 bD m bi) 4-1 C's 4.1 IZ. Cld 4-1 Cld C', 4-1 '4-J 101 Ici cn Cld bD C,d Cld bD bfi Z cn (4-4 4-J Cld 00 C'd Ici 142, b cn C'd (4-4 Cld m 4-1 C'd C>,dCld m C/) Q v ; Cld 10 Cd CL4 v 4- Cld C'd > ;.4 4Cn v 4-1 W bo cld > ZZ 00 10 OC) cld 4-1 > C's 84 41 bfi C,,d Cld 'n Cd 4 k > 4-1 u ct Cld 10M 4-1 4-1 Cd P-4 C-Id > 0 w Cd C/) (4-4 Cld 4-( >, tn 7q bD r. Cld > z Cd Cd w w Ci C>ld 4-J 1 41 Cld CZ This content downloaded from 169.237.90.178 on Mon, 14 Sep 2015 20:39:51 UTC All use subject to JSTOR Terms and Conditions 76 JUNE 151 MAMMALIAN PLAY 2001 bi S :!~~~~~~~~c O 5 , 'A 'AC' ? . (4 S~~~~~~~~~~4 b n~~~~. g o ,r 0 _ a u > Q~~~~~~~~~~~~~C *g b.>00 t X X, ) , . _Z This content downloaded from 169.237.90.178 on Mon, 14 Sep 2015 20:39:51 UTC All use subject to JSTOR Terms and Conditions THE QUARTERLY REVIEW OF BIOLOGY 152 VOLUME C~~~~~~~~~~~~~~~~~~~~~~~~~~~~~C ~ 00 ~ ~ ~ C' o C' C ~ C ~~c 2S~~~~~~~~~~~S ~C~3C H ~~ H ~ C~ -e~~~~~~~~~~~~~~~~~~~~~~~~~' - ~~ ~ ~ ~ ~ ~ ~ ~ c~~~~~~~~~~~~~~~~~~~~~~~C a) m u C)H 0. >~~~~~~~0?> ' CC UO~~e Su ~~~~~~~~~~~~CF R C.- C uL4-4 Cld 5 CA~~~~~~~~~~ C*~~~~~~~~~~~EC i~~~~lE - 0 C0 $ OQ~AC. This content downloaded from 169.237.90.178 on Mon, 14 Sep 2015 20:39:51 UTC All use subject to JSTOR Terms and Conditions 76 JUNE 153 MAMMALIAN PLAY 2001 S~~~~~~~~~~~~~~~~~~~~' _ = IVz 00~~~~~~~~~~~~~~~~~~~~~~~~~~~~~0 C's NN~~~~~~~~~~~~~~~~~~~~~~~~' =~~~~~~~~~~~~~~~~~~~~~~' .-.:q0U =~~~~~~~~~~~~~~~~ O ;~~~~~~~~~~~~~~~~~' CS C' o~~~~~~~~~~~~~f 'C : c f This content downloaded from 169.237.90.178 on Mon, 14 Sep 2015 20:39:51 UTC All use subject to JSTOR Terms and Conditions THE QUARTERLY REVIEW OF BIOLOGY 154 .~~~~~~~. C O E t VOLUME X S o~~~~~~~~~~~~~b W g D ~~~~~~C4 DB 8 , z~~~~~~~~~~&~~~~~~~~~~~~~~E v a ? d This content downloaded from 169.237.90.178 on Mon, 14 Sep 2015 20:39:51 UTC All use subject to JSTOR Terms and Conditions 76 JUNE 2001 MAMMALIAN PLAY playhypothesisagainstthe above fourhypotheses. We discusstheexistingempiricalsupport forour predictions,and suggestfutureresearch needed to investigatethe validityof each hypothesis. SUPPORT FOR THE "TRAINING THE UNEXPECTED" FOR HYPOTHESIS AND ITS ALTERNATIVES FUNCTIONAL CONSEQUENCES OF PLAY In this section, we refer to correlational studies that have investigatedwithin-species naturalvariationin playand laboratorystudies thathave examined the effectsof playdirectly through experimental manipulation of the amount of play. There is some correlational support for Prediction 1 of the "trainingfor theunexpected" hypothesisin human studies. Pellegrini (1995) found that the ability of young boys to switchactivitiesduring roughand-tumbleplay is correlated withtheirlater scores on a testof social problem-solvingability.Thus, experience gained from switching between activitiesin play may be associated withgreaterbehavioralplasticity in unexpected or novel circumstances.Similarly,Saunders et al. (1999) found a correlation between playfulnessand coping skillsin preschoolers. To our knowledge,our predictionthatplay experience increases movement flexibility when recoveringfromawkwardpositions has not been tested.On the other hand, thereare experimental data that support our prediction that the abilityto cope emotionallywith challengingsituationsis compromised byplay deprivation in the juvenile period. Most of these data stem from laboratoryrats (Rattus norvegicus) deprived of social playvia isolation during the fourthand fifthweek of life and testedlater (between 85 and 120 days of age) in a varietyof social and nonsocial challenge tests.Potegal and Einon (1989) showed that these ratswere more prone to attack and injure another rat after receiving an electric shock fromthe floor grid. This resultcan be interpretedas showing thatsocial play deprivation produced a strongerreaction to an unexpected and unidentifiable aversive situation. Van den Berg et al. (1999a) found that ratspreviouslyisolated as juveniles undulyescalated conflictswhen confrontedwitha resident rat in their territory,and had higher 155 plasma corticosteroneand epinephrine concentrations,compared withnonisolated controls.Rats isolated as juveniles also had a reduced motivationfor social contact with an alien rat during the first10 minutes afterbeing placed in a neutral arena, although this trendtended to reversein the second 10 minutes (Hol et al. 1999). In other studies, rats isolated asjuveniles have displayedbehavioral hyperactivityin open novel environments (Gentsch et al. 1988), slowerhabituationrates in an open field (Einon et al. 1978), and greateravoidance of open spaces in a preference test (da Silva et al. 1996). The disadvantageof thesestudiesis thatisolation deprives the juvenile not only of play but also of other formsof contactand sensory stimulationfrom companions. However, the findingthat the daily provisionof a drugged (nonplayful) partnerduring the isolation period did not prevent the effectsof isolation, whereas providinga normal play partnerdid prevent these effects(Einon et al. 1978; Potegal and Einon 1989), suggeststhatthe consequences of social isolation may indeed be due largely to play deprivation. As a whole, these resultssuggest that play deprivationresultsin increased fearand uncertaintyin novel environments,and more escalated aggressive behavior towardsconspecificsin serious conflicts,as predicted from our functional hypothesis. Bycontrast,thereis littledirectevidence for the adaptive effectsof play predicted by Barber's surplus energy hypothesis.The prevention of obesitythroughheat production does not seem plausible,giventhatenergyexpenditure in play is generally low. Even playful young kittens(Felis catus) allocate only 4% of theirenergyexpenditureto play(Martin1984). Furthermore,there are examples of intense playin situationswhen the preventionof obesitywas not an adaptive goal. Kittenswhose milksupplywas experimentally limitedshowed more play than normal (Martin and Bateson 1985), and young harborseals (Phoca vitulina) were observed to play most at the time of fat accumulation (Renouf 1993). The motor training/sensitiveperiod hypothesisin itsstrongform(playduringa sensitive period is necessary for normal behavioral development) is rebuttedby the factthat,de- This content downloaded from 169.237.90.178 on Mon, 14 Sep 2015 20:39:51 UTC All use subject to JSTOR Terms and Conditions 156 THE QUARTERLY REVIEW OF BIOLOGY VOLUME 76 spitesuppressionofplaydue toveryharshnat- (Gentschet al. 1988; Potegal and Einon 1989). ural conditions or laboratorymanipulation, These resultssuggestthat,contraryto the monormalprey-catch- tor training/sensitiveperiod hypothesis,the animalsdisplayqualitatively ing and social behavior (Baldwin and Baldwin behavioral effectsof play deprivationare not 1974, 1976; Bekoff1976; Caro 1988; Sommer (solely) mediated by permanent effectson and Mendoza-Granados 1995). Even in itsless cerebellar synaptogenesis and muscle fiber strongform (play experience during a sensi- differentiation. tiveperiod affectsthe formand efficiencyof KINEMATICS AND STRUCTURE OF PLAY species-typicalmovements), there is contrathat the dictoryevidence. Caro (1980) found The ubiquitous presence of self-handicapstimulation of extra object play in kittens ping positions and movements in play sethroughprovisionof suitable objects did not quences is themajorpredictionofour hypothealtersubsequentpredatorybehavior.Playdepri- sisabout theformofplaybehavior (Prediction vationin rat isolation studies has generallyre- 3). Self-handicappingelements need not resulted in differencesin the quantity,rather semble any "serious" adult or juvenile motor than quality,of adult social behavior. For ex- pattern,as theycan mimicmovementsthatocample, prior isolation of ratsincreased the la- cur withoutan animal's active contribution. tencyand decreased the frequencyof social On the contrary,the motor training/sensitive behavior elements,but did not alter theirba- period hypothesis implies that play movesic formand sequential patterning (van den ments are similar to "serious" adult moveBerg et al. 1999b). Given the similarityof cer- ments because optimal permanent tuning of