Content uploaded by Márta Gácsi
Author content
All content in this area was uploaded by Márta Gácsi on Nov 11, 2018
Content may be subject to copyright.
Playing styles and possible causative factors in dogs’
behaviour when playing with humans
Lilla To
´th
a,
*,Ma
´rta Ga
´csi
a
,Jo
´zsef Topa
´l
b
,A
´da
´m Miklo
´si
a
a
Department of Ethology, Eo
¨tvo
¨s Lora
´nd University, 1117 Budapest, Pa
´zma
´ny Pe
´ter se
´ta
´ny 1./C., Hungary
b
Institute for Psychological Research, Hungarian Academy of Sciences, Hungary
Accepted 27 February 2008
Available online 8 April 2008
Abstract
Individual differences and causative factors could modify the behaviour of dogs in object related
games played with a human partner. In a two-by-two within-subject design we observed 68 family
dogs’ behaviour when playing two different types of games (ball game and tugging) with two different
play partners (owner or unfamiliar experimenter) in order to categorize each dog’s playing style. In all
four conditions we have measured the following behavioural variables: tendency for possession,
willingness to retrieve, behaviours related to fear/avoidance and aggression, and occurrence of play
bows. We also calculated the relative duration of time when the dogs oriented ‘‘outwards’’ from the play
situation to the other non-interacting person (owner or experimenter) during a session. Than we
examined the effect of six factors on dog–human play behaviour: the familiarity of the play partner, the
type of the game, the dogs’ gender, age and breed, and the duration of daily active interaction between
dog and owner.
We used factor analysis to unfold the relationship among the recorded variables and revealed three
factors that accounted for 63% of the total variance. On Factor 1 variables measured in tugging with both the
experimenter and owner had high loadings (labelled ‘‘Motivation for tugging’’). Factor 2 contained all
variables measuring fear and avoidance (Fear/Avoidance), and Factor 3 consisted of variables measured in
ball game (Motivation for ball game). The cluster analysis of the dogs’ individual factor scores classified
them into 5 significantly different groups on the basis of their tendency to be involved in playing with a ball,
a rag and to show fear/avoidance during the tests. Moreover, the gender (but not the age or breed) of the dogs
and the duration of the daily active interaction with the owner had significant effects on the distribution of
dogs between the cluster groups. These results suggest that in play situations the behaviour of well
socialized family dogs is influenced more by their motivation to play and to a certain extent by the level of
www.elsevier.com/locate/applanim
A
vailable online at www.sciencedirect.com
Applied Animal Behaviour Science 114 (2008) 473–484
* Corresponding author. Tel.: +36 30 553 9194; fax: +36 1 381 2180.
E-mail address: tlilla@gmail.com (L. To
´th).
0168-1591/$ – see front matter #2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.applanim.2008.02.011
fear than by the familiarity of the play partner or their possible general tendencies for cooperative or
competitive behaviours.
#2008 Elsevier B.V. All rights reserved.
Keywords: Dog; Social play; Ball game; Tug-of-war; Factor and cluster analysis
1. Introduction
Social play represents one of the most complex interactions between two individuals. This
complexity can be revealed by identifying at least two levels of behavioural organisation. At the
lower level, play is based on actions borrowed from various behavioural contexts (i.e. predatory,
agonistic, and sexual). At the higher level, however, interacting partners need to cooperate in
order to achieve their common goal of playing together (Mitchell and Thompson, 1991).
Dogs (Canis familiaris) offer a good model for the investigation of play behaviour, because
they display a high level of such activities. Observations from various canid species describe
social play where interacting juvenile animals use their own or the other’s body as a focus of the
game. Dogs living with humans often use objects as means of play when playing with other dogs,
not just humans (Hubrecht and Serpell, 1993; Rooney et al., 2000; Bauer and Smuts, 2007).
Given the variability of play behaviour in wild canids and the use of overlapping behavioural
repertoire for various aspects of behaviour, the origin of such object-related play between
humans and dogs is not clear.
Recently, evidence has been accumulating that individuals differ consistently in their
behavioural tendencies and their behaviour in one context is correlated with the behaviour in
multiple other contexts. Thus, personalities are not only found in humans but also in a wide
range of other animal species (Wolf et al., 2007), however, they are often referred to as coping
styles (Overli et al., 2007), or behaviour syndromes (Sih et al., 2004). A number of reviews
have already addressed the question of personality in dogs from several different perspectives,
such as behavioural genetics (Ruefenacht et al., 2002), comparative psychology (e.g. Jones
and Gosling, 2005) or practical application (Taylor and Mills, 2007; Diederich and Giffroy,
2006). In the Dog Mentality Assessment test battery (Svartberg and Forkman, 2002)five
factors were obtained (playfulness, curiosity/fearlessness, chase-proneness, sociability and
aggressiveness), and play behaviour turned out to be consistent in the two slightly different
‘‘play’’ test units.
