Content uploaded by Nicola J Rooney
Author content
All content in this area was uploaded by Nicola J Rooney on Mar 26, 2014
Content may be subject to copyright.
Links Between Play and Dominance
and Attachment Dimensions
of Dog–Human Relationships
Nicola J. Rooney and John W. S. Bradshaw
Anthrozoology Institute
University of Southampton
It is often claimed that certain behavioral problems in domestic dogs can be triggered
by the games played by dog and caregiver (owner). In this study, we examine possible
links between the types of games played and dimensions of the dog–owner relation-
ship that are generally considered to affect such problems. Fifty dog–owner partner-
ships were filmed during 3-min play sessions in which the owner was allowed to
choose the games played. All partnerships then undertook a 1-hr test designed to mea-
sure elements of behavior commonly ascribed to “dominance” and “attachment.”
Principal components analysis of the data produced 2 dominance-related factors
(Amenability and Confident Interactivity) and 4 factors describing aspects of attach-
ment (Nonspecific Attention Seeking, Preference for Owner, Preference for Unfamil
-
iar Person, and Separation-Related Behavior). Amenability, in particular, varied sig
-
nificantly between breeds. In the study, we then compared types of games played to
each of these factors. Dogs playing rough-and-tumble scored higher for Amenability
and lower on Separation-Related Behavior than did dogs playing other types of
games. Dogs playing tug-of-war and fetch scored high on Confident Interactivity.
Winning or losing these games had no consistent effect on their test scores. If the dog
started the majority of the games, the dog was significantly less amenable and more
likely to exhibit aggression. The results suggest that how dogs play reflects general at
-
tributes of their temperament and relationship with their owner. This study provides
no evidence that games play a major deterministic role on dominance dimensions of
dog–human relationships, but the results suggest that playing games involving con
-
siderable body contact may affect attachment dimensions.
JOURNAL OF APPLIED ANIMAL WELFARE SCIENCE, 6(2), 67–94
Copyright © 2003, Lawrence Erlbaum Associates, Inc.
Requests for reprints should be sent to John W. S. Bradshaw, Department of Clinical Veterinary Sci
-
ence, University of Bristol, Langford, BS40 5DU, England. E-mail: j.w.s.bradshaw@bristol.ac.uk
The potential benefits of dog caretaking depend on the formation of a successful
relationship between dog and caregiver (owner). Frequently, people acquire a
dog whose behavior does not meet their expectations or fit in with their lifestyle.
Although some owners tolerate suboptimal relationships, many do not, and their
dogs are surrendered or even euthanized. In the United Kingdom, behavioral
problems are among the most common reasons for owners abandoning their
dogs (Podberscek, 1997). In the United States, an estimated 7.5 to 9 million
dogs per year are euthanized because of unacceptable behavior (Anderson &
Foster, 1988). To prevent such welfare problems and reduce the number of part
-
nerships ending tragically, we need to build a thorough understanding of the dy
-
namics of the dog–human relationship.
In the popular dog literature, there are numerous claims of factors that can af
-
fect dog–human relationships (Appleby, 1997; O’Farrell, 1992; Rogerson, 1992).
However, there are few data to substantiate these claims. Recent studies have
started to investigate influences on dog–human relationships but have relied
mainly on owners’ reports (Goodloe & Borchelt, 1998). Useful for obtaining large
samples, this methodology is less objective than assessment by a third party. Al-
though case studies of problematic dog–human relationships are numerous (An-
derson & Foster, 1988; Overall, 1997), there have been very few attempts to
measure apparently healthy relationships objectively.
The terms attachment and dominance often are used to describe elements of
dog–human relationships, but there are no universal definitions of these terms
(Drews, 1993; Cairns, 1972; Overall, 1997), particularly as applied to interspecific
relationships. For this article, we define dominance as the tendency of the dog to
assert priority of access to resources and attachment as the tendency of the dog to
seek and maintain contact with the owner. In our experimental studies, we have
found that both dominance (Rooney & Bradshaw, 2002) and attachment (Rooney,
1999) in dog–human relationships are multidimensional. In this article, we de
-
scribe a test procedure, modified from these two prior studies, which we use to
measure dog–owner relationships within the home environment. In devising these
tests, we have made as few a priori assumptions about what constitutes dominance
and attachment as possible but have taken a post hoc approach, measuring the
dogs’ behavioral responses in a variety of situations and then applying exploratory
statistical analysis to discover the underlying dimensions. Having obtained mea
-
sures of these dimensions, we then investigated their connections with the types of
games played by dog and owner.
It is commonly asserted that play between dog and owner has significant effects
on dominance dimensions of their relationship. Dog trainers and companion ani
-
mal behavior counselors warn that allowing a dog to win uncontrolled games such
as tug-of-war can increase the likelihood that the dog will attempt to become domi
-
nant over the owner (O’Farrell, 1992; Rogerson, 1992); on this basis, controlled
tug-of-war is used as therapy to correct behavioral problems (Appleby, 1997). We
68
ROONEY AND BRADSHAW
refer to this as the dominance enhancement theory. However, questionnaires
(Goodloe & Borchelt, 1998) and experimental studies of Labrador (Rooney 1999)
and Golden Retrievers (Rooney & Bradshaw, 2002) have found no evidence for
these postulated effects.
In this article, we further investigated the effects of dog–owner play. We used
videotaped play sessions and questionnaire surveys to measure the play that occurs
between 50 dogs and their owners. We then explored whether features of that play
bear any significant relationship to the measured dimensions of the dog–owner re
-
lationship. In particular, we test whether
1. Dogs who play specific game types (tug-of-war) differ significantly in their
relationship with their owners from dogs who do not play those games.
2. Dogs who play uncontrolled games (which they are allowed to win) with
their owners differ in dominance aspects of their relationship compared to
dogs who play controlled games.
3. The amount of play occurring between dog and owner correlates to dimen-
sions of their relationship.
In addition to play, numerous other factors are commonly believed to affect
dog–human relationships. These include characteristics of the dog and features
of the owner’s behavior toward the dog. We used a verbal questionnaire to in-
vestigate three of the most common assertions:
1. Dogs who have continual access to toys often become dominant (Rogerson,
1992).
2. Allowing dogs to sleep in the owner’s bedroom can lead to overattachment
(Appleby, 1997).
3. Dogs who are frequently allowed to initiate social interactions (including
play) behave with increased dominance toward their owner (O’Farrell,
1992).
METHODS
Subjects
Fifty dog owners (35 women and 15 men) were recruited through personal con
-
tacts and posters in veterinary surgeries in Southampton (Hampshire, Southern
England) and the Wirral (Cheshire, Northwest England). All had owned a single
dog for at least 1 year and were willing to be filmed. The dogs were 29 males
and 21 females; 52% of the males and 76% of the females were neutered. They
ranged in age from 20 months to 14 years, with a median of 7 years (25th per
-
centile = 4.5; 75th percentile = 9). There were 19 crossbred dogs and 31 pure
-
ATTACHMENT DIMENSIONS OF DOG–HUMAN RELATIONSHIPS 69
bred/pedigree dogs representing 17 different breeds. When classified according
to Kennel Club categories (Sylvester, 1984)—crossbreeds were categorized ac
-
cording to the breed that they most closely resembled as agreed by experimenter
and owner—there were 17 gundogs, 16 working dogs, 10 terriers, 3 hounds, 2
toy dogs, and 2 utility dogs.
