Content uploaded by Adam Miklosi
All content in this area was uploaded by Adam Miklosi
Content may be subject to copyright.
Journal of Comparative Psychology
1998, Vol. 112, No. 3,219-229
Attachment Behavior in Dogs
A New Application of Ainsworth's (1969) Strange Situation Test
József Topál, Á
dám Miklósi, Vilmos Csányi,
Fifty-one owner-dog pairs were observed in a modified version of M. D. S. Ainsworth's
(1969) Strange Situation Test. The results demonstrate that adult dogs
patterns of attachment behavior toward the owner. Although there was considerable variability
in dogs' attachment behavior to humans, the authors did not find any effect of gender, age,
living conditions, or breed on most of the behavioral variables. The human-dog relationship
was described by means of a factor analysis in a 3-dimensional factor space:
A cluster analysis revealed 5 substantially different classes of
dogs, and dogs could be categorized along the secure-insecure attached dimensions of
Ainsworth's original test. A dog's relationship to humans is analogous to child-parent and
chimpanzee-human attachment behavior because the observed behavioral phenomena and the
classification are similar to those described in mother-infant interactions.
Although the construct of attachment was first used to
explain the affectional bond that develops between a human
infant and its caregiver (Bowlby, 1958), this concept has
been elaborated for behavioral phenomena that are fundamen-
tal in social species and has been approached in a number of
ways over the years. It was considered a hypothetical factor
that ties individuals together (Lorenz, 1966) or a behavior
system that results in one individual seeking and maintain-
ing proximity to another individual (Bowlby, 1972). Cohen
as a special affectional relation-
ship between two individuals that is specific in its focus and
endures over time. This relationship is based on dependency
between individuals that becomes evident through behav-
ioral preferences (Wickler, 1976).
The theories of attachment behavior range from the
psychoanalytic approach (Freud, 1946) through the different
learning theories (Cairns, 1966; Gewirtz, 1972; Hoffman &
Ratner, 1973; Solomon & Corbit, 1973) to the ethological
model of attachment (Ainsworth, 1969, 1972; Bowlby,
1958, 1969; for a review, see Rajecki, Lamb, & Obmascher,
1978). The ethological approach uses the term
evolutionary and developmental contexts and emphasizes
that attachment systems are neurobiological structures that
have been shaped by the normal environment and that
function best in that environment (Bowlby, 1958; Kraemer,
1992). The ultimate function of attachment may be to defend
against predation (Bowlby, 1969) or to obtain necessary
Miklósi, Vilmos Csányi,
Department of Ethology,
We would like to thank
Krisztina Soproni, Márta
Szima Naderi for their assistance during the observations and
analysis. This study was supported by Grant T-016-013/1995 from
the Hungarian Foundation for Basic Research and by a grant from
Masters Food, Hungary.
Correspondence concerning this article should be addressed to
József Topál or
Department of Ethology,
University, Jávorka S. U. 14, H-2131
mail may be sent to
Copyright 1998 by the American Psychological Association,
resources that are provided by the caregiver. Nevertheless,
perhaps attachment has no distinct function but simply
evolved as a consequence of a close relationship between
two individuals (Gubernick, 1981).
Attachment is an organizational construct (J. P Connel &
Goldsmith, 1982), a product of maturation that always
denotes a one-to-one relationship with a particular other and
manifests itself in different behaviors (Sears,
Novlis, & Sears, 1953). Although the operational criteria of
attachment have been developed from research on humans
and other primates, they can be applied to other species.
Attachment presumes (a) the ability to discriminate and
respond differentially to the object of attachment (i.e., the
secure-base effect), (b) a preference for the attachment
figure (e.g., proximity and contact seeking and maintenance
of proximity), and (c) a response to separation from and
reunion with the attachment figure that is distinct from
responses to others (Crnic, Reite, & Shucard, 1982; Guber-
nick, 1981; Rajecki et al., 1978).
One of the most important methodological approaches
with respect to the assessment of attachment is the Strange
Situation Test elaborated by Ainsworth and Wittig (1969).
This laboratory procedure was originally designed (Ains-
worth, 1969) to examine the balance of attachment and
exploratory behaviors under conditions of low and high
stress. Researchers hypothesized that during the experimen-
conditions, the attachment behavior is activated by
separation from and reunion with the attachment figure (see
Ainsworth, Blehar, Waters, & Wall, 1978, for an assessment
of the human infant-parent attachment). Infant responses to
this situation are customarily classified as fitting into one of
three overall patterns of behavioral organization:
infant shows signs of missing the parent upon separation,
greets the parent actively upon reunion, and then settles and
returns to play; identified as Category B);
(the infant shows little or no distress at separation from the
parent and actively avoids and ignores the parent upon
reunion; Category A); or
(the infant is
highly distressed by separation and seeks for contact on
reunion but cannot be settled by the parent and may show
strong resistance; Category C). Recently, Main and Solomon
(1990) described an additional insecure attachment pattern
and called it the
(Category D). Because
the Strange Situation paradigm is based on the evolutionary
approach of attachment behavior, some authors have success-
fully adapted it to ethological studies. For example, research-
ers have found that during separation from and reunion with
cagemates, chimpanzees reacted similarly to children (Bard,
1983, 1991; Miller, Bard, Juno, & Nadler, 1986) and that not
only conspecifics but also humans could serve as attachment
figures for young chimpanzees (Miller, Bard, Juno, &
With the exception of primate studies, only a limited
number of studies have explored the animal to human
attachment. One of the most promising subjects for investi-
gating this topic is the dog
The origin of
the dog-human relationship dates back through 10,000 years
of domestication. The dog's ancestor was originally a social
species (see, e.g., Serpell, 1995). Dogs were selectively bred
not only for "sociocognitive abilities" and for "attachment
to humans" (Millot, 1994), but also for "infantile" features
(Coppinger et al., 1987). The
relationship is in
some sense similar to the parent-child relationship (i.e.,
asymmetrical and dependency based), and people are apt to
consider their dog as a child substitute (Collis, 1995). As
attachment theory presumes, "Dogs seem to miss their
owners during an absence and will appear excited upon the
owner's return" (New, 1995, p. 25). Dogs seem to be
innately responsive to humans so that not even strict
punishment conditions can extinguish the proximity seeking
of pups to a handler (Fisher, 1955). The dog's orientation to
humans is also supported by Pettijohn, Wont, Ebert, and
Scott (1977), who found that separation distress in puppies
was greatly reduced when a human being was nearby.
