Is the dog-human relationship an attachment bond?

IS THE DOG–HUMAN RELATIONSHIP
AN ATTACHMENT BOND? AN OBSERVATIONAL STUDY
USING AINSWORTH’S STRANGE SITUATION
by
EMANUELA PRATO-PREVIDE 1,3) , DEBORAH MARY CUSTANCE 2) ,
CATERINA SPIEZIO 1) and FRANCESCA SABATINI1,4)
(1 Istituto di Psicologia, Università di Milano, Italy; 2 Goldsmith College, Department of
Psychology, University of London)
(Acc. 24-I-2003)
Summary
Ainsworth’s ‘strange situation’ procedure was used to investigate the dog (Canis familiaris)
– human relationship. 38 adult dog-owner pairs were observed in an unfamiliar room, introduced to a human stranger and subjected to four short episodes of separation. The procedure
and behavioural analyses were as similar as possible to those used in studying human infants,
except for the inclusion of an extra separation period in which the dogs were left alone in
the room with articles of clothing belonging to the owner and stranger. A secure base effect
was suggested by the fact that the dogs accepted to play with the stranger more in the presence of their owner than during his or her absence. They also explored more in the presence
of their owner, but this appeared to be due to diminishing curiosity over time rather than a
secure base effect. The dogs also exhibited a range of attachment behaviours, i.e. search and
proximity seeking behaviours when separated from their owner, including following, scratching and jumping up on the door, remaining oriented to the door or the owner’s empty chair
and vocalising. They also greeted their owner more enthusiastically and for longer durations
compared to the stranger. Finally, they contacted the owner’s clothing more often and for
3)
Corresponding authors address: Prof. E. Prato-Previde, Istituto di Psicologia, Università di
Milano, Via T. Pini 1, 20134 Milano, Italy, e-mail: [email protected]
4) This research was supported by funds from Università di Milano to Emanuela PratoPrevide. We are grateful to Marcello Cesa-Bianchi and Marco Poli for allowing us to carry
out the work in the Psychology Institute of Università di Milano. We thank Barbara Rotta for
her invaluable help in data collection and scoring, Clara Palestrini for helping in running the
experiment, Marco Colombetti for reading and commenting on the preliminary draft of the
paper. Finally, a special thank to Tipota, a female mongrel, for being our Ž rst pilot subject
and to all the owners and dogs that participated as volunteers.
© Koninklijke Brill NV, Leiden, 2003
Behaviour 140, 225-254
Also available online -
226
PRATO-PREVIDE, CUSTANCE, SPIEZIO & SABATINI
longer durations compared to the stranger’s clothing and spent more time next to the owner’s
chair when the owner’s objects were present. Hence, the dogs’ behaviour in the strange situation was very similar to that reported in human infants and chimpanzees. However, despite
conducting detailed behavioural analyses, the order effects inherent in Ainsworth’s procedure
prevented the study from providing conclusive evidence that the dog–human bond constitutes
an attachment.
Keywords: dog, attachment, affectional bonds, Ainsworth, Strange Situation.
In recent years, there has been increasing interest in human-dog relationships and literature on this topic has  ourished (e.g. Serpell, 1995;
Menanche, 1998). Most studies have either explored the issues concerning
attachment of humans to their pet dogs (Voith, 1985; Scott, 1992; Endenburg, 1995; Collis & McNicholas, 1998), the effect of dogs on their owner’s
well-being and health (Rowan & Beck, 1994; Hart, 1995; Wilson & Turner,
1998), or the relationship between the owner’s behaviour and their dogs’
behavioural problems (Voith et al., 1992; O’Farrell, 1995; Jagoe & Serpell,
1996). Only a few studies have investigated the nature of the dog’s affectional tie with its owner, and to our knowledge, the literature contains only
one empirical study on this topic (Topal et al., 1998).
The sparcity of empirical studies on the dog–human bond is surprising
given that dogs have been part of human society for longer than any other
domestic species (Clutton-Brock, 1999) and they are by far the most popular
species of companion animal (Hart, 1995). All modern dog breeds were domesticated as wolves over 10,000 years ago (Serpell, 1995; Clutton-Brock,
1999) and subsequently subjected to an intense process of artiŽ cial selection.
It has been suggested that artiŽ cial selection in dogs has not only favoured
speciŽ c anatomical and behavioural traits, but also promoted general socialisation (Kretchmer & Fox, 1975) and the predisposition to form attachments
with humans (Millott, 1994).
The owner-dog relationship resembles the parent-child bond in many
ways. People tend to view and treat their dogs as child-like and many behaviour patterns in dogs, like those of children, seem especially designed to
elicit care (Voith, 1985; Askew, 1996). Askew (1996) argued that our pets
display evolutionary modiŽ cations of ancestral behaviours that function to
elicit human parental care. In his opinion, the behaviour of modern pet owners towards their dogs not only resembles human parental behaviour, but
it actually is parental behaviour directed towards the members of another
species.
DOG–HUMAN ATTACHMENT
227
Among the various theories aimed to explain mother-infant attachment in
humans, the ethological-evolutionary perspective (Bowlby, 1958, 1969/1982;
Ainsworth, 1964, 1967, 1969) is probably most compatible with a comparative approach. Bowlby’s original theory was inspired by work on non-human
animals and hence it was rooted in biology with a concern for the adaptive value of attachment and other related behavioural systems (Bretherton,
1992). According to this view, social animals have evolved a predisposition
to form attachment bonds, especially infants with their mothers.
Attachment can be considered as a particular kind of affectional bond.
Throughout the lifespan, individuals can form a variety of different affectional bonds that are not attachments. Thus, in establishing whether a relationship constitutes an attachment, it is important to distinguish it from other
types of affectional bonds. All affectional bonds (including attachments) endure over time, involve a speciŽ c individual and are emotionally signiŽ cant;
furthermore, affectionally bonded individuals tend to maintain proximity and
contact and become distressed when involuntary separation occurs (Cassidy,
1999). “There is, however, one criterion of attachment that is not necessarily
present in other affectional bonds: the experience of security and comfort
obtained from the relationship with the partner, and yet the ability to move
off from the secure base provided by the partner, with conŽ dence to engage
in other activities” (Ainsworth, 1989, p. 711). Hence, in establishing whether
the dog–human relationship constitutes an attachment, one must demonstrate
the major elements common to all affectional bonds and in addition provide
unambiguous evidence of a secure base effect.