Procedure
The Visit
The experimenter (Nicola Rooney) visited each dog owner in his or her own
home once. The visit lasted approximately 1 to 1.5 hr. When possible, it took place
at the dog’s usual feeding time and when the owner and dog were alone. When
other family members were present (12 cases), they were requested to remain in a
different part of the house and to avoid interrupting the procedure or attracting the
dog’s attention. The visit followed the protocol outlined in Figure 1.
70
ROONEY AND BRADSHAW
Acclimatization
10 mins
Feeding *
2mins
Play session *
3mins
Questionnaire
15 mins
Testing
30 mins
FIGURE 1 Outline of the pro
-
tocol for the visit to the volunteer
dog owner’s home including ap
-
proximate durations of each
stage. *Filmed via camera held
by experimenter; all other parts
filmed by camera on tripod.
Acclimatization
None of the dogs had met the experimenter previously, so there was a 10-min
familiarization period. This allowed the owner to relax and the dog to acclima
-
tize to the experimenter and the equipment. During this time, the experimenters
chatted to the owner, petted their dogs if they approached, and set up the equip
-
ment. They also outlined the protocol to the owners. The most commonly used
room was selected as the filming location (usually the living room but in two
cases the kitchen). A video camera (Sony Video 8 CCD–TR370E; Sony Corpo
-
ration, Japan) with a wide angle lens was mounted on a tripod and positioned to
allow the maximum floor area to be viewed while minimizing the risk of dam
-
age by the dog. This setup was used to film the questionnaire and testing sec
-
tions of the protocol. The same camera was handheld by the experimenter to
film the play session.
Play Session
Owners were instructed to play with their dogs for approximately 3 min. They
were urged to play as they would normally, to ignore the experimenter, and to start
and end the session in their usual way. The owner determined the games played
and the precise session length. The experimenter remained seated and filmed the
session by hand.
Questionnaire
The questionnaire was composed of 23 questions designed to obtain informa
-
tion about normal play routines and ownership style. The experimenter sat oppo
-
site to and at least 2 m away from the owner. She asked the questions,
standardizing presentation and making a written record of the owner’s responses.
The Petting test (described following) was carried out concurrently.
Testing
The testing procedure was designed to assess attachment and dominance di
-
mensions of the dogs’ relationships with their owners. It was based on tests de
-
veloped during experimental studies (Rooney, 1999; Rooney & Bradshaw,
2002) but modified to measure the dog’s behavior toward the owner (instead of
an experimenter), to ensure that the test was easily replicable by different dog
owners, was unaffected by the size and shape of the room, and accommodated
the presence of an additional person (the experimenter). The test had 16 compo
-
ATTACHMENT DIMENSIONS OF DOG–HUMAN RELATIONSHIPS 71
nents, 5 designed specifically to assess attachment dimensions, 10 to assess
dominance dimensions, and 1 for the assessment of both. They subsequently are
referred to separately as the Attachment Test and Dominance Test. We describe
the procedure during each component next.
Attachment Test Components
A1. Petting: This test took place while the questionnaire was being answered.
The owner and experimenter were seated at least 2 m apart. Owners were instructed
to let their dogs roam freely and to pet them only if they approached. Dogs who ap
-
proached the experimenter were petted moderately. The first 10 min were analyzed.
A2. Ignoring: This test lasted for 3 min. The owner was instructed to read si
-
lently and not to interact with the dog in any way. The experimenter did the same.
If the dogs approached the owner or experimenter they were ignored.
A3. Novel object: The experimenter encouraged the dog to approach her to be
petted. She then took a motorized child’s toy (Motorized Crazy Shaker, Ertl Pre-
school, Bolingbrook, IL) from a bag, turned it on, and placed it on the ground half-
way between the owner and herself. The toy vibrated and moved along the floor.
After 20 sec, it was turned off and placed out of the dog’s reach.
A4. Call response: The experimenter crouched on the floor and petted the dog.
The owner was seated opposite and called the dog by name. The call was repeated
until the dog responded or for a maximum of 30 sec. This component was repeated
once and also formed part of the Dominance Test.
A5. Separation: Owners were asked to behave toward their dogs as they usu-
ally did when leaving them alone: to give commands, put on coats, and collect
keys. The owners then left the room and if possible the house, closing the door and
leaving the experimenter and dog alone. The experimenter remained seated for 1
min, petting any dog who approached. She then exited via the same door as the
owners and remained outside with the owners, out of sight and earshot of the dogs,
for a further 3 min.
A6. Return: The experimenter reentered the room and greeted the dogs moder
-
ately. She sat down and petted any dogs who approached. Thirty seconds later, the
owners reentered and greeted their dogs in their usual manner.
Dominance Test Components
D1. Food removal: This component was carried out after the play session
and before the questionnaire. This ensured all dogs were similarly satiated and
prevented begging from interfering with their behavior during the remaining test
components. The owners followed their normal feeding routine; but once the
dog had been given food and had commenced eating, the owner removed the
bowl. The owner used the necessary force to stop the dog feeding and to pick up
72
ROONEY AND BRADSHAW
the food bowl, holding it above the dog’s head for 10 sec before returning it to
the floor. The owner then instructed the dog to “Leave” using verbal commands
and/or physically blocking to prevent the dog from touching the food for 20 sec.
The dog was then allowed to resume eating. This component was repeated, but
the second time, when the food was replaced on the floor only a single verbal
command was given.
D2. Sit: The owner instructed the dog to sit. If there was no response, the com
-
mand was repeated until there was a response or for a maximum period of 30 sec.
This component was repeated once.
D3. Forced down: While their dogs were sitting, the owners placed a hand on
the dogs’ shoulders and attempted to push them gently into a lying position. This
continued for a maximum of 20 sec or until the dog lay down. This component was
repeated once.
D4. Rise: After the dogs had been forced down, the owners instructed them to
“Stay” and took five steps backward, remaining facing their dogs until they rose or
for a maximum of 1 min.
D5. Approach: The experimenter crouched on the floor and petted the dog. The
owner then approached the dog at walking pace and halted when 10 cm away. This
was repeated once.
D6. Lying: The owners lay on their back on the floor and remained motionless
for 30 sec.
D7. Grooming: The owners groomed their dogs lightly for 1 min. They were
encouraged to groom the entire surface, including the head.
D8. Toy removal: The owners were given a small plastic squeaky toy, which
they gave to their dog. After 20 sec, the owner attempted to retrieve the toy. They
initially commanded the dog, but if this failed, they applied the necessary force to
retrieve the toy. This was carried out twice.
D9. Toy regain: After retrieving the toy, the owners held it out of the dog’s
reach for 10 sec. They then placed it on the floor in front of the dog and instructed
the dog to “Leave.” If the dog was still obeying after 20 sec, they were praised and
allowed to regain the toy. This component was repeated after each toy removal.
D10. Shout: The owner crouched down and petted the dog for 10 sec before ris
-
ing suddenly and shouting “No.”
All Attachment and Dominance Test components were carried out within a sin
-
gle session. They were presented in a standard order: D1, A1 (during question
-
naire), A2, A3, A4, D2, D3, D4, A5, A6, D5, D6, D7, D8, D9, and D10.