On the other hand, because the evidence for using a
mother as a secure base is poor for dogs (Elliot & Scott,
1961; Frederickson, 1952; Ross, Scott, Cherner, & Denen-
berg, 1960) and the puppies' separation distress can be
reduced also by nonconspecifics (Cairns & Werboff, 1967;
Pettijohn et al., 1977), attachment behavior in dogs has been
questioned by some researchers (Rajecki et al., 1978).
However, it is important to distinguish the "attachment"
behavior of puppies that still have not developed individual
relationships with their owners from the attachment of adult
dogs that have more individualized bonds to their human
masters. The calming effect that the presence of a human has
on puppies might simply be the result of general distress
reduction by a supposed conspecific and thus has no direct
relation to a dog's later attachment to a person.
most researchers assume that the accep-
tance of human beings as conspecifics and the motherlike
and effective security-providing role of humans for dis-
tressed puppies are results of domestication. For 10,000
years, artificial selection in dogs favored socialization with
humans as if they were conspecifics (Kretchmer & Fox,
1975). The genetic changes regarding a dog's capacity for
conspecific recognition might have played a key role in this
process that resulted in a preference for humans. A dog's
preparedness for forming a bond with humans, a bond that is
rooted in the evolutionary past, would be a prerequisite for
the development of attachment between a particular person
(the owner) and the dog itself.
Apart from some questionnaire studies that addressed the
psychological features of the dog-human bond (e.g., Barker
& Barker, 1988; Cox, 1993; Voith, Wright, & Danneman,
1992), to date there have been no experimental studies that
aimed to give a behavioral description of dog-human attach-
ment. We may suppose that for dogs in strange situations, as for
children, it is not only the separation from the attachment figure
(owner) but the reunion with her or him that activates the dogs'
attachment behavior. So the application of Ainsworth's
(1969) Strange Situation Test could provide useful informa-
tion regarding the owner-dog relationship. Additionally, by
using analogous methods, researchers can see how the
similarity of observed behavioral phenomena and the organi-
zational system of owner-dog interactions could provide a
useful model for human infant attachment.
The purposes of this article are to demonstrate adult dogs'
attachment behavior toward humans, to describe the human-
dog relationship by an ethological method used for evalua-
attachment, and to study the similarity
of owner-dog relationships in the form of mother-child
Fifty-one owner-dog pairs volunteered for our experiment from
kennel clubs in the vicinity. In the group of owners, there were 31
women and 20 men whose ages ranged from 13 to 60 years
8.5 years). In the group of dogs
there were 28 males and 23 females whose ages ranged from 1 to
= 3.12 ± 0.40 years). The dogs were from 20 different
pure breeds: Belgian Shepherd
= 17), English Setter
= 3), Staffordshire Terrier (n = 2), Hungarian
(n = 2), Irish
(n = 2), German Shepherd (n = 2),
Briard (n = 1), Bobtail (n = 1), Caucasian Shepherd (n = 1),
Czech Wolf (n = 1), Doberman Pinscher (n = 1), Great Dane
(n = 1), Husky (n = 1), Irish Setter (n = 1), Giant Schnauzer
(n = 1), Golden
(n = 1),
(n = 1), Newfound-
land (n = 1), and Spaniel (n = 1). There were also mixed-breed
dogs in our sample
= 7). According to the American Kennel
Club's (A.K.C.'s) classification, the dogs fit into five groups:
= 11), nonsporting dogs (n = 3), working dogs
(n = 27),
= 3), and mixed breeds
= 7). (The A.K.C.
has divided recognized breeds into six main groups on the basis of
behavioral characteristics rather than phylogeny.)
The basic experimental setup and the protocol were as similar as
possible to that of Ainsworth et al. (1978). The novel environment
was a relatively empty rectangular room (6 m longX
2.5 m high) containing two chairs. At one end of the room
(opposite the door), there were toys for dogs on the floor. The
14.5-min procedure consisted of an introductory episode and seven
experimental episodes. The behavior of the dogs was videotaped
and analyzed later.
Experimental Episodes of the Strange Situation
Introductory episode (30 s).
The observer introduces the owner
and dog to the experimental room and leaves.
Episode 1 (2 min): owner and dog.
The owner is a nonpartici-
pant while the dog explores. After 1.5 min, a signal (a knock on the
wall) is given to the owner who stimulates play.
Episode 2 (2 min): stranger, owner, and dog.
A stranger enters
and sits down. After 30 s, she initiates conversation with the owner.