We used Ainsworth’s Strange Situation procedure to test the hypothesis
that the dog–human relationship constitutes an affectional bond that can be
characterised as an infantile-like attachment. The Strange Situation procedure involves conducting controlled observations of a subject’s response to
being placed in an unfamiliar room, introduced to an unfamiliar adult (the
stranger) and subjected to three short episodes of separation from the attachment Ž gure. Although devised to test human infants, the Strange Situation
procedure has also proved an effective tool for studying affectional bonds
and attachment in other species (Bard, 1991; Topal et al., 1998; M.A. Shunik, pers. comm.).
Under natural conditions, the safety of the infants of many species would
be greatly compromised by even short periods of separation from speciŽ c
family members, particularly the mother. Therefore, in a typical infantile
228
PRATO-PREVIDE, CUSTANCE, SPIEZIO & SABATINI
system of secure attachment, all incidences of separation from the attachment Ž gure (particularly in unfamiliar circumstances) would be expected to
cause distress in the infant and attempts to re-establish proximity. Indeed,
Ainsworth & Bell (1970) deŽ ned attachment behaviour as those behaviours
aimed at promoting and restoring proximity and contact such as following,
approaching, vocalising, clinging and crying. Although the presence of another individual with whom the infant has an affectional bond (i.e. a same
age play mate) may provide some comfort, the child will not usually be fully
comforted or secure until she or he is reunited with the attachment Ž gure.
It is only after reunion with the attachment Ž gure that one would expect the
infant to feel sufŽ ciently secure to engage in non-attachment related behaviours such as exploration or play.
Recently, Topal et al. (1998) used the Strange Situation procedure to explore the relationship between adult dogs and humans. They used a shortened
version of the test in which each episode lasted two minutes and collected
a relatively small number of behavioural variables to compare the dogs’ responses to their owner versus a stranger. They found that the dogs explored
and played more in the presence of the owner compared to the stranger.
Furthermore, the dogs greeted their owners more enthusiastically than the
stranger upon reunion episodes and stood by the door more when separated
from their owner versus the stranger. On the basis of these Ž ndings, Topal et
al. concluded that the dog-owner relationship conforms to an attachment.
Although Topal et al.’s study constitutes a valuable contribution to the
Ž eld, in our opinion their conclusions about the nature of the dog–human
bond are somewhat premature. We believe that Topal et al.’s basic behavioural analyses were not sufŽ ciently detailed to be able to distinguish attachment from a general affectional bond. Topal et al.’s Ž ndings clearly showed
that dogs exhibit a strong preference for their owner over a stranger. However, all affectional bonds involve exhibiting a preference for a speciŽ c individual, a Ž gure who is not interchangeable with anyone else (Ainsworth,
1989). To establish the presence of an attachment, one must focus much more
on the behavioural categories that indicate security-, proximity- and comfortseeking. Topal et al. collected almost no data that would indicate distress
or protest during separation, such as vocalising or searching. They reported
results from only one category of behaviour, standing by the door, which
might indicate searching. However, a whole suite of attachment and search
behaviours have been reported in children including crying, following the
DOG–HUMAN ATTACHMENT
229
mother to the door, standing by the door, orientation to the door from a distance, pounding upon the door, trying to open the door, going to the mother’s
empty chair or simply looking at it (Ainsworth & Bell, 1970; Ainsworth et
al., 1978). These behaviours convey information on the level of activation
of the attachment behavioural system and on the degree to which separation
causes the infants distress and activates a proximity seeking response.
According to Ainsworth (1989), the secure base effect is the primary
factor that distinguishes an attachment from other types of affectional bonds.
Thus, the most serious omission in Topal et al.’s analysis is the lack of
evidence relating to the secure base effect. The fact that the dogs played
more with their owner rather than the stranger does not necessarily indicate
a secure base effect. It could simply indicate a preference for the owner over
the stranger. Similarly, the fact that the dogs explored more in the presence
of the owner than the stranger could be due to an order effect in Ainsworth’s
procedure rather than a secure base effect. The dogs were left alone with the
owner for the Ž rst two minutes of the procedure. By the time the stranger
entered the room and particularly by the time the dogs were left alone with
the stranger, they had already been given plenty of opportunity to explore the
room and satisfy their curiosity. In their study, Topal et al. lumped together
all the data on the dogs’ behaviour in the presence of the owner versus
the stranger providing little information on how a given behaviour changed
from one episode to another. Such comparisons are extremely important for
discerning possible order effects.
The Strange Situation procedure provides three different means by which
one can identify a secure base effect. First, play and exploration in human
infants has been found to become depressed when they are in the presence
of just the stranger and when alone, but recovers after reunion with the
mother. Second, although human infants tend to explore the room and toys
to a consistently high degree across the Ž rst episode of the procedure, they
often cease and return immediately to their mother’s side upon the entrance
of the stranger. The presence of strangers is a source of some anxiety to
young infants and thus the mother provide a secure base from which to
view this potentially fear inducing stimulus. Third, some human infants are
sufŽ ciently conŽ dent when in the presence of their mother to engage in play
with the stranger. However, when their mother leaves, without the security
she provides, most infants refuse to continue playing. We investigated to
what extent adult dogs exhibit similar safe base effects.
230
PRATO-PREVIDE, CUSTANCE, SPIEZIO & SABATINI
The present study differed from Topal et al.’s work in a number of important respects. First, we conducted much more detailed behavioural analyses
that concentrated on many of the key behavioural measures for attachment
as used by Ainsworth & Bell (1970) and Ainsworth et al. (1978). Second,
the behaviour of the dogs in the presence of the stranger and when alone
was analysed separately and, as in studies on human infants, these qualitatively different conditions were compared. Third, the dogs’ behavioural
changes across episodes were evaluated. Fourth, we followed the traditional
procedure used in studies on human infants, which incorporate episodes of
3-minutes duration. Finally, we included an extra phase at the end of the standard Ainsworth procedure in which the owner and stranger left an article of
clothing and their shoes in the room while the dog was left alone for a further
three minutes. This extra episode aimed to test the degree to which the dogs
would orient to their owner’s objects and the degree to which they seemed
to use these objects as a source of comfort. We hypothesised that if the dogs
were attached to their owners, we would Ž nd evidence of secure base effects,
proximity seeking and maintenance, search behaviours and comfort seeking
that are functionally similar to those observed in human infants.