The owners were instructed before each component and also prompted when
necessary. At the outset, it was stressed that they could refuse to perform any of the
components and were free to pet and interact with their dogs between components.
The dog’s behavior was filmed using the video camera and written records were
made of any subtle behaviors observed during testing.
ATTACHMENT DIMENSIONS OF DOG–HUMAN RELATIONSHIPS 73
Behavioral Measures
The behavior of the dog during each phase of the visit was transcribed from the
videotapes.
Play Session
We recorded all the game types that were played during the play session. Ten
owners used more than five games, so only these games were used in subsequent
analysis (Table 1). The total play duration was recorded, and subjective ratings
were made for the dog’s involvement and the interactivity of the play session
(Rooney, 1999). There also were five variables that described the extent to which
the owner and dog controlled the play session (Table 2).
Testing
Preliminary observations of the videoed behavioral tests were used to select
variables. The selection criteria were that the variable must describe a behavior
pattern exhibited by and showing variation in at least 10% of subjects. When ap-
propriate, rarer behaviors were combined into categories to reach the 10% thresh-
old. During the six Attachment Test components, 38 variables were measured that
described the dog’s behavior toward the owner and the experimenter.
During the Dominance Test, 50 variables were measured. Of these, 29 were
specific to individual test components. A further 21 behavior patterns could occur
at various stages throughout the test; therefore, we measured the number of test
components out of a maximum of 11 in which the dog performed the pattern.
74
ROONEY AND BRADSHAW
TABLE 1
Definitions of Dog–Human Game Types
Game Type Definition
Fetch Object is repeatedly thrown by person and retrieved by dog
Rough-and-tumble Partners wrestle, involving no focal object, and there is a high level of contact
between the players
Tug-of-war Two partners simultaneously pull on a single object, each apparently aiming
to gain sole possession
Chase Partners reciprocally chase and run away from one another
Keep-away One player possesses an object and tries to retain it, often while enticing their
partner to pursue
Data Analysis
Data Reduction
Five partnerships failed to complete one of the test components because of the
owner’s choice or physical disabilities. Three owners could not conduct the Toy
Removal test (D8) because their dogs were not interested in toys; none of these
dogs were reported as aggressive toward their owners. One owner could not com
-
plete the Food Removal test because that dog would not eat; this dog had been ag
-
gressive in the past but not over food and not currently. The missing data were
replaced with the mean value calculated using the remainder of the sample. The
variables measured during Attachment and Dominance Tests were treated as dis
-
crete data sets. Principal components analysis (PCA) without rotation was applied
twice to each data set to maximize parsimony (Cooley & Lohnes, 1971). The ini
-
tial application served to reduce the number of variables, and the second identified
the relevant factors.
ATTACHMENT DIMENSIONS OF DOG–HUMAN RELATIONSHIPS 75
TABLE 2
Behavioral Variables Measured During the Play Session
Variable Description Units
General play
Involvement Subjective rating of the dog’s involvement or
commitment to playing during the play session
(Rooney, 1999)
Scale: 1 to 5
Interactivity Subjective rating for the degree of interactivity or
reciprocity between the players during the play
session (Rooney, 1999)
Scale: 1 to 5
Play duration Time for which the dog plays with the owner over 3 min
session: 0 < 160 sec; 1 ≥ 160 sec
0/1
Play control
Proportion of dog
wins
Proportion of object competitions following which the
dog maintains possession of the object
Prop
Dog winner Session winner as derived from proportion of dog wins:
0 = proportion of dog wins < 0.66; 1 = proportion dog
wins ≥ 0.66 (“dog win”)
0/1
Dog possession
duration
Duration for which the dog has sole possession of the
object
Sec
Toys removed Focal toy is retained by the owner at the end of the play
session
0/1
Command
frequency
Total number of commands given by the owner to the
dog
Freq
Note. Prop = proportion; sec = seconds; freq = frequency.
Attachment Test.
Preliminary PCA identified six potentially interpretable
factors (as identified from a scree plot of eigenvalues). The key variables (those
with a loading factor of greater than half the maximum loading factor, irrespective
of sign) comprising these factors were examined. Any of the 38 measured variables
that (a) were not highly loaded in any of the factors or (b) always occurred in the
same factors as another variable, the accompanying variable having a higher load
-
ing factor in the first factor, were eliminated from further analysis. Two variables
were eliminated for reason (a) and 6 for reason (b). A final PCA was run on the re
-
maining 30 variables (Table 3). This produced seven factors with eigenvalues over
76
ROONEY AND BRADSHAW
TABLE 3
Behavioral Variables Measured During the Attachment Test Components Excluding
Measures Eliminated During Data Reduction
Test Component/Variable Description Unit
A1. Petting
Owner interaction
duration
Time dog spends in contact with, or < 50 cm from
and attending to, owner
Sec
Owner approach
frequency
Number of times dog approaches to within 50 cm of
owner
Freq
Owner resting distance Distance from owner that dog sits/lies for the longest
time
Meter
Owner furthest distance Furthest distance from owner that dog reaches Meter
Owner toy presentations Number of times dog contacts owner with toy Freq
Owner—lick, paw,
nuzzle, jump
Number of times dog contacts owner via lick, paw,
nuzzle, or jump
Freq
Experimenter interaction
duration
Time dog spends in contact with, or < 50 cm from
and attending to experimenter
Sec
Experimenter approach
frequency
Number of times dog approaches to within 50 cm of
experimenter
Freq
Experimenter resting
distance
Distance from experimenter that dog sits/lies for the
longest time
Meter
Experimenter furthest
distance
Furthest distance from experimenter that dog reaches Meter
Experimenter toy
presentations
Number of times dog contacts experimenter with toy Freq
Experimenter—lick,
paw, nuzzle, jump
Number of times dog contacts experimenter via lick,
paw, nuzzle, or jump
Freq
Duration object interest Time for which dog attends to object Sec
A2. Ignoring
Owner attention duration Time dog spends in contact with or < 50 cm from
and attending to owner
Sec
Owner vocalizations Number of vocalizations dog makes while facing
towards owner
Freq
Experimenter attention
duration
Time dog spends in contact with or < 50 cm from
and attending to experimenter
Sec
(continued)
1.0, which jointly explained 65% of the initial variance in the data. Factors 5 and
above were composed primarily (over 80%) of variables that also were key vari
-
ables in earlier factors. The latter three factors were not interpreted, and only Fac
-
tors 1 to 4 (Nonspecific Attention Seeking; Preference for Owner; Separa
-
tion-Related Behavior; Preference for Unfamiliar Person) were retained (Tables 4
to 7).
Dominance Test.