At the 2nd-min mark, the stranger approaches the dog and tries to
stimulate playing. At the end of this episode, the owner leaves as
unobtrusively as possible, but the dog's leash remains on the chair.
Episode 3 (2 min): stranger and dog.
This is the first separa-
tion episode. The stranger's behavior is geared to that of the dog.
During the 1st min, the stranger tries to engage the dog and keep
him or her out of the door by playing. If the dog is not ready to play,
the stranger tries to engage the dog by petting. At the 2nd
min-mark, the stranger stops playing. If the dog initiates petting, it
Episode 4 (2 min): owner and dog.
This is the first reunion
episode. The owner approaches the closed door and calls the dog.
The owner opens the door and pauses a moment to allow the dog to
respond. The owner greets and comforts the dog. Meanwhile, the
stranger leaves. After 2 min, the owner leaves and says to the dog
"stay here." The leash is left on the chair.
Episode 5 (2 min): dog alone.
This is the second separation
Episode 6 (2 min): stranger and dog.
This is a continuation of
the second separation. The stranger enters and gears her behavior to
that of the dog. During the 1 st min, the stranger tries to engage the
dog and keep him or her out of the door by playing. If the dog is not
ready to play, the stranger tries to engage the dog by petting. At the
2nd min-mark, the stranger stops playing. Petting is permitted if it
is initiated by the dog.
Episode 7 (2 min): owner and dog.
This is the second reunion
episode. The owner opens the door and pauses a moment before
greeting the dog, giving him or her an opportunity to respond
Then the owner greets and comforts the dog.
Meanwhile, the stranger leaves.
To conduct the Strange Situation Test in a standard manner, we
gave several instructions (see Appendix) to the stranger (who was
the same woman in all cases). The owners did not know anything
about the real goals and the hypotheses of the study in advance;
they were informed that this study was to examine the exploratory
behavior of the dogs in a strange situation.
Observations and Behavior Categories
Two trained observers analyzed the 51 videotaped sessions using
eight behavior categories. Each behavior listed was scored for both
owner and stranger. Recorded variables were as follows: explora-
tion in the presence of the owner
and in the presence of the
stranger (EXPS), playing in the presence of the owner (PLYO) and
in the presence of the stranger (PLYS), passive behaviors in the
presence of the owner (PASO) and in the presence of the stranger
(PASS), physical contact with the owner (CONTO) and with the
stranger (CONTS), and standing by the door in the presence of the
owner (SBYO) and in the presence of the stranger (SBYS). The
relative percentage of the time spent with these behaviors was
established, and the relative duration of each behavioral variable
was summed across Episodes 1-7 for the statistical analysis.
We also analyzed the greeting behavior of the dogs toward the
owners during the reunion episodes (Episodes 4 and 7) and toward
ATTACHMENT BEHAVIOR IN DOGS
the entering stranger (Episodes 2 and 6). Greeting was character-
ized by proximity of, contact seeking by, and contact maintenance
of the dogs toward the entering owner (COSO, DCONTO, and
DELO) and toward the stranger (COSS, DCONTS, and DELS).
Interobserver agreement was assessed by means of parallel
coding of 20% of the total sample (10 strange situation sessions).
Behavior was point sampled every 10 s (for assessing confidence
for Elements 1-5), and the greeting episodes were evaluated
separately for assessing confidence. We assessed agreement in two
ways: percentage agreement and Cohen's kappa, a statistic that
corrects for chance agreement (Martin & Bateson, 1986). The
descriptions for behavior categories and the kappa and percentage
scores are given in Table 1.
Analysis of Data
We recorded behavioral data continuously during observations,
and we calculated the relative percentage of the time spent in each
behavior. Although some of the variables had a normal distribution,
in eight cases the transformation of raw data was necessary for
methods. To achieve normality, we had to
perform a square-root transformation for PLYS, PASO, PASS,
CONTS, SBYO, DELO, and DCONTS and a log transformation
The behavior that the dogs exhibited in the presence of owner
and the stranger was compared using two-tailed
tests. We studied
the correlation pattern of the dog's behavior in the strange situation
using a factor analysis, which was also used to get theoretical
dimensions (superordinate variables) thought to account for indi-
vidual differences in a set of behaviors observed in the Strange
Situation Test (J. P Connel & Goldsmith, 1982).
We then reanalyzed all the behavioral variables by cluster
analysis to classify the individuals according to their strange
situation behavior and to establish categories for the dog-human
relationship. In previous human studies, researchers also explored
the quantitative consistency of the Ainsworth (1969) system using
cluster analysis, and they found that the traditional A, B, and C
classifications (see above) were more or less relevant (D. B.
Connel, 1976; Gardner & Thompson, 1983) to the distinct clusters.
Nevertheless, the categorization along the secure versus insecure
dimension has seemingly more predictive power than the A, B, and
C groups (Arend, Gove, & Sroufe, 1979; Matas, Arend, & Sroufe,
1978; Waters, Wippman, & Sroufe, 1979).
we analyzed the effects of independent variables (i.e.,
the owner's sex and the dog's gender and breed) on the strange
situation behavior using three-way analyses of variance (ANOVAs)
and correlation analysis (the dog's age and the number of family
members). We also analysed the effect of breed differences on the
strange situation behavior using an ANOVA and an F test for
equality of variances on a matched-pair sample (a homogeneous
and a heterogeneous subgroup of dogs).