Method
Subjects
Thirty-eight volunteer owner-dog pairs formed a convenience sample, recruited both by personal contact and from advertisements distributed within the University of Milan. The dog
owners comprised 10 men and 28 women whose ages ranged from 18 to 74 years (mean D
38.8 years, 13.5 SD). 30 owners (78.95%) had previous experience of dog ownership and 25
(65.79%) owned only one dog at the time the study took place.
The sample of dogs consisted of 19 males and 19 females whose ages ranged from 1 to 10
years (mean D 4.51 years, 2.46 SD). Twenty dogs were pedigrees and 18 were mixed-breeds
or ‘mongrels’. Thirteen dogs (11 mongrels an 2 pure breed) had been adopted from rescue
centres and their age of adoption varied from less than 3 months (3 subjects) to 3-6 months
(3 subjects), 6-12 months (2 subjects), 1-4 years (5 subjects). The remaining 25 dogs had
been reared in the same family home from puppy-hood and their age at acquisition ranged
between 50 days and 3 months. The pedigree dogs included 2 German Shepherd Dogs, 3
Labrador Retrievers, 2 Siberian Huskies, 1 Fox Terrier, 1 Dachshund, 2 Boxers, 2 Shih Tzus,
3 Bernese Mountain Dogs, 1 Yorkshire Terrier, 1 Beagle, 1 German Wirehaired Pointer, and
1 Italian Hunting Hound.
All the dogs were kept exclusively for companionship and lived within the human household. None of the dogs were working dogs, guard dogs or highly trained beyond basic obedience. All the dogs were accustomed to being taken out of doors with their owner and encountering human strangers. As far as their owners were aware, none of the dogs had ever shown
DOG–HUMAN ATTACHMENT
231
aggression towards a human. All the owners reported that their dogs were used to staying
alone at home and could remain alone without problems.
Experimental setting and materials
The dogs were tested in a standardised and controlled ‘strange’ environment, as similar as
possible to that described by Ainsworth & Bell (1970). Figure 1 is an aerial plan of the
unfamiliar environment. It comprised a relatively bare room (3:00£ 5:30 m) at the Institute of
Psychology of the University of Milan. The room was equipped with two chairs (one labelled
as ‘stranger’ and the other as ‘owner’), a selection of dog toys (one large and one small ball,
a plastic bottle, a rope pull-toy, and a squeaky toy), a water bowl and two videocameras
(Sony Handycam Video HI 8). One of the two videocameras was connected with a monitor
positioned in an adjacent room, so that each session could be independently observed. To
minimise extraneous noise, testing was conducted each weekend between 1430 and 1830
hours, over a period of three months.
Procedure
Pre-experimental phase (approximately 10 minutes in duration): The owner and dog were
met and escorted to a waiting room. The procedure was brie y described to the owner and
the owner was asked to Ž ll in a questionnaire providing background information on his or
her dog. The speciŽ c goal of the study was not disclosed at this stage. Instead, the owner
was told that the aim of the study was to investigate the exploratory behaviour of dogs in
an unfamiliar environment. After gaining the owner’s permission to videorecord behaviour
during the experimental phase, the two video cameras were activated and both the dog and
owner were led to the experimental room.
Fig. 1.
Experimental setting. The experimental room was divided into 5 different areas:
door (I), bowl (II), owner (III), stranger (IV), and toys (V).
232
PRATO-PREVIDE, CUSTANCE, SPIEZIO & SABATINI
Experimental phase: The experimental protocol was adapted from Ainsworth & Bell
(1970). The procedure lasted approximately 27 minutes and consisted of an introductory
episode followed by 8 three-minute experimental episodes (Fig. 2, episodes 1-8). The introductory episode and the Ž rst seven experimental episodes were those typical of the Strange
Situation. An additional eighth episode was included to study the dogs’ response to the presence of objects or articles of clothing left behind by the stranger or owner. The episodes were
as follows:
Introductory episode: Owner, Dog and Experimenter (approximately 30 s). The owner was
guided to his or her chair, and asked to sit and complete the questionnaire. The experimenter
took the dog’s lead and left the room.
Episode 1: Owner and Dog (3 min). The owner sat quietly Ž lling in the questionnaire and,
as instructed, only interacted with the dog (i.e. by petting or talking to it) if it speciŽ cally
sought attention. The dog was free to explore the room.
Episode 2: Owner, Dog and Stranger (3 min). The stranger, who was always played by
the same woman and who had never met the dog before, entered the room and sat quietly
for 1 minute. She conversed with the owner for the second minute, and then during the third
minute she approached the dog and attempted to stimulate play by throwing a ball or offering
a pull-toy. At the end of this episode the owner left the room unobtrusively.
Episode 3: Stranger and Dog (3 min, 1st separation episode). The stranger continued to
play with the dog if it was willing; if it was inactive or distressed, the stranger attempted to
distract it with play or by providing verbal and tactile comfort.
Episode 4: Owner and Dog (3 min, 1st reunion episode). The owner entered the room,
paused momentarily after opening the door to allow the dog to initiate interaction and then,
as instructed, greeted and/or comforted the dog in the same manner as after returning from
work or shopping. The stranger quietly exited the room as soon as she was conŽ dent that
she would cause the least possible disruption to the dog and owner’s reunion. The owner had
been told that she or he was free to play with the dog throughout the episode. At the end of
the 3 minutes, a telephone in the room rang on a low volume alerting the owner to leave the
room with a verbal farewell.
Episode 5: Dog alone (3 min, 2nd separation episode). The dog was left alone for three
minutes, but was constantly observed by the owner and researchers on the monitor in the
adjacent room. The owner was instructed to remain silent throughout this period, but retained
the right to terminate the episode if she or he believed the dog was becoming too distressed
(none of the owners felt the need to execute this right).
Episode 6: Stranger and Dog (3 min, 3rd separation episode). The stranger entered the
room paused momentarily after opening the door to allow the dog to initiate interaction and
greeted the dog if it was willing by petting its head and back. She then followed the same
protocol as in episode 3. Toward the end of the episode, the stranger removed her shoes
and placed them near her chair, and also left an article of clothing (equivalent to the article
selected by the owner) on the chair before leaving.