Comparing dog–human play with dog–dog interactions Rooney et al. (2000) found that dogs
were more likely to give up in competition, to show and present the toy to the human partner, they
were more interactive, less likely to possess the toy in the games when they interact with humans,
in contrast to playing with nonspecifics. It is not clear, however, whether the revealed tendencies
are consistent irrespective of the human partner’s identity, or the familiarity of the person had a
significant effect on the behaviour of the dogs. We may assume that in case of well socialized
family dogs in this procedure the observed tendencies in play behaviour would not depend on the
familiarity of the playing partner (Mitchell and Thompson, 1991b).
An other important, but unexplored factor of the dog–human play is the type of the game
(object retrieval or tugging) or the focal object which could also have an effect on the mode of the
play behaviour of dogs. We suggest that depending on the dog’s individual bias towards
possessive or object-sharing behaviours, some of them might participate more likely in ‘‘tug-of-
war’’ while others would prefer to play with the ball.
L. To
´th et al. / Applied Animal Behaviour Science 114 (2008) 473–484474
The duration of the daily interactions with the owner might also influence the behaviour of
dogs, as well as past experiences and the effects of training. For example, increased retrieving and
tugging in dog–owner partnerships could possibly be a result of these owners’ spending more
time teaching their dogs to perform these behaviours.
The gender, breed and age of the dog can also be causative factors. Possible gender differences
can be due to biological differences in the playfulness or a sex bias in the amount of training or
play devoted to male vs. female dogs. It would be a rather complex task to fully unfold the effect
of breed, taking that there are so many, so we decided to compare the behaviour of a single breed
(Belgian shepherd) to a mixed breed group. Moreover, if play behaviour has some connection
with full maturity or dominance relationships one could predict young dogs to show different
behaviour patterns compared to older ones.
In sum the purpose of the present study was to investigate factors affecting the individual
differences in the behaviour of dogs playing with humans such as: (1) the familiarity of the
playing partner (owner and an unfamiliar experimenter), (2) the type of the game (ball and
tugging), (3) the daily active interaction between owner and dog, (4) gender, (5) age and (6)
breed.
We applied factor analysis for the evaluation of our data to reduce the number of dependent
variables and to arrive at a smaller number of independent derived factors that can explain the
most variability in the original variables. This method is appropriate to get theoretical dimensions
(superordinate variables) thought to account for individual differences in a set of behaviours
observed in the dog–human play. Factor analysis followed by cluster analysis, has previously
been used in the study of complex behaviours of dogs (Hart and Miller, 1985; Topa
´l et al., 1998;
Bradshaw and Goodwin, 1998). Cluster analysis is a proper method to classify the individuals
according to their behaviour patterns when playing with humans and to establish categories for
the individuals with most similar behaviour patterns.
2. Methods
2.1. Subjects
Sixty-eight adult dogs from several breeds (30 males and 38 females; mean age: 4.9 3.2 years, range
1–12) took part in our experiment. These subjects were part of a larger sample participating in a test-series
designed to evaluate the personality characteristics of pet dogs (Kubinyi et al., 2006). Our subjects were
chosen from a larger sample to make balanced groups. The dogs were from 13 different breeds: Belgian
Shepherd (Tervueren (n= 18) + Groenandael (n= 17) + Malinois (n= 1)), Rough Collie (n= 5), German
Shepherd (n= 4), Mudi (n= 4), Golden Retriever (n= 3), Sheltie (n= 2), West Highland White Terrier
(n= 1), Hovawart (n= 1), Labrador Retriever (n= 1), Doberman Pinscher (n= 1), Miniature Pinscher
(n= 1), Hungarian Vizsla (n= 1), Rhodesian Ridgeback (n= 1), and the sample contained also mongrels
(n= 7).
To study the possible effect of breed differences on the dogs’ behaviour we selected two subgroups from
the 68 subjects. Subgroup 1 was a homogeneous group; it included the 36 Belgian Shepherds (Tervueren,
Groenandael, Malinois). Subgroup 2 included 32 dogs of different breeds and also the mongrels. Subgroups
were balanced for gender and age.
The sample was also divided into two age categories; young (12–24 months, n= 26) and old (>24
months, n= 42), which were also balanced for gender and breed.
Subjects were recruited by phone from our ‘Family Dog Project’ database. These dogs can be classified
as ‘pet dogs’ because they live in the house or in the garden, their owners regularly walk them and/or they
often take part in some training class in a dog school (basic obedience, agility). The owners’ (59 women and
9 men) age ranged between 14 and 60 years.