The Dominance Test data were analyzed in the same
way. Preliminary PCA identified five factors. Three of the measured variables
were not key variables in any of these factors, and 5 variables occurred only in
the same factors as another variable, the accompanying variable having a higher
TABLE 3 (Continued)
Test Component/Variable Description Unit
A3. Novel object
Avoid Dog withdraws from object 0/1
Investigate Dog approaches or contacts object 0/1
Vocalization Dog vocalizes while facing toward object 0/1
Owner response level Degree of dog’s response toward owner: 0 (none); 1
(look); 2 (approach); 3 (contact)
Scale: 0 to 3
Experimenter response
level
Degree of dog’s response towards experimenter: 0
(none); 1 (look); 2 (approach); 3 (contact)
Scale: 0 to 3
First approach owner Person dog approaches first: 0 (none); 1
(experimenter); 2 (owner)
Categoric
A4. Call
Response lag Latency of dog’s approach to owner following first
call
Sec
A5. Separation
Accomp. near door
duration
Time dog spends < 50 cm from exit door while
experimenter is present
Sec
Time with experimenter Time dog spends < 50 cm from and oriented towards
experimenter while owner is absent
Sec
Alone attend door
duration
Time dog spends oriented toward and attending to
exit door while experimenter and owner are absent
Sec
Alone near door duration Time dog spends < 50 cm from exit door while
experimenter and owner are absent
Sec
Alone vocalization
frequency
Number of vocalizations dog makes while
experimenter and owner are absent
Freq
Alone contact door
frequency
Number of times dog contacts exit door while
experimenter and owner are absent
Freq
A6. Return
Experimenter greeting
intensity
Subjective rating of dog’s greeting of experimenter:
0 (none); 1 (moderator); 2 (intense)
Scale: 1 to 3
Note. Sec = seconds; freq = frequency; accomp. = accompanied by experimenter.
TABLE 5
Key Variables and Their Loadings on Attachment Factor 2: Preference for Owner
Positively Loaded Variables Negatively Loaded Variables
Test
Component Variable
Loading
Factor
Test
Component Variable
Loading
Factor
Petting Experiemnter
resting distance 0.73
Petting Experiementer
interaction duration –0.67
Novel
object
First approach
owner 0.67
Call Response lag –0.45
Novel
object
Owner response
level 0.65
Ignoring Experimenter
attention duration –0.42
Petting Duration object
interest 0.44
Petting Owner interaction
duration 0.43
Petting Owner toy
presentations 0.43
Petting Experimenter
furthest distance 0.39
Petting Owner approach
frequency 0.37
TABLE 4
Key Variables and Their Loadings on Attachment Factor 1: Nonspecific Attention Seeking
Positively Loaded Variables Negatively Loaded Variables
Test
Component Variable
Loading
Factor
Test
Component Variable
Loading
Factor
Petting Experimenter approach
frequency 0.87
None
Experimenter toy
presentations 0.86
Owner approach frequency 0.83
Owner toy presentations 0.74
Duration object interest 0.57
78
TABLE 7
Key Variables and Their Loadings on Attachment Factor 4: Preference for Unfamiliar
Person (Experimenter)
Positively Loaded Variables Negatively Loaded Variables
Test
Component Variable
Loading
Factor
Test
Component Variable
Loading
Factor
Petting Experimenter—lick,
paw, nuzzle, jump 0.49
Call Response lag –0.48
Petting Experimenter
interaction duration 0.47
Separation Alone contact door
frequency –0.44
Petting Owner resting
distance 0.46
Separation Alone near door
duration –0.41
Novel
object
Experimenter
response level 0.35
Petting Owner interaction
duration –0.35
Petting Owner furthest
distance 0.35
Return Experimenter
greeting intensity –0.31
Separation Alone attend door
duration –0.31
Novel
object
Investigate –0.26
TABLE 6
Key Variables and Their Loadings on Attachment Factor 3: Separation-Related Behavior
Positively Loaded Variables Negatively Loaded Variables
Test
Component Variable
Loading
Factor
Test
Component Variable
Loading
Factor
Separation Accomp. near door
duration 0.76
Separation Time with
experimenter –0.67
Ignoring Owner vocalizations 0.56
Ignoring Owner attention
duration 0.50
Petting Owner resting distance 0.41
Note. Accomp. = accompanied by experimenter.
79
loading factor in the first factor. These 8 variables were eliminated before sub
-
sequent analysis. Final PCA on the remaining 42 dominance variables (Tables 8
and 9) produced 7 factors with eigenvalues greater than 1.0, jointly explaining
51% of the original data variance. Factors 3 and above were composed primar
-
ily (over 70%) of variables featuring as key variables in earlier factors and were
therefore rejected. Factors 1 (Amenability) and 2 (Confident Interactivity) were
retained for further analysis (Tables 10 and 11).
80
ROONEY AND BRADSHAW
TABLE 8
Behavioral Variables Measured During Specific Components of the Dominance Test
Excluding Measures Eliminated During Data Reduction
Test
Component/Variable Description Unit
D1. Food removal
Response
intensity
Strategy dog uses to regain food: 0 (none); 1 (skirt); 2
(push)
Scale: 0 to 2
Restrain force Force used by owner to remove food from dog: 0 (one
command); 1 (multiple command); 2 (gesture); 3
(contact); 4 (restraint); 5 (not removed)
Scale: 0 to 5
One command eat Time for dog to resume feeding following second food
replacement with one command
Sec
A4. Call
Response lag Latency for dog to approach owner following first call
(mean of 2 trials)
Sec
Approach gait Gait of dog’s approach to owner: 0 (none); 1 (walk); 2
(trot); 3 (run); 4 (bound)
Scale: 0 to 4
Indirect approach Dog takes nondirect route to owner 0/1
D2. Sit
Sit latency Latency for dog to sit following first command (mean of
two trials)
Sec
Orientate away Dog sits facing away from owner 0/1
D3. Forced down
Number comply Number of times owner successfully forces dog into lying
position (out of two trials)
Freq
Comply latency Time owner takes to force dog into lying position (mean of
two trials)
Sec
D4. Rise
Rise latency Latency for dog to rise from lie after forced down Sec
Reapproach gait Gait of dog’s approach to owner after forced down: 0
(none); 1 (walk); 2 (trot); 3 (run); 4 (bound)
Scale: 0 to 4
D5. Approach
Response
positivity
Degree of dog’s response to owner’s approach: 0
(retreat); 1 (unreactive); 2 (approach); 3 (jump up)
Scale: 0 to 3
(continued)
Hypothesis Testing
The six factors (four attachment and two dominance) were used as dependent
variables. Because they were derived from observation, they are referredtoas “mea
-
sured” dependent variables. Additionally, two “reported” dependent variables were
derived from the owners’ questionnaire responses. Owners were asked if their dogs
had ever exhibited aggression defined as “any exhibition of growling (not playful),
baring teeth, snapping or biting” toward them and when it had last occurred. If they
described aggression as ongoing, they scored 1 for the categoric (0/1) variable “cur
-
rent aggression.” They also were asked to describe the form the aggression took.
Their responses were ranked 0 (none),1(growling),2(baring teeth),3(snapping),
or 4 (biting). This ordinal variable was named “aggression intensity.”
