Available online at www.s
m
u
AU
niv
an
development
In this study, we videotaped social play within litters of play (Burghardt 2005). Animals may play to learn valu-
ARTICLE IN PRESS
ANIMAL BEHAVIOUR, 20
doi:10.1016/j.anbehaStreet, University of Michigan, Ann Arbor, MI 48109-1043, U.S.A.unexpected situations (Spinka et al. 2001) or as a way
to assess their own capabilities relative to conspeci?cs
(Thompson 1998; Smith et al. 1999; Palagi et al. 2004).
Although all of these hypotheses seem reasonable, it is
extremely dif?cult to determine the reproductive bene?ts
of social play. In a longitudinal study of social play in
Correspondence: C. Ward, Department of Psychology, 530 Church
(email: rameses@umich.edu). E. B. Bauer is at the Department of Ani-
mal Programs, Smithsonian National Zoological Park, NZP-Lion/Tiger
MRC: 5507, P.O. Box 37012, Washington, D.C. 20013-7012, U.S.A.domestic dogs to examine its role in the development of
social relationships. We studied four litters of different
breeds (one was a mixed-breed litter), and we continued
observations on one litter for longer (40 weeks) than
most previous studies have done. To our knowledge, this
able social skills (Biben 1998) or to strengthen (Bekoff
1984) or test social bonds (Zahavi 1977; Pozis-Francois
et al. 2004), as training for cognitive (Bekoff 1984; Spinka
et al. 2001) or motor development (Byers 1998), to de-
velop the emotional ?exibility needed for dealing withSocial play is play directed at a conspeci?c (Bekoff & Byers
1981; Fagen 1981; Bekoff 1984) and in canids includes
behaviours such as chasing and play-?ghting games,
mounting behaviour (i.e. mimicking copulatory behav-
iour) and inhibited biting (Bekoff 1974; Burghardt 2005).
is the ?rst study in domestic dogs to compare systemati-
cally numerous aspects of social play (such as play-partner
preferences, role reversals and self-handicapping) both
across time and across different litters.
Hypotheses abound concerning the functions of socialrelationships within litters. We collected data when the puppies were between 3 and 40 weeks of age,
but collection times varied by litter. We divided data analysis into three time periods to coincide approx-
imately with critical periods in the early social development of dogs. Early play-partner preferences were
associated with preferences in later time periods, and the tendency for puppies to prefer speci?c partners
increased over time. Play did not conform to 50e50 symmetry of roles between partners, which some
researchers claim is necessary to sustain play. In the later juvenile period (time 3), dogs who engaged in
high rates of offense behaviours (e.g. chasing, forcing partners down) also initiated play at higher rates,
implying that winning during play may become more important as puppies mature. Self-handicapping
behaviours were positively associated with play signalling, suggesting that, like play signals, self-handicap-
ping may function to indicate playful intent. In mixed-sex dyads, males initiated play, engaged in offense
behaviours, and self-handicapped more than females. Females were more likely to initiate with females
across all time periods, but males were more likely to initiate with males only in time 3. We discuss results
from mixed- and same-sex interactions with reference to inter- and intrasexual competition. The types of
offense and self-handicapping behaviours displayed were similar across litters, suggesting that the expres-
sion of these behaviours may follow a similar ontogeny in puppies in general.
 2008 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
Keywords: asymmetrical play; Canis lupus familiaris; domestic dog; partner preferences; self-handicapping; socialWe videotaped behaviour in four litters of domesti0003e3
Please c
Anim. Bc dogs to explore social play and the development of(Received 30 November 2007; initial acceptance 27 February 2008;
?nal acceptance 13 May 2008; published online - - -; MS. number: A07-00006)Partner preferences and asym
domestic dog, Canis lup
CAMILLE WARD* , ERIKA B. B
*Department of Psychology, U
yDepartment of Animal Programs, Smithsoni1
472/08/$34.00/0  2008 The Association for the Stu
ite this article in press as: Camille Ward et al., Partner preferences and asym
ehav. (2008), doi:10.1016/j.anbehav.2008.06.004ciencedirect.com
etries in social play among
s familiaris, littermates
ER? & BARBARA B. SMUTS*
ersity of Michigan, Ann Arbor
National Zoological Park, Washington, D.C.
08, --, --e--
v.2008.06.004dy of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
metries in social play among domestic dog, Canis lupus familiaris, littermates,
ARTICLE IN PRESS
2multiple litters of wild meerkats (also social carnivores),
Sharpe (2005a, b, c) concluded that several of the most
frequently proposed functions of social play were not
supported by the data, but she did not propose any alter-
native explanations for why young meerkats play. Only
one study has linked play directly to survival (brown
bear cubs, Ursus arctos, that play more in their ?rst sum-
mer are more likely to survive until the end of the next
summer), but the mechanisms involved remain un-
known (Fagen & Fagen 2004).
Evidence for some species indicates that mammals play
more and initiate play more often with individuals they
can dominate during play (Owens 1975; Biben 1986,
1998). Even so, experiencing subordinate positions during
play may also confer long-term bene?ts because individ-
uals gain the opportunity to practice defensive strategies
that could later be used during an actual ?ght. Also, indi-
viduals who consistently maintain a dominant position
during play may ?nd themselves at a loss for willing
play partners (Biben 1998). Some researchers contend
that in order for play to occur, both participants must
win an equal proportion (50%) of play encounters (Bekoff
2001). This is called the 50e50 rule (Aldis 1975; Pellis &
Pellis 1998). Individuals can balance the time spent in
dominant and subordinate roles during play by self-
handicapping and reversing roles. Self-handicapping
occurs when individuals place themselves in disadvan-
taged positions or situations that could make them more
vulnerable to attack by ?opponents? (Fagen 1981; Bekoff
& Allen 1998; Spinka et al. 2001; Bauer & Smuts 2007).
For example, red-necked wallabies, Macropus rufogriseus
banksianus, self-handicap when playing with younger
partners by standing in a defensive, ?at-footed posture
and pawing rather than sparring (Watson & Croft 1996),
thereby allowing younger partners to gain a competitive
advantage. Role reversals occur when animals change
dominant and subordinate positions during play ?ghting
(Balfour 1987; Biben 1998; Burghardt 2005) or when ani-
mals dominant outside of play assume subordinate roles
during play (Bauer & Smuts 2007). In some cases, the op-
portunity to play may be more important than winning.
For example, among squirrel monkeys, Saimiri sciureus
(Biben 1998), and hamadryas baboons, Papio hamadryas
hamadryas (Pereira & Preisser 1998), stronger individuals
?soften? their play style as a mechanism to promote play
relations when choice of play partners is restricted.
Of the studies mentioned above, only Bauer & Smuts
(2007) published quantitative data on adherence to the
50e50 rule, which is why additional data (like those pro-
vided here) are important. Bauer & Smuts (2007) found
that adult dogs do not play in accordance with the pre-
dictions of the 50e50 rule, although the degree of diver-
gence from egalitarian play varies among dyads (only
5.4% of the 55 dyads tested displayed 50e50 symmetry
during play, whereas 21.8% displayed complete asymme-
try). The 50e50 rule predicts that the dominant member
of a dyad will self-handicap more (to bring the probabil-
ity of winning for the subordinate closer to 50%), but
they found the reverse effect. The 50e50 rule also pre-
ANIMAL BEHAVIOUR, --, -dicts that dominance relationships outside of play will
be absent (or at least very relaxed) during play, but they
Please cite this article in press as: Camille Ward et al., Partner preferences and asym
Anim. Behav. (2008), doi:10.1016/j.anbehav.2008.06.004found that dominance relationships were still evident
during play. Although it is not yet known why dog dyads
display such wide variation in degrees of asymmetry dur-
ing play, the study by Bauer & Smuts (2007) makes it
clear that ?fair play? (Bekoff 2001) is not necessary for
play to occur.
Because social play often includes motor patterns used
in predatory, mating and agonistic contexts, play signals
have evolved to help animals convey playful intentions
(West 1974; Bekoff & Allen 1998). For example, Bekoff
(1995) found that among captive coyotes, Canis latrans,
wolves, Canis lupus, and domestic dogs, the play bow sig-
nal was more likely to occur immediately preceding or fol-
lowing behaviours that could most easily be
misinterpreted as real aggression, such as when an animal
shook the head while biting another. In adult domestic
dogs, the individual within a dyad who showed more
self-handicapping behaviour also play-signalled more of-
ten (Bauer & Smuts 2007).
For a given species, sex differences in play should be
more prominent when adult females and males differ in
physical characteristics (e.g. body size), behavioural char-
acteristics (e.g. involvement in hunting activity, intraspe-
ci?c ?ghting, territorial defence) or social preferences (e.g.
formation of strong bonds with same- versus opposite-sex
individuals; Maestripieri & Ross 2004; Burghardt 2005).