L. To
´th et al. / Applied Animal Behaviour Science 114 (2008) 473–484 475
2.2. Experimental design
Observations were carried out from July 2003 to September 2004 at the Department of Ethology, Eo
¨tvo
¨s
Lora
´nd University in Budapest (64 dogs) and in another Hungarian town, Debrecen (4 dogs). The
experiments were conducted in a relatively empty room (in Budapest 6 m 3 m; in Debrecen
5m5 m) that was unfamiliar for the subjects. Before the test the owners were asked how much time
they spent daily with the dog in active interaction (play, training, walking, etc.). During all tests the
unfamiliar female experimenter, the owner, and the camera-woman were present.
2.3. Procedure
All dogs participated in 4, 1-min-long play sessions with 30 s breaks between them. In the first two
sessions the dogs played with the experimenter, in the last two they played with their owner. There were two
different types of game sessions: a ball game and a tugging with each partner. In case of all dogs the order of
the testing episodes was the same: (1) ‘‘ball game’’ with the experimenter, (2) ‘‘tugging’’ with the
experimenter, (3) ‘‘ball game’’ with the owner, (4) ‘‘tugging’’ with the owner. We applied this fixed test
order because these subjects were a sub-sample of a larger project on personality testing, so individual
evaluation and comparability were of great importance. To ensure, however, that group-level analysis can
also be applied, we ran a control test to look for possible order effects. Ten dogs were tested with the same
protocol in a balanced order for playing partner and game type. Control subjects were chosen to make
balanced groups.
We asked the playing partners (experimenter and owners) to play the ‘‘usual’’ way with the rag and the
ball. Both of them were instructed to play as intensively as possible, but at the same time, to adjust their
behaviour to the dog’s reactions. The only restrictions were the following: (1) the human play partners were
instructed to throw both the ball and the rag minimum once, encouraging the dog to fetch it, and (2) to try to
take the object (both the ball and the rag) from the dog’s mouth. The human partner always tried to take the
object by asking for it kindly at first, but if the dog did not lose hold of it he/she gave increasingly firm orders
and tried to take the object by grabbing it and finally by gently opening the mouth of the dog. Compliance
with the protocol was controlled by the experimenter.
During the ‘‘ball game’’ tests the human partners could use one or more of the following play items: 2
compact rubber balls (5–7 cm in a diameter) and 2 tennis balls. In the ‘‘tugging’’ sessions the human partners
could use a rag (one of two towelling ropes, 20 and 40 cm long with knots on both ends) for inducing play
behaviour. During the play sessions with one partner the other human stood still turning towards the dog and
did not interfere.
2.4. Variables and data analysis
The tendency for possession, the willingness to retrieve, behaviours related to fear/avoidance and
aggression, and the occurrences of play bows were measured. We also calculated the relative duration of
time when the dog oriented ‘‘outwards’’ from the play situation to the other non-interacting person (owner or
experimenter) during a session. The occurrence of the characteristic playing signal, the bow was so sparse
during both games and with both partners (with the experimenter: 3 times both in ball game and in tugging,
with the owner: 5 times in ball game and 2 times in tugging), that no meaningful statistical analysis could be
done for this variable. We experienced a similar result with the behaviours connected to aggression.
Considering the playful characteristics, the rather rare occurrence and the uneven distribution of these
behaviours, we did not use the scores of aggression for further analysis. For the detailed description of the
analysed behavioural variables see Table 1.
Based on the answers for the question ‘‘How much time do you spend daily with the dog in active
interaction (play, training, walk)?’’ we classified two groups: less than 1 h interaction per day (n= 30), or
more than 1 h interaction per day (n= 36). We excluded two dogs from this classification since not their
owners but other member of the family was in active interaction with them.
L. To
´th et al. / Applied Animal Behaviour Science 114 (2008) 473–484476
All tests were video recorded and analysed later by one of the experimenters (L.T.) by watching the
video. Interobserver agreement between her and a naı
¨ve observer was assessed by comparing their parallel
coding of the behaviour categories on the same video records and the evaluation of the 20% (15 dogs) of the
whole sample. The following Cohen Kappa results were obtained: ‘‘Willingness to retrieve’’: 0.85;
‘‘Possessivity’’: 0.85; ‘‘Fear’’: 0.89; ‘‘Orientation outwards’’: 0.94; ‘‘Bow’’: 0.82. The high Kappa values
show that there was a good agreement between the independent observers.
2.5. Statistical analysis
We analysed the tests for controlling possible order effects of the playing partners (experimenter or
owner) and the type of the toy (ball or rag) using Wilcoxon signed-ranks test.