Episode 7: Owner and Dog (3 min, 2nd reunion). The owner entered the room, paused
momentarily after opening the door, then greeted the dog as though he or she had just returned
from work or shopping. The stranger left the room unobtrusively. At the end of third minute,
the telephone rang again alerting the owner to remove his or her shoes and to place them near
his or her chair and to leave an article of clothing (such as a scarf) on the chair before exiting
the room with a verbal farewell.
DOG–HUMAN ATTACHMENT
Fig. 2.
The eight experimental episodes from camera 2.
233
234
PRATO-PREVIDE, CUSTANCE, SPIEZIO & SABATINI
TABLE 1. Behavioural categories recorded in the Strange Situation Procedure
Behavioural
category
DeŽ nition
Mutually exclusive categories
Exploration Activity directed toward physical aspects of the environment, including snifŽ ng, close visual inspection, distal visual inspection,and gentle oral examination such as licking
Passive
Sitting, standing or lying down without any obvious orienbehaviour
tation toward the physical or social environment
Locomotion Walking, pacing or running around without exploring the
environment or playing
Individual
Any vigorous or galloping gaited behaviour directed toplay
ward a toy when clearly not interacting with the owner
or stranger; including chewing, biting, shaking from side
to side, scratching or batting with the paw, chasing rolling
balls and tossing using the mouth
Social play
Any vigorous or galloping gaited behaviour performed
when interacting with the owner or stranger; including running, jumping, active physical contact and chasing toys
Following
Following the person around the room or to the door
Approach
Withdraw
Oriented to
door
Oriented to
person
Oriented to
chair
Scratch the
door
Other
behaviours
Approaching while clearly visually oriented to the person,
either spontaneously or when called
Avoiding interaction with the owner or the stranger by either moving away, very clearly turning away or looking
away
Staring Ž xedly at the door, either when close to it or from
a distance
Staring Ž xedly at the owner or stranger, regardless of
whether the behaviour was reciprocated
Staring Ž xedly at the owner’s or stranger’s empty chair
All active behaviours resulting in physical contact with the
door, including scratching the door with the paws, jumping
on the door, pulling on the door handle with the forelegs or
mouth
Any activity not included in the behavioural catalogue,
such as self-grooming self-scratching or drinking
Non mutually exclusive categories
Physical
Being in physical contact with owner or stranger
contact with
person
Sampling
method
Point sampling
Point sampling
Point sampling
Point sampling
Point sampling
Point sampling
& frequency
Point sampling
& frequency
Point sampling
& frequency
Point sampling
Point sampling
Point sampling
& frequency
Point sampling
& frequency
Point sampling
Point sampling
235
DOG–HUMAN ATTACHMENT
TABLE 1. (Continued)
Behavioural
category
DeŽ nition
Sampling
method
Physical
contact with
chair
Contact
with objects
Vocalising
Being in physical contact with owner’s or stranger’s chair
Point sampling
& frequency
Being in physical contact with the owner’s or stranger’s
shoes or clothing
Any kind of vocalisation, i.e. barking, growling, howling,
whining: vocalisations were recorded as bouts
Point sampling
& frequency
Point sampling
& frequency
Greeting
behaviour
All greeting behaviours toward the entering owner or
stranger, such as approaching, tail-wagging, jumping,
physical contact
Duration & intensity
Episode 8: Dog and Objects (3 min, 4th separation and 3rd reunion). The dog was left
alone and monitored as in episode 5. At the end of the third minute, the owner returned and
greeted and/or comforted the dog until the experimenter arrived and ofŽ cially terminated the
procedure.
Immediately after each session, the experimental room, water bowl and toys were washed
using a non-toxic, weakly scented disinfectant.
Data collection and analysis
The behaviour of each dog during the experimental episodes (ep. 1-8) was recorded on
videotapes and each session was analysed by two trained observers recording a total of 17
behavioural categories. These categories were selected after a preliminary analysis of the
videotapes and on the basis of previous work based on the strange situation procedure (i.e.
Ainsworth & Bell, 1970; Ainsworth et al., 1978; Topal et al., 1998).
The dogs’ behaviour during each episode was Ž rst scored using 5-second point sampling
(instantaneous sampling). This method of time sampling was chosen because it is a way of
condensing information, thereby making it possible to record several different categories of
behaviour simultaneously and with accuracy when the exact duration of a given behaviour
is not strictly necessary. If the sample interval is short enough, point sampling is accurate in
estimating duration of behavioural states and gives a good approximation of the proportion
of time spent performing a behaviour pattern. However, instantaneous point sampling is not
suitable for recording discrete events of short duration and rare behaviour patterns. Therefore, in a second round of coding we used continuous recording to collect frequencies for a
number of discrete events. Table 1 contains a list of the different behavioural categories, their
deŽ nitions and the type of sampling method employed for data collection.
The greeting behaviour of the dogs towards the owner (ep. 4 and 7) and the stranger
(ep. 2 and 6) was recorded both quantitatively (duration of greeting in seconds before the
resumption of another behaviour) and qualitatively(in terms of an intensity score) considering
the Ž rst minute of each reunion episode. The greeting intensity score was as follows: if no
greeting behaviour occurred, the score was 0 (no greeting); if the dog initiated approach, but
236
PRATO-PREVIDE, CUSTANCE, SPIEZIO & SABATINI
then showed a mild greeting with limited physical contact and tail wagging, the score was
1 (mild greeting); if physical contact was full and the greeting behaviour was intense (i.e.
jumping up on the owner, vocalising and strongly wagging the tail) the score was 2 (intense
greeting).
Interobserver reliability was assessed by means of independent parallel coding of a random sample of 4 videotaped sessions (i.e. 10% of the total number of dogs) using percentage
agreement: agreement on frequencies and greeting behaviour was evaluated separately. Percentage agreement was always more than 96%. Since, histograms revealed that the majority
of the data were not normally distributed, all statistical tests were non-parametric and twotailed.
Results
The results presented re ect average tendencies rather than individual differences. There were clear individual differences among the dogs, but a detailed
exposition of these is not within the scope of the present paper. Since, there
was a relatively small sample of dogs, subjects with extreme experiences had
the potential to skew the overall results. Hence, the analyses were carried out
using non-parametric tests based on the median rather than the mean, which
tends to minimise this kind of distortion. The results have been arranged into
four categories: secure base effects, proximity seeking, search behaviours
and comfort seeking. D refers to dog, S to stranger and O to owner.