Dogs' Behavior in the Presence of the Owners Versus
The dogs tended to play more (PLYO vs. PLYS:
t = 5.4, p < 0.0001)
and spent more time exploring
t = 2.5,p =0.013) in the presence of
their owners (Figure 1). Passive behaviors and physical
contact did not show significant differences (PASO vs.
PASS: t = 1.5, p = ns;
CONTO vs. CONTS:
t = 1.5,
Behavioral Variables Observed in the Strange Situation Test
= exploration in the presence of owner; EXPS = exploration in the presence of
stranger; PLYO = playing with owner present; PLYS = playing with stranger present; PASO =
passive behavior in presence of owner; PASS = passive behavior in presence of stranger; CONTO =
physical contact with owner; CONTS = physical contact with stranger; SBYO = standing by door
with owner present; SBYS = standing by door with stranger present; COSO = contact seeking with
entering owner; COSS = contact seeking with entering stranger; DELO = delay of contact seeking
with owner; DELS = delay of contact seeking with stranger; DCONTO = duration of physical
contact while greeting entering owner; DCONTS = duration of physical contact while greeting
p = ns,
respectively). During the separation episodes (i.e.,
owner absent), the dogs stood by the door more than when
the owner was present (SBYS vs. SBYO:
t = 10.7,
p < 0.0001). Furthermore, dogs showed higher levels of
contact seeking toward the entering owner compared with
(COSO vs. COSS: t = 7.2, p < 0.0001).
the case of the former, we also noticed a shorter delay of
contact seeking (DELO vs. DELS:
t = 7.0, p < 0.0001)
Relative duration of playing (left panel) and exploration (right panel) behaviors in the
presence of the owner versus the stranger. *p < .02.
Exploration: activity directed toward nonmovable aspects of the envi-
ronment, including sniffing, distal visual inspection
ning), close visual inspection, or oral examination;
Playing: any vigorous, toy- or social partner-related behavior,
including running, jumping, or any physical contact with toys (chew-
ing, biting); PLYO and PLYS.
Passive behaviors: sitting, standing, or lying down without any orienta-
tion toward the environment; PASO and PASS.
Physical contact; CONTO and CONTS.
Stand by the door: the time spent close to the door (<1 m) with the
face oriented to the exit; SBYO and SBYS.
The score of contact seeking; that is, the sum of the following scores:
approach initiation (+1); full approach, characterized by physical con-
tact (+2); any sign of avoidance behavior (-1); COSO and COSS.
Delay of contact seeking: the amount of time (in s) from the moment
of the opening of the door to the first sign of approaching behavior;
DELO and DELS. (If approach was not recorded, DELO or DELS was
considered to be the duration of full episode, or 120 s.)
Duration of physical contact while greeting; DCONTO and DCONTS.
ATTACHMENT BEHAVIOR IN DOGS
and a longer duration of first physical contact
DCONTS: t(50) = 5.9, p < 0.0001; see Figure 2).
Factor analysis performed on 14 behavioral variables
(PASO, PASS, PLYO, PLYS, CONTO, CONTS,
SBYS, COSO, COSS, DELO, DELS, DCONTO,
DCONTS) resulted in three rotated factors (varimax rota-
tion, eigenvalue > 1.5) that accounted for 57% of the total
variability (29%, 15%, and 13% respectively). For the first
factor, five behavioral variables were represented by high
loadings (> 0.55). Individuals that scored high on this factor
behaved passively (i.e., they did not play and spent long
amounts of time exhibiting passive behaviors in the presence
of the stranger compared with the owner) and strove for
physical contact with the owner. So, this factor related to the
stressfulness of the strange situation and can be referred to as
Because the second factor is character-
ized by long-lasting physical contact with the stranger
and by a high level of contact
seeking toward the entering stranger (COSS), it can be
Scores of contact seeking toward the entering owner versus the stranger (upper right
panel) and relative duration (in seconds) of different behaviors: stand by the door in the presence of
the owner versus the stranger (upper left panel), delay of contact seeking toward the entering owner
versus the stranger (lower left panel), and physical contact with the owner versus the stranger while
greeting (lower right panel). *p < .001.
referred to as the
Acceptance of the Presence of the Stranger.
The third one is characterized by the high level of contact
seeking toward the owner with low (if any) delay of
approach and by permanent greeting contact with the
entering owner (COSO, DELO, and DCONTO). So this
factor is related to the
relationship and can be
referred to as the factor
analysis are given in Table 2.
Using the same variables as the factor analysis, we
calculated a hierarchical cluster analysis. Visual examina-
(Figure 3) revealed that the dogs
could be divided into three separate groups, and Group 1 and
Group 3 each consisted
2 subgroups (la, lb and 3a, 3b).
dogs in each group was as follows: la
(n = 5),
(n = 9),
(n = 7).
dogs into these groups was further
supported by the results
post hoc ANOVA tests on the
behavioral variables using the groups as independent vari-
We found significant differences among the groups in
Factor Loadings of Behavioral Variables
playing with owner present;
with stranger present;
passive behavior in presence of
passive behavior in presence of stranger; CONTO =
physical contact with owner; CONTS = physical contact with
stranger; SBYO = standing by door with owner present; SBYS =
standing by door with owner absent; COSO = contact seeking with
entering owner; COSS = contact seeking with entering stranger;
DELO = delay of contact seeking with owner; DELS = delay of
contact seeking with stranger; DCONTO = duration of physical
contact while greeting entering owner; DCONTS = duration of
physical contact while greeting entering stranger. Boldface indi-
cates behavioral variables with high loadings
on the three
The result of the cluster analysis for Groups la, lb,
3a, and 3b. * = dogs of homogeneous group (Belgian Shepherds);
x = dogs of heterogeneous group.
the 16 observed behavioral variables. Furthermore, the
post hoc Duncan multiple-range test showed significant
ranges for a given dependent variable (this result revealed a
characteristic behavior pattern for the groups). According to
this latter analysis, the level
any behavioral variable could be
low (L), medium (M), or high (H) in a given group (Table 3).