L. To
´th et al. / Applied Animal Behaviour Science 114 (2008) 473–484 477
Table 1
Short description of behavioural variables coded in both types of play sessions (ball game and tugging) and with both
partners (experimenter and owner)
Behavioural variables and abbreviations Values of the variable
Willingness to retrieve RET
score (0–2)
0: The dog never brings the object to the human partner or there is
not any physical contact between the dog and the object (e.g. the
dog does not hold or chew the toy at least once in its mouth)
1: The dog does not always bring the object back or although it
always starts to move towards the human with the object in
its mouth, the human cannot get the object without approaching
the dog
2: The dog always brings the object back
Possessivity POS score (0–3) 0: The dog does not hold or chew the object in its mouth
1: The human can take the object from the dog’s mouth without
any sign of aggression or struggle
2: There are visible signs of wrangle when the human tries to
take the object or the dog shows avoidance with the object in
its mouth, but finally the human can take it from the dog
3: The human is unable to take the object from the dog
during the session
Fear FEA score (0–2) Signs of fear/avoidance: the dog’s tail is retracted between its
hind legs, avoids or recoils, seeks support (e.g. hides behind
human’s leg), crouches, trembles, and attempts to flee
0: No signs of fear can be observed during the play session
1: The dog shows at least one of the behaviours described above
once or just for a short time during the play session
2: The dog shows at least one of the behaviours described above
for a longer period (more than 2 s) during the play session
Orientation outwards (%)
ORIO
The duration of the orientation ‘‘outwards’’ is orientation to the other
human partner with which the dog was not playing at the time
(e.g. orientation to the owner during the play with the experimenter
and reverse). This was measured in seconds and as the duration
of the test sessions varied slightly we calculated the relative percentage
of the time spent engaged in this behaviour. (The value was divided by
the duration of the whole episode and multiplied by 100)
In the result section we indicated the human partner (i.e. experimenter or owner) and the type of toy (i.e. ball or rag) with
relevant letters before the abbreviation of the variables. (For example, experimenter rag POS: possession was coded while
the dog played with experimenter with the rag).
For further analyses multivariate methods (factor analysis followed by cluster analysis) were also used to
unfold the correlational pattern of the behavioural variables and to determine the behaviours that have major
effects on the individual differences. Factor analysis (complemented with Varimax rotation) was performed
on 16 variables: four behaviour categories (RET, POS, FEA, ORIO) were recorded in all (4) conditions.
Items were required to have a minimum factor loading 0.5. We also calculated the individual factor scores
to reveal the position the individuals on the different factors. The factor scores were standardized using z-
transformation to make the scores comparable on different factors. (After the standardisation all scores have
a value between 0 and 1.) These standardized individual factor scores had normal distributions, so
parametrical tests could be used for the analyses. After the factor analysis, we applied a hierarchical
cluster analysis for the classification of the individuals depending on their standardized individual factor
scores. One-way analysis of variance (with Bonferroni post hoc test) was used to test the distribution of
standardized individual factor scores among the clusters.
We analysed the distribution of the gender, breed, age and the two groups of interaction categories among
the clusters with x
2
-test. We used x
2
-test to compare the distribution of gender, age and breed between the
groups characterized by different levels of interaction with the owner (less vs. more than 1 h daily
interaction).
For all statistical tests the SPSS 10 statistical package was used.
3. Results
First we analysed the control experiment looking for possible order effects, comparing their
results in the same episodes of the different orders, but neither the order of the type of the game
(ball game or tugging), nor the order of the playing partners (experimenter or owner) caused
significant difference in any of the behaviour variables of the dogs (all p>0.05).
L. To
´th et al. / Applied Animal Behaviour Science 114 (2008) 473–484478
Table 2
Factor loadings (correlation between the factors and the behavioural variables) following Varimax rotation
Behavioural variables Factor 1 Factor 2 Factor 3
Motivation for
tugging (42.0%)
Fear/Avoidance
(11.9%)
Motivation for
ball game (9.4%)
Exp. rag POS 0.83 0.26 0.14
Exp. rag RET 0.71 0.11 0.29
Exp. rag ORIO 0.72 0.26 0.01
Own. rag POS 0.78 0.33 0.23
Own. rag RET 0.78 0.24 0.13
Own. rag ORIO 0.53 0.35 0.06
Exp. ball FEA 0.12 0.70 0.26
Exp. rag FEA 0.25 0.84 0.22
Own. ball FEA 0.22 0.77 0.22
Own. rag FEA 0.22 0.82 0.13
Exp. ball POS 0.48 0.28 0.66
Exp. ball RET 0.35 0.23 0.71
Exp. ball ORIO 0.01 0.12 0.20
Own. ball POS 0.29 0.29 0.77
Own. ball RET 0.26 0.34 0.78
Own. ball ORIO 0.29 0.20 0.60
The factor pattern yielded three factors (Eigenvalues >1.5) that account for 63% of the total variance. Afore all the
abbreviated behavioural variables (capital) we used two letters: the first one referring to the person who played with the
dog (exp: experimenter, own: owner), the second one referring to the game-type (ball or rag). POS; Possessivity, RET;
Willingness to retrieve, ORIO; Orientation outwards, FEA; Fear/Avoidance. Factor loadings (behavioural variables)
belong to the certain factor signify bold values.