Among canids, males and females show little size dimor-
phism and engage in similar roles (Derix et al. 1993;
McLeod & Fentress 1997), and some research has sug-
gested no sex differences in play style (Bekoff 1974; Biben
1983; Bauer & Smuts 2007). However, among infant
domestic dogs, sex differences in play have been reported
(Lund & Vestergaard 1998; Pal 2008).
In domestic dogs, social and sexual play (e.g. mounting)
?rst emerge during the socialization period that begins at
3 weeks and ends at approximately 12 weeks of age
(Freedman et al. 1961; Scott & Fuller 1965). During this
time, puppies learn social skills and form bonds with other
dogs (Lindsay 2000), and play-partner preferences may
begin to form. Following the socialization period, the
juvenile period lasts from approximately 12 weeks to 6
months or later (until sexual maturity; Scott & Marston
1950; Scott & Fuller 1965). During this time, play con-
tinues to be common.
Some research suggests that in domestic and wild
canids, play contributes to the formation of dominance
relationships within litters (domestic dogs: Scott & Fuller
1965; Bekoff 1972; wild red foxes: Meyer & Weber
1996). However, no one, to our knowledge, has systemat-
ically quanti?ed the development of dominance relations
among littermates in wolves or domestic dogs during nat-
urally occurring social interactions. In our study, virtually
every intraspeci?c social interaction puppies had (other
than resting in body contact or snif?ng) occurred in the
context of play. Therefore, we made no attempt to analyse
dominance relationships per se, although we did examine
role asymmetries during play (see below).
We examined the development of social relationships
during play among littermates in three ways. First, we
examined play-partner preferences in puppies, diversity in
choice of partners and the stability of preferences over time.
metries in social play among domestic dog, Canis lupus familiaris, littermates,
Second, we examined asymmetries in social play. Speci?-
cally, we determinedwhether play conformed to the 50e50
rule; we examined the relationship between various play
behaviours, including play initiations, ?offense behaviours?
(similar to ?attacks and pursuits? in Bauer & Smuts 2007),
self-handicapping and play bows; and we determined
whether play behaviours varied by sex. Last, we examined
how the individual actions that comprised offense and
self-handicapping behaviours varied by litter and time.
METHODS
Subjects
riods
riods
associated with the development of social behaviours
collected in the adjacent yard, a 506-m2 fenced area. For
all litters, living areas were large enough to allow puppies
to move and play freely.
During time period 2 (Table 1), owners of puppies from
litters 1 and 2 brought them back to the breeders? homes
once or twice a month and placed them together for a cou-
ple of hours in the outdoor enclosures described above
(extended observations of litters 3 and 4 were not possi-
ble). This allowed us to collect longitudinal data on litter-
mate play behaviour starting at 10e11 weeks through 23
weeks of age. For litter 1, all six littermates attended four
sessions. For litter 2, ?ve puppies attended three sessions,
and all six puppies came to the remaining session.
Finally, for litter 1, we continued once-a-month obser-
vations from 27 to 40 weeks of age (time period 3, Table 1)
tify-
Litter Breed Birth date* Sex compositiony Ob
F
M
M
M
ARTICLE IN PRESS
WARD & SMUTS: PLAY AMONG DOMESTIC DOG SIBLINGS 31 Shepherd mix 14 Feb 2004 3 \
3 _
2 Labrador retriever 21 Apr 2005 2 \
4 _
3 Doberman pincher 22 Apr 2004 1 \
2 _
4 Malamute 4 May 2005 4 _Table 1. Summary of domestic dog litters(Scott & Fuller 1965) and also to equate data collected
across litters. Time 1 included the socialization period,
time 2, the late socialization and early juvenile period,
and time 3, the later juvenile period.
For all litters, we observed puppies in the breeders?
homes from 3 to 7e8 weeks of age (time period 1; see
Table 1). During time 1, litters 1e3 were housed indoors in
one-room enclosures of approximately 7.2 m2. Litters 2
and 3 also had free access to outdoor fenced areas 99
and 72 and 99 m2 in size, respectively. Litter 4 was ken-
nelled outdoors in a 4-m2 enclosure, but data wereWe divided the data collection into three time pe
(Table 1) to coincide approximately with critical peWe observed four litters of domestic dogs (three pure-
bred litters and one mixed-breed litter; Table 1). All dams
lived in ordinary households, and all puppies whelped
naturally (no caesarean births) at home. The dams suckled
the puppies and weaned them prior to placement in per-
manent homes. Breeders supplemented nursing with solid
foods starting around 4 weeks of age. Puppies from litters
2e4 remained sexually intact for the duration of the
study. With the exception of one male, all of the puppies
from litter 1 were spayed or neutered, starting at 21 weeks
of age.
Data Collection*Only surviving puppies are listed.
yOne puppy (female) from litter 1 died shortly after birth, and two pupp
zObservations were collected over three time periods.
Please cite this article in press as: Camille Ward et al., Partner preferences and asym
Anim. Behav. (2008), doi:10.1016/j.anbehav.2008.06.004Behavioural Coding
We coded data from videotapes into a Microsoft Excel
(version 2003) spreadsheet. We coded data only for play
bouts that involved mutual, social play lasting for at least
2 s. Based on the log survivorship analysis (Martin &
Bateson 1993), we considered a subsequent play bout
between a focal and the same play partner to be indepen-
dent if the interval separating successive bouts was at least
1 min.
To investigate play initiations and partner preferences,
we coded play-partner identities, the puppy who initiated
play, how play was initiated (see Bekoff 1972) and the
servation dates Time periodz Age range of puppies (weeks)
ebeNov 2004 1 3e8
2 11e23
3 27e40
ayeOct 2005 1 3e8
2 10e23
ayeJun 2004 1 3e7
ayeJun 2005 1 3e83, were marked by the breeder with nail polish in iden
ing locations on their bodies.under the conditions described above. All littermates
attended two sessions and ?ve attended one session.
For litters 1e3 (hereafter referred to as ?focal litters?) for
all time periods that applied, we conducted 5-min focal
samples on each puppy randomly selected during a given
session. We videotaped focal animals and those who
interacted with them using Canon ZR50 and Canon
ZR95 digital video cameras. We had limited access to litter
4 and therefore decided to maximize data by videotaping
play on an ad libitum basis (Altmann 1974). For identi?ca-
tion, puppies wore coloured collars or, in the case of litteries (both females) from litter 2 died shortly after birth.
metries in social play among domestic dog, Canis lupus familiaris, littermates,
times that play began and ended with each partner. To
ant
fo
g s
?s b
he
use
rou
aw
ith
sad
k m
m a
and
lay
ear
ctl
ARTICLE IN PRESS
4investigate behaviours that involved asymmetric roles
(hereafter ?asymmetries?), we divided them into (1) offense
behaviours and (2) self-handicapping behaviours (cf.
Bauer & Smuts 2007). For these and other behaviours
(e.g. play bows) we used a detailed ethogram (Table 2)
based on prior research on domestic dogs (Bekoff 1972;
Abrantes 1997; Bauer & Smuts 2007) and wolves (Schenkel
1967; van Hooff & Wensing 1987).Table 2. Ethogram of asymmetric behaviours in play
Behaviour
Offense behaviours Behaviours used to maintain a domin
Bite shake D* bites Sy and shakes head back and
Chase D runs after S with a least two runnin
Chin over D places the underside of chin over S
S?s shoulders, but sometimes over S?s
Forced down D uses physical force or contact to ca
a moving, standing or sitting position
Mount D rears up (keeping hindlegs on the g
spine with curved front legs and forep
Muzzle bite D places mouth around S?s muzzle
Over D sits on, stands over, or lies over S w
Self-handicapping Behaviours that place an actor in a di
Muzzle lick S licks on or around D?s muzzle. A lic
Voluntary down S drops completely to the ground fro
without D?s physical enforcement. D
Play signal Used to begin play or to maintain a p
Play bow Dog is crouched down, touching or n
end high in the air. Orientation is dire
*D: dog in dominant or winning position.
yS: dog in subordinate or losing position.
ANIMAL BEHAVIOUR, --, -Data Analysis
Play-partner preferences
For focal litters, we calculated a play-partner preference
(PPP) score for each animal in each litter using the
following index (cf. Thompson 1996),
Iij ?