Secure base effects
Exploration
According to the ethological approach, an attachment Ž gure functions as a
secure base and thereby allows the child to move off and engage in other
behaviour, especially exploration. Figure 3 shows how exploratory behaviour
varied in successive episodes. Exploration declined sharply from episode 1
when the dog was alone with the owner to episode 2 when also the stranger
was present (Wilcoxon test: z D 4:28, p < 0:0001/ remaining low in
episode 3, when the dog was alone with the stranger, and in the following
episodes. A comparison between the overall proportion of sample points
spent exploring in episodes characterised by owner (O) presence (1, 4, 7),
stranger (S) presence (3, 6) or by isolation (D; 5, 8) revealed a signiŽ cant
difference across these situations (Friedman test: O+D vs S+D vs D, Â 2 D
32:42, p < 0:0001/. Overall, the dogs explored more in the presence of their
DOG–HUMAN ATTACHMENT
Fig. 3.
237
Proportion of time spent by the dogs exploring and passive across the 8 episodes.
O = owner; D = dog; S = stranger.
owner than with the stranger or alone (SNK test for multiple comparisons
(Glantz, 1997): O+D vs S+D, p < 0:05I O+D vs D, p < 0:05I S+D vs D,
p D NS).
However, it is possible that these differences in exploration were due to
an order effect of diminishing curiosity over time rather than to a secure
base effect. When the dogs were Ž rst introduced to the room, they were in
the presence of their owner and the most intense exploration occurred in the
Ž rst minute. Although there was a signiŽ cant difference in the amount of
exploration between the third minute of episode 1 and the Ž rst minute of
episode 2 (Wilcoxon test: z D 2:25, p D 0:025/, there was also a signiŽ cant
difference between the Ž rst and second minute (Wilcoxon test: z D 2:49,
p D 0:012/ and a decrease between the second and third minute (Wilcoxon
test: z D 1:919, p D 0:055/ of episode one. Figure 4 clearly indicates that
exploration rapidly and steadily declined across the Ž rst four minutes of the
procedure.
Overall males (N D 19) were slightly more explorative than females
(N D 19) and in particular engaged in exploratory behaviour signiŽ cantly more during the Ž rst episode (Mann-Whitney test: ep. 1, z D 2.250,
p D 0:024). No differences in exploration were found comparing abandoned
(N D 13) and non-abandoned (N D 25) dogs.
238
Fig. 4.
PRATO-PREVIDE, CUSTANCE, SPIEZIO & SABATINI
Proportion of sample points spent in exploration across the Ž rst 5 minutes.
Response to stranger’s entrance
Only 7 out of 38 dogs (18,4%) remained close to their owner or returned
to their owner’s side on the entrance of the stranger, observing her from a
distance. Most of the dogs (81,6%) did not seem at all wary of the stranger: a
number of them greeted her or approached her, others ignored her continuing
the activities they were engaged in.
Passive behaviour
If the owner operated as a secure base, this might have allowed the dogs to
feel sufŽ ciently calm to engage in passive behaviour when in their owner’s
presence. Passive behaviour (i.e. sitting, standing or lying down without obviously focussing on some aspect of the environment) was higher in episode
1 and 2 and when the dogs were alone (ep. 5, 8; Fig. 3). During episodes 1
and 2 the subjects tended to stay close to the owner, whereas during episodes
5 and 8 most passive behaviour occurred in the door area. Comparison between the overall proportion of sample points spent passive in the episodes
characterised by owner presence (1, 4, 7), stranger presence (3, 6) or by isolation (5, 8) revealed signiŽ cant differences (Friedman test: O+D vs S+D vs
D, Â 2 D 6:77, p D 0:034/. Dogs were more passive with their owner and
when alone than with the stranger (SNK test for multiple comparisons: O+D
vs S+D, p < 0:05I D vs S+D, p < 0:05I O+D vs D, p D NS). Overall
females (N D 19) tended to be more passive than males (N D 19) but differences were signiŽ cant only in the Ž rst episode (Mann-Whitney test: ep. 1,
DOG–HUMAN ATTACHMENT
239
z D 2.213, p D 0:026). No differences in passive behaviour were found
comparing abandoned (N D 13) and non-abandoned (N D 25) dogs. During isolation episodes non-abandoned dogs tended to stay more passive than
abandoned ones (Mann-Whitney test: ep. 5-8, z D 1.879, p D 0:06/.
Play
Individual play occurred rarely: only 4 out of 38 subjects (10.5%) played
with toys in episodes 1 and 2, and 1 subject did so when left alone in episode
5. In episode 2.66% of the subject accepted to play with the stranger, whereas
only 55.3% played, at least to some extent, in episode 3, and 39.5% did
so in episode 6. A comparison between the proportion of sample points
spent playing with the stranger in episodes 2 (last minute) and 3 showed
that play decreased after the owner left the room (Wilcoxon test: ep. 2 vs
ep. 3, z D 2:072, p D 0:038/. Figure 5 shows the proportion of sample
points spent by the dogs playing with the owner (ep. 4 and 7) and with
the stranger (ep. 3 and 6): a comparison between the overall proportion of
sample points spent playing with the owner and the stranger in these episodes
showed that dogs played signiŽ cantly more with their owner (Wilcoxon test:
O+D vs S+D, z D 2:08, p D 0:037/. Social play with the owner clearly
decreased on the second reunion episode (Wilcoxon test: ep. 4 vs ep. 7,
Fig. 5.
Proportion of time spent in social play with the owner (ep. 4, 7) versus the stranger
(ep. 3, 6). O = owner; D = dog; S = stranger.
240
PRATO-PREVIDE, CUSTANCE, SPIEZIO & SABATINI
z D 3:468, p D 0:0005/ and declined signiŽ cantly on the second encounter
with the stranger (Wilcoxon test: ep. 3 vs ep. 6, z D 3:563, p D 0:0004/. We
found no sex differences in play behaviour with the owner (ep. 4 and 7) and
the stranger (ep. 3 and 6), whereas abandoned dogs played less than nonabandoned ones with the stranger (Mann-Whitney test: ep. 3-6, z D 1. 967,
p D 0:049/.