3.1. Factor analysis
The factor analysis of the coded variables revealed three factors that accounted for 63% of the
total variance (Table 2). Behavioural variables measured in ‘‘tugging’’ with both the
experimenter and the owner had high loadings on the first factor. Individuals with high values
on this factor showed marked interest in playing with the rag: they retrieved it but at the same
time also tried to possess it during the tests, and did not orient outwards (at the passive human) for
long. This factor was named as ‘‘Motivation for tugging’’.
The second factor, characterized by variables indicating fear and avoidance in all four test
situations was labelled as ‘‘Fear/Avoidance’’.
Similarly to the first factor, behavioural variables recorded in the ‘‘ball game’’ situation
with both the experimenter and owner had high loadings on the third factor (and only here), so
it was labelled as ‘‘Motivation for ball game’’. On this factor there were all but one (exp. ball
ORIO) behavioural variables observed in episodes where the toy object was a ball. In both
cases ‘‘Willingness to retrieve’’ and ‘‘Possessivity’’ showed high positive loadings, however,
‘‘Orientation outwards’’ was always represented with negative loadings. It is interesting to
note that irrespectively of the type of game the structures of both motivation factors are very
similar.
3.2. Hierarchical cluster analysis
The standardized individual factorial variables (‘‘Motivation for tugging’’, ‘‘Fear/Avoidance’’
and ‘‘Motivation for ball game’’) of the tested dogs were used to group them by a hierarchical
cluster analysis. Visual examination of the dendrogram revealed that the dogs could be divided
into five separate groups at the rescaled distance of 15. Since the 5th cluster consisted of only two
individuals we excluded it from the further analyses. (As Fig. 1 shows, these dogs represented a
rather rare combination of high motivation for tugging without any tendency for engaging in
retrieving the ball.) The analysis of variance showed significant differences among the four
L. To
´th et al. / Applied Animal Behaviour Science 114 (2008) 473–484 479
Fig. 1. The 5 clusters represent different patterns of the three factors. We used ‘‘labels’’ for the clusters on the figure, see
details in the text. Comparing the mean of the individual factor scores among the clusters significantdifferences are signed
by different letters in case of each factor. We used different letter types for reason of simpler readability: capitals show the
differences among clusters in case of the factor ‘‘Motivation for tugging’’, lowercase in case of the factor ‘‘Fear/
Avoidance’’, and italics (lowercase) in case of the factor ‘‘Motivation for ball game’’.
clusters in the case of all three factors; ‘‘Motivation for tugging’’: F(3,65) = 57.4, p<0.01;
‘‘Fear/Avoidance’’: F(3,65) = 59.0, p<0.01; ‘‘Motivation for ball game’’: F(3,65) = 28.9,
p<0.01 (Fig. 1).
The Bonferroni post hoc comparisons showed that each group represented a distinctive pattern
with regard to play behaviour. There are significant differences between clusters 1 vs. 2, 3, and 4
in factor ‘‘Motivation for tugging’’ ( p<0.01 in all three cases) while clusters 2, 3 and 4 do not
differ from each other (all p>0.05). In the factor ‘‘Fear/Avoidance’’ cluster 1 is different from 2
and 4 ( p<0.01, p<0.01 and p= 0.49 in case of cluster 3), while cluster 4 differs also from 2
and 3 clusters ( p<0.01 in both cases), while clusters 2 and 3 do not differ from each other
(p= 0.92). In the case of factor ‘‘Motivation for ball game’’ all clusters are significantly different
from each other ( p<0.05) with the exception that clusters 3 and 4 do not differ significantly
(p= 0.09).