Bij
Bi=?k 1?
where k is the total number of puppies in a litter, Bij is the
number of play bouts initiated by the ith puppy with the
jth puppy as the recipient and Bi is the total number of
play bouts initiated by the ith puppy. A puppy who initi-
ated with all other partners an equal number of times
would have all Iij score equal to 1.0. Scores were greater
than 1.0 in situations in which individuals initiated play
with some individuals more often than with others. Fol-
lowing Thompson (1996), we used PPP scores of 2.0 to
de?ne ?strong partner preferences?. We did not calculate
PPP scores for litter 2, time 2 (because on most occasions,
only ?ve of the six puppies were present).
We used rowwise matrix correlation tests (Kr; MatMan
software package with 10 000 permutations; Hemelrijk
1990a, b; de Vries 1993) to examine several aspects of
play-partner preferences. Matrices were constructed sepa-
rately by litter and time period for litters 1 and 2 (although
Please cite this article in press as: Camille Ward et al., Partner preferences and asym
Anim. Behav. (2008), doi:10.1016/j.anbehav.2008.06.004we calculated PPP scores for litter 3, sample size was too
small, N ? 3, to conduct matrix analyses (Hemelrijk
1990a), and, as mentioned above, we did not collect focal
data on litter 4).
To determine if PPP scores were reciprocal (i.e. if puppy
A preferred puppy B as a play partner, did B also tend to
prefer A?), we compared a data matrix of PPP scores (with
actors in rows and recipients in columns) with a second
matrix that was a transposition of the ?rst. To test whether
PPP scores were correlated across time periods for litter 1,
we created a PPP matrix (as described above) for each time
Definition
or superior position over a partner
rth while maintaining a hold on S
trides while S runs or trots away from D
ack, usually right behind the neck or near
ad
S to drop completely to the ground from
nd) to place forelegs on S?s back. D has a rounded
s to grasp S?s torso. Pelvic thrusting may or may not be present
at least 25% of D?s torso over S?s torso
vantaged or inferior position
ay or may not be accompanied by nudging
moving, standing or sitting position
S must be interacting when S goes down
ful mood during a play bout
ly touching forelimbs to the ground with rear
y towards play partnerperiod and compared matrices. Finally, to test if mean
play-bout lengths were longer with preferred partners, we
compared the PPP matrix with a matrix that contained
mean play-bout lengths for each dyad (the total amount
of time each dyad played divided by the number of play
bouts for that dyad; Martin & Bateson 1993).
To testwhether thenumber of stronglypreferred partners
for each puppy in litter 1 increased across the next two time
periods, we conducted a McNemar test in SAS version 9.1
(SAS Institute, Cary, North Carolina, U.S.A., 2003).
We also tested whether mean bout length varied by time
period or dyadic sex composition (femaleefemale versus
femaleemale versus maleemale) with a general linear
mixed model (GLMM) in SAS (SAS Institute, 2003). We
controlled for litter effects and repeated observations on
dyads within a litter across time. By including litter in the
model, we accounted for variables that contributed to
interlitter variation. For this and all subsequent GLMMs,
all two-way interaction terms were included in the initial
model, but the least signi?cant interactions were sequen-
tially dropped using a backwards elimination procedure.
Only signi?cant interactions or those showing trend
effects were included in the ?nal models. For all GLMM
procedures, we tested for normality and homogeneity of
variances, as appropriate, and transformed the dependent
variable when necessary to meet the assumptions.
metries in social play among domestic dog, Canis lupus familiaris, littermates,
a GLMM to determine if the degree of asymmetry differed
by time or dyadic sex composition. If neither time nor sex
ARTICLE IN PRESS
5composition was signi?cant, we pooled the time and sex
composition data and ran a second GLMM to determine
if the degree of asymmetry differed from 0. We controlled
for litter effects and repeated observations on dyads
within a litter across time.
Relationship between play behaviours. We ran GLMMs to
determine: (1) if offense behaviour rates were associated
with initiation rates (dependent variable), (2) if self-
handicapping rates were associated with offense behav-
iour rates (dependent variable), and (3) if self-handicap-
ping and offense behaviour rates (dependent variables, cf.
Bauer & Smuts 2007) were associated with play bow rates.
We included time as a ?xed effect, and for these and all
other GLMMs that follow, we controlled for litter effects
and repeated observations on puppies within a litter
across time. We did not test for associations between initi-
ations and self-handicapping or initiations and play bows
because scatter plots suggested no relationships.Realized diversity
Following Thompson (1996), we used the Shannone
Weaver diversity index (Shannon & Weaver 1949) to mea-
sure the diversity in choices of play partners for each
puppy in all focal litters. The index (R) is
R?
P
piln

l=pi

ln?n?
where pi is the probability of a puppy initiating play with
the ith puppy in the litter and n is the number of available
play partners. The index ranges from 0 to 1, and R is max-
imized when a puppy initiates play with all possible part-
ners with equal frequency and minimized when a puppy
initiates with only one partner. To test whether realized
diversity scores varied by sex or time, we ran a GLMM
with sex and time as ?xed effects. We controlled for litter
effects and repeated observations on puppies within a litter
across time periods.
Asymmetries
Testing the 50e50 rule. To examine whether play con-
formed to the 50e50 rule, we ?rst calculated asymmetry
(or conversely, symmetry) in play as follows. The number
of ?wins? for individual A in a dyad (e.g. AB dyad
consisting of puppies A and B) equalled the number of
offense behaviours by A directed at B plus the number of
self-handicapping behaviours (Table 2) by B directed
towards A. B?s wins were calculated similarly. Next, we cal-
culated the proportion of wins for A as the number of wins
for A divided by the total number of wins for both A and
B. We calculated the proportion of wins for B in the same
way. We subtracted the smaller proportion of wins from
the larger proportion to obtain a measure of the degree
of asymmetry in each dyad. We normalized this measure
with an arcsine-root transformation. The closer the value
was to 0, the more symmetrical the play. Next, we ranWe calculated rates of behaviours (e.g. initiations, of-
fenses, self-handicapping andplay bows) at the dyadic level
Please cite this article in press as: Camille Ward et al., Partner preferences and asym
Anim. Behav. (2008), doi:10.1016/j.anbehav.2008.06.004for each puppy in each time period as the number of times
each puppy in a dyad performed one of the behaviours
divided by the total time that dyad was observed playing.
For example, if A and B played together for 100 s in time 1
andAdisplayed 20 offense behaviours and B 10, the offense
ratewouldbe0.20 forA and0.10 for B (recall that initiations
occurred only once per play boutdat the startdwhereas of-
fense behaviours, self-handicapping and play bows could
occurmultiple timeswithin eachbout).Weusedmean rates
of behaviours for eachpuppywithin each timeperiod in the
following GLMM analyses.
Variation in play behaviour as a function of dyadic sex com-
position. We tested whether initiation, offense, self-
handicapping or play bow rates (dependent variables)
varied by sex or time in mixed-sex (FM) and same-sex
(FF versus MM) dyads using GLMMs as described above.
We also compared rates of female behaviours when
females were playing with other females (ff dyads) versus
when they were playing with males (fm dyads), and
similarly, we compared rates of male behaviours in
maleemale (mm) dyads versus maleefemale (mf) dyads.
Note that fm and mf dyads have the same composition;
we use lower- versus upper-case abbreviations to remind
the reader which sex?s behaviour we are measuring (the
one listed ?rst in mixed-sex dyads) in the relevant
analyses. For example, when measuring the rate of offense
behaviours in an ?fm? dyad, we calculated only the rate for
females when playing with males (see below). In contrast,
when measuring the rate of offense behaviours in an FM
dyad, we calculated rates for both sexes.
For ff and fm dyads, we obtained difference scores for
each dependent variable (e.g. rates of initiations, offense
behaviours, self-handicapping andplaybows) separately by
subtracting the rates of female-to-female behaviours from
the rates of female-to-male behaviours. Initially, we used
GLMMs to determine if difference scores varied by time. If
not, we pooled observations across time periods and reran
themodels to test for behavioural differences in general.We
followed a similar procedure for mm and mf dyads.
All statistical tests were two-tailed, unless otherwise
speci?ed, and a was set at 0.05. However, if a was between
0.05 and 0.10, we report it as a nonsigni?cant trend.
Frequency of offense and self-handicapping behaviours. We
determined the proportions of offense and self-handicap-
ping behaviours averaged across all dyads for each litter and
time period as follows. First, we determined the rates of
individual offense behaviours (Table 2) separately for each
dyad and time period by dividing the frequency of individ-
ual offense behaviours by the total time played for a given
dyad. We averaged the rates of individual offense behav-
iours by dyad across dyads to obtain a mean rate for each
of the behaviours. Rates of self-handicapping were deter-
mined similarly. We calculated the proportions of offense
and self-handicapping behaviours based on mean rates.