tain play movementsand elements of sexual muscle typesand cerebellar synapsescan only behaviorin male rats (Pellis 1993), adult male be achieved ifthe "right"movementsare persexual behavior motor patternsshould be af- formed during the sensitiveperiod. The selffectedbyjuvenile playdeprivation,according assessmenthypothesisis also at variance with to the motor training/sensitiveperiod hy- the occurrence of self-handicapping movepothesis.However,van den Berg et al. (1999a) ments because animals should "play to win" found thatthe capacityto performsexual mo- (Thompson 1998:197); thatis, theyshould try tor acts was intactin ratsisolated as juveniles. to achieve success in a playbout withmaximal Also, theirintromission/mountratio,whichis effortbecause, otherwise,theirself-assessment reduced in exhausted or inexperienced male will be inaccurate. According to the surplus rats,was unaffected;thustheircopulating skill resource model, "imperfect" elements are was comparable to controls. possible in play. They should stillbear some hypoth- resemblance to adult "serious" behavior eleThompson's (1998) self-assessment esis predictsthatreduced play experience re- ments,however,because play is, according to sultsin an impaired abilityto assess risk,costs, that model, a behavioral bridge between the and benefits.The effectsof social isolation on periods ofjuvenile dependence and adult reemotional coping abilityin rats are compati- sponsibilities(Burghardt 1988:130). ble withthisprediction as well as our own. In supportof our hypothesis,thereis ample In regard to the timingof beneficial effects evidence thatself-handicapping-in thesense of play (Prediction 2), Byers (1998) argued thatanimals deliberatelydo not exertfullconthat play has nonnegligible energy and sur- trolover theirmovementsor activelyput themvivalcostsforjuveniles,but no benefitsforju- selvesinto disadvantageous positions-is omvenile survivalhave been documented; there- nipresent in play. First,a prominent feature fore the adaptive benefitsof play must occur ofplaybehaviorelementsis thattheyare exaglater in life. However, studies that investigate gerated (Bekoffand Byers1981; Fagen 1981; the impact of play experience on short and Burghardt 1984). Playful running and aplong-termrates of mortalityfrom different proaches have been described as "bouncy"in causes are lacking. As to the permanency of domestic dogs (Canis familiaris),wild canids, play effects,rodent play deprivation studies and piglets (Sus scrofa) (Bekoff 1974; Biben show thatmanyeffectson behavior can be re- 1983; Newberryet al. 1988); playfulsocial inversed or alleviated bylater social experience teractions as "capriciously exaggerated" in This content downloaded from 169.237.90.178 on Mon, 14 Sep 2015 20:39:51 UTC All use subject to JSTOR Terms and Conditions JUNE 2001 MAMMALIAN PLAY chimpanzees (Pan troglodytes) (Spijkerman et al. 1996); hops as "frisky" in punares (Thrichomysapereoides)(Wilson and Kleiman 1974; Thompson 1998) and "frantic"in rats (Panksepp 1998); torsomovementsas 'jerky"in bighorn sheep (Ovis canadensis) (Berger 1980); manipulation of objects as "boisterous"in cats (Felis catus) (Hall 1998); and water play as "ebullient" in harbor seals (Phoca vitulina) (Renouf 1993). These adjectivesare shortcuts for sayingthat the play elements are too disproportional, too distorted,too fast,too uncontrolled,or too quicklyrepeated in comparison with their "serious" counterparts and, hence, are less suited to achieving the apparent proximategoal of those movements,such as escape, effectiveobject handling, or overpoweringan opponent. Second, vigorous rotational elements are among themostfrequentelementsin playand are prominentlyfeaturedin the play descriptions of many species. These movements include head shakes, tosses and jerks, neck and torso twists,heel kickingwithlateral rotation, high speed turns,somersaults,and pirouettes (Mutller-Schwartze 1968; Gentry1974; Wilson and Kleiman 1974; Byers 1977; Fagen and George 1977; Berger 1980; Biben 1983; Pellis and Pellis 1983; Byers 1984; Rasa 1984; Gomendio 1988; Rothstein and Griswold 1991; Renouf 1993; Miller and Byers 1998; Byers 1999). In a systematicstudy,Byers (1984) observed head shakes, head jerks, or neck twists in the play of all 12 ungulate familiesinvestigated; theywere the second most widely occurring type of behavior found in play after running. Most rotatoryactions involve highspeed angular movement of the head and, hence, impairsensoryand spatial orientation. Third, play movementshave more degrees of freedom than other typesof behavior (Golani 1992; Fontaine 1994). For instance,mammals usuallydo not use rotationin the frontal plane in their"serious"behaviorbut oftenadd thisdegree of freedomto theirmovementsin play. Primatesuse head tiltsto communicate playfulmotivation(Sade 1973), and ungulates oftenkicktheirheels to one side duringlocomotoryplay. Baldwin and Baldwin (1974) reported thatsquirrelmonkeys(Saimirisciureus) signal play intent by looking between their legs. We may speak about deliberate kine- 157 matic plasticityin play that mimics the "enforced kinematicplasticity"that mightbe inflictedby an external force. These positions are self-handicappingbecause the animal sees the environmentfromunusual angles and is placed in bilaterallyasymmetricpositionsthat limitkinematicoptions in the next move. Fourth, animals in play often performobject manipulation or social interactionsin disadvantageous positions. Differentspecies of primatesexhibit play behavior while hanging bytheirhands,feet,or tails(Baldwinand Baldwin 1974; Aldis 1975), or they replace the usual quadrupedal walking withunstable bipedal locomotionduringplay (Fontaine 1994). Young animals ofmanypredatorymammalian species exhibitplaybehaviors (sociallyor with objects) while lying on their side on the ground (Fagen 1981). Fifth,use of self-handicappingactions by largeror strongeranimals thatengage weaker partnersin play has been reported in, forexample, five species of macropodoids (MacM. rufogriseus, M. ruropusparryi,M. robustus, fus, Thylogalebillardierii)(Watson and Croft 1996; Watson 1998), Africanelephants (Loxodontaafricana)(Moss 1988:163), coyotes(Canis latrans) (Bekoff,unpublished observations), squirrel monkeys (S. sciureus)(Biben 1989), hamadryas baboons (Papio hamadryas)(Pereira and Preisser 1998), and chimpanzees (P. troglodytes) (Mendoza-Granados and Sommer 1995). Our second structuralpredictionis thatplay is sequentiallyvariable (Prediction 4). High sequential variabilityof play has been documented in several species. Bekoffand Byers (1981) reported thatsequences of play in infant canids were significantlymore variable than sequences of agonistic behavior. Newberryet al. (1988), observingpiglets (S. scrofa) in a seminaturalenclosure, identifiedsix play markers (elements occurring only in play) and found that all 22 remaining behavioral elements in the ethogram occurred at least once just before or just afterone or more of the play markers.Rasa (1984) observed that dwarfmongooses (Helogale undulata) switch veryfrequentlybetween behavior representativeof preycapture,social interaction,enemy avoidance, and exploration during play, and Gentry(1974) reported similarvariabilityin This content downloaded from 169.237.90.178 on Mon, 14 Sep 2015 20:39:51 UTC All use subject to JSTOR Terms and Conditions 158 THE QUARTERLY REVIEW OF BIOLOGY the play of Stellar sea lions (Eumetopiasjubatus). Where locomotory, social, and object play occur in the repertoireof a species, the relativefrequencies of these differentmodes of play may change during ontogeny (Newberryet al. 1988), during the course of a day (Byers 1977), or even within a play session 1968; Rasa 1984). Never(MfXller-Schwartze theless,elements of these threemodes of play (or at leastthoseof locomotoryand social play) are often intermixedwithina single bout of play (Inhelder 1955; Muller-Schwartze1968; Gentry 1974; Aldis 1975; Biben 1983; Pellis and Pellis 1983; Newberryet al. 1988; Pedersen et al. 1990; Mendoza-Granados and Sommer 1995; Vanderschuren et al. 1997; Thompson 1998). thepresence