Because individual factor scores were also calculated, we
used them for comparison
cluster analysis groups. This
comparison resulted in significant differences for all three
factorial variables (see, Table 4 and Figure 4), providing
further evidence for the homogeneity
the dogs within a
Independent Variables on Strange Situation
We analyzed the observed behavioral variables and the
three factor scores
the dogs by a three-way ANOVA using
the independent variables recorded in this sample (i.e., the
owner's sex, and the dog's gender and breed), but they had
no significant effect on these variables.
Stand by the door
Score of contact seeking
Delay of contact seeking
Duration of physical contact
The effects of the dog's age and the number of family
members on the strange situation behavior were analyzed by
correlation analysis. Only the number of family members
correlated significantly with some of the variables (SBYS:
-0.38, PASS: 0.43, p < 0.01
in each). That is, dogs living in
large families tended to spend less time close to the door and
showed more passive behaviors in the presence of the
The Effect of Breed Differences on Strange
To study the effect of breed on the dog's behavior, we
selected two subgroups from the
subjects. Subgroup 1
was a homogeneous group; it included the 17 Belgian
was a heterogeneous group; it
included 17 dogs of different breeds (a German Shepherd, an
Irish Setter, a Giant Schnauzer, a Briard, a Staffordshire
Golden Retriever, a Newfoundland, an Irish
Wolfhound, a Caucasian Shepherd, a Laika, a Hungarian
a Bobtail, and
dogs of mixed breed). The two
groups were matched regarding the owners' sex and the sex
of the dogs. Subgroups 1 and
were also balanced for all of
the other independent variables except for breed. When we
Behavior Patterns of the Five Cluster Groups
ATTACHMENT BEHAVIOR IN DOGS
Dashes indicate that these variables cannot be divided into
statistically different levels.
= exploration in the presence of
owner; EXPS = exploration in the presence of stranger; PLYO =
playing with owner present; PLYS = playing with stranger present;
PASO = passive behavior in presence of owner; PASS = passive
behavior in presence of stranger; CONTO = physical contact with
owner; CONTS = physical contact with stranger; SBYS =
standing by door with owner absent; COSO = contact seeking with
entering owner; SBYO = standing by door with owner present;
COSS = contact seeking with entering stranger; DELO = delay of
contact seeking with owner; DELS = delay of contact seeking with
stranger; DCONTO = duration of physical contact while greeting
entering owner; DCONTS = duration of physical contact while
greeting entering stranger. L = low;
= medium; and H
value of a given variable, which are significant ranges established
by post hoc Duncan range test. ANOVA = analysis of variance.
Factorial Patterns of the Five Cluster Groups
L = low; M
= medium; and
= high value of a given
which are significant ranges established by post hoc
Duncan range test. ANOVA = analysis of variance.
compared Subgroups 1 and
neither the mean ages,
F(1, 33) = 1.18, ns,
nor the average number of
5.3, F(1, 33) = 0.22, ns,
We analyzed the effect of breed differences on the strange
behavior of the dogs using an ANOVA that
compared the homogeneous and the heterogeneous groups.
variables differed significantly; that is, the
Belgian Shepherds had a lower level of contact seeking
toward the entering stranger, COSS:
F(1, 50) = 7.40, p =
01, and they spent more time in close proximity to their
F(1, 50) = 10.40, p < .01.
means of the dogs' individual factor scores, there were no
significant differences between the Belgian Shepherds and
the mixed group.
Nevertheless, we decided that comparing the variance
scores of the observed variables might demonstrate more
clearly the effect of breed differences on the strange situation
behavior. To determine whether Belgian shepherds are a
more homogeneous group than the mixed group, we used the
test of homogeneity of variance. With the exception of
F(16,16) = 3.78, p < .001,
there were no significant
differences in variances of the behavioral variables and the
values are as follows (p >
PASO = 1.42, PASS = 1.42, CONTO = 0.16,
1.77, COSO =
COSS = 1.33,
The aim of this study was to investigate the human-dog
relationship by means of Ainsworth's
(1969) Strange Situa-
tion Test. As the results show, the experimental conditions of
the test proved to be effective in activating the attachment
behavior of owner-dog dyads, despite the fact that our
participants were all physiologically adults, and attachment
behavior is usually regarded as a feature of childhood, as a
part of parent-offspring interactions. The observed attach-
ment behavior of adult dogs toward owners is presumably
the result of 10,000 years of domestication. During this time,
dogs' dependency was increased by artificial selection, and
thus long-lasting, caregiver-receiver relations between hu-
mans and dogs could be formed by way of socialization
L L L
EXPS L L
CSÁNYI, AND D
Means of individual factor scores of the Anxiety (top panel), Acceptance (middle panel),
and Attachment (bottom panel) factors in the five cluster groups.
during an individual's life. Another possible explanation for
the attachment behavior could be a tendency on the part of
human breeders to select dogs that behave in social situa-
tions similarly to humans, especially children. The result of
such a process is the domesticated dog that simulates many
human (infant) behavior patterns such as attachment.