Proximity seeking
Approach
The dogs approached both the owner and the stranger in the different
episodes both spontaneously and when called. In episode 2 the stranger was
approached more than the owner (Wilcoxon test: z D 2:44, p D 0:014/I
however, a comparison between episodes 4 and 7 (presence of the owner)
and 3 and 6 (presence of the stranger) showed that the dogs approached
their owner signiŽ cantly more than the stranger (Wilcoxon test: z D 2:341,
p D 0:02/. While approach behaviour toward the owner was comparable on
episodes 4 and 7, approach behaviour toward the stranger increased remarkably in episode 6 (Wilcoxon test: ep. 3 vs ep. 6, z D 3:071, p D 0:002/.
Withdraw
Of course, withdraw or avoidance is the antithesis of proximity seeking. It is
included here to offer a direct contrast to approach. None of the dogs clearly
avoided interaction with its owner whereas 3 subjects (2 of them abandoned)
showed strong avoidance of the stranger especially during episodes 3 and 6.
Avoidance responses occurred when the stranger tried to interact with the
dog or to comfort him/her and varied in intensity from looking away to
turning away to moving away while staring at the stranger.
Oriented to person
Orientation to a person may have different functions, one of which could
be monitoring proximity. The dogs paid attention to both the owner and
stranger. In episode 2, their attention was focused signiŽ cantly more on
the stranger (Wilcoxon test: z D 4:285, p < 0:0001/. A comparison between episodes 4 and 7 (presence of the owner) and 3 and 6 (presence
of the stranger) showed that they spent more time oriented to their owner
(Wilcoxon test: O+D vs S+D, z D 2:98, p D 0:003/.
DOG–HUMAN ATTACHMENT
241
Following
Following the owner to the door was scored for episodes 2, 4 and 7 when
the owner left the room. Most of the dogs (86.8%) followed their owner to
the door at some point during the procedure, and only 5 subjects (13.2%)
never showed following behaviour. In particular, 36.8% of the dogs followed
their owner to the door all three times that she or he exited the room, 28.9 %
followed on two occasions and 21.1% only one time. On the contrary only 4
dogs (10.5%) followed the stranger to the door when she left and did it only
one time.
Vocalising
It has been suggested that vocalising might function to promote proximity
maintenance or resumption, in addtion to indicating distress. The majority
of the dogs (86.8%) vocalised during the strange situation although to a
different extent. Figure 6 reports the mean number of vocalisation bouts
(barking, howling and whining) in successive episodes; vocal behaviour was
low in the episodes characterised by the presence of the owner (1, 4, 7)
increasing during separation episodes (3, 6 and 5, 8). Vocalisations increased
when the owner left the room and the dogs remained alone with the stranger
(Wilcoxon test: ep. 2 vs ep. 3: z D 2:046, p D 0:04/. A comparison of vocal
behaviour across conditions showed that the mean number of vocalization
Fig. 6.
Mean number of vocalisation bouts across the 8 episodes. O = owner; D = dog; S =
stranger.
242
PRATO-PREVIDE, CUSTANCE, SPIEZIO & SABATINI
bouts varied remarkably (Friedman test: O+D vs S+D vs D, Â 2 D 30:38,
p < 0:0001/. Vocalisations increased signiŽ cantly during social isolation
(ep. 5, 8) but not signiŽ cantly in the presence of the stranger (ep. 3, 6:
SNK test for multiple comparisons: O+D vs S+D, p D NS; O+D vs D,
p < 0:05I D+S vs D, p < 0:05/. Dogs vocalised signiŽ cantly less on the
second isolation episode (ep. 8) when they remained alone in the presence of
objects (Wilcoxon test: z D 2:55, p D 0:01/.
Search behaviours
Scratch door
24 out of 38 dogs (63.2%) scratched the door and jumped on it during separation episodes. This behaviour occurred mainly when the dogs remained alone
(ep. 3, 6: 36.3% of scratching bouts; ep. 5, 8: 61.6% of scratching bouts). A
comparison between the two separation conditions (ep. 3, 6 and 5, 8) showed
that the number of scratching bouts increased signiŽ cantly during social isolation (Wilcoxon test: S+D vs D, z D 2:170, p D 0:03/I it decreased the
second time the dogs remained with the stranger or alone (Wilcoxon test: ep.
3 vs ep. 6: z D 2:654, p D 0:008I ep. 5 vs ep. 8: z D 2:825, p D 0:005/.
Oriented to door
Oriented to the door was interpreted as search behaviour by Ainsworth et
al. (1978). As shown in Fig. 7, this behaviour varied in successive episodes,
increasing dramatically during the separation episodes (ep. 3, 5, 6, 8). The
overall proportion of sample points spent staring at the door in episodes
characterised by owner presence (1, 4, 7), stranger presence (3, 6) or by
isolation (5, 8) differed remarkably (Friedman test: O+D vs S+D vs D, Â 2 D
64:225, p < 0:0001/I the dogs’ attention was oriented towards the door far
more when the owner was absent or the dog was alone in the room (SNK test
for multiple comparisons: O+D vs S+D , p < 0:05I O+D vs D, p < 0:05I
S+D vs D, p < 0:05/. This behaviour increased sharply from episode 2
to episode 3 when the owner left the room (Wilcoxon test: ep. 2 vs ep. 3:
z D 4.816, p < 0:0001/I the dogs remained more oriented towards the door
on the second encounter with the stranger (Wilcoxon test: ep. 3 vs ep. 6:
z D 2:286, p D 0:022/ but stared at the door less on the second isolation
episode (Wilcoxon test: ep. 5 vs ep. 8: z D 2:241, p D 0:025/.
DOG–HUMAN ATTACHMENT
Fig. 7.
243
Proportion of time spent by the dogs oriented to the door across the 8 episodes. O =
owner; D = dog; S = stranger.
Male dogs (N D 19) tended to stay more oriented towards the door than
females (N D 19) during separation episodes (ep. 3, 6 and 5, 8) and stared
at the door more when in the presence of the stranger (Mann-Whitney test:
ep. 3-6, z D 2:016, p D 0:044). Abandoned dogs were not more oriented to
the door than non-abandoned during isolation (ep. 5, 8) but tended to stare at
the door more in the presence of the stranger and in particular in episode 3
when their owner left the room for the Ž rst time (Mann-Whitney test: ep. 3,
z D 1:917, p D 0:055).