ofself-handicapInterestingly, ping elements in play is implied by the most widelyused definitionof play: "playis all motor activitythat appearspurposeless" (Bekoff and Byers 1981; Bekoff 1984; Pellegrini and Smith 1998; Byers1999). This definitionis intuitivelyappealing despite the inclusion of a subjectivejudgment. Play does appear purposeless, but not because it has no function. The "purposelessness" of play is strikingbecause some movementsare inconsistentwith the apparent purpose of the behavior thatimmediatelyprecedes or followsthe playfulbehavioralelements.For example, ifplayfulrunning bypigletsis viewed as trainingforescape fromdanger, it seems pointless to performa deliberateflop-overaftera fewmetersofscampering. However, this behavior does make sense ifthe animals are preparing themselves forthe possibilityofa fallwhilemakinga rapid directional change-duringflight. SOCIAL ASPECTS OF PLAY Of the evaluated hypotheses,three (those of Burghardt,Barber, and Byers) do not give specificpredictions about the choice of partners in play or related social aspects. The selfassessmentand "trainingforthe unexpected" hypothesesdo offerpredictions about social play,but these predictions are opposing. Acis an imcording to our hypothesis,familiarity portant criterionfor choice of play partners hy(Prediction6), whereas theself-assessment pothesis suggeststhat playmatesare selected strictlyaccording to their physicalmatch. In VOLUME 76 polycotous species, young often prefer siblings as playmates(Dobao et al. 1985; Holmes 1994). Several studies also document the tendency of young mammals to play preferentiallywithyoungstersof theirage (e.g., Berger 1980; Byers 1980; Rothstein and Griswold 1991; Mendoza-Granados and Sommer 1995; Thompson 1996). However,itis oftenunclear whether this preference is due to familiarity or a match in body size. Studies thatexplicitlyidentifythe role of reversus body size match latedness/familiarity give conflictingresults.For example, in sable niger),relatedindividuals antelope (Hippotragus were not preferredif theywere not matched in size (Thompson 1996), whereas in bighorn sheep (Ovis canadensis)and chimpanzees (P. familiarityaffectedthe choice of troglodytes), playmatesmore than age (Berger 1980; Markus and Croft 1995). Differences between studiesmaybe related to the degree of differbetween ence in bodysize, age, and familiarity available play partners,and methodological differencesin quantifyingchoice of playpartners. More long-termstudies are needed to identifythe prevalentpatternover an arrayof species. Such observationsshould also demonstratewhat happens when the strengthof two previouslymatched partners diversifies.Achypothesis,the cording to the self-assessment animals will switchtheirpreferenceto bettermatched partners, whereas we predict that theywillcontinue to playwitheach other and use self-handicappingto compensate for the difference(Prediction 8). The "trainingforthe unexpected" and selfassessment hypotheses give two additional contrastingpredictionsabout social aspects of play.Regardingplayinitiation(Prediction 7), the available evidence tends to support the self-assessmenthypothesisin that the larger partner more often initiatesplay or there is no effectof body size (Owens 1975; Pellis and Pellis 1991; Mendoza-Granados and Sommer 1995; Biben 1998; Thompson 1998). In some species, however,younger animals are more often the initiators (Gomendio 1988) or weaker individuals initiate more escalated formsof contactplay (Watson 1998). Regarding the contagious nature of play (Prediction 9), social play often involves more than two animals (Gentry 1974; Pedersen et al. 1990; This content downloaded from 169.237.90.178 on Mon, 14 Sep 2015 20:39:51 UTC All use subject to JSTOR Terms and Conditions JUNE 2001 MAMMALIAN PLAY 159 is a conventional wisdom that sick, injured, depressed, or angryanfrightened,frustrated, imals do not play,littledirectevidence is available. Quantitativestudies could be used to assess whether animals are willing to expend more timeor energyforan opportunityto play in relation to their level of well-being (e.g., through demand curve methodology; Dawkins 1990). The "trainingfortheunexpected" hypothesis predictsthatthereis a specificneuroendocrinological basis of play thatis similaracross differentplaymodes and across various mammalian species (Prediction 13). On the contrary, Burghardt'ssurplusresourcemodel, with its claim about disparate origin and lack of a general functionforplay, implies thatdifferent neuroendocrinologicalsubstratesand processes underlie differentplay patterns.Thus far,the evidence collected on laboratoryrats suggests that play may specificallytap brain mechanisms involved in somatosensoryperception, emotional experience, and arousal control. First,thereis evidence thatregions of brain thatprocess certaintypesofsomatosensoryinformationare particularlyrelevantfornormal playto occur (Siviy1998:233-234). Cells in the parafascicularnucleus of the thalamus (PF), havinglarge receptivefieldsand high thresholds for excitation,receive somatosensoryinformation of a very diffusenature, whereas cells in the ventrobasal thalamic complex (VTC), having small receptivefields and low thresholds, receive somatosensory information of a verydetailed nature. Lesions to the PF substantiallyreduce "pinning,"an element used in quantifyingrat social play,while having minimaleffectson otherbehavior.Bycontrast,lesions to theVTC have a minimaleffect on play. PF activityalso increases substantially duringplay episodes, as measured by the c-fos Panksepp (1998:288) technique. Interestingly, reportsthatfullbody ticklingwas more effecMECHANISMS NEUROPSYCHOLOGICAL tivein inducing 50 kHz chirpingvocalizations INVOLVED IN PLAY in rats (presumablyconnected to play mood) We predict thatplay is stronglysuppressed than tickling of individual parts of the rat by negative emotional states (Prediction 11), body. Taken together,these findingssuggest whereas the self-assessmenthypothesis pre- that nonspecific reticular nuclei involved in whole-body dicts that animals should "play to win," im- the perception of high-intensity plyingthat theyshould be prepared to over- sensations, like those in the PF, specifically come negative emotions in play. Although it mediate play urges. This is in agreementwith Loranca et al. 1999), supporting our prediction over Thompson's prediction that play should be dyadic. In domestic piglets,triadic play amounted to about 20% of all interactions (M Spinka, unpublished observations). Our prediction that play signals are often behavior (Prederivedfromself-handicapping diction 10) is supported by many examples. Self-handicappingpostures, such as lyingon side or looking between the legs, and movementssuch as playbows,roll-overs,pivots,and head tosses,are used especiallyoftenimmediately before vigorous locomotory and social play actions (Geist 1963; Miller-Schwartze 1968; Baldwin and Baldwin 1974; Wilson and Kleiman 1974; Pellis and Pellis 1983; Gomendio 1988; Newberryet al. 1988; Rothsteinand Griswold1991; Bekoff1995; Watson and Croft 1996; Miller and Byers1998); thustheyhave a specificmetacommunicativefunction(Bekoff 1975, 1995). They affectthe behavior of the receivinganimal in such a waythatthe sender is able to sustain the playfulinteraction despite the vigorous action thatfollows. There is a dramaticdifferencebetween the movementsand posturesused to signal intent to attack and those used to signal intent to play. Threatening animals displaytheirbroad side, use frontallysymmetricpositions, and performsymmetricalmovements.Given that are reliable indicators body size and symmetry of physical fitness (M0ller 1998), these displays express signaler condition. In contrast, movementsand postures theself-handicapping used to signalplayfulintenthave elementssimilar to thoseused byweak,tired,subordinate,or even sick,injured, or developmentallyhandicapped animals-lowering down, decreasing apparent body size, gaping, shakingthe head, and twistingasymmetrically.Compare the threatstanceof dogs (C.familiaris) standing-tall withthe "play bow" signal