RESULTS
WARD & SMUTS: PLAY AMONG DOMESTIC DOG SIBLINGSWe analysed 7.10 h of dyadic play from four litters of
puppies consisting of 39 dyads and 1200 total play bouts.
metries in social play among domestic dog, Canis lupus familiaris, littermates,
of
and
times combined.
atrix
ntly
related to mean play-bout length for puppies in litters 1
ime:
ntly
between times 1 and 2 (P ? 0.796; Fig. 1).
F2,74 ? 3.88, P ? 0.025; sex: F2,74 ? 0.510, P ? 0.605). To
explore this time effect further, we ran a second GLMM
pooled across sex composition while retaining time as
0
0.2
0.4
0.6
0.8
1
1 2 3
Time periods
M
ea
n
 r
ea
li
ze
d
 d
iv
er
si
ty
 s
co
re
a
b
a
Figure 1. Realized diversity scores across three time periods. Values
are mean estimates and error bars indicate SE for litters 1 (time
periods 1e3), 2 (time period 1) and 3 (time period 1). Columns
with a letter in common are not significantly different (P > 0.05).
Age ranges (in weeks) of puppies: time 1, 3e8; time 2, 10e23;
time 3, 27e40.
0
0.1
0.2
0.3
0.4
0.5
1 2 3
Time periods
D
eg
re
e 
of
 a
sy
m
m
et
ry
Figure 2. The degree of asymmetry across three time periods. Values
are mean estimates and error bars indicate SE for litters 1 (time
periods 1e3), 2 (time periods 1 and 2), 3 (time period 1) and 4
(time period 1). To obtain these values, we ran a GLMM with the
degree of asymmetry (based on nontransformed data) as the depen-
dent variable and time as an independent variable. We controlled for
the random effects of litter and repeated observations on dyads
ARTICLE IN PRESS
6Asymmetries
Testing the 50e50 rule(P ? 0.0008) and time 2 (P ? 0.004; GLMM: t
F2,21 ? 8.32, P ? 0.002). Scores did not differ signi?caand 2, time 1 (rowwise matrix tests: litter 1: Kr ? 1,
P ? 0.830; litter 2: Kr ? 8, P ? 0.351). However, PPP
scores and mean bout length were positively related for lit-
ter 1 in time 2 (rowwise matrix test: Kr ? 23, P ? 0.012)
and in time 3 (Kr ? 18, P ? 0.079, trend effect). For all lit-
ters combined, mean bout length varied by time period
but not by dyadic sex composition (GLMM: time:
F2,76 ? 4.20, P ? 0.019; dyadic sex composition:
F2,76 ? 0.30, P ? 0.744). Play bouts were shorter in time
3 compared to times 1 and 2 (time 1 versus time 2:
P ? 0.172; time 1 versus time 3: P ? 0.065; time 2 versus
time 3: P ? 0.005; time 1: X estimate  SE ? 16.377 
3.641 s; time 2: 19.934  3.966 s; time 3: 9.990 
4.549 s).
PPP scores in times 2 and 3 were positively correlated
with scores in time 1 for litter 1 (rowwise matrix tests:
times 1 and 2: Kr ? 16, P ? 0.060; times 1 and 3: Kr ? 19,
P ? 0.048), but scores in times 2 and 3 were not signi?-
cantly related (Kr ? 12, P ? 0.18).
The number of strongly preferred partners (i.e. PPP
scores  2.0) increased over time. In times 1 and 2, only
one puppy from litter 1 had a strongly preferred partner
(PPP scores ranged from 0 to 2.5 in time 1 and from 0 to
2.2 in time 2), but in time 3, all six puppies had one strong
preference (McNemar?s test: c21 ? 5:00, P ? 0.025; PPP
scores ranged from 0 to 3.76). In time 1 for litters 2 and
3, none of the puppies had strongly preferred play part-
ners (PPP scores ranged from 0.32 to 1.8 for litter 2 and
from 0.71 to 1.3 for litter 3), although they did appear
to initiate play with particular puppies more often than
with others (PPP scores > 1.0).
Realized Diversity
Variation in realized diversity scores was not attribut-
able to sex (GLMM: F1,21 ? 0.92, P ? 0.348), but scores
were lower in time 3 compared with scores in time 1P ? 0.852) or for litter 2 puppies in time 1 (rowwise m
test: Kr ? 0, P ? 1.000). PPP scores were not signi?caPlay-Partner Preference Scores
PPP scores were not reciprocal for puppies from litter 1
in any time period (rowwise matrix tests: time 1: Kr ? 7,
P ? 0.454; time 2: Kr ? 14, P ? 0.200; time 3: Kr ? 2,13.15  5.13 bouts, and we coded an average
5.22  0.71 min of play per dyad across all littersThe average number of play bouts per dyad (X SD) was
ANIMAL BEHAVIOUR, --, -The overall degree of symmetry in play did not vary by
sex composition, but it decreased over time (GLMM: time:
Please cite this article in press as: Camille Ward et al., Partner preferences and asym
Anim. Behav. (2008), doi:10.1016/j.anbehav.2008.06.004a ?xed effect. This analysis showed that littermates as
a whole did not conform to the 50e50 rule during play
in any time period (GLMM: time 1: t76 ? 7.15,
P < 0.0001; time 2: t76 ? 8.52, P < 0.0001; time 3:
t76 ? 7.18, P < 0.0001; all one-tailed), but symmetry was
higher in time 1 relative to time 2 (P ? 0.015) and time
3 (P ? 0.024). For litter 1, there was no signi?cant differ-
ence in symmetry between times 2 and 3 (P ? 0.693).
Figure 2 shows the degree of asymmetry across time
periods.
0.6within a litter across time. Age ranges (in weeks) of puppies: time
1, 3e8; time 2, 10e23; time 3, 27e40.
metries in social play among domestic dog, Canis lupus familiaris, littermates,
Relationship between play behaviours
The relationship between rates of initiations and offense
behaviours was positive in all time periods, but it was only
signi?cant in time 3 (P < 0.0001; GLMM: offense behav-
iours: F1,28 ? 17.8, P ? 0.0002; time: F2,28 ? 1.92,
P ? 0.165; offense behaviours*time: F2,28 ? 6.16,
P ? 0.006). Neither self-handicapping nor play bows
were related to offense behaviours in any time period
(GLMM: self-handicapping: F1,30 ? 0.04, P ? 0.848; time:
F2,30 ? 0.96, P ? 0.393; GLMM: play bows: F1,30 ? 1.35,
P ? 0.254; time: F2,30 ? 1.23, P ? 0.305). However, play
bows were positively associated with self-handicapping
across all time periods (GLMM: play bows: F1,30 ? 9.85,
P ? 0.004; time: F2,30 ? 0.40, P ? 0.677).
Variation in play behaviour as a function
of dyadic sex combination
Rates of play initiation. In mixed-sex dyads, males initi-
ated play more often than their female partners (Table 3).
This result could re?ect higher rates of play initiations by
males in general, but comparisons of male and female ini-
tiation rates in same-sex dyads showed no differences ex-
cept in time period 3, when MM rates of initiation were
higher than FF rates (Table 3).
Females initiated play more often with other females
than with males, and results did not vary with time period
(Table 3; ff versus fm dyads). However, male initiation rates
did vary by time period. In time 3,males initiated playmore
often with other males than with females, but in times 1
and 2, males were just as likely to initiate play with females
as they were to initiate with other males (Table 3).
Offense behaviour rates. In mixed-sex dyads, males dis-
played offense behaviours towards females more often
than the reverse (Table 4). In same-sex dyads, males and
females displayed offense behaviours at similar rates
across all time periods (Table 4).
Rates of offense behaviours for both females and males
were stable across time periods. Females displayed of-
fense behaviours at similar rates whether playing with
females (ff) or males (fm), but males displayed offense
behaviours slightly more often when playing with
females (mf) than when playing with other males (mm;
trend effect; Table 4).
Self-handicapping rates. In mixed-sex dyads, males self-
handicapped more often than females (Table 5); however,
males and females self-handicapped at similar rates in
same-sex dyads (Table 5).
Self-handicapping rates did not vary by sex across time
periods. Females self-handicapped at similar rates whether
playing with females (ff) or males (fm), and the same was
true for males (Table 5).
versus fm
ARTICLE IN PRESS
WARD & SMUTS: PLAY AMONG DOMESTIC DOG SIBLINGS 7Male initiationsz: mm versus mf
Time period 1 14 1.02 0.327
Time period 2 14 0.62 0.543
Time period 3 14 2.18 0.047 mm>mf
FM: femaleemale; FF: femaleefemale; MM: maleemale; ff: female
rates with females; fm: female rates with males; mm: male rates
with males; mf: male rates with females. T1: time 1; T2: time 2;
T3: time 3.