As was the case with children and human-reared chimpan-
zees (Bard, 1991), the observed behavioral changes in
owner-dog dyads fulfilled the operational criteria of attach-
ment (Gewirtz, 1972; Rajecki et al., 1978). A dog's use of
the owner as a secure base and its specific reaction to
separation from and reunion with the owner are clear
manifestations of the attachment of dogs to
This secure-base effect was revealed by the dogs' in-
creased exploration and by more frequent playing in the
presence of the owners (caregivers), just as in children and
chimpanzees. In the separation episodes, dogs stood at the
door for considerable lengths of time; the fact that this
behavior was not reduced by the presence of the stranger
suggests dogs' strong preference for their owners in stress
This reaction seems to be analogous to the
"searching response" described in young children (Ains-
worth, 1969), monkeys (Kaufmann & Rosenblum, 1969),
and chimpanzees (Bard, 1991) that was interpreted as an
effort to maintain the attachment-comfort bond (Gewirtz,
1961; Rheingold, 1961; Walters & Parke, 1964). The dogs'
specific reaction to reunion with the owner was active and
mmediate contact seeking and a tendency toward contact
maintenance for the returning owner (COSO, DELO, and
DCONTO) but not the stranger.
None of the independent variables (age, sex of humans,
and gender of dogs) seems to account for the dogs'
considerable behavioral variability in the Strange Situation
However, dogs living in large families exhibited less
proximity-seeking behavior toward the owner (SBYS) and
tended to behave more passively (PASS) in the situation.
This effect can be attributed to the differences in socializa-
tion of particular dogs because in larger families, pets
probably form multiple attachments to some members of the
On this basis, these dogs will show less clinging
behavior toward the owner.
Moreover, the breed-specific differences, usually re-
garded as a major source of behavioral variability in dogs
(Scott & Fuller, 1965), also had only a slight effect on the
behavior of the dogs in the test. The comparison of mixed
breed (heterogeneous) versus one purebred (homogeneous)
group (Belgian Shepherds) showed only small differences.
Because cluster analysis did not rank the individuals of
different breeds into groups of distinct categories, perhaps
the variability of the attachment behavior among breeds is
However, with regard to the behavioral parameters
used in this study, one breed might be found to show a
different kind of attachment compared with other breeds.
This possibility underlines the necessity of the use of a
multivariate approach because it reduces the effect of
existing breed differences on the evaluation of attachment.
As was found in other studies of applied factor analysis
P Connel & Goldsmith, 1982; Miller et al., 1990), the
strange situation behavior was influenced by different vari-
ables, including the dogs' reaction to a separation from the
owner, the unfamiliar environment that was more or less
stressful for dogs, and the dogs' responsiveness to the
stranger. The results of our factor analysis support the notion
that the strange situation behavior could be explained by
major hypothetical variables, including the dogs'
reaction to separation from the owner (Factor 3: Attach-
ment), the unfamiliar environment (Factor 1: Anxiety) that
was more or less stressful for dogs, and the dogs' responsive-
ness to the stranger (Factor 2: Acceptance). In contrast with
earlier attempts to validate traditional A, B, or C categoriza-
tion (avoidant, secure, or resistant) by post hoc cluster
analysis (D. B. Connel, 1976; Gardner & Thompson, 1983),
we used this multivariate analysis to establish categories of
dog-human relationships. The results of the exploratory
cluster analysis showed that dogs could be separated into at
least three or at most five major groups (see Figure 3).
ATTACHMENT BEHAVIOR IN DOGS
Although these groups differed with respect to many behav-
ioral variables, the difference in the factorial variables was
more pronounced. Starting from the principle that factorial
variables can be divided into three statistically distinct
categories (L, M, or H) we found five distinguishable groups
that differed by at least one factorial variable from each
other. Still, our findings support the view of J. P Connell and
Goldsmith (1982) that group classification is best perceived
as representing an underlying continuum in three dimensions.
Dogs that belonged to Cluster Group 1 were characterized
by low anxiety in the stressful situation, but they differed in
the interrelationship between their acceptance and attach-
ment. The low levels of attachment to the owners in Group
1 a contrast with their medium level of acceptance of the
stranger. Dogs in Group lb showed higher levels of attach-
ment to the owner than acceptance of the stranger, suggest-
ing that the former was more preferred for reducing stress in
these dogs. Although one might suspect there was a differ-
ence in socialization of the dogs in the two subgroups, at the
present time we have no evidence for this.
Cluster Group 2 was characterized by high anxiety related
to a stressful situation, high levels of acceptance regarding
the separation, and high levels of contact seeking and
contact maintenance toward the entering owner. These dogs
seemed not to distinguish between owner and stranger,
which could have been the result of their marked stress
during separation episodes. In sum, these dogs expressed the
most extreme behavior of all subjects observed.
Cluster Group 3 was characterized by medium levels of
anxiety and acceptance. There was a clear separation in two
subgroups (3a and 3b) of the level of attachment to the
Whereas dogs in Group 3a showed low levels of
attachment toward the owner, Group 3b was characterized
by significantly higher attachment behavior in this situation.
One might suppose well-socialized dogs should belong into
this latter group (3b). We suspect that the dogs of Group 3a
tended to avoid close -contact with the owner and the
stranger, which was a more effective strategy for them in
reducing stress during separation episodes.