in which the front part of the body is lowered (Bekoff1977). This content downloaded from 169.237.90.178 on Mon, 14 Sep 2015 20:39:51 UTC All use subject to JSTOR Terms and Conditions 160 THE QUARTERLY REVIEW OF BIOLOGY our hypothesisthatplaytrainsanimalsforcoping with unexpected movements and positions rather than perfectingspecific skilled movements. Second, in the rewarddomain, opioids appear to be involved in play in a specificway. For instance,low doses of morphine increase playand opiate antagonistsreduce it,although the same manipulations have opposite effects on the desire for general social interaction (Panksepp 1998:293). Ligands of ,-opioid receptors influence social behavior elements that occur in play but not those unrelated to play (Vanderschuren et al. 1995b). Also, neonatal cocaine exposure affects subsequent learning based on a play rewardin a different way than learning based on a food reward (Willfordet al. 1999). Third, play is dependent on monoaminergic systemsin a distinctway (reviewedby Siviy 1998). In the anticipatoryand warm-upphases of a play bout, the ventrotegmentaldopaminergic systemis involved.Once a playbout is in full stride, high noradrenergic activityenhances play. On the other hand, serotonergic activitymustbe low forplay to occur. Thus, forrats,evidence is accumulatingthat play might,indeed, employ the brain in a specificway.However,we do not knowyetwhether similarmechanisms are active during play in other mammals because almost no studiesfocused on brain mechanisms underlyingplay have been conducted in other species. The "trainingforthe unexpected" hypothesis predictsthatplayactivityresultsin measurable changes in somatosensory,motor, and emotional brain centers (Prediction 14). The motortraining/sensitive period hypothesisimplies thatchanges should be in motor control areas, especiallyin the cerebellum,ratherthan in emotional centers.The opposite prediction follows from the self-assessmenthypothesis. Depriving young rats of social play through isolation during the age when playis mostfrequent resultedin long-termchanges in p-opioid receptorsin basolateral amygdalaand bed nucleus of striaterminals,and in K-opioidreceptorsin severalcorticalareas (van den Berg et al. 1999c). Rats reared in isolation also have permanentlyaltered levels of dopamine, norepinephrine,and serotonin (reviewedbySiviy 1998:235). These monoamines are important VOLUME 76 for coordinating an organism's response to stress.Thus, it appears thatjuvenile play does alter the neurochemistryof brain emotional centers. More evidence is needed, however, thatshould include species other than the rat and should more preciselydistinguishwhether the effectsare specificallydue to playdeprivationor some more general aspectsofthesocial isolation procedure. Although the specificeffectsof play on motor centers have not been tested,it is known thatin cerebellar Purkinje cell axons, the selection ofwhich synapseswill be retained is experience dependent and occursat about theage whenratsplaymost (Byers 1998). In sum, the limited available evidence suggests that somatosensory,emotional, and motorcentersof the brain are affectedbyplay experience. ENVIRONMENTAL CONDITIONS FOR PLAY The "trainingforthe unexpected" hypothesis predicts more locomotoryplay in species thatlive in veryvariable or changing environments (Prediction 15). The surplus resource model suggeststhatanimals livingin more energyefficientenvironmentsshould playmore. The surplus energyhypothesispredictsmore playin species thatlivein habitatswithsurplus food. Examples can be found of species that fiteach of these predictions. Thus, bighorn sheep (0. canadensis) living in variable hilly terrain,harbor seals (P. vitulina)livingin an energy-efficient waterenvironment,and Hanuman langurs (Presbytis entellus)living in a food-richhabitatare all highlyplayful(Berger 1980; Renouf 1993; Sommer and MendozaGranados 1995). Systematicstudiesacross numerous, taxonomicallycloselyrelated species habitatsare needed to deterlivingin different mine which of these predictionsholds true. Our hypothesispredicts that a change in habitat or in a significantenvironmentalfeatureshould stimulateplay (Prediction 16). On the contrary,Burghardt's surplus resource model predictsmore playin an unstimulating environment.There is evidence that animals tend to play in new,rare,or more demanding local habitats.The stimulatoryeffectof fresh snow or shallowwateron playin dogs and humans is well known.Berger (1980) found that bighorn sheep (0. canadensis)livingin British Columbia's Chilcotin hills spent about one This content downloaded from 169.237.90.178 on Mon, 14 Sep 2015 20:39:51 UTC All use subject to JSTOR Terms and Conditions JUNE MAMMALIAN PLAY 2001 fourthof their total play time in small sand bowls at the base of the hills. Byers(1977) reported thatSiberian ibex kids (Capra siberica) preferredto play on sloped rather than flat ground. De Waal (1996:47) described rhesus monkeys(Macaca mulatta)as beingverypreoccupied bya giantverticalwheel thattheycould spin rapidly,release, and thus catapult themselvesmetershigh.It iswell knownthatcaptive animals can be induced to play by the introduction offreshbedding into theiraccommodation or bymovingthem temporarilyinto an alternativeenvironment(e.g., Pedersen et al. 1990; Wood-Gush and Vestergaard 1991; de Passille et al. 1995;Jensen et al. 1998). Mildly dangerous prey, sudden noise, or windy weather can stimulate play as well (MullerSchwartze 1968; Biben 1979; Prescott 1985; Newberryet al. 1988). Onlyour hypothesismakes a specificpredictionin regardsto thesafetyoftheplayenvironment (Prediction 17). Interestingly, iflaboratoryratsare keptin dim light(whichis natural for them), exposure to intense light suppresses play but not other social behaviors (Vanderschuren et al. 1995a; Knutson et al. 1998). This makes sense when viewedfroman evolutionaryperspective,giventhatplayactivitybywild ratsduringperiods of high visibility would probably increase their risk of predation. Our prediction thatanimals should play only when the environmentis relativelysafe ties in withour prediction thatvigilance is reduced during play (Prediction 12), and with the observation of Biben et al. (1989) that squirrelmonkeymothers(S. sciureus)increased their level of vigilance when their offspring were at play. ONTOGENETIC DISTRIBUTION OF PLAY Thompson (1998:192, 200) hypothesized that the information gained through selfassessmentin play facilitatesnormal development in differingenvironments,therebyimplyingthatplay should decline when animals reach adulthood. We predict, more specifically,thatplayshould startat low levelsin neonates, peak during the juvenile period, and decline later (Predictions 18 and 20). This prediction holds true for the vast majorityof species examined (e.g., Biben 1983; Prescott 1985; Gomendio 1988; Newberryand WoodGush 1988; Renouf 1993; Byers1998). 