Examination of the dependent variables revealed that they were not normally
distributed; therefore, all subsequent statistics are nonparametric, and all descrip
-
tions of data are in the form of medians and quartiles. A range of independent vari
-
ables was selected, each of which was hypothesized to affect relationship
dimensions. These described demographics, features of dog–owner play, and
ATTACHMENT DIMENSIONS OF DOG–HUMAN RELATIONSHIPS 81
TABLE 8 Continued
Test
Component/Variable Description Unit
D6. Lying
Ignore Dog shows no response to owner lying down 0/1
Step over legs Dog steps over owner’s legs 0/1
Stand on/over Dog stands on or over owner’s torso: 0 (never); 1 (head
over); 2 (body over)
Scale: 0 to 2
D7. Grooming
Stance height Posture assumed by dog for most of 60 sec: 0 (lie); 1 (sit);
2 (stand); 3 (move)
Scale: 0 to 3
Time tolerated Duration dog tolerates grooming without escape attempt Sec
D8. Toy removal
Force required Force required for owner to retrieve toy from dog: 0
(command); 1 (contact); 2 (contact and command); 3
(pull); 4 (strong tug); 5 (will not surrender)
Scale: 0 to 5
Brings toy Dog brings toy to owner when instructed to leave 0/1
D9. Toy Regain
Regain latency Time for dog to pick up toy following removal (mean of
two trials)
Sec
D10. Shout
Stance height Posture dog assumes after shout: 0 (lie); 1 (sit); 2 (stand);
3 (move)
Scale: 0 to 3
Change posture Dog changes posture immediately after shout 0/1
Note. Sec = seconds; freq = frequency.
ownership style that had been measured during the play session (Table 2) and de
-
rived from the questionnaire responses (Table 12). To explore the effect of each in
-
dependent variable on the dependent variables, four nonparametric statistical tests
were used (Spearman Rank Correlation, Kruskal–Wallis, Mann–Whitney U, Con
-
tingency, and Fisher’s Exact tests; SPSS Version 10).
RESULTS
Play Behavior
The most common observed game type was keep-away, which was played by 36
of the 50 partnerships (Figure 2). Involvement and interactivity scores both
ranged from 1 to 5, with medians (Mdn) of 3.9 (25th percentile = 2.5, 75th per
-
centile = 4.6) and 3.6 (25th percentile = 2.5, 75th percentile = 4.5), respectively.
The median play duration was 158 sec (25th percentile = 97, 75th percentile =
82
ROONEY AND BRADSHAW
TABLE 9
Behavioral Variables Measured Over All 11 Dominance Test Components
Variable Description Units
Bark Dog barks at owner Number
Climb up Dog jumps up placing forepaws on owner Number
Display Dog displays inguinal region Number
Ears pricked Dog’s ears are pricked up Number
Ears low Dog’s ears are low or back Number
Growl Dog growls at owner Number
Head in lap Dog places head in owner’s lap Number
Lick lips Dog licks lips Number
Lick owner Dog contacts owner with tongue Number
Low body position Dog exhibits low posture Number
Mouth owner Dog places mouth around owner’s hand Number
Nuzzle Dog contacts owner with its head Number
Paw Dog raises paw or places paw on owner Number
Playful Dog reacts playfully Number
Roll Dog rolls over on floor Number
Sit Dog sits uncommanded Number
Tail height Average of tail positions during each component:
1 (tucked) to 5 (vertical)
Scale: 1 to 5
Tail movement Dog wags tail Number
Tail tucked Dog’s tail is tucked between its legs Number
Note. Number = the number of test components during which the behavior occurred.
TABLE 10
Key Variables and Their Loadings on Dominance Factor 1: Amenability
Positively Loaded Variables Negatively Loaded Variables
Test
Component Variable
Loading
Factor
Test
Component Variable
Loading
Factor
All Roll 0.68 All Ears pricked –0.52
All Low body
position 0.65
Food
removal
Response
intensity –0.52
All Ears low 0.64 Call Response lag –0.52
All Display 0.62 Toy removal Force required –0.51
Food removal One command eat 0.61 Forced down Comply latency –0.43
All Tail tucked 0.60 Call Indirect approach –0.43
All Head in lap 0.51 All Mouth owner –0.41
Toy regain Regain latency 0.43 All Climb up –0.39
Rise Rise latency 0.44 All Tail height –0.35
Grooming Time tolerated 0.44
Toy removal Brings toy 0.41
All Lick owner 0.41
TABLE 11
Key Variables and Their Loadings on Dominance Factor 2: Confident Interactivity
Positively Loaded Variables Negatively Loaded Variables
Test
Component Variable
Loading
Factor
Test
Component Variable
Loading
Factor
All Tail movement 0.65 Call Response lag –0.46
Call Approach gait 0.61 Lying Ignore –0.33
All Playful 0.55 All Tail tucked –0.33
All Lick owner 0.53
Lying Stand on/over torso 0.53
Grooming Stance height 0.53
Rise Reapproach gait 0.51
All Sit 0.49
All Tail height 0.48
All Bark 0.46
All Paw 0.46
Approach Response positivity 0.44
83
180), and 25 dogs played for 160 sec or more and were classified as “long play
-
ers.” The dogs observed to play fetch were significantly younger (Mdn age = 5.3
years; 25th percentile = 3.0, 75th percentile = 8.6) than nonplayers (Mdn age=8
years; 25th percentile = 7, 75th percentile = 11; Mann–Whitney U = 164, p <
.01). No further variables were affected by age.
84
ROONEY AND BRADSHAW
TABLE 12
Variables Extracted From the Questionnaire and Used As Independent Variables
for Testing Relationships With Attachment and Dominance Dependent Variables
Variable Description Units
Demographics
Dog age Age of dog at time of visit Years
Dog breed type Kennel Club category of dog breed or breed most
closely resembled
6 categories
Pure breed Reported purebred or pedigree dog 0/1
Dog sex Sex of dog: 0 (male); 1 (female) 0/1
Owner gender Gender of owner: 0 (male); 1 (female) 0/1
Play behavior
Play frequency No. of dog owner play sessions per week Frequency
Dog initiates When asked how they start play, owner volunteers that
dog usually starts play sessions
0/1
Owner’s behavior
Daily time together Time owner spends with dog in an average day Minutes
Daily interaction
time
Time owner spends interacting with dog in an average
day
Minutes
Current training time Time spent training dog per week: 0 (none); 1 (< 10
min); 2 (10 to 30 min); 3 (>30 min)
Scale
Sleeping location Dog’s usual night sleeping place: other room; locked in
other room; owner’s bedroom; owner’s bed
4 categories
FIGURE 2 Number of dog–owner
partnerships (out of 50) observed to
play each game type during the play
session.
Although the games played appear to be influenced by breed type, insuffi
-
cient numbers of some breed categories made the relationship statistically
untestable. Female dogs scored higher for involvement in the play session (Mdn
= 4.5; 25th percentile = 2.9, 75th percentile = 5.0)) than did male dogs (Mdn =
3.5; 25th percentile = 2.5, 75th percentile = 4.5; U = 194, p < .05). Male owners
were more likely to engage in rough-and-tumble play (67%) than were female
owners (37%), but this trend did not reach statistical significance (Fisher’s Exact
test; p = .07).
Thirty-nine owners incorporated commands within their play, giving an aver
-
age of 6 (25th percentile = 1, 75th percentile = 14.5) per session. Seven partner
-
ships did not include a toy in their game; therefore, the remaining play control
variables were measured on a reduced sample of 43. The proportion of dog wins
varied from 0 to 1 with a median of 0.17 (25th percentile = 0, 75th percentile =
0.5). Seven partnerships were classified as dog winners(> 0.66 dog wins). The dog
possession duration ranged from 1 to 173 sec, with a median of 78 (25th percentile
= 27, 75th percentile = 105). Eighteen owners removed the toy at the end of the
play session, whereas 25 dogs had continual access to the focal toy.