Oriented to chair
As Fig. 8 shows the dogs glanced at both the owner’s and stranger’s empty
chairs during the different separation episodes; however, they were signiŽ cantly more oriented to their owner’s chair (Wilcoxon test: ep. 3, 6, Owner
vs Stranger, z D 2:084, p D 0:037I ep. 5, 8, Owner vs Stranger, z D 2:926,
p D 0:003/. Glancing at the owner’s chair occurred almost exclusively when
the owner was absent (95.5% of all occurrences) and was signiŽ cantly higher
when the dogs were alone in the room (Wilcoxon test: S+D vs D, z D 3:60,
p D 0:0003/I this behaviour decreased the second time they remained with
the stranger (ep. 3 vs ep. 6: z D 3:353, p D 0:0008/ and increased signiŽ cantly the second time they remained alone with objects (Wilcoxon test:
z D 2:10, p D 0:035/.
244
PRATO-PREVIDE, CUSTANCE, SPIEZIO & SABATINI
Fig. 8. Percentage of events in which the dogs were oriented or in contact with the owner’s
versus the stranger’s chair during separation episodes. O C O = oriented to owner’s chair;
O C S = oriented to stranger’s chair; C C O = physical contact with owner’s chair; C C S =
physical contact with stranger’s chair.
Comfort seeking
Physical contact with chair
The dogs made contact with both the owner’s and stranger’s empty chairs
during the different episodes; however, when in the presence of the stranger
(ep. 3, 6) and when alone (ep. 5, 8) they were signiŽ cantly more in contact
with their owner’s chair (Wilcoxon test: ep. 3, 6, Owner vs Stranger, z D
3:354, p D 0:0008I ep. 5, 8, Owner vs Stranger, z D 2:481, p D 0:013I
Fig. 8). Three different types of contact with the chair were observed: putting
the head on the chair and snifŽ ng it for a while, sitting or lying down with
the body in contact with the chair legs, and climbing on the chair remaining
seated on it: these last two type of contact were almost exclusively with the
owner’s chair.
Contact with objects
71% of the dogs made contact with the objects (shoes and piece of clothing)
during episode 8 and mainly with those of the owner (65.2% of all contacts).
A comparison between the number of contacts with owner’s and stranger’s
objects showed that the dogs made signiŽ cantly more contacts with the objects left by their owner (Wilcoxon test: z D 2:267, p D 0:023/. Contact
DOG–HUMAN ATTACHMENT
245
with objects ranged from snifŽ ng them to maintaining physical contact with
them (i.e. sitting on the clothes on the owner’s chair): physical contact was
directed to only the owner’s objects. Furthermore, a comparison between the
proportion of sample points spent by the dogs in the owner’s area (Area 3,
see Fig. 1) during social isolation (ep. 5 and 8) showed that the amount of
time spent next to the owner’s chair increased signiŽ cantly when the owner’s
object were present (Wilcoxon test: ep. 5 vs ep. 8, z D 3:026, p D 0:002/.
Physical contact with person
The dogs sought and maintained physical contact with both the owner and
the stranger during the different episodes. However, a comparison between
the proportion of sample points spent in physical contact with the owner (ep.
1, 4, 7) and with the stranger (ep. 3, 6) showed that the dogs spent more time
in contact with their owner (Wilcoxon test: z D 2:31, p D 0:021/. During
separation episodes the dogs, although to a different extent, accepted to stay
in physical contact with the stranger and to be petted: physical contact was
signiŽ cantly higher in episode 6 (Wilcoxon test: ep. 3 vs ep. 6, z D 4:424,
p < 0:0001/. However, when in contact with the stranger they were much
more oriented to the door than when in contact with their owner (Wilcoxon
test: ep. 3, 6 vs ep. 4, 7, z D 4:259, p < 0:0001/.
Greeting
Greeting behaviour was scored during the Ž rst minute of episodes 2 and 6
(entrance of the stranger) and 4 and 7 (reunion with the owner). Only 2 out
of 38 subjects showed no greeting towards both the owner and the stranger,
7 subjects (18.4%) never greeted the stranger and 29 (76.3%) greeted, at
least to some extent, both the owner and the stranger. Dogs showed considerably more intense levels of greeting towards the entering owner compared
with the stranger. As Fig. 9 shows, differences in overall greeting duration
towards the owner (ep. 4, 7) and the stranger (ep. 2, 6) were highly significant (Wilcoxon test: z D 2:99, p D 0:002/ and the duration of greeting
behaviour did not decrease on the second reunion with the owner (Wilcoxon
test: ep. 4 vs ep. 7, p D NS) and on the second encounter with the stranger
(Wilcoxon test: ep. 2 vs ep. 6, p D NS). Greeting towards the owner was
also more intense: most subjects showed no greeting (36.8%) or mild greeting (50%) when the stranger entered the Ž rst time, whereas the majority
246
Fig. 9.
PRATO-PREVIDE, CUSTANCE, SPIEZIO & SABATINI
Duration of greetings directed toward the entering owner (ep. 4, 7) versus the
stranger (ep. 2, 6). O = owner; D = dog; S = stranger.
(60.5%) showed intense greeting on the Ž rst reunion with the owner. Similarly, after being left completely alone in episode 5, and even though the
stranger entered Ž rst (ep. 6), the dogs still showed more intense greetings
towards their owner when she or he returned three minutes later at the start
of episode 7 (Sign test: p < 0:05/. Differences in greeting intensity were
assessed comparing the scores of each dog during episodes 2 and 4 (Ž rst encounter) and episodes 6 and 7 (second encounter). In both cases, greetings
towards the owner were signiŽ cantly more intense (Sign test: p < 0:05/.
Discussion
The aim of this study was to investigate in some detail the nature of the dog–
human social relationship by observing owner-dog pairs under controlled
conditions using a modiŽ ed version of the strange situation procedure developed by Mary Ainsworth. This procedure seemed particularly suitable for
testing dogs as it reproduces situations that dogs are likely to encounter in
their everyday life, such as being in a new environment, meeting a stranger,
and being separated from their owner for short periods. Overall, our Ž ndings
clearly indicate that the dog–human relationship is a strong affectional bond,
but they provide only rather limited evidence that it constitutes an attachment.