161 Burghardt'ssurplusresource model and Byers's sensitive period hypothesis imply that there should be no play (or verylimitedplay) in adult mammals. Yet it is now well documented that adults of many species do play (Breuggeman 1978; Biben 1979; Berger 1980; Fagen 1981; Byers1984; Rasa 1984; Zucker et al. 1986; Coppinger and Smith 1989; Enomoto 1990; Pedersen et al. 1990; Barber 1991; Pellis et al. 1993; Renouf 1993; Hall 1998; Millerand Byers1998; Watson 1998). Barber's surplus energyhypothesispredicts adult play in species that use brown adipose tissue for thermoregulation,such as hibernators, but thisprediction does not explain whyplay occursin,forexample, adult kangaroos,rats,primates, and pinnipeds, and whyno adult play has been documented in hedgehogs (Erinaceus europaeus) (Fagen 1981). We expect to find adult play,especially in more intelligent species faced withthe unpredictablebehavior of conspecificsand, possibly,withrelativelyintelligent predators or prey. More data are needed to assesswhetherthisis a robustmammalian trend. Two of the alternative hypotheses differ from the "trainingfor the unexpected" hypothesisin theirpredictionsabout when, during the juvenile period, play should occur. Where we predict that play will be most frequent during the period of rapid allometric growth (Prediction 19), Barber's surplus energy hypothesispredicts that play will peak when the consummation of solid food becomes prevalent.The sensitiveperiod hypothesis of Byerspredictsontogenetic parallelism betweenplay,on the one hand, and cerebellar synaptogenesisand muscle fiber differentiation on the other.This has been found in mice (Mus musculus),rats,and cats (F. catus) (Byers and Walker 1995). However, furtheranalysis of the literatureis needed to assess the level of supportforthe differentpredictionsabout the precise timingof the peak in play during thejuvenile period. We predictthatdifferentmodes of playwax and wane at different ages withinthe same species (Prediction 21), whereas the surplus energyand sensitiveperiod hypothesesimplyan ontogeneticallyparallel course forall formsof play. Available evidence appears to favorthe former prediction (Gentry 1974; Burghardt This content downloaded from 169.237.90.178 on Mon, 14 Sep 2015 20:39:51 UTC All use subject to JSTOR Terms and Conditions 162 THE QUARTERLY REVIEW OF BIOLOGY 1998b). Our more specificprediction thatlocomotoryplay will precede social and sexual play during ontogenyalso appears to be supported (Berger 1980; Berry and Signoret 1984; Prescott 1985; Gomendio 1988; Newberryand Wood-Gush 1988; Newberryet al. 1988; Haynes 1997). In spotted hyenas (Crocuta crocuta),by contrast,social play precedes locomotoryplay (Drea et al. 1996). This exception is in accord withour prediction that playhas immediatebenefits,giventhatwithinden aggression among hyena siblings precedes locomotion outside the den. Similardevelopmental patternsare observed in coyotes (Canis latrans) and red foxes (Vulpesvulpes) (Bekoff1974). Withrespectto predictionsabout sex differences in play,Barber's surplusenergyhypothesis predictsnone; the optimal formforadaptive energyloss should be the same for both sexes. The sensitiveperiod hypothesisimplies that sexual dimorphism of play, where present, will occur throughout the sensitiveperiod. We predict that the sexual dimorphism in play will increase withage (Prediction 23) because, presumably, the serious situations for which play provides trainingbecome increasinglydisparate in the two sexes and because thebenefitsofplayare partiallyimmediate. Examples of pronounced, moderate, and nonexistent sex differencesin play can be found in theliterature(e.g., Byers1977; Biben 1983; Prescott1985; Caro 1988; Pedersen et al. 1990; Renouf 1993; Watson 1998). However, a majorityof the available data supports the prediction of increasing sexual dimorphism in play with increasing age (Gentry 1974; Meaney et al. 1985; Newberryand Wood-Gush 1988; Newberryet al. 1988; Spijkerman et al. 1996; Haynes 1997; Smith et al. 1998). Available data also suggest that this differenceis not dependent on concurrentlydivergingsex steroidlevels (Meaney et al. 1985). VOLUME 76 cies withmoderate body size and thermogenically active brown adipose tissue should be the most playful.In the only systematicstudy published so far,Byers(1999) found thatplay in Australianmarsupialsis associated withrelativebrain size but not withbody size or metabolic rate; Ortega and Bekoff(1987) reported similartrendsin birds. DIRECTIONSFORFUTURERESEARCH On the whole, our hypothesisappears to fare better than the four alternativesthatwe have evaluated. This assessmentmustbe considered preliminary,however,because: (i) the evidence is fragmentary,(ii) there may be other relevant papers that we have not reviewed, and (iii) it was difficultto assess data fromdifferentsources using uniformcriteria because ofvariationsin methodologyand emphasis.Consequently,further efforts are needed to elucidate whichofthemanypredictionsdiscussed above are generallysupported across mammalian taxa. The most directevidence forany of the putative functional hypotheses of play could come from controlled experimentsin which the amount ofplayexperience is manipulated and the predictedeffectson behavior,physiology, adult skill and/or coping, survival,and reproduction are recorded. The difficulty withthese studiesis thatplay deprivationusually removes other sources of stimulationas well,as, forexample,when animalsare socially isolated during the period when theywould play the most. Thus, it is unclear whetherthe resultingeffectsare due specificallyto the lack ofplay.More refinedmethods,such as the use of drugged play partners, employing short daily periods of play in isolates (Potegal and Einon 1989), or using a housing environment thatinhibitsplayare promising,although typicallyrestrictedto species amenable to studyin laboratorysettings. Exciting progress has been achieved in asPHYLOGENETIC DISTRIBUTION OF PLAY sessing the neurophysiological and neuroWhere we predictmore playin species with chemical basis of play (Vanderschuren et al. highlydeveloped cognitiveabilitiesand, hence, 1997; van den Berg et al. 1999c), and promisa high encephalization quotient (Prediction ing investigationshave been made into the 24), Burghardt'ssurplusresource model pre- emotionalprocessesassociatedwithplay (Pankdicts that moderatelysized species with high sepp 1998). However,fora more complete unmetabolic ratesshould playthe most.Barber's derstandingof the relationshipbetween play surplus energy hypothesissuggests that spe- behavior,the associated cognitive,emotional, This content downloaded from 169.237.90.178 on Mon, 14 Sep 2015 20:39:51 UTC All use subject to JSTOR Terms and Conditions JUNE MAMMALIAN PLAY 2001 and motivationalprocesses, and the underlying brain machinery,considerable conceptual and experimental difficultiesmust be surmounted. we suggestthatsubGiven these difficulties, stantial progress in understanding the function of mammalian play can be achieved by focusing on the ontogenetic, phylogenetic, and ecological distributionofplay,and on sex differencesin play. Widelyscattereddata can be drawn together,combined with available strategies,and sibinformationon life-history mittedto detailed analyses to verifyquantitativelywhich of the trendspredicted by the differenthypotheses are supported within and across mammalian taxa. FINAL REMARKS THE" ONE FUNCTION-MANY FUNCTIONS PROBLEM REVISITED 163 (Siegel 1995). typeslike herbivory/carnivory Highly diverse functionsfor REM sleep have been proposed, including the maintenance of a high,stablebrain temperature(Wehr 1992), ensuring recoveryfrom sleep (Vertes 1986), memory and/or sensorimotor experience processing (Winson 1993; Ribeiroet al. 1999), compensation for brain processes that take place during non-REM sleep (Benington and Heller 1994), promoting brain development (Mirmiran1995), and reinforcingpsychological individuation (Jouvet1998). Despite this diversity offormsand possible functions,REM sleep is considered a homologous traitthat originatedveryearlyin mammalian historyor even in reptilian ancestors of mammals (Siegel et al. 1998), probably to fulfilla substantial, although not yet fully elucidated, originalfunction.In ourview,thereare strong indications that mammalian play may have a similarphylogenetichistory. Our hypothesisprovides a plausible explanation about whyplay,specifically,could have diversifiedinto such a rich arrayof formswith various derived functions.According to the constituentsof play presented in this review, play should resultin flexible reactions to unexpected situations. Play may thus result in more diverse behavior both within and betweenindividuals.In other words,individuals may,throughplay,findand develop theirown personal methods for coping with different kindsofunexpectedmisfortunes. Indeed, qualitativedifferencesbetween individualsand betweenlittersarise duringjuvenileplay (Wilson and Kleiman 1974; Lund and Vestergaard 1998). Thus, playmayincrease phenotypicbehavioralvariabilitywithinpopulations, which, in turn,mayopen the door to the Baldwin Effect,a process by which novel variations become geneticallyfixed (Baldwin 1896; Parisi and Nolfi 1996). In this way, play may have acquired novel kinematicforms,causal links, developmental