Questionnaire Responses
The reported frequency of play sessions varied between 0 and 42 per week, with
a median of 7 (25th percentile = 7, 75th percentile = 21). When asked how they
usually started play, 29 of the 50 owners said that the dog was often the initiator.
Owner’s Behavior
Owners spent between 4 and 24 hr per day (Mdn = 13 hr; 25th percentile = 7,
75th percentile = 17.5)) with their dog. An average of 1.75 hr (25th percentile =
1.5, 75th percentile = 2.5) of this time was spent interacting. However, the major
-
ity (72%) of owners currently devoted no time to training their dog, and only 14%
spent more than 30 min per week training. When asked about the dog’s usual
sleeping position, 45% of the dogs were reported to be locked out of the owner’s
room at night, 14% voluntarily slept in another room, 10% usually slept in the
owner’s room, and 31% slept on the owner’s bed.
Dog’s Behavior
Reported aggression levels ranged from 0 to 4 with a median of 1 (25th percen
-
tile = 0, 75th percentile = 1.3). Twelve owners reported their dog as showing cur
-
rent aggression.
ATTACHMENT DIMENSIONS OF DOG–HUMAN RELATIONSHIPS 85
What Affected Attachment Dimensions of the Dog–Owner
Relationship?
Demographics
Preference for Owner was negatively correlated to the dog’s age (ρ = 0.29, df
= 49, p < .05). Preference for Unfamiliar Person was the only attachment dimen
-
sion to vary significantly according to breed type (Kruskal–Wallis test; χ
2
=
13.1, df =5,p < .05). Crossbred dogs scored significantly higher for Preference
for Owner (Mdn = 0.3; 25th percentile = –0.5, 75th percentile = 0.9) than did
purebred dogs (Mdn = –0.3; 25th percentile = –0.8, 75th percentile = 0.6; U =
196, p < .05). The attachment dimensions were not significantly affected by the
dog’s sex or the owner’s gender.
Play Behavior
Dogs observed to play rough-and-tumble scored lower for Separation-Related
Behavior (Mdn = –0.6; 25th percentile = –0.9, 75th percentile = 0.2) than did
nonplayers (Mdn = 0.4; 25th percentile = –0.6, 75th percentile = 1.1; U = 210, p
= .05). Interactivity scores were significantly correlated to scores for Preference
for Unfamiliar Person (ρ = 0.36, df = 49, p = .01). Dogs classified as long play-
ers scored higher for Preference for Owner (Mdn = 0.3; 25th percentile = –0.4,
75th percentile = 1.0) than did short players (Mdn = –0.5; 25th percentile = –1.1,
75th percentile = 0.5; U = 196, p < .05). Involvement scores and other game
types played showed no significant relationships to any of the attachment di-
mensions.
Owner’s Behavior
Neither daily time together nor the daily interaction time showed significant
correlation to the attachment dependent variables (ρ < 0.2, df = 49, p > .18). No at
-
tachment dependent variables were affected by the dog’s sleeping location
(Kruskal–Wallis; χ
2
< 5.0, df = 3, p > 0.2).
What Affected Dominance Dimensions of the Dog–Owner
Relationship?
Reports of current aggression showed no significant relationship to the measured
(U > 150, p > .05) dominance variables.
86
ROONEY AND BRADSHAW
Demographics
Confident Interactivity scores decreased with the age of the dog (ρ = –0.39, df =
49, p = .005). The remaining dependent variables were unaffected by age.
Amenability scores varied significantly between the breed types
(Kruskall–Wallis; χ
2
= 14.7, df =5,p = .01). Crossbred dogs obtained higher
scores for Amenability (Mdn = 0.4; 25th percentile = –0.2, 75th percentile = 0.9)
than did purebred dogs (Mdn = –0.1; 25th percentile = –1.2, 75th percentile = 0.2;
U = 189, p < .05). Extracted dominance variables did not vary with the sex of the
dog (U > 210, p > .05) or the owner (U > 181, p > .1).
Play Behavior
Observed play session.
Dogs observed to play specific game types dif
-
fered from nonplayers in dominance dimensions. Players of tug-of-war scored sig-
nificantly higher for Confident Interactivity (Mdn = 0.4; 25th percentile = –0.5,
75th percentile = 1.2) than did nonplayers (Mdn = –0.4; 25th percentile = –0.7, 75th
percentile = 0; U = 197, p < .05). Similarly, players of fetch obtained higher scores
for Confident Interactivity (Mdn = 0.2; 25th percentile = –0.5, 75th percentile =
1.1) than did nonplayers (Mdn = –0.4; 25th percentile = –0.7, 75th percentile =
–0.1; U = 192, p < .05) and players of rough-and-tumble scored higher for
Amenability (Mdn = 0.2; 25th percentile = –0.1, 75th percentile = 0.9) than dogs
who did not play this game type (Mdn = –0.2; 25th percentile = –1.2, 75th percentile
= 0.3; U = 207, p < .05). Involvement in the play session was highly correlated to
Confident Interactivity (ρ = 0.43, df = 49, p < .005), and Interactivity in the play ses-
sion correlated to Amenability (ρ = 0.30, df = 49, p < .05). Also, dogs described as
long players scored higher for Confident Interactivity (Mdn = 0.3; 25th percentile =
–-0.5, 75th percentile = 1.3) than did short players (Mdn = –0.3; 25th percentile =
–0.9, 75th percentile = 0; U = 197, p < .05).
None of the control variables measured during the play session showed sig
-
nificant relationships to the measured dominance variables or the reported ag
-
gression level. However, current aggression was more common in dogs who
took part in a dog winner play session (Fisher’s Exact test; p < .01). Five of the
seven dog-win sessions involved currently aggressive dogs (of the nine aggres
-
sive dogs playing with toys). Command frequency had no effect on any of the
dependent variables.
Questionnaire variables.
The reported play frequency did not signifi
-
cantly affect the dominance variables. The variable “dog initiates” showed very
ATTACHMENT DIMENSIONS OF DOG–HUMAN RELATIONSHIPS 87
significant associations to two dependent variables. Dogs who were reported to
initiate play scored lower for Amenability (Mdn = –0.2; 25th percentile = –1.2,
75th percentile = 0.3; N = 29) than dogs who did not initiate play (Mdn = 0.2;
25th percentile = 0, 75th percentile = 1; N = 21; U = 162, p = .005). Dogs who ini
-
tiated play also were more likely to exhibit current aggression (Fisher’s Exact
test; p < .01). Of the 12 currently aggressive dogs, 11 were reported to initiate
play (Figure 3).
Owner’s Behavior
The time currently devoted to training did not affect any of the measured domi
-
nance variables.
DISCUSSION
Attachment Test
We are aware of no previous attempts to quantify attachment dimensions of a
dog–owner relationship within the home setting. Our test yielded four meaning-
ful attachment factors, confirming that attachment is a multidimensional con-
struct (Cairns, 1972), although given our sample size (50), we cannot be
confident that the factors we have identified would be adequate to describe all
dog–owner relationships. However, they are in broad agreement with those we
have found in similar studies (Rooney, 1999). The attachment factors were af-
fected by demographics. Preference for Owner decreased with age and was
higher in crossbred dogs, which is consistent with the higher prevalence of sepa
-
88
ROONEY AND BRADSHAW
FIGURE 3 Relationship be
-
tween dogs who initiate play ses
-
sions and those who exhibit current
aggression.
ration disorders in crossbreeds found in some studies (Takeuchi, Ogata, Houpt,
& Scarlett, 2001), although not all (Bradshaw, McPherson, Casey, & Larter,
2002). Preference for Unfamiliar Person was significantly associated with breed
type, suggesting that characteristics of the dog have an important role in deter
-
mining this attachment dimension.