Regarding the total sample, all of the five cluster groups
and subgroups are represented by a considerable proportion
of subjects (10%-35%). In other words, all these groups
represent a relatively common form of attachment behavior,
but the level of expression was influenced by the stressful-
ness of the strange situation.
Dogs in Group lb seemed to attach only to the owner.
Dogs of Groups 2 and 3b both seemed to bond to humans,
not just to a particular person (medium and high scores of
Acceptance and Attachment). However, they differed in
sensitivity to stress, and thus their reactions toward the
owner also differed: The extremely anxious dogs (Group 2)
showed more expressed attachment behavior and acceptance
than the less anxious ones (Group 3b).
Dogs in two groups (la and 3a) avoided the owner (low
Attachment scores). The difference between them was the
level of anxiety. Compared with Group 2, the relative
owner-avoidant behavior of nonanxious Group l a does not
unconditionally mean a lack of attachment on the part of
these dogs because we would not expect the activation of
TOPÁL, MIKLÓSI, CSÁNYI, AND
attachment behavior under a condition
Group 3a that were more susceptible to stress, but in
this fact behaved in a very avoidant manner, can be
referred to as the nonattached group
our sample. The
moderately anxious dogs
Group 3a seemed to bond to
humans in general but not to the owner.
our subjects were more or less socialized
dogs living in families, we did not find dogs that reject
humans in general (low Attachment,
suppose that extremely unsocialized dogs (feral dogs or
some dogs from shelters) would prove to be more avoidant
these categories gives us an impor-
tant tool for examining the effect
the genetic influences (breed differences) on the attachment
dogs to owners. Furthermore, we are now able to measure
the temporal stability
relationship and its
development. All these issues should be the targets for future
Ainsworth, M. D. S. (1969). Object relations, dependency and
attachment: A theoretical review of the infant-mother relation-
Child Development, 40,
M. D. S. (1972). Attachment and dependency: A
comparison. In J. L. Gewirtz (Ed.),
Attachment and dependency
(pp. 97-137). Washington DC: Winston.
Ainsworth, M. D. S., Blehar, M. C., Waters, E., & Wall, S. (1978).
Patterns of attachment: A psychological study of the strange
Hillsdale, NJ: Erlbaum.
Ainsworth, M. D. S., & Wittig, B. A. (1969). Attachment and
exploratory behavior of one-year olds in a strange situation. In
Determinants of infant behavior (Vol.
111-136). London: Methuen.
Arend, R., Gove, F. L., & Sroufe, L. A. (1979). Continuity of
individual adaptation from infancy to kindergarten: A predictive
study of ego-resiliency and curiosity in preschoolers.
K. A. (1983). The effect of peer separation in young
(Pan troglodytes). American Journal of Primatol-
Bard, K. A. (1991). Distribution of attachment classifications in
American Journal of Primatology, 24,
Barker, S. B., & Barker, R. T. (1988). The
Closer than family ties?
Journal of Mental Health Counseling,
Bowlby, J. (1958). The nature of the child's tie to his mother.
International Journal of Psychoanalysis, 39,
Bowlby, J. (1969).
Attachment and loss: Vol. I. Attachment.
York: Basic Books.
Bowlby, J. (1972).
Middlesex, England: Penguin
Cairns, R. B. (1966). Attachment behavior of mammals.
cal Review, 73,
Cairns, R. B., & Werboff, J. (1967, November 24). Behavior
development in the dog: An interspecific analysis.
Cohen, L. J. (1974). The operational definition of human attach-
Psychological Bulletin, 81,
Collis, G. M. (1995, September).
Health benefits of pet ownership:
Attachment vs. psychological support.
Paper presented at
mals, health and quality of
Seventh International Confer-
ence on Human Animal Interactions,
Connel, D. B. (1976).
Individual differences in infant attachment:
An investigation into stability, implications and relationships to
structure of early language development.
dissertation, Syracuse University, Syracuse, NY.
Connel, J. P., & Goldsmith, H. H. (1982). A structural modeling
approach to the study of attachment and strange situation
behaviors. In R. N. Emde & R. J. Harmon (Eds.),
ment of attachment and
York: Plenum Press.
Coppinger, R. J., Glendinning, E., Torop, C., Matthay, C., Suther-
M., & Smith, C. (1987). Degree of behavioral neoteny
differentiates canid polymorphs.
Cox, R. P. (1993). The human/animal bond as a correlate of family
functioning.Clinical Nursing Research, 2,
Cynic, L. S., Reite, M. L., & Shucard, D. W. (1982). Animal models
of human behavior: Their application to the study of attachment.
In R. N. Emde & R. J. Harmon (Eds.),
The development of
attachment and affiliative systems (pp.
& Scott, J. P (1961). The development of emotional
distress reactions to separation in puppies.
Journal of Genetic
Fisher, A. E. (1955).
The effects of differential early treatment on
the social and exploratory behavior of puppies.
doctoral dissertation, Pennsylvania State University, University
E. (1952). Perceptual homeostasis and distress
vocalization in the puppy.
Journal of Personality, 20,
Freud, A. (1946). The psychoanalytic study of infantile feeding
Psychoanalytic Study of the Child, 2,
Gardner, W. P., & Thompson, L. A. (1983, June).
A cluster analytic
evaluation of the strange situation classification system.
presented at the meeting of the Society for Research in Child
Development, Detroit, MI.
Gewirtz, J. L. (1961).
analysis of the effects of normal
stimulation, privation and deprivation on the acquisition of
social motivation and attachment. In B. M.
nants of infant behavior (pp.