pathways,and beneficialfunctionsin variouslineages whilestillretainingits basic nature. While the core of our hypothesisis thatplay in mammalshas one originalfunction,namely to trainforunexpected situations,clearlyplay behaviorin itsextantvarietycan fulfillvarious other functionsin individualspecies and animals that differin age and sex. To illustrate this "one ancestral-manyderived functions" possibility,Byers (1984:44-45) compared the diversificationof mammalian play to the evolution of the pentadactylforelimb.Although the quantitativemodificationsof the mammalian manus are impressive(e.g., the prolongadetion of fingersin bats), the basic five-digit sign is retained. Yet the ancestralfunctionof walkinghas been complemented, or even replaced, by diverse new functionssuch as digging in anteatersand swimmingin dolphins. Similarly,once play was established in mammalian behavioral design,itdiversifiedin individual taxa to fulfillvarious specificfunctions. A perhaps closer phylogeneticparallel can be drawnbetween mammalian playand mammalian rapid eye movement (REM) sleep, whichis also omnipresentin mammalsbut absent in ectothermicvertebrates.Its modifications, including specializations such as unihemisphericsleep in dolphins (Tobler 1995), A MULTIFACETED APPROACH TO PLAY: are impressive. Its amount and distribution THE PUZZLES OF PLAY varysubstantiallyacross mammalian taxa and From our major hypothesisabout the funcit is correlated with such ecological variables as a safesleeping place, life-history parameters tion ofplay,we have derivedanotherfourspeand niche cifichypotheses:thatself-handicappingis insuch as precociality/altriciality, This content downloaded from 169.237.90.178 on Mon, 14 Sep 2015 20:39:51 UTC All use subject to JSTOR Terms and Conditions 164 THE QUARTERLY REVIEW OF BIOLOGY tegralto play,thatplayis stimulatedbynovelty and unpredictability,that play is associated with a specificemotional state,and that play presupposes high cognitivecapacity.Our hypothesis about function (i.e., ultimate causation) is complemented by hypotheses about how thisfunctionis achieved (i.e., proximate causation in termsofboth physicaland mental processes involved in play), and why play is more predominant in some animals than in others (i.e., due to phylogeneticconstraints). Strictlyspeaking, none of the four derived hypothesesis a logical necessityfor our core functionalhypothesisto be true.On one hand, it may seem thatpresentingthese hypotheses independentlywould be a more appropriate approach; on the otherhand, function,proximatemechanisms,and phylogeneticconstraints are intertwinedin numerous ways,and knowledge of these interactionsis essentialfora full VOLUME 76 understandingof behavior (Curio 1994). This is especially true for play. Play is a robust behavioral phenomenon with several puzzling features,such as its kinematicand sequential structure,and ontogenetic and phylogenetic distribution.Anyfunctionalhypothesisabout playmustbe comprehensiveenough to encompass these features,and most previouslypublishedhypothesesabout playhave attemptedto do so. For instance, Burghardt's surplus resource model attemptsto incorporate physiological, life-history, and ecological conditions thatsupportthe evolutionofplay,and the sensitiveperiod hypothesisof Byersabout a neurodevelopmentalmechanism is closelylinked to the motor trainingfunctional hypothesis. This is whyour proposal, although centered on function,is a complex theoryabout the natureofplay.We hope thatour new conceptual frameworkwilllead to enhanced understanding of thisfascinatingbehavior. REFERENCES Adams N, Boice R. 1989. Developmentof dominance in domestic rats in laboratoryand seminatural environments.BehaviouralProcesses19:127-142. Aldis 0. 1975. Play Fighting.New York: Academic Press. Allen C, BekoffM. 1997. SpeciesofMind: ThePhilosophyand Biologyof Cognitive Ethology.Cambridge (MA): MIT Press. Apanius V. 1998. Stressand immune defense. Pages 133-153 in Advancesin theStudyofBehavior,Volume27: Stressand Behavior,edited byA P M0ller, M Milinski,and P J B Slater. San Diego (CA): Academic Press. Araba B D, Crowell-DavisS L. 1994. Dominance relationships and aggression of foals (Equus caballus). AppliedAnimalBehaviourScience41:1-25. BaldwinJD. 1986. Behavior in infancy:exploration and play. Pages 295-326 in Comparative Primate Biology,Volume2, Part A: Behavior,Conservation, and Ethology, edited by G Mitchell andJ Erwin. NewYork: Alan R. Liss. BaldwinJ D, BaldwinJ I. 1974. Exploration and social playin squirrelmonkeys(Saimiri). American Zoologist14:303-315. BaldwinJ D, BaldwinJ I. 1976. Effectsof food ecology on social play: a laboratory simulation. ZeitschriftFur 40:1-14. Tierpsychologie BaldwinJM. 1896. A new factorin evolution.American Naturalist30:441-451, 536-553. Barber N. 1991. Playand energyregulationin mammals. Quarterly ReviewofBiology66:129-147. BarrettL, Dunbar R I M, Dunbar P. 1992. Environmentalinfluenceson playbehaviourin immature gelada baboons. AnimalBehaviour44:111-115. BekoffM. 1974. Social play and play-solicitingby infantcanids. AmericanZoologist14:323-340. BekoffM. 1975. The communication of play intention: are play signals functional? Semiotica15: 231-239. Bekoff M. 1976. Animal play: problems and perspectives.Pages 165-188in Perspectives inEthology, edited by P P G Bateson and P H Klopfer. New York: Plenum Publishing. BekoffM. 1977. Social communication in canids: evidence for the evolution of a stereotyped mammalian display.Science197:1097-1099. BekoffM. 1984. Social play behavior. BioScience34: 228-233. Bekoff M. 1995. Play signals as punctuation: the structureof social play in canids. Behaviour132: 419-429. BekoffM, Allen C. 1998. Intentional communication and social play: how and whyanimals negotiate and agree to play. Pages 97-114 in Animal Play:Evolutionary, Comparative and EcologicalPeredited byM BekoffandJ A Byers.Camspectives, bridge: Cambridge UniversityPress. BekoffM, ByersJA. 1981. A criticalreanalysisof the ontogeny of mammalian social and locomotor play:an ethological hornet's nest.Pages 196-337 in BehavioralDevelopment: The Bielefeld InterdisciplinaryProject,edited by K Immelmann et al. Cambridge: Cambridge UniversityPress. This content downloaded from 169.237.90.178 on Mon, 14 Sep 2015 20:39:51 UTC All use subject to JSTOR Terms and Conditions JUNE 2001 MAMMALIAN PLAY 165 BekoffM, ByersJA, editors. 1998. AnimalPlay:Evoevolution of animal minds. Evolutionand Cognition5:115-123. lutionary,Comparativeand EcologicalPerspectives. ByersJA. 1977. Terrain preferencesin the play beCambridge: Cambridge UniversityPress. BeningtonJH, Heller H C. 1994. Does the function haviourof Siberian ibex kids. ZeitschriftFur Tierpsychologie 45:199-209. of REM sleep concern non-REM sleep or waking? Progress in Neurobiology 44:433-449. ByersJA. 1980. Playpartnerpreferencesin Siberian ibex, Capra ibexsiberica.ZeitschriftFur TierpsycholoBergerJ.1980. The ecology,structureand function ofsocial playin Bighorn sheep (Ovis canadensis). gie53:23-40. ByersJ A. 1984. Play in ungulates. Pages 43-65 in JournalofZoology192:531-542. BerryM, SignoretJ-P. 1984. Sex play and behavPlayinAnimalsand Humans,edited byP K Smith. ioural sexualization in the pig. Reproduction, NuOxford: Basil Blackwell. trition, Development 24:507-513. ByersJA. 1998. Biological effectsoflocomotor play: Biben M. 1979. Predation and predatoryplay begettinginto shape, or somethingmore specific? Pages 205-220 in AnimalPlay:Evolutionary, Comhaviour of domestic cats. Animal Behaviour27: edited byM BeparativeandEcologicalPerspectives, 81-94. koffandJ A Byers.Cambridge: Cambridge UniBiben M. 1983. Comparative ontogenyof social beversityPress. haviour in three South American canids, the maned wolf,crab-eatingfox and bush dog: im- ByersJA. 1999. The distributionof play behaviour among Australian marsupials.JournalofZoology plications for sociality. Animal Behaviour 31: 814-826. 