The Link Between Rough-and-Tumble Play
and Attachment Dimensions
This study did not detect any relationship between attachment dimensions
and the quantity of play, but the types of games played were important. The
dogs observed to play rough-and-tumble games scored significantly lower for
Separation-Related Behavior than did nonplayers. Because this is a correlational
study, in isolation it provides no evidence that rough-and-tumble is the cause of,
and not a result or correlate of, reduced Separation-Related Behavior. However,
based on this and a previous study (Rooney, 1999), we suggest that there may be
a causal relationship between the playing of rough-and-tumble and attach-
ment-related behavior. We suggest that this game may inhibit the exhibition of
separation-related behavior and therefore may be beneficial for the dog. The
means by which this occurs is unknown, but one possibility is that the physical
contact between dog and owner during rough-and-tumble play fulfils the dog’s
need for contact that they no longer need to seek; hence, the exhibition of Con-
tact Seeking (Rooney, 1999) and Separation-Related Behavior (this study) are
decreased. A similar relationship between play and contact is seen in rats (Rattus
norvegicus), as deprivation of social contact leads to an increased drive for
rough-and-tumble play (Ikemoto & Panksepp, 1992).
Dominance Test
This home-based Dominance Test was adapted from a test developed in an ex
-
perimental setting (Rooney & Bradshaw, 2002). Our test yielded two meaning
-
ful factors—Amenability and Confident Interactivity—that equated well to the
factors produced in experimental studies. Our test, therefore, may be useful,
with modifications, in the home as well as the laboratory context. As for attach
-
ment, however, given our sample size (50) we cannot be confident that the fac
-
tors we have identified would be adequate to describe all dog–owner relation
-
ships. However, they are in broad agreement with those we have found in
similar studies (Rooney, 1999; Rooney & Bradshaw, 2002).
Amenability is a desirable quality in a dog–human relationship. Low scores for
Amenability can be problematic because they describe dogs who refuse to leave
ATTACHMENT DIMENSIONS OF DOG–HUMAN RELATIONSHIPS 89
their food or their toys or who do not respond to a call. Confident Interactivity, al
-
though including many behaviors that may be described as dominant, also is a desir
-
able trait in moderation. A dog scoring high for Confident Interactivity is likely to be
playful, lick the owner frequently, and approach quickly when called. Therefore, we
suggest that factors that decrease Amenability are potentially more harmful than
those that increase Confident Interactivity.
Dissociation Between Dominance and Aggression
We found no significant relationship between reported aggression incidence
and the measured dominance dimensions. This is likely to be because aggression is
multicausal; it can be a result of fear or pain (Overall, 1997) as well as being domi
-
nance related. Presumably, many of the dogs reported to show aggression did so
for reasons other than dominance.
Effects of Demography
Characteristics of the dogs strongly affected the dominance dimensions. Confi-
dent Interactivity decreased with age (p = .005), and Amenability varied with breed
type (p =.01).Theseresults suggest that biological factors play a major role in the de-
termination of dominance dimensions of dog–human relationships, and the effects
of external factors are comparatively weak. Thus, games and owners’ behavior may
only modify dominance tendencies that are already predetermined.
Does Playing Tug-of-War Enhance Dominance?
Tug-of-war players scored higher for Confident Interactivity than did
nonplayers. Although this could be interpreted as supporting the dominance en
-
hancement theory, further evidence suggests it does not. Dogs who played fetch in
the play session also scored higher for Confident Interactivity than did nonplayers.
A link between fetch and confidence was found in previous experimental studies
when a group of Labrador Retrievers who received regular games of fetch in
-
creased significantly in confidence (Rooney, 1999). Fetch is human controlled,
therefore the dominance enhancement theory would predict players of fetch to
have lower scores for Confident Interactivity or higher scores for Amenability.
Our results do not support these predictions.
Previous experimental studies have failed to identify any significant domi
-
nance-related effects (Rooney & Bradshaw, 2002), and links between dominance
and tug-of-war in this study are indiscernible from those between dominance and
fetch. We therefore conclude that playing tug-of-war does not substantially alter
dominance dimensions of the dog–human relationship.
90
ROONEY AND BRADSHAW
Effects of Other Game Types on Dominance Dimensions
Dogs who were observed to play rough-and-tumble games scored higher for
Amenability than nonplayers. Amenability is a desirable trait; therefore, this link
does not support O’Farrell’s (1992) warnings that rough-and-tumble should be
avoided. Other experimental studies have shown no significant effects of
rough-and-tumble on dominance dimensions (e.g., Rooney, 1999), which implies
this link is correlative and not causal.
Control the Game, Control the Dog?
Although it is often advised that tug-of-war games should be avoided, a crucial
factor in the effects of games is claimed to be which partner (human or dog) wins
or controls the game (McBride, 1995). A previous experimental study (Rooney &
Bradshaw, 2002) failed to find any differences in dominance-related behavior of
dogs after winning and losing tug-of-war games.
In this study, we examined the effect of control variables (including the propor-
tion of competitions won, the command frequency, and the possession duration)
during play. None of these variables affected the measured dominance dimen-
sions. However, there was a significant trend for dogs reported to be currently ag-
gressive to win over two thirds of competitions during the play session. This
appears to support the link between winning games and aggression, but there is no
evidence that this aggression was dominance related. Although five of the seven
dogs who were allowed to win at play were aggressive, they did not differ signifi-
cantly in dominance dimensions to the control sample. Also, we cannot be sure
which factor is the cause and which the effect. It is equally plausible that owners
who are aware of the dog’s aggressive tendencies allow them to win possession of
toys rather than winning causing the aggression. Longitudinal studies may prove
useful in resolving this issue.
Popular Theories or Common Myths?
This study allowed preliminary examination of a number of popular theories,
and we now review the extent to which each is supported by these results.
Do Dogs Who Have Continual Access to Toys Become
Dominant?
Rogerson (1992) claimed that dogs who have continual access to toys learn to
dictate interactions and often become dominant over their owners. We found no
ATTACHMENT DIMENSIONS OF DOG–HUMAN RELATIONSHIPS 91
difference in dominance dimensions of those dogs who retained toys at the end of
the observed games and those dogs whose owners retained the toys.
Does Allowing Dogs to Sleep in the Bedroom Lead
to Overattachment?
The importance of the dog’s sleeping location is stressed by many behavioral
counselors (Appleby, 1997). In her longitudinal study, McPherson (1998) was un
-
able to detect any effects of sleeping location on the development of separa
-
tion-related problems in Labrador Retrievers, Border Collies, or a sample of
re-homed dogs of many breeds. Within this mixed-breed sample, we too found no
association between sleeping location and any of the attachment dimensions of the
dog–owner relationship.
Does Interaction Initiation Affect Dominance?
It is claimed that dogs who are frequently allowed to initiate social interactions
(including play) behave with increased dominance toward their owner (O’Farrell,
1992). This study provides strong evidence for this claim. Dogs who were reported
to initiate play frequently were found to score lower for Amenability and were
more likely to exhibit aggression.