*Rates of initiations in mixed-sex and same-sex dyads in relation to
sex of initiator and time period. GLMMs controlling for litter and re-
peated observations of dyad{litter} across time.
yTest of the difference in female initiation rates between ff and fm
dyads and the difference in male initiation rates between mm and
mf dyads. GLMMs controlling for litter and repeated observations
of dyad{litter} across time. GLMM was pooled across time periods
for female initiations owing to nonsignificant time effects (GLMM:Table 3. Initiation rates analyses
Source of variation df
F* or ty
value P
Direction of
effect
Mixed-sex dyads (FM)
Initiator sex 1, 27 23.98 <0.0001 M>F
Time period 2, 27 0.44 0.649
Same-sex dyads (FF versus MM)
Initiator sex 1, 27 1.49 0.233
Time period 2, 27 5.50 0.009 T3>T1, T2
Initiator
sex*time period
2, 27 4.33 0.023 MM>FF
in T3
Female initiations: ff 1 15.98 0.039 ff>fmF2,9 ? 0.96, P ? 0.418).
zFor male initiations, results were not pooled across time because of
a trend effect with time (GLMM: F2,14 ? 2.78, P ? 0.096).
Please cite this article in press as: Camille Ward et al., Partner preferences and asym
Anim. Behav. (2008), doi:10.1016/j.anbehav.2008.06.004Play bow rates. In mixed-sex dyads, males and females
play bowed at similar rates (Table 6). The same was true for
females compared with males in same-sex dyads (Table 6).
Play bow rates did not vary by sex across time periods.
Females play bowed at similar rates whether playing with
Table 4. Offense behaviour rates analyses
Source of variation df
F* or
ty value P
Direction
of effect
Mixed-sex dyads (FM)
Sex 1, 27 7.79 0.009 M>F
Time period 2, 27 0.14 0.870
Same-sex dyads (FF versus MM)
Sex 1, 29 1.63 0.212
Time period 2, 29 3.21 0.055
Female offense
behaviours: ff versus fm
1 0.92 0.525
Male offense
behaviours:
mm versus mf
2 3.19 0.086
FM: femaleemale; FF: femaleefemale; MM: maleemale; ff: female
rates with females; fm: female rates with males; mm: male rates
with males; mf: male rates with females.
*Rates of offense behaviours in mixed-sex and same-sex dyads in
relation to sex and time period. GLMMs controlling for litter and
repeated observations of dyad{litter} across time.
yTest of the difference in rates of offense behaviours by females in ff
and fm dyads and the difference in rates by males in mm and mf
dyads. GLMMs controlling for litter and repeated observations of
dyad{litter} across time. GLMMs were pooled across time periods
for female and male offense behaviours owing to nonsignificant
time effects (GLMM: females: F2,9 ? 1.39, P ? 0.297; males:
F2,14 ? 1.96, P ? 0.177).
metries in social play among domestic dog, Canis lupus familiaris, littermates,
were
, b).
Table 5. Self-handicapping rates analyses
Source of variation df
F* or
ty
value P
Direction
of effect
Mixed-sex dyads (FM)
Self-handicapper sex 1, 27 4.87 0.036 M>F
Time period 2, 27 0.04 0.960
Same-sex dyads (FF versus MM)
Self-handicapper sex 1, 29 0.13 0.723
Time period 2, 29 0.92 0.442
Female
self-handicapping:
ff versus fm
1 1.17 0.449
Male
self-handicapping:
mm versus mf
2 1.01 0.418
FM: femaleemale; FF: femaleefemale; MM: maleemale; ff: female
rates with females; fm: female rates with males; mm: male rates
with males; mf: male rates with females.
*Self-handicapping rates in mixed-sex and same-sex dyads in rela-
tion to sex and time period. GLMMs controlling for litter and re-
peated observations of dyad{litter} across time. GLMMs were
pooled across time periods for female and male self-handicapping
ARTICLE IN PRESS
ANIMAL BEHAVIOUR, --, -8behaviours owing to nonsignificant time effects (GLMM: females:
F2,9 ? 0.51, P ? 0.615; males: F2,14 ? 1.35, P ? 0.291).peated observations of dyad{litter} across time.
yTest of the difference in female self-handicapping rates between ff
and fm dyads and the difference in male self-handicapping rates be-
tween mm and mf dyads. GLMMs controlling for litter and re-females (ff) or males (fm), and the same was true for males
(Table 6).
Frequency of offense and self-handicapping behaviours
Rates of individual offense and self-handicapping be-
haviours were similar across litters, so we pooled the data
et of
in-
Table 6. Play bow rates analyses
Source of variation df F* or ty value P
Mixed-sex dyads (FM)
Sex 1, 27 0.80 0.379
Time period 2, 27 2.24 0.126
Same-sex dyads (FF versus MM)
Sex 1, 29 2.29 0.141
Time period 2, 29 1.47 0.246
Female play bows: ff versus fm 1 0.88 0.542
Male play bows: mm versus mf 2 0.78 0.516
FM: femaleemale; FF: femaleefemale; MM: maleemale; ff: female
rates with females; fm: female rates with males; mm: male rates
with males; mf: male rates with females.
*Play bow rates in mixed-sex and same-sex dyads in relation to sex
and time period. GLMMs controlling for litter and repeated obser-
vations of dyad{litter} across time.
yTest of the difference in female play bow rates between ff and fm
dyads and the difference in male play bow rates between mm
and mf dyads. GLMMs controlling for litter and repeated observa-
tions of dyad{litter} across time. GLMMs were pooled across time
periods for female and male play bows owing to nonsignificant
time effects (GLMM: females: F2,9 ? 0.77, P ? 0.490; males:
F2,14 ? 2.01, P ? 0.171).
Please cite this article in press as: Camille Ward et al., Partner preferences and asym
Anim. Behav. (2008), doi:10.1016/j.anbehav.2008.06.004creased over time as diversity in partner choice decreased,
and by time 3, each puppy had one strongly preferred play
partner. PPP scores were consistent between early and late
time periods for litter 1.
These results indicate that puppies begin to form
differentiated relationships very early in life. Detailed
information on play partner preferences in other animals
is rare, but preferences for speci?c play partners occur in
infant sable antelope, Hippotragus niger (Thompson 1996)
and juvenile Belding?s ground squirrels, Spermophilus
beldingi (Nunes et al. 2004). Like the littermates in this
study, sable calves played with a variety of partners early
on, but later (9e12 weeks) developed stable preferences
with a subset of strongly preferred playmates. However,
unlike young sable antelope, puppies did not show recip-
rocal play partner preferences, a ?nding we discuss later.
Asymmetries
Testing the 50e50 rule
Puppies did not adhere to the 50e50 rule in any time
period; however, play was more symmetrical between very
young littermates and became less so as the puppies
matured. Red fox littermates of approximately 4e8 weeks
of age (three litters consisting of two trios and one pair)
also showed considerable asymmetry during social playconcentrated on initiating play with a smaller subs
individuals. Their preferences for speci?c partnersAlso in time 2, mounts appeared for the ?rst time (5% of
offense plus self-handicapping behaviours in both litters),
and the proportion of bite shakes decreased (Fig. 3b) In
time 3, mounts became the most common offense behav-
iour, followed by chases, and bite shakes disappeared
entirely (Fig. 3c). Muzzle bites and chin overs (Table 2)
were a small proportion of offense behaviours across all
time periods.
In time 1, nearly 100% of self-handicapping behaviours
involved voluntary downs (Fig. 3a; Table 2), but by time 2,
muzzle licks had grown to 44% of all self-handicapping
behaviours (in addition to voluntary downs; Fig. 3b). In
time 3, 75% of self-handicapping was accounted for by
muzzle licks (performed by one female; Fig. 3c).
DISCUSSION
Partner Preferences and Diversity
In time periods 1 and 2, puppies in all focal litters
initiated play with most of the other individuals in their
litter. However, as they matured, puppies in litter 1In times 1 and 2, forced downs and overs (Table 2)
the two most common offense behaviours (Fig. 3aacross litters and report results by time period. Offense
behaviours were much more common than self-handicap-
ping behaviours across all time periods, accounting for
94% of all behaviours (offense and self-handicapping
behaviours combined; Table 2) during time 1 (Fig. 3a),
91% of behaviours during time 2 (Fig. 3b) and 84% of be-
haviours during time 3 (Fig. 3c).(the proportion of ?wins? by the fox who won most often
within each dyad varied from 62 to 82%; values derived
metries in social play among domestic dog, Canis lupus familiaris, littermates,
MVoluntary down
bite
Mount
5%
ARTICLE IN PRESS
9Bite shake
13%
Chin over
2%
Chase
8%
Forced down
36%
Over
35%
Chin over
1% Forced down
36%
Chase
12%
Over
33%
Muzzle 
<1%
(a)
(b)from Table 2; Meyer & Weber 1996). Bauer & Smuts (2007)
reported that adult dogs did not conform to the 50e50
rule during dyadic play, and they found that play in adult
dogs re?ects the existing dominance structure outside of
play. The same is true for adult meerkats (Sharpe 2005a).