247:349-356. Biben M. 1989. Effectsof social environment on ByersJA, Walker C. 1995. Refiningthe motor training hypothesisfor the evolution of play. Ameriplay in squirrel monkeys (Saimiri sciureus):recan Naturalist146:25-40. solvingHarlequin's dilemma. Ethology 81:72-82. Biben M. 1998. Squirrel monkeyplayfighting: mak- Caro T M. 1980. The effectsof experience on the predatorypatternsof cats. Behavioraland Neural ing the case fora cognitivetrainingfunctionfor Biology29:1-28. play. Pages 161-182 in AnimalPlay:Evolutionary, and EcologicalPerspectives, edited by Caro T M. 1988. Adaptive significanceof play: are Comparative M BekoffandJ A Byers.Cambridge: Cambridge we gettingcloser. Trendsin Ecology& Evolution 3:50-53. UniversityPress. Biben M, SymmesD, Bernhards D. 1989. Vigilance Chalmers N R, Locke-HaydonJ. 1984. Correlations Jouramong measures of playfulnessand skillfulness during play in squirrel monkeys.American nal ofPrimatology in captivecommon marmosets(Callithrixjacchus 17:41-49. Bjorklund D F, Brown R D. 1998. Physicalplay and jacchus). Developmental Psychobiology 17:191-208. cognitivedevelopment: integratingactivity, cog- Coppinger R P, SmithC K. 1989. A model forunderstandingthe evolution of mammalian behavior. nition, and education. Child Development69: edited by 604-606. Pages 335-374 in CurrentAlammalogy, H Genoways.New York: Plenum Press. Breuggeman J A. 1978. The functionof adult play in free-rangingMacaca mulatta.Pages 169-191in Curio E. 1994. Causal and functional questions: SocialPlay in Primates, edited byE 0 Smith.New how are they linked? Animal Behaviour47:999York:Academic Press. 1021. da Silva N L, FerreiraV M M, Carobrez A D, Morato BurghardtG M. 1984. On the originsof play. Pages G S. 1996. Individual housing fromrearingmod6-41 in Play in Animalsand Humans,edited by P K Smith. Oxford: Basil Blackwell. ifies the performance of young rats on the ele&Behavior vated plus-mazeapparatus. Physiology BurghardtG M. 1988. Precocity,play,and the ecto60:1391-1396. therm endotherm transition-profound reorganization or superficialadaptation? Pages 107- Dawkins M S. 1990. From an animal's point of 148 in HandbookofBehavioralNeurobiology, edited view-motivation, fitness,and animal welfare. Behavioraland Brain Sciences13:1-61. byE M Blass. New York: Plenum Publishing. BurghardtG M. 1998a. The evolutionaryoriginsof de Passille A M, RushenJ,MartinF. 1995. Interpretingthe behaviour of calves in an open-fieldtest: play revisited:lessons fromturtles.Pages 1-26in a factoranalysis.AppliedAnimalBehaviourScience AnimalPlay:Evolutionary, Comparative andEcologedited byM BekoffandJ A Byers. 45:201-213. icalPerspectives, de Waal F. 1996. GoodNatured:The OriginsofRight Cambridge: Cambridge UniversityPress. and Wrongin Humans and OtherAnimals.CamBurghardtG M. 1998b. Play. Pages 725-735 in ComA Handbook, edited by G bridge (MA): Harvard UniversityPress. parative Psychology: Greenberg and M M Haraway. New York: Gar- Dobao M T, RodriganezJ, Silio L. 1985. Choice of land Publishing. companions in social playin piglets.AppliedAnimalBehaviourScience13:259-266. BurghardtG M. 1999. Conceptions of play and the This content downloaded from 169.237.90.178 on Mon, 14 Sep 2015 20:39:51 UTC All use subject to JSTOR Terms and Conditions 166 THE QUARTERLY REVIEW OF BIOLOGY Drea C M, HawkJE, GlickmanS E. 1996. Aggression decreases as playemergesin infantspottedhyaenas: preparation forjoining the clan. AnimalBehaviour51:1323-1336. Einon D F, Morgan M J, Kibbler C C. 1978. Brief periods ofsocialization and laterbehavior in the rat.DevelopmentalPsychobiology 11:213-225. EisenbergJ F. 1981. TheMammalianRadiations:An AnalysisofTrendsinEvolution,Adaptation,and Behavior.Chicago: Universityof Chicago Press. Enomoto T. 1990. Social play and sexual behavior of the bonobo (Pan paniscus) withspecial reference to flexibility. Primates31:469-480. Fagen R. 1981. Animal Play Behavior.Oxford: OxfordUniversityPress. Fagen R M, George T K. 1977. Play behavior and exercise in youngponies (Equus caballus). BehavioralEcologyand Sociobiology 2:267-269. Fontaine R P. 1994. Playas physicalflexibility training in fiveceboid primates.Journal ofComparative Psychology 108:203-212. Fraser D, Duncan IJ H. 1998. "Pleasures", "pains" and animal welfare:towarda natural historyof affect.AnimalWelfare 7:383-396. GeistV. 1963. On the behaviour of the NorthAmerican moose (AlcesalcesandersoniPetrason 1950) in BritishColumbia. Behaviour20:378-415. GentryR L. 1974. The development of social behavior throughplay in the Stellarsea lion. American Zoologist14:391-403. Gentsch C, LichtsteinerM, FrischknechtH R, Feer H, SiegfriedB. 1988. Isolation-induced locomotorhyperactivity and hypoalgesia in ratsare prevented by handling and reversedby resocializa& Behavior43:13-16. tion. Physiology Golani I. 1992. A mobilitygradientin the organization of vertebratemovement: the perception of movement through symbolic language. Behavioraland Brain Sciences15:249-266. Gomendio M. 1988. The development of different types of play in gazelles-implications for the nature and functionsof play. AnimalBehaviour 36:825-837. Groos K. 1898. ThePlayofAnimals.NewYork: D. Appleton. Hall S L. 1998. Object play by adult animals. Pages 45-60 in Animal Play: Evolutionary,Comparative edited byM Bekoffand andEcologicalPerspectives, JAByers.Cambridge:CambridgeUniversityPress. Haynes M R. 1997. Playbehavior injuvenile domestic cattle (Bos taurus):sex differencesand developmental changes. [Masters Thesis]. College Park (MD): Universityof Maryland. Holmes W G. 1994. The development of littermate preferencesin juvenile Belding's ground squirrels. AnimalBehaviour48:1071-1084. Hol T, van den Berg C L, van ReeJ M, SpruijtB M. 1999. Isolation duringthe playperiod in infancy VOLUME 76 decreases adult social interactionsin rats.BehavioralBrainResearch100:91-97. Inhelder E. 1955. Object play in an Indian rhino (Rhinoceros unicornis).Zeitschrift Fr Tierpsychologie 12:88-144. Jensen M B, VestergaardK S, Krohn C C. 1998. Play behaviour in dairycalves kept in pens: the effect of social contact and space allowance. Applied AnimalBehaviourScience56:97-108. Jerison H. 1973. EvolutionofBrain and Intelligence. New York:Academic Press. Jouvet M. 1998. Paradoxical sleep as a programming system.JournalofSleepResearch7 Supplement 1:1-5. Knutson B, BurgdorfJ,PankseppJ. 1998. Anticipation of play elicitshigh frequencyultrasonicvocalizations in young rats.Journalof Comparative Psychology 112:65-73. Loranca A, Torrero C, Salas M. 1999. Development of play behavior in neonatally undernourished rats.Physiology & Behavior66:3-10. LundJ D, Vestergaard K S. 1998. Development of social behaviour in fourlittersof dogs (CanisfaActaVeterinaria miliaris). Scandinavica39:183-193. Markus N, CroftD B. 1995. Play behaviour and its effectson social development of common chimpanzees (Pan troglodytes).Primates36:213-225. Martin P. 1984. The time and energycosts of play behaviour in the cat. ZeitschriftFir Tierpsychologie 64:298-312. MartinP, Bateson P. 1985. The influenceof experimentallymanipulatinga component ofweaning on the development ofplayin domesticcats.AnimalBehaviour33:511-518. MeaneyMJ, StewardJ,BeattyWW. 1985. Sex differences in social play:the socialization ofsex roles. Pages 1-58 in Advancesin theStudyofBehaviour, Volume15, edited byJ S Rosenblatt et al. New York: Academic Press. Mendoza-Granados D, Sommer V. 1995. Play in chimpanzees of the Arnhem zoo-self-serving compromises. Primates36:57-68. Miller M N, ByersJ A. 1998. Sparring as play in young pronghorn males. Pages 141-160 in AnimalPlay:Evolutionary, Comparative and Ecological edited by M Bekoffand J A Byers. Perspectives, Cambridge: Cambridge UniversityPress. MirmiranM. 