The nature of the link between initiation of play and dominance dimensions (in-
cluding aggression) is not clear. O’Farrell (1992) maintains that interaction initia-
tion by the dog is the cause of increased dominance, but there are alternative
explanations:
1. Play initiation may reflect a dominance relationship that already exists.
Dominant hamsters, Mesocricetus auratus (Pellis & Pellis, 1993) and squirrel
monkeys, Saimiri sciureus (Biben, 1998) more frequently initiate play with subor
-
dinate conspecifics. Dogs may similarly instigate play with their owners only when
they are sufficiently dominant to do so.
2. Allowing dogs to initiate play may reflect a general ownership trait. Dogs
who determine the timing of play sessions may also dictate other aspects of their
routine; for example, they may be allowed access to all areas of the house, de
-
mand attention, and beg for food. The owner may reinforce the dog’s status in a
great number of ways, and responding to play initiation is just one symptom of
this.
3. When owners initiate play, they usually perform human–dog play signals
(Rooney, Bradshaw, & Robinson, 2001). Play sessions that are initiated by the dog
are likely to involve reduced human–dog signaling. The link between dog initia
-
tion and decreased Amenability (and increased aggression) may result from this
92
ROONEY AND BRADSHAW
reduced metacommunication, allowing games to assume competitive conse
-
quences.
Play As a Window on Dog–Human Relationships
This study points to some factors that may play a deterministic role in dog–hu
-
man relationships, for example, rough-and-tumble games. However, it also iden
-
tifies a number of features of a dog’s play behavior with the owner, which corre
-
lates to quantifiable aspects of the relationship. These include the following:
1. Interactivity during the play session was positively correlated to Preference
for Unfamiliar Person and to Amenability.
2. Dogs who were long players showed higher Preference for Owner scores.
3. Involvement during the play session was highly correlated with Confident
Interactivity.
It is unlikely that these characteristics of play are the cause of their associated
relationship dimensions. It is more likely that features of an established
dog–owner relationship determine the way they play or that play behavior, and re-
lationship dimensions are controlled by the same overriding traits. Play behavior
reflects relationship patterns in children (Clark, Wyon, & Richards, 1969), squirrel
monkeys (Biben, 1998), and rats (Smith, Field, Forgie, & Pellis, 1996) and, we
suggest, also in dog–human relationships. Because quantifiable aspects of play
correlate to both attachment and dominance dimensions of dog–human relation-
ships, examination of play sessions may give an insight into a dog–owner relation-
ship. This suggests that play has the potential, with further research, to be used as a
probe in the assessment of dog–human relationships.
ACKNOWLEDGMENTS
We thank WALTHAM and the BBSRC for financial support of the project. We
thank the 50 kind volunteers who were filmed with their dogs. Our thanks to
Rob Atkinson, Emily Blackwell, Rachel Casey, Sarah Lowe, and Peter Hepper
for comments on the drafts of the manuscript.
REFERENCES
Anderson, R. K., & Foster, R. E. (1988). Understanding your dog. In R. K Anderson & R. J. Holmes
(Eds.), Dogs: Companions or nuisances: Public seminars 22 (pp. 29–35). Princes Highway,
Australia: Werribee Veterinary Clinical Center.
ATTACHMENT DIMENSIONS OF DOG–HUMAN RELATIONSHIPS
93
Appleby, D. (1997). Ain’t misbehavin’: A good behaviour guide for family dogs. Bristol, England:
Broadcast.
Biben, M. (1998). Squirrel monkey playfighting: Making the case for a cognitive training function for
play. In M. Bekoff & J. A. Byers (Eds.), Animal play: Evolutionary, comparative and ecological
perspectives (pp. 161–182). Cambridge, England: Cambridge University Press.
Bradshaw, J. W. S., McPherson, J. A., Casey, R. A., & Larter, I. S. (2002). Aetiology of separa
-
tion-related behaviour in domestic dogs. Veterinary Record, 151, 43–46.
Cairns, R. B. (1972). Attachment and dependency: A psychobiological and social learning synthesis. In
J. L. Gewritz (Ed.), Attachment and dependency (pp. 29–80). Washington DC: Winston.
Clark, A. H., Wyon, S. M., & Richards, M. P. M. (1969). Free-play in nursery school children. Journal of
Child Psychology and Psychiatry, 10, 205–216.
Cooley, W. W., & Lohnes, P. R. (1971). Multivariate data analysis (pp. 102–105). New York: Wiley.
Drews, C. (1993). The concept and definition of dominance in animal behavior. Behaviour, 125,
282–313.
Goodloe, L. P., & Borchelt, P. L. (1998). Companion dog temperament traits. Journal of Applied Animal
Welfare Science, 1, 303–338.
Ikemoto, S., & Panksepp, J. (1992). The effects of early isolation on the motivation for social play in ju
-
venile rats. Developmental Psychobiology, 25, 261–274.
McBride, A. (1995). The human–dog relationship. In I. Robinson (Ed.), The Waltham book of human–ani
-
mal interactions: Benefits and responsibilities of pet ownership (pp. 99–112). New York: Pergamon.
McPherson, J. A. (1998). Aetiology, characterisation and prediction of separation induced behaviour in
the domestic dog. Unpublished doctoral thesis, University of Southampton, Southampton, England.
O’Farrell, V. (1992). Manual of canine behaviour (2nd ed.). Cheltenham, England: British Small Ani-
mal Veterinary Association.
Overall, K. L. (1997). Clinical behavioral medicine for small animals. St. Louis, MO: Mosby.
Pellis, S. M., & Pellis, V. C. (1993). Influence of dominance on the development of play fighting in pairs
of male Syrian Golden hamsters (Mesocricetus auratus). Aggressive Behavior, 19, 293–302.
Podberscek, A. L. (1997). Illuminating issues of companion animal welfare through research into hu-
man–animal interactions. Animal Welfare, 6, 365–372.
Rogerson, J. (1992). Training your dog. London: Popular Dogs Publishing.
Rooney, N. J. (1999). Play behaviour of the domestic dog (Canis familiaris), and its effects on the
dog–human relationship. Unpublished doctoral thesis, University of Southampton, Southhampton,
England.
Rooney, N. J., & Bradshaw, J. W. S. (2002). An experimental study of the effects of play upon the
dog–human relationship. Applied Animal Behaviour Science, 75, 161–176.
Rooney, N. J., Bradshaw, J. W. S., & Robinson, I. (2001). Do dogs respond to play signals given by hu
-
mans? Animal Behaviour, 61, 715–722.
Smith, L. K., Field, E. F., Forgie, M. L., & Pellis, S. M. (1996). Dominance and age-related changes in
the play fighting of intact and post-weaning castrated male rats (Rattus norvegicus). Aggressive Be
-
havior, 22, 215–226.
Sylvester, P. (1984). The Reader’s Digest illustrated book of dogs. London: Reader’s Digest Associa
-
tion.
Takeuchi, T., Ogata, N., Houpt, K. A., & Scarlett, J. M. (2001). Differences in background and outcome
of three behavior problems of dogs. Applied Animal Behaviour Science, 70, 297–308.
94 ROONEY AND BRADSHAW