The studies mentioned above, however, did not pro-
vide data on changes in the degree of asymmetry over
time. In young rats paired together, symmetry in play
decreased over time as stable dominance relationships
developed (Panksepp 1981). Like rats (Panksepp 1981;
Pellis & Pellis 1991), dogs may use play to establish stable
social relationships and test their place in the existing so-
cial structure of a group (Bekoff 1972). If so, then the in-
creased asymmetry over time among littermates may
simply re?ect dominance relationships formed during
the juvenile period, making puppy play more like that
of adult dogs in this respect (cf. Bauer & Smuts 2007).
Domestic dogs, like wild meerkats (Sharpe 2005a), do
not use aggression (distinct from play ?ghting) to estab-
lish stable dominance relationships in the ?rst few weeks
of life, in contrast to some other social carnivores (cap-
tive coyotes: Bekoff 1974; spotted hyaenas, Crocuta
Bite shake
4%
Force
1
Chin over
1%
Chase
27%
Over
6%
Muzzle bite
1%
Mount
37%
(c)
Figure 3. Proportions of all offense and self-handicapping behaviours for (
1). Large pie includes offense behaviours and self-handicapping; small p
ranges (in weeks) of puppies: time 1, 3e8; time 2, 10e23; time 3, 27e
Please cite this article in press as: Camille Ward et al., Partner preferences and asym
Anim. Behav. (2008), doi:10.1016/j.anbehav.2008.06.004Self-
handicapping
9%
Muzzle lick
5%uzzle bite
<1%
Muzzle lick
<1%
Voluntary down
6%
Self-
handicapping
6%
WARD & SMUTS: PLAY AMONG DOMESTIC DOG SIBLINGScrocuta: Wahaj & Holekamp 2006). As far as we know, re-
searchers have not addressed the functional signi?cance
of interspeci?c variation in the presence or absence of
early ?ghting to establish stable dominance relationships
within litters.
Relationships between play behaviours
In time period 3, rates of initiations were directly related
to rates of offense behaviours. This ?nding suggests that as
puppies mature, winning or being in the dominant
position during play becomes more important. Similarly,
young male squirrel monkeys (Biben 1998), baboons
(Owens 1975), and laboratory rats (Hole 1988) preferred
partners they could ?dominate? during play. It seems likely
that in dogs, as in rats, male bison Bison bison, and many
primates, individuals become increasingly aware of intra-
group status as they mature, and play becomes more com-
petitive (Fagen 1981; Rothstein & Griswold 1991; Smith
et al. 1999). Preference for partners who can be dominated
could help explain why PPP scores were not reciprocal (see
above).
Self-
handicapping
16%
4%
Voluntary down
4%
d down
2%
Muzzle lick
12%
a) time 1 (litters 1e4), (b) time 2 (litters 1 and 2) and (c) time 3 (litter
ie shows proportions of specific self-handicapping behaviours. Age
40.
metries in social play among domestic dog, Canis lupus familiaris, littermates,
ing were less successful than mating attempts that elicited
female cooperation (Ghosh et al. 1984; Pal et al. 1999). In
same-sex dogs (Borchelt 1983; Sherman et al. 1996), same-
sex partners may play to practice threat and appeasement
ARTICLE IN PRESS
10summary, play with females may allow males to practice
both cooperative and competitive mating strategies in
a less serious context.
In contrast to our results, Bauer & Smuts (2007) found
no sex differences in attacks/pursuits (cf. offense behav-
iours this study) or self-handicapping. Possible discrep-
ancies in the ?ndings may be related to the choice of
subjects. Bauer & Smuts (2007) studied unrelated adultSimilar to the ?ndings of Bauer & Smuts (2007) for adult
dogs, we found no association between offense behaviours
and either self-handicapping or play bows, but high rates
of self-handicapping were related to high rates of play
bows across all time periods. This suggests that self-
handicapping and play signalling may function together
to encourage play (Bauer & Smuts 2007).
Variation in play behaviour as a function
of dyadic sex combination
Mixed-sex partners. In mixed-sex dyads, males initiated
play more often than females across all time periods (Table
3). With one possible exception (MM dyads initiated more
often than FF dyads in time 3), these results were not just
an artefact of males being more active in general. Similar
to our ?ndings, Lund & Vestergaard (1998) reported that
males initiated social play with females more often than
expected by chance in four litters of domestic dogs ages
3 to 8 weeks. Male free-ranging dog puppies studied in
India engaged in social play more frequently than female
puppies; however, it is not clear whether these male
puppies were playing with other males, other females or
both (Pal 2008).
In addition to initiating at higher rates than females,
males in FM dyads also displayed self-handicapping more
often than females and, as indicated above, self-handicap-
ping correlates with play signalling and appears to
function to maintain play. Perhaps playing with females
provides opportunities for males to learn characteristics of
female behaviour and gain competence in interactions
with them. If so, it could translate into greater male
reproductive success later in life via female mate prefer-
ences. Pal et al. (1999) found that in a population of free-
ranging dogs in India, oestrous females selectively allowed
some males to mate and even solicited them but avoided
other males or even attacked them if they tried to mount.
Beach & LeBoeuf (1967) also reported female mating pref-
erences in captive dogs.
In FM dyads, male puppies displayed offense behaviours
more often than females (Table 4). In food competition
tests, male puppies also tended to dominate females by
15 weeks (Scott & Fuller 1965). Learning to dominate fe-
males may have reproductive payoffs as well. For example,
in Indian free-ranging dogs, males forced copulations on
unwilling, often sexually inexperienced, oestrous females
who responded with a combination of aggressive, submis-
sive, and escape postures. However, attempts to force mat-
ANIMAL BEHAVIOUR, --, -dogs, whereas we examined social relationships between
young littermates. Ways in which age and kinship might
Please cite this article in press as: Camille Ward et al., Partner preferences and asym
Anim. Behav. (2008), doi:10.1016/j.anbehav.2008.06.004signals that ritualize aggression. Practicing aggressive com-
ponents within a safe range of intensity could limit the
occurrence of overt aggression later on (Lindsay 2005).
In adult pet dogs, ?ghts between females lead to more se-
rious injuries than ?ghts between males or ?ghts between
mixed-sex dogs (Sherman et al. 1996). The early onset and
consistent preference for females to initiate play with
other females may be one way females learn to moderate
same-sex aggression.
In contrast to some of our ?ndings, Lund & Vestergaard
(1998) reported that male and female puppies (littermates
up to 8 weeks of age) did not prefer same- over mixed-sex
play partners. However, Lund & Vestergaard (1998) did
not limit their observations to dyadic interactions, as in
the current study, but also included triadic interactions,
which could have in?uenced their ?ndings in ways that
remain to be investigated.
Additionally, the increased rates of same-sex initiations
recorded in our study were not simply an artefact of play-
bout lengths varying by dyadic sex combination (e.g.
males initiated with males more often than they initiated
with females in time 3 because MM play-bout lengths
were shorter than FM bout lengths in time 3; Table 3).
Mean bout lengths did not vary by dyadic sex composi-
tion across any time period in the current study.
Preferences for play with same-sex partners have been
reported in other species, including juvenile male bison
(Rothstein & Griswold 1991), rats of both sexes (Laviola &
Terranova 1998) and juvenile male Belding?s ground squir-
rels (Nunes et al. 2004).
Individual offense and self-handicapping behaviours
Some behaviours were more common early in develop-
ment (e.g. forced downs and overs) and became less
common as puppies matured. Conversely, other behav-
iours were less common early in development but became
more common with time (e.g. mounts). These ?ndings
suggest that the style and possibly the function of play,
even between the same play partners, may change over
time. This is clearly the case for male (but not female) rats,
who, at puberty, shift the type of defensive strategy used
with other males as they begin to establish dominance
relationships (Smith et al. 1996). It is not clear, however,in?uence intersexual play behaviours remain to be
investigated.
Same-sex partners. The preference for same-sex play
initiations in males and females during various time
periods (Table 3) suggests that play may serve as training
for intrasexual competition between same-sex littermates.
Like wolves (Mech 1970; Packard 2003), both female and
male domestic dogs form intrasexual dominance relation-
ships (Pal et al. 1998), and play may function in the for-
mation of these dominance relationships (Bekoff 1972).