1995. The functionof fetal/neonatal rapid eye movementsleep. BehaviouralBrainResearch69:13-22. M0ller A P. 1998. Developmental instabilityas a general measure of stress.Pages 181-213 in Advances in theStudyofBehavior,Volume27: Stress and Behavior,edited by A P M0ller, M Milinski, and PJ B Slater.San Diego (CA): Academic Press. Moss C. 1988. ElephantMemories:Thirteen Yearsin the Lifeofan ElephantEamily. NewYork: FawcettColumbine. This content downloaded from 169.237.90.178 on Mon, 14 Sep 2015 20:39:51 UTC All use subject to JSTOR Terms and Conditions JUNE 2001 MAMMALIAN PLAY 167 MuAller-Schwartze D. 1968. Playdeprivationin deer. PrescottJ.1985. Play and the development of indiBehaviour31:144-162. vidual and social behaviorin theAmerican bison NewberryR C, Wood-Gush D G M. 1988. Develop(Bison bison).BiologyofBehaviour10:261-276. ment of some behaviour patternsin piglets un- Rasa 0 A E. 1984. A motivationalanalysisof object der semi-naturalconditions.AnimalProduction 46: play in juvenile dwarfmongooses (Helogale un103-109. dulata rufula).AnimalBehaaviour 32:579-589. NewberryR C, Wood-Gush D M G, HallJ W. 1988. Renouf D. 1993. Playin a captivebreeding colonyof Playfulbehaviour of piglets.BehaviouralProcesses harbour seals (Phoca vitulina)-constrained by 17:205-216. timeor byenergy.Journal ofZoology 231:351-363. OrtegaJC, BekoffM. 1987. Avianplay:comparative Ribeiro S, Goyal V, Mello C V, Pavlides C. 1999. evolutionaryand developmentaltrends.Auk 104: Brain gene expression during REM sleep de338-341. pends on priorwakingexperience. Learningand Owens N W. 1975. A comparison of aggressiveplay Memory 6:500-508. and aggressionsin free-livingbaboons Papio anRothsteinA, GriswoldJG. 1991. Age and sex preferubis.AnimalBehaviour23:755-765. ences forsocial partnersbyjuvenile bison bulls, PankseppJ. 1998. Affective Neuroscience: TheFoundaBisonBison.AnimalBehaviour41:227-237. tionsofHuman and AnimalEmotions. Oxford: OxSade D S. 1973. An ethogramforrhesusmonkeys.I. fordUniversityPress. Antitheticalcontrastin posture and movement. Parisi D, Nolfi S. 1996. The influence of learning American JournalofPhysicalAnthropology 38:537on evolution.Pages 419-428 in AdaptiveIndividu542. als in EvolvingPopulations:Modelsand Algorithms, Saunders I, SayerM, Goodale A. 1999. The relationedited by R K Belew and M Mitchell. Reading ship between playfulness and coping in pre(MA): Addison-Wesley. Pedersen J M, Glickman S E, Frank L G, Beach F school children: a pilot study.American Journal A. 1990. Sex differencesin the play behavior of ofOccupationalTherapy53:221-226. immature spotted hyenas, Crocuta crocuta.Hor- Siegel J M. 1995. Phylogenyand the function of monesand Behavior24:403-420. REM sleep. BehaviouralBrainResearch69:29-34. PellegriniA D. 1995. Boys' rough-and-tumbleplay Siegel JM, Manger P R, Nienhuis R, Fahringer H and social competence: contemporaneous and D. 1998. Monotremesand the evoM, PettigrewJ longitudinal relations. Pages 107-126 in TheFulution ofrapid eye movementsleep. Philosophical tureofPlay Theory: A Multidisciplinary Inquiryinto TransactionsoftheRoyalSocietyofLondon B 353: theContributions ofBrian Sutton-Smith, edited by 1147-1157. A D Pellegrini.Albany (NY): State Universityof SiviyS M. 1998. Neurobiological substratesof play New York Press. behavior: glimpses into the structureand funcPellegrini A D, Smith P K. 1998. Physical activity tion of mammalian playfulness.Pages 221-242 play: the nature and functionof a neglected asin AnimalPlay:Evolutionary, Comparative and Ecopect of playing. ChildDevelopment 69:577-598. edited by M Bekoffand J A logicalPerspectives, Pellis S M. 1993. Sex and the evolution of playfightByers.Cambridge: Cambridge UniversityPress. ing: a reviewand model based on the behavior SmithL K, Forgie M L, Pellis S M. 1998. Mechanisms of muroid rodents. Play Theory Research1:55-75. underlyingthe absence of the pubertal shiftin Pellis S M, Pellis V C. 1983. Locomotor-rotational the playfuldefense of female rats.Developmental movementsin the ontogenyand playof the laboratoryrat Rattusnorvegicus. 33:147-156. Developmental PsychoPsychobiology biology 16:269-286. Sommer V, Mendoza-Granados D. 1995. Play as inPellis S M, Pellis V C. 1991. Role reversalchanges dicator of habitatquality:a field studyof langur during the ontogeny of play fightingin male monkeys(Preslytis entellus).Ethology 99:177-192. rats: attack vs. defense. Aggressive Behavior17: Spencer H. 1898. The PrinciplesofPsychology. New 179-189. York: Appleton. Pellis S M, Pellis V C, McKenna M M. 1993. Some Spijkerman R P, Dienske H, van HooffJ A R A M, subordinates are more equal than others: play Jens W. 1996. Differencesin variability,interacfightingamongst adult subordinate male rats. tivityand skillsin social play of young chimpanBehavior19:385-393. Aggressive zees livingin peer groups and in a large family M M Pereira E, Preisser C. 1998. Do strongprimate zoo Behaviour133:717-739. group. players "self-handicap"during competitive soStampsJ. 1995. Motor learning and the value of facial play?Folia Primatologica 69:177-180. miliar space. AmericanNaturalist146:41-58. Potegal M, Einon D. 1989. Aggressivebehaviors in adult ratsdeprived ofplayfighting experience as Thompson KV. 1996. Play-partnerpreferencesand the functionof social play in infantsable antejuveniles. Developmental Psychobiology 22:159-172. lope, Hippotragusniger.Animal Behaviour 52: Power T G. 2000. Playand Explorationin Childrenand Animals.New York: Lawrence Erlbaum. 1143-1155. This content downloaded from 169.237.90.178 on Mon, 14 Sep 2015 20:39:51 UTC All use subject to JSTOR Terms and Conditions 168 THE QUARTERLY REVIEW OF BIOLOGY Thompson K V. 1998. Self assessment in juvenile play. Pages 183-204 in AnimalPlay:Evolutionary, Comparativeand EcologicalPerspectives, edited by M BekoffandJ A Byers.Cambridge: Cambridge UniversityPress. Tobler I. 1995. Is sleep fundamentallydifferentbetween mammalian species? Behavioraland Brain Sciences69:35-41. van den Berg C L, Hol T, van ReeJ M, SpruijtB M, EvertsH, KoolhaasJ M. 1999a. Playis indispensable for an adequate development of coping with social challenges in the rat. Developmental Psychobiology 34:129-138. van den Berg C L, van ReeJ M, SpruijtB M. 1999b. Sequential analysisofjuvenile isolation-induced decreased social behavior in the adult rat.Physiologyand Behavior67:483-488. van den Berg C L, van ReeJ M, SpruijtB M, Kitchen I. 1999c. Effectsofjuvenile isolation and morphine treatmenton social interactionsand opioid receptorsin adultrats:behaviouraland autoradiographic studies. EuropeanJournalofNeuroscience 11:3023-3032. Vanderschuren LJ MJ, Niesink RJ M, SpruijtB M, van ReeJ M. 1995a. Influence of environmental factorson social play behavior ofjuvenile rats. and Behavior58:119-123. Physiology Vanderschuren LJ MJ, Niesink RJ M, SpruijtB M, van Ree J M. 1995b. [L- and K-opioid receptormediated opioid effectson social playinjuvenile rats.EuropeanJournalofPharmacology 276:257-266. Vanderschuren LJ MJ, Niesink RJ M, van ReeJ M. 1997. The neurobiology of social play-behavior in rats.Neuroscience and Biobehavioral Reviews21: 309-326. VOLUME 76 VertesR P. 1986. A life-sustaining functionforREM sleep: a theory.Neuroscience and Biobehavioral Reviews10:371-376. von Holst D. 1998. The concept of stress and its relevance for animal behavior. Pages 1-131 in Advancesin theStudyofBehavior,Volume27: Stress and Behavior,edited by A P M0ller, M Milinski, and PJ B Slater.San Diego (CA): Academic Press. Watson D M. 1998. Kangaroos at play: play behaviour in the Macropodoidea. Pages 61-95 in AnimalPlay:Evolutionary, and Ecological Comparative Perspectives, edited by M Bekoffand J A Byers. Cambridge: Cambridge UniversityPress. Watson D M, CroftD B. 1996. Age-related differences in playfightingstrategiesof captive male red-neckedwallabies(Macropusrufogriseus banksianus).Ethology 102:336-346. Wehr TA. 1992. A brain-warming functionforREM sleep. Neuroscience and BiobehavioralReviews16: 379-397. WillfordJA, Segar T M, Barron S. 1999. The behavioral response to noveltyis altered in ratsneonatallyexposed to cocaine. DevelopmentalPsychobiology35:343-351. Wilson S C, Kleiman D G. 1974. Eliciting play: a comparativestudy.American Zoologist 14:341-370. WinsonJ. 1993. The biology and functionof rapid eye movementsleep. CurrentOpinionin Neurobiology3:243-248. Wood-Gush D G M, VestergaardK. 1991. The seeking of noveltyand itsrelation to play.AnimalBehaviour42:599-606. Zucker E L, Dennon M B, Puleo S G, Maple T L. 1986. Play profilesof captive adult orangutans: a developmental perspective.Developmental Psychobiology 19:315-326. This content downloaded from 169.237.90.178 on Mon, 14 Sep 2015 20:39:51 UTC All use subject to JSTOR Terms and Conditions