213-299). London: Methuen.
Gewirtz, J. L. (1972). Attachment, dependence, and a distinction in
terms of stimulus control. In J. L. Gewirtz (Ed.),
139-177). Washington, DC: Winston.
Gubemick, D. J. (1981). Parent and infant attachment in mammals.
In D. J.
& P H. Klopfer (Eds.),
Parental care in
243-300). London: Plenum Press.
Hoffman, H. S., & Ratner, A. M. (1973). A reinforcement model of
mprinting: Implications for socialization in monkeys and men.
Psychological Review, 80,
Kaufmann, I. C., & Rosenblum, L. A. (1969). Effects of separation
from mother on the emotional behavior of infant monkeys.
Annals of the New York Academy of Sciences, 159,
Kraemer, G. W (1992). A psychobiological theory of attachment.
Behavioral Brain and Sciences, 15,
Kretchmer, K. R., & Fox, M. W. (1975). Effects of domestication
on animal behavior.
The Veterinary Record, 96,
Lorenz, K. (1966).
New York: Harcourt, Brace &
Main, M., & Solomon, J. (1990). Procedures for identifying infants
as disorganized/disoriented during the Ainsworth Strange Situa-
tion. In M. T. Greenberg, D. Cicchetti, and E. M. Cummings
Attachment in the preschool years: Theory, research and
121-160). Chicago: University of Chicago
Martin, P, & Bateson, P. (1986).
Measuring behavior. Cambridge,
England: Cambridge University Press.
Matas, L., Arend, R. A., & Sroufe, L. A. (1978). Continuity of
adaptation in second year: The relationship between quality of
attachment and later competence.
Child Development, 49,
Miller, L. C., Bard, K. A., Juno, C. J., & Nadler, R. D. (1986).
Behavioral responsiveness of young chimpanzees
to a novel environment.
Folia Primatologica, 47,
Miller, L. C., Bard, K. A., Juno, C. J., & Nadler, R. D. (1990).
Behavioral responsiveness to strangers in young chimpanzees
(Pan troglodytes). Folia Primatologica, 55,
Millot, J. L. (1994). Olfactory and visual cues in the interaction
systems between dogs and children.
Behavioural Process, 33,
New, J. C., (1995, September). Quality of life of companion
Animals, health and quality of life: Seventh Interna-
tional Conference on Human Animal Interactions (pp.
Pettijohn, T. F.,
Wont, T. W., Ebert, P D., & Scott, J. P. (1977).
Alleviation of separation distress in 3 breeds of young dogs.
Developmental Psychobiology, 10,
Rajecki, D. W., Lamb, M. E., & Obmascher, P. (1978). Toward a
general theory of infantile attachment: A comparative review of
aspects of the social bond.
Behavioral and Brain Sciences, 3,
Rheingold, H. L. (1961). The effect of environmental stimulation
upon social and exploratory behavior in the human infant. In
In leaving during reunion episodes, the stranger must be
unobtrusive and never interfere with the reunion (i.e., say nothing
to the owner or dog, do not move between them, and leave quietly).
If necessary, the stranger can wait to exit.
2. The stranger should never position herself between the dog
and the owner, especially during reunions.
3. The stranger should never sit in the owner's chair.
When playing, the stranger should take her cue from the dog
and do something similar.
5. In Episodes 3 and 6, if the dog is upset, the stranger should try
to reassure it by petting and then distract it with toys.
At the end of Episodes 3 and 6, the stranger should never be
ATTACHMENT BEHAVIOR IN DOGS
Instructions to the Stranger
Determinants of infant behaviour (pp.
143-171). London: Methuen.
Ross, S., Scott, J. P, Chemer, M., & Denenberg, V H. (1960).
Effects of restraint and isolation on yelping in puppies.
Scott, J. P., & Fuller, J. L. (1965).
Genetics and the social behavior
of the dog.
Chicago: University of Chicago Press.
Sears, R. R., Whiting, J. W. M., Novlis, V., & Sears, P S. (1953).
Some child rearing antecedents of aggression and dependency in
Genetic Psychology Monographs, 47, 135-247.
Serpell, J. (1995).
The domestic dog.
Cambridge, England: Cam-
bridge University Press.
Solomon, R. L., & Corbit, J. D. (1973). An opponent-process
theory of motivation II: Cigarette addiction. Journal of Abnor-
mal Psychology, 81,
Voith, V L., Wright, J. C., & Danneman, P J. (1992). Is there a
relationship between canine behavior problems and spoiling
activities, anthropomorphism, and obedience training?
Animal Behavior Science, 34,
Walters, R. H., & Parke, R. D. (1964). Social motivation, depen-
dency, and susceptibility to social influence. In L. Berkowitz
Advances in experimental social psychology (Vol. 1, pp.
232-276). New York: Academic Press.
Wippman, J., & Sroufe, L. A. (1979). Attachment,
positive affect, and competence in the peer group: Two studies in
Child Development, 50,
W. (1976). The ethological analysis of attachment:
motivational and sociophysiological aspects.
Zeitschrift für Tierpsychologie, 42, 12-28.
playing or interacting with the dog so that the dog is not distracted
when the owner returns.
7. In Episode 2, if the owner is not in her or his chair, the stranger
may remind the owner to move to the chair.
8. The stranger must learn to remain calm in the presence of very
distressed dogs and must not feel distressed if she cannot calm the
Received February 10, 1997
Revision received December 12, 1997
Accepted December 18, 1997 v