Because dominance con?icts generally occur betweenwhether puppies show sharp discontinuities in frequen-
cies of different play behaviours, as is the case for rats
metries in social play among domestic dog, Canis lupus familiaris, littermates,
ARTICLE IN PRESS
11(Smith et al. 1996) and domestic cats (Barrett & Bateson
1978; Caro 1981), but future work could address this area.
Even with a small sample of four litters comprising
various breeds, litter sizes, and sex ratios, we found
consistent patterns of partner preferences, role asymme-
tries, and sex differences (for example, in rates of offense
behaviours). These results suggest that domestic dog
littermates, in general, show important similarities in
social development. It would be interesting to determine
whether these general patterns evolved during domesti-
cation or were inherited from wolf ancestors, but compa-
rable data are not available for wolves.
Despite the similarities we found across litters, studies of
other species suggest that important interlitter variations
in play behaviour are likely to occur. Play behaviour varies
among litters as a function of litter size, litter sex ratio, and
nutritional state or weight (Caro 1981; Laviola & Alleva
1995; D?Eath & Lawrence 2004; Nunes et al. 2004). Such
interlitter variability is also correlated with major differ-
ences in adult behaviours, including aggressiveness, sexual
behaviour, maternal behaviour and behavioural, hor-
monal and neurochemical responses to stress and novelty
(Sharpe et al. 1973; Namikas & Wehmer 1978; Laviola &
Terranova 1998; D?Eath & Lawrence 2004). Experimental
studies in rodents manipulating these variables show
that in at least some cases, interlitter variation in size, sex
ratios and play-?ghting behaviours are not only correlated
with, but actually cause, differences in adult behaviours
(Laviola & Terranova 1998; Pellis & Pellis 2007).
It is no surprise that social interactions early in life affect
adult behaviour. However, we know very little about why
particular early experiences affect adult behaviour in the
ways they do. Most of the data relating developmental
variation to adult behaviour come from laboratory rodents
or domestic piglets reared in environments different from
those of their wild counterparts, which limits our ability
to analyse these relationships from an adaptive perspec-
tive. We think that domestic dogs afford an exceptional
opportunity to pursue such questions because different
breeds have undergone systematic selection for adult dif-
ferences in behaviour (including social behaviour, e.g.
hounds and sled dogs must be able to work peacefully
with conspeci?cs in groups and terriers, as rodent hunters,
tend to work as individuals). Future research on domestic
dog littermates can help to: (1) reveal how speci?c selec-
tion pressures alter early social experiences and (2) explain
why some aspects of social development remain similar
across litters and across breeds. Such studies are likely to
advance our theoretical understanding of relationships
between evolutionary and ontogenetic processes.
Acknowledgments
We thank Monika Dressler, Lynne Coleman, Mary West-
off, Mara Markov and Pamela Fusco for allowing us to
videotape their litters. We thank Kathy Welch for her
statistical consultation, numerous undergraduate students
for assistance with data collection and coding (especially
Jane Na and Sarah Alessi) and Theresa Lee, Bobbi Low,
Patricia McConnell and John Mitani for their reviews of
the manuscript. This study was supported by funding
Please cite this article in press as: Camille Ward et al., Partner preferences and asym
Anim. Behav. (2008), doi:10.1016/j.anbehav.2008.06.004from the Horace H. Rackham School of Graduate Studies
at the University of Michigan, the Psychology Department
at the University of Michigan, and the Center for the
Education of Women at the University of Michigan. The
research presented here was approved by the University
Committee on the Use and Care of Animals at the
University of Michigan, Approval 8988.
References
Abrantes, R. 1997. Dog Language. Napierville, Illinois: Wakan Tanka
Publishers.
Aldis, O. 1975. Playfighting. New York: Academic Press.
Altmann, J. 1974. Observational study of behavior: sampling
methods. Behaviour, 49, 227e267.
Balfour, A. D. 1987. Social relationships in litters of domestic dogs
(Canis familiaris). Ph.D. thesis, University of Guelph.
Barrett, P. & Bateson, P. 1978. The development of play in cats.
Behaviour, 66, 106e120.
Bauer, E. B. & Smuts, B. B. 2007. Cooperation and competition
during dyadic play in domestic dogs, Canis familiaris. Animal
Behaviour, 73, 489e499.
Beach, F. A. & LeBoeuf, B. J. 1967. Coital behaviour in dogs. I.
Preferential mating in the bitch. Animal Behaviour, 15, 546e558.
Bekoff, M. 1972. Development of social interaction, play, and
metacommunication in mammals: an ethological perspective.
Quarterly Review of Biology, 47, 412e434.
Bekoff, M. 1974. Social play and play-soliciting by infant canids.
American Zoologist, 14, 323e340.
Bekoff, M. 1984. Social play behavior. Bioscience, 34, 228e233.
Bekoff, M. 1995. Play signals as punctuation: the structure of social
play in canids. Behaviour, 132, 419e429.
Bekoff, M. 2001. Social play behavior: cooperation, fairness, trust,
and the evolution of morality. Journal of Consciousness Studies, 8,
81e90.
Bekoff, M. & Allen, C. 1998. Intentional communication and social
play: how and why animals negotiate and agree to play. In: Animal
Play: Evolutionary, Comparative, and Ecological Perspectives (Ed. by
M. Bekoff & J. A. Byers), pp. 97e114. Cambridge, U.K.: Cam-
bridge University Press.
Bekoff, M. & Byers, J. A. 1981. A critical reanalysis of the ontogeny
of mammalian social and locomotor play: an ethological hornet?s
nest. In: Behavioral Development: the Bielefeld Interdisciplinary Pro-
ject (Ed. by K. Immelmann, G. W. Barlow & L. Petrinovich), pp.
296e337. New York: Cambridge University Press.
Biben, M. 1983. Comparative ontogeny of social behaviour in
three South-American canids, the maned wolf, crab-eating fox
and bush dogdimplications for sociality. Animal Behaviour, 31,
814e826.
Biben, M. 1986. Individual- and sex-related strategies of wrestling
play in captive squirrel monkeys. Ethology, 71, 229e241.
Biben, M. 1998. Squirrel monkey play-fighting: making the case for
a cognitive training function for play. In: Animal Play: Evolutionary,
Comparative, and Ecological Perspectives (Ed. by M. Bekoff & J. A.
Byers), pp. 161e182. Cambridge, U.K.: Cambridge University Press.
Borchelt, P. L. 1983. Aggressive behavior of dogs kept as compan-
ion animals: classification and influence of sex, reproductive status
and breed. Applied Animal Ethology, 10, 45e61.
Burghardt, G. M. 2005. The Genesis of Animal Play: Testing the
Limits. Cambridge, Massachusetts: MIT Press.
Byers, J. A. 1998. Biological effects of locomotor play: getting into
shape, or something more specific? In: Animal Play: Evolutionary,
WARD & SMUTS: PLAY AMONG DOMESTIC DOG SIBLINGSComparative, and Ecological Perspectives (Ed. by M. Bekoff & J. A.
Byers), pp. 205e220. Cambridge, U.K.: Cambridge University Press.
metries in social play among domestic dog, Canis lupus familiaris, littermates,
ARTICLE IN PRESS
ANIMAL BEHAVIOUR, --, -12Caro, T. M. 1981. Sex differences in the termination of social play in
cats. Animal Behaviour, 29, 271e279.
D?Eath, R. B. & Lawrence, A. B. 2004. Early life predictors of the
development of aggressive behaviour in the pig. Animal Behaviour,
67, 501e509.
Derix, R., van Hooff, J., de Vries, H. & Wensing, J. 1993. Male and
female mating competition in wolves: female suppression vs. male
intervention. Behaviour, 127, 141e174.
Fagen, R.1981.Animal Play Behavior. NewYork:OxfordUniversity Press.
Fagen, R. & Fagen, J. 2004. Juvenile survival and benefits of play
behaviour in brown bears, Ursus arctos. Evolutionary Ecology
Research, 6, 89e102.
Freedman, D. G., Elliot, O. & King, J. A. 1961. Critical period in
social development of dogs. Science, 133, 1016e1017.
Ghosh, B., Choudhuri, D. K. & Pal, B. 1984. Some aspects of the
sexual behaviour of stray dogs, Canis familiaris. Applied Animal
Behaviour Science, 13, 113e127.
Hemelrijk, C. K. 1990a. Models of, and tests for, reciprocity, unidir-
ectionality and other social interaction patterns at a group level.
Animal Behaviour, 39, 1013e1029.
Hemelrijk, C. K. 1990b. A matrix partial correlation test used in
investigations of reciprocity and other social interaction patterns
at group level. Journal of Theoretical Biology, 143, 405e420.
Hole, G. 1988. Temporal features of social play in the laboratory rat.