Dogs in cluster 1 (n= 36) were characterized by relatively high individual factor scores on
both ‘‘Motivation for tugging’’ and ‘‘Motivation for ball game’’, and a moderate level on
‘‘Fear/Avoidance’’, so cluster 1 can be labelled as a ‘‘reactive’’ group. Dogs in cluster 2
engaged preferably only in the ball game, thus this cluster can be described as a ‘‘ball game
preference’’ group, consisting of individuals with high scores on the factor ‘‘Motivation for
ball game’’ and low scores on ‘‘Motivation for tugging’’ and ‘‘Fear/Avoidance’’. Dogs in
clusters3and4didnotshowexplicitmotivation for interactive games; they were
characterized by relatively low scores of both ‘‘Motivation for tugging’’ and ‘‘Motivation for
ball game’’. Individuals in cluster 3 (n= 12) seemed to be rather passive showing little fear
and also not much affinity to play any games, so cluster 3 could be labelled as a ‘‘passive
avoidant’’ group. A relative minority of dogs showed definite signs of fear/avoidance while
engaging in both games just on medium level (cluster 4: n= 10). Thus, cluster 4 was labelled
as a ‘‘fearful’’ group.
The distribution of the males and females was significantly different among the cluster-groups
(x
2
(3) = 10.4; p<0.05), with the most males in the ‘‘reactive’’ group (Fig. 2). There were also
differences depending on the duration of daily interaction with the owner (x
2
(3) = 13.7;
p<0.05). Dogs with more active interaction with their owners were mainly sorted into the
‘‘reactive’’ and ‘‘ball game preference’’ groups (Fig. 3). The distribution of Belgian Shepherd
and non-Belgian groups (x
2
(3) = 1.2; p= 0.8) and the age groups (x
2
(3) = 1.1; p= 0.8) did not
differ across the cluster groups.
L. To
´th et al. / Applied Animal Behaviour Science 114 (2008) 473–484480
Fig. 2. Percentage distribution of males and females within clusters. We used ‘‘labels’’ for the clusters on the figure, see
details in the text.
Analyzing the owners’ answers we found that the Belgian Shepherds had more daily
interaction with their owners (x
2
(1) = 11.0; p<0.05) than non-Belgians. There were no
significant differences in the daily interaction between males and females (x
2
(1) = 2.5; p= 0.12)
or the age categories (x
2
(1) = 2.8; p= 0.09).
4. Discussion
In this study, we have observed the play behaviour of well-socialized pet dogs with both their
owner and an unfamiliar play-partner. We used the same behavioural criteria for describing both
types of games (ball game and tugging) and we examined the effect of these factors on dog–
human play behaviour: the familiarity of the play partner (owner or unfamiliar experimenter), the
type of the game, the gender, age and breed of the dog and the duration of daily active interaction
between dog and owner.
Our data shows that family dogs are better distinguished in terms of their reaction to different
toys than to different humans, which could be explained by the familiarity of the situations in case
of well-socialized dogs. As mainly the human partners initialized the games by verbal
communication and body language, even dogs less experienced in tug-of-war games could be
ensured of the playful context of the situation. This might also account for the relatively rare
occurrence of the standard species-specific play signal, the bow and also for the lacking or the
playful nature of aggressive behaviours.
In our sample about half of the dogs engaged in both types of games whilst there were
‘‘specialists’’ who showed a preference for one type of game. Importantly, these results
contradict the widely held view that there is a general tendency in a dog either to transfer
objects to humans after taking possession of them or to engage in tug-of-war. In this case the
appropriate variables (tendency to retrieve or to be possessive) should have defined distinct
factors representing this aspect of their behaviour. Even if the factors define the two game
types, high loadings on ‘‘Willingness to retrieve’’ variables and low on ‘‘Possessivity’’
variableswouldbeexpectedinballgameandthe opposite tendencies in tugging. On the
contrary, however, we have found that separate factor variables have emerged for the
different types of games with high and positive loadings for both ‘‘Willingness to retrieve’’
and ‘‘Possessivity’’.
L. To
´th et al. / Applied Animal Behaviour Science 114 (2008) 473–484 481
Fig. 3. Percentage distribution of dogs within clusters depending on the duration of active interaction with their owners.
We used ‘‘labels’’ for the clusters on the figure, see details in the text.
Our observations also point to the importance of individual differences in play behaviour that
could have been influenced by learning and experience. Dogs that received more playful
interaction with their owner were less likely to show fear during play in an unfamiliar place,
moreover, these dogs show either stronger motivation to play tugging or decreased inhibition.
Previous experiences probably increase the predictability of the play partner’s behaviour for the
dog, fostering the use of mutually accepted behaviour rules to avoid conflicts (Mitchell and
Thompson, 1991). Dogs without such experiences are more prone to misunderstanding humans’
enticement and for lack of clear play signals they respond showing avoidance or fear. It is likely
that over time dog and owner develop a routine of games, but dogs do not generalise these
behaviour routines to other, functionally different situations. The habitual play of games is also
supported by the observation that the dogs in this sample did not discriminate between the owners
and the unfamiliar play partners. However, it should be emphasized that owners often teach their
dogs to retrieve, so the playing of ball game may simply be a trained response, and not the
manifestation of any underlying behavioural tendencies. Additionally, due to more daily
interaction with the owner dogs could be not only more familiar with the games but also more
active and stimulated.