DOG–HUMAN ATTACHMENT
247
Ainsworth (1989) stated that the secure base effect was a primary factor
in identifying an attachment bond. Three measures of the secure base effect
have been identiŽ ed: (1) play and exploration become depressed when in
the presence of just the stranger and when alone, but recover after reunion
with the mother; (2) infant cease playing or exploring on the entrance of
the stranger and return to their mother’s side; and (3) infants will sometimes
play with the stranger in their mother’s presence, but not her absence. In our
study there was an insufŽ cient amount of individual play in the dogs to use
this as a measure of the secure base effect. In addition, only 18.4% of the
dogs remained close to their owner or returned to their owner’s side on the
entrance of the stranger. Most of the dogs did not seem at all wary of the
stranger and hence her presence did not appear to activate the attachment
system.
The dogs’ exploratory behaviours consisted mainly of locomotory exploration. Other forms of exploration reported in human infants and chimpanzees were either rare (i.e. visual) or physically difŽ cult for dogs to perform (i.e. manipulatory). Exploration was higher in the presence of the owner
(see also Topal et al., 1998). Moreover, similar to what is reported for human
infants and hand-reared chimpanzees (Ainsworth & Bell, 1970; Ainsworth
et al., 1978; Bard, 1991), exploration occurred mainly in episode 1, when
the dogs were with their owner, and a signiŽ cant decline was observed from
episode 1 to episode 2 with the entrance of the stranger.
At Ž rst glance, the above pattern of results suggests that the owner provided a secure base for exploration and that the entrance of the stranger negatively affected exploratory behaviour, indicating stranger’s wariness or fear.
However, by the end of the third minute of episode 1 most of the dogs had
either Ž nished exploring the room or were doing so to a much lesser extent
than before. In addition to there being a signiŽ cant difference between the
last minute of episode 1 and the Ž rst minute of episode 2, there was also a
signiŽ cant decline in exploration across the whole of the Ž rst episode. Thus
the overall proŽ le of exploration during the Ž rst Ž ve minutes of the procedure
suggests that these differences could be due to an order effect of reduced curiosity over time. Therefore, the dogs’ exploratory behaviour did not supply
convincing evidence of a secure base effect.
In accordance with Topal et al.’s results, our sample of dogs engaged in
social play signiŽ cantly more in the presence of the owner. However, as with
exploration, this Ž nding does not necessarily indicate a secure base effect.
248
PRATO-PREVIDE, CUSTANCE, SPIEZIO & SABATINI
Instead, it might suggest that the owner was a preferred playmate. If the dogs
had engaged in individual play more in the presence of the owner rather than
the stranger, there would have been better evidence of a secure base effect.
However, very few dogs engaged in individual play of any kind, whether
with or without their owner.
Rather more indicative of a secure base effect was the Ž nding that the
proportion of social play with the stranger was higher in the presence of
the owner (i.e. during the last minute of episode 2) than when the dogs
were left alone with the stranger in episode 3. This Ž nding suggests that in
episode 2 the presence of the owner provided a sufŽ cient sense of security to
promote play with the stranger. Although this decrease of play in the absence
of the owner might be partly due to fatigue or reduced interest, the overall
behaviour of the dogs during episode 2 and 3 suggests that in episode 3 the
attachment behavioural system was activated and interfered with play: in
particular, play became discontinuous and a variety of behaviours indicating
distress, protest and searching appeared (i.e. vocalising, running to the door
or looking at it continuously). Play behaviour with the owner was intense
again in episode 4, thus suggesting that the decrease in play in the absence
of the owner in episode 3 and the differences in the propensity to play with
the owner versus the stranger were unlikely to be due to fatigue or satiation.
In general the dogs showed little distress in the pre-separation episodes.
In fact, some of them were so deeply engrossed in play with the stranger in
the last minute of episode 2 that they did not even notice their owner leaving
the room. However, once they did notice their owner’s absence, most dogs
showed distress and protest during separation and engaged in a variety of
behaviours indicating that they were searching for the absent owner either
actively or by orienting to the last place in which she or he was seen. Our
observations during separation episodes (ep. 3, 6, 5 and 8) show that the
dogs’ search behaviours closely resembled those described in human infants
and chimpanzees and included following the owner to the door, scratching at
and jumping up on the door, pulling on the door handle with the forelegs or
mouth, remaining oriented to the door, going to the owner’s empty chair or
looking at it from a distance.
Both human infants and chimpanzees have been reported to show alarm
and distress in terms of crying during separation episodes. In the presence
of the owner (ep. 1, 4 and 7), and when both the owner and the stranger
were in the room (ep. 2) the dogs’ vocal behaviour was limited and consisted
DOG–HUMAN ATTACHMENT
249
mainly of attention seeking, play soliciting and greeting vocalisations. In
contrast, during separation episodes whining and barking increased, reaching
the highest levels during social isolation (ep. 5 and 8). Vocal behaviour was
especially activated by being left alone and was reduced by the presence of
the stranger. This suggests that for dogs separation from the owner in an
unfamiliar environment was distressing, but not as distressing as being left
completely alone. The decline of vocal behaviour in episode 6, together with
the increase of physical contact with the stranger suggests that, in general,
the dogs did not Ž nd the stranger frightening per se and she was able to
provide some comfort. Nevertheless, a number of dogs continued to whine
even when in physical contact with the stranger. Some accepted petting, but
remained oriented to the door most of the time, others did not vocalise but
showed an increase in search behaviours such as going to the owner’s chair
or looking at it. Therefore, the friendly stranger was not able to provide
sufŽ cient comfort to curtail search and protest behaviours during the owner’s
absence. The continuation of distress and search behaviour by the dogs when
they were left alone with the stranger, but not the owner, is consistent with
the pattern of attachment behaviours described by Ainsworth & Bell (1970)
in human infants.
Separation from the owner promoted proximity and contact seeking behaviour upon reunion (ep. 4 and 7). During reunion with the owner, proximity seeking behaviour included greeting responses, approaching, and following the owner around the room. Greeting toward the entering owner was
signiŽ cantly different from greeting toward the entering stranger in terms of
duration and quality and the stranger was approached signiŽ cantly less in
comparison to the owner.
Avoidance or withdrawal behaviour was seldomly exhibited by the dogs.
In fact, it was recorded in only three dogs (all of whom had been abandoned
at some time in their life) and only in response to the stranger. Human infants
exhibit resistant behaviours, in which they resist being picked up or held and
struggle to be put down again. None of the owners attempted to pick up their
dogs and nothing that could be characterised as resistance was observed.
Although the dogs visually oriented to both the owner and the