Ethology, 78, 1e20.
van Hooff, J. A. R. A. M. & Wensing, J. A. B. 1987. Dominance and
its behavioral measures in a captive wolf pack. In: Man and Wolf
(Ed. by H. Frank), pp. 219e252. Dordrecht: Dr W. Junk Publishers.
Laviola, G. & Alleva, E. 1995. Sibling effects on the behavior of
infant mouse litters (Mus domesticus). Journal of Comparative
Psychology, 109, 68e75.
Laviola, G. & Terranova, M. L. 1998. The developmental psychobi-
ology of behavioural plasticity in mice: the role of social experi-
ences in the family unit. Neuroscience and Biobehavioral Reviews,
23, 197e213.
Lindsay, S. R. 2000. Handbook of Applied Dog Behavior and Training:
Adaptation and Learning. Ames: Iowa State University Press.
Lindsay, S. R. 2005. Applied Dog Behavior and Training: Procedures
and Protocols. Ames, Iowa: Blackwell Scientific.
Lund, J. D. & Vestergaard, K. S. 1998. Development of social
behaviour in four litters of dogs (Canis familiaris). Acta Veterinaria
Scandinavica, 39, 183e193.
McLeod, P. J. & Fentress, J. C. 1997. Developmental changes in the
sequential behavior of interacting timber wolf pups. Behavioural
Processes, 39, 127e136.
Maestripieri, D. & Ross, S. R. 2004. Sex differences in play among
western lowland gorilla (Gorilla gorilla gorilla) infants: implications
for adult behavior and social structure. American Journal of Physical
Anthropology, 123, 52e61.
Martin, P. & Bateson, P. 1993. Measuring Behaviour. Cambridge,
U.K.: Cambridge University Press.
Mech, L. D. 1970. The Wolf: the Ecology and Behaviour of an Endan-
gered Species. New York: Doubleday.
Meyer, S. & Weber, J. 1996. Ontogeny of dominance in free-living
foxes. Ethology, 102, 1008e1019.
Namikas, J. & Wehmer, G. 1978. Gender composition of the
litter affects behaviour of male mice. Behavioral Biology, 23,
219e224.
Nunes, S., Muecke, E., Sanchez, Z., Hoffmeier, R. R. & Lancaster,
L. T. 2004. Play behavior and motor development in juvenile
Belding?s ground squirrels (Spermophilus beldingi). Behavioral
Ecology and Sociobiology, 56, 97e105.Owens, N. W. 1975. Social play behaviour in free-living baboons,
Papio anubis. Animal Behaviour, 23, 387e408.
Please cite this article in press as: Camille Ward et al., Partner preferences and asym
Anim. Behav. (2008), doi:10.1016/j.anbehav.2008.06.004Packard, J. M. 2003. Wolf behavior: reproductive, social, and intelli-
gent. In: Wolves: Behavior, Ecology, and Conservation (Ed. by L. D.
Mech& L. Boitani), pp. 35e65. Chicago: University of Chicago Press.
Pal, S. K. 2008. Maturation and development of social behaviour
during early ontogeny in free ranging dog puppies in West Bengal,
India. Applied Animal Behaviour Science, 111, 95e107.
Pal, S. K., Ghosh, B. & Roy, S. 1998. Agonistic behaviour of free-
ranging dogs (Canis familiaris) in relation to season, sex and age.
Applied Animal Behaviour Science, 59, 331e348.
Pal, S. K., Ghosh, B. & Roy, S. 1999. Inter- and intra-sexual
behaviour of free-ranging dogs (Canis familiaris). Applied Animal
Behaviour Science, 62, 267e278.
Palagi, E., Cordoni, G. & Borgognini, T. 2004. Immediate and
delayed benefits of play behaviour: new evidence from chimpan-
zees (Pan troglodytes). Ethology, 110, 949e962.
Panksepp, J. 1981. The ontogeny of play in rats. Developmental
Psychobiology, 14, 327e332.
Pellis, S. M. & Pellis, V. C. 1991. Role reversal changes during the
ontogeny of play fighting in male rats: attack vs. defense. Aggres-
sive Behavior, 17, 179e189.
Pellis, S. M. & Pellis, V. C. 1998. Structureefunction interface in the
analysis of play. In: Animal Play: Evolutionary, Comparative, and Eco-
logical Perspectives (Ed. by M. Bekoff & J. A. Byers), pp. 115e140.
Cambridge, U.K.: Cambridge University Press.
Pellis, S. M. & Pellis, V. C. 2007. Rough-and-tumble play and the
development of the social brain. Current Directions in Psychological
Science, 16, 95e98.
Pereira, M. E. & Preisser, M. C. 1998. Do strong primate players
?self-handicap? during competitive social play? Folia Primatologica,
69, 177e180.
Pozis-Francois, O., Zahavi, A. & Zahavi, A. 2004. Social play in
Arabian babblers. Behaviour, 141, 425e450.
Rothstein, A. & Griswold, J. G. 1991. Age and sex preferences for
social partners by juvenile bison bulls, Bison bison. Animal Behav-
iour, 41, 227e237.
Schenkel, R. 1967. Submission: its features and function in wolf and
dog. American Zoologist, 7, 319e329.
Scott, J. P. & Fuller, J. L. 1965. Genetics and the Social Behavior of the
Dog. Chicago: University of Chicago Press.
Scott, J. P. & Marston, M. 1950. Critical periods affecting the devel-
opment of normal and mal-adjustive social behavior of puppies.
Journal of Genetic Psychology, 77, 25e60.
Shannon, C. E. & Weaver, W. 1949. The Mathematical Theory of
Communication. Urbana: University of Illinois Press.
Sharpe, L. L. 2005a. Play fighting does not affect subsequent
fighting success in wild meerkats. Animal Behaviour, 69,
1023e1029.
Sharpe, L. L. 2005b. Play does not enhance social cohesion in
a cooperative mammal. Animal Behaviour, 70, 551e558.
Sharpe, L. L. 2005c. Frequency of social play does not affect dispersal
partnerships in wild meerkats. Animal Behaviour, 70, 559e569.
Sharpe, R. M., Morris, A. & Wyatt, A. C. 1973. The effect of the sex
of littermates on the subsequent behaviour and breeding perfor-
mance of cross-fostered rats. Laboratory Animals, 7, 51e59.
Sherman, C. K., Reisner, I. R., Taliaferro, L. A. & Houpt, K. A. 1996.
Characteristics, treatment, and outcome of 99 cases of aggression
between dogs. Applied Animal Behaviour Science, 47, 91e108.
Smith, L. K., Field, E. F., Forgia, M. L. & Pellis, S. M. 1996. Dom-
inance and age-related changes in the play-fighting of intact and
post-weaning castrated male rats (Rattus norvegicus). Aggressive
Behavior, 22, 215e226.
Smith, L. K., Fantella, S. L. N. & Pellis, S. M. 1999. Playful defensive
responses in adult male rats depend on the status of the unfamiliar
opponent. Aggressive Behavior, 25, 141e152.
metries in social play among domestic dog, Canis lupus familiaris, littermates,
Spinka, M., Newberry, R. C. & Bekoff, M. 2001. Mammalian play:
training for theunexpected.Quarterly Reviewof Biology,76, 141e168.
Thompson, K. V. 1996. Play-partner preferences and the function of
social play in infant sable antelope, Hippotragus niger. Animal
Behaviour, 52, 1143e1155.
Thompson, K. V. 1998. Self-assessment in juvenile play. In: Animal
Play: Evolutionary, Comparative, and Ecological Perspectives (Ed.
by M. Bekoff & J. A. Byers), pp. 183e204. Cambridge, U.K.: Cam-
bridge University Press.
deVries,H.1993. The rowwise correlationbetween twoproximityma-
trices and the partial rowwise correlation. Psychometrika, 58, 53e69.
Wahaj, S. A. & Holekamp, K. E. 2006. Functions of sibling aggres-
sion in the spotted hyaena, Crocuta crocuta. Animal Behaviour, 71,
1401e1409.
Watson, D. M. & Croft, D. B. 1996. Age-related differences in play-
fighting strategies of captive male red-necked wallabies (Macropus
rufogriseus banksianus). Ethology, 102, 336e346.
West, M. 1974. Social play in the domestic cat. American Zoologist,
14, 427e436.
Zahavi, A. 1977. Testing of a bond. Animal Behaviour, 25,
246e247.
ARTICLE IN PRESS
WARD & SMUTS: PLAY AMONG DOMESTIC DOG SIBLINGS 13Please cite this article in press as: Camille Ward et al., Partner preferences and asym
Anim. Behav. (2008), doi:10.1016/j.anbehav.2008.06.004metries in social play among domestic dog, Canis lupus familiaris, littermates,