Males and females did not distribute evenly in the clusters: more males were in the ‘‘reactive’’
and less in the ‘‘preference for ball’’, ‘‘passive avoidant’’ and ‘‘fearful’’ group. It could be
assumed that males receive more training, this way are more experienced in playing games, but
we did not find such a relationship comparing the duration of the daily active interaction between
the genders. Thus, we suggest that gender differences refer to temperament differences in males
and females, as was revealed by Svartberg (2001) in case of the ‘‘Boldness’’ factor.
The distribution of young and older dogs in the clusters was not different, which may follow
from the interacting effect of two factors; younger dogs may be more reactive and/or show more
fear or avoidance in unfamiliar situations. The even distribution of breed groups can have a
similar reason, as the Belgian shepherd is known as a highly reactive and sometimes shy breed.
Less time spent in active interaction with the owner in the Belgian shepherd group might be
explained by the fact that there are more individuals used for show/breeding purposes in this
group.
Games can be characterized as being either cooperative or competitive (object-sharing or
object possessing) or a mixture of these which also often can be observed in dogs’ dyadic play
interactions (Bauer and Smuts, 2007). In tug-of-war game the competitive (possessing)
tendencies might be more dominant than in ball game, while in each type of play the cooperation
between dogs and humans plays a significant role. As recent work on dog personalities described
cooperability or competitive tendencies as background behavioural factors (Svartberg and
Forkman, 2002), one could ask about the relationship between these traits and play behaviour.
One way to interpret our results is that each type of game has its own (probably biologically also
determined) set of rules, and to play the game efficiently the rules should be adhered to
independently from the fact whether they are competitive or cooperative. This suggests that
playing such games actually suppresses some types of individual differences in the behaviour of
dogs, and therefore the behaviour in such games might not be a good indicator for personality
measures like cooperativeness or sociability.
Earlier it has been assumed that competitive games increase agonistic tendencies in the
behaviour, suggesting an effect of play activity on later sociability with partners (McBride,
1995), although Rooney and Bradshaw (2002, 2003) found no evidence that competitive games
increased competitiveness. On the basis of our results one could also assume just the opposite
case; ‘‘cooperability’’ and ‘‘competitiveness’’ of an individual might determine the type of game
L. To
´th et al. / Applied Animal Behaviour Science 114 (2008) 473–484482
it would prefer to play. It is more likely that in some dogs the tendency to be dominant has a
genetic origin that gains expression through interaction with the social environment. It has been
questioned even on an ethological basis whether dogs generalize from play experiences to the
dominant/submissive nature of social relationships (Rooney et al., 2001).
One can even assume that no such direct relationship exists between competitive behaviour in
tug-of-war game and striving after dominance because play signals, in our case the human
partners’ play signals, clearly distinguish games from real competitive situations.
5. Conclusion
Present study pointed to individual differences in the dog behaviour during ball game and
tugging. The effect of several factors, which could modify the dog–human play, was revealed.
The behaviour of family dogs was influenced more by the type of the game (and to a certain
extent by the level of fear) than by their familiarity to the play partner or their willingness to
retrieve and tendency for being possessive. The gender and the duration of the daily active
interaction with the owner also had a significant effect on the play behaviour. We assume that
over time dog and owner develop a routine of games, and dogs do not generalise these behaviour
routines to other, functionally different situations.
Acknowledgements
This study was supported by a grant of the Hungarian Foundation for Basic Research (OTKA
T049615), by EU (FP6 NEST 012787) and by the Bolyai Foundation of the Hungarian Academy
of Sciences. We are very grateful to the two referees for giving valuable comments on a previous
version of this manuscript. We would like to express our gratitude to Linda Tama
´si for being kind
enough to correct the language of our manuscript. We thank E
´va Pe
´ch for her assistance in the
tests and are also grateful to Antal Do
´ka, Borba
´la Gyo
˝ri and Csaba Molna
´r for their valuable help.
We would like to thank all the owners and dogs who devoted some of their leisure time to this
study.
References
Bauer, E.B., Smuts, B.B., 2007. Cooperation and competition during dyadic play in domestic dogs. Canis familiaris.
Anim. Behav. 73 (3), 489–499.
Bradshaw, J.W.S., Goodwin, D., 1998. Determination of behavioural traits of pure-bred dogs using factor analysis and
cluster analysis; a comparison of studies in the USA and UK. Res. Vet. Sci. 66, 73–76.
Diederich, C., Giffroy, J.M., 2006. Behavioural testing in dogs: a review of methodology in search for standardisation.
Appl. Anim. Behav. Sci. 97, 51–72.
Hart, B.L., Miller, M.F., 1985. Behavioural profiles of dog breeds. J. Am. Vet. Med. A 186, 1175–1180.
Hubrecht, R.C., Serpell, J.A., 1993. Influence of housing conditions on the behaviour and welfare of dogs. Appl. Anim.
Behav. Sci. 35, 293.
Jones, A.C., Gosling, S.D., 2005. Temperament and personality in dogs (Canis familiaris): a review and evaluation of past
research. Appl. Anim. Behav. Sci. 95, 1–53.
Kubinyi, E., To
´th, L., He
´jjas, K., Sasva
´ri-Sze
´kely, M., Topa
´l, J., Ga
´csi, M., Miklo
´si, A
´., 2006. Relation between dopamin
D4 receptor gene and the attentional focus during play behaviour in dogs. In: Third European Conference on
Behavioural Biology—Physiological Mechanisms in Behavioural Ecology. Belfast, September 4–6th (abstract).
McBride, A., 1995. The human–dog relationship. In: Robinson, I. (Ed.), The Waltham Book of Human–Animal
Interaction: Benefits and Responsibilities of Pet Ownership. Pergamon, Oxford, pp. 99–112.
L. To
´th et al. / Applied Animal Behaviour Science 114 (2008) 473–484 483
Mitchell, R.W., Thompson, N.S., 1991. Projects, routines, and enticements indog–human play. In: Bateson, P.P.G.,Klopfer,
P.H. (Eds.), Perspectives in Ethology. Plenum Press, New York, London, (Chapter 6), pp. 189–216.
Mitchell, R.W., Thompson, N.S., 1991b. The effects of familiarity on dog–human play. Anthrozoo
¨s 4, 24–43.
Overli, O., Sorensen, C., Pulman, K.G.T., Pottinger, T.G., Korzan, W., Summers, C.H., Nilsson, G.E., 2007. Evolutionary
background for stress-coping styles: relationships between physiological, behavioral, and cognitive traits in non-
mammalian vertebrates. Neurosci. Biobehav. Rev. 31 (3), 396–412.
Rooney, N., Bradshaw, J.W.S., 2002. An experimental study of the effects of play upon the dog–human relationship. Appl.
Anim. Behav. Sci. 75, 161–176.
Rooney, N.J., Bradshaw, J.W.S., 2003. Links between play and dominance and attachment dimensions of dog–human
relationships. J. Appl. Anim. Welf. Sci. 6, 67–94.
Rooney, N.J., Bradshaw, J.W.S., Robinson, I.H., 2000. A comparison of dog–dog and dog–human play behaviour. Appl.
Anim. Behav. Sci. 66, 235–248.
Rooney, N.J., Bradshaw, J.W.S., Robinson, I.H., 2001. Do dogs respond to play signals given by humans? Anim. Behav.
61, 715–722.
Ruefenacht, S., Gebhardt-Henrich, S., Miyake, T., Gaillard, C., 2002. A behaviour test on German Shepherd dogs:
heritability of seven different traits. Appl. Anim. Behav. Sci. 79 (2), 113–132.
Sih, A., Bell, A., Johnson, J.C., 2004. Behavioral syndromes: an ecological and evolutionary overview. Trends Ecol. Evol.
19 (7), 372–378.
Svartberg, K., 2001. Shyness-boldness predicts performance in working dogs. Appl. Anim. Behav. Sci. 79, 157–174.
Svartberg, K., Forkman, B., 2002. Personality traits in the domestic dog (Canis familiaris). Appl. Anim. Behav. Sci. 79,
133–155.
Taylor, K.D., Mills, D.S., 2007. The development and assessment of temperament tests for adult companion dogs. J. Vet.
Behav. Clin. Appl. Res. 1 (3), 94–108.
Topa
´l, J., Miklo
´si, A
´., Csa
´nyi, V., Do
´ka, A., 1998. Attachment behavior in dogs (Canis familiaris): a new application of
Ainsworth’s (1969) strange situation test. J. Comp. Psychol. 112 (3), 219–229.
Wolf, M., van Doorn, S.G., Leimar, O., Weissing, F.J., 2007. Life-history trade-offs favour the evolution of animal
personalities. Nature 447 (7144), 581–584.
L. To
´th et al. / Applied Animal Behaviour Science 114 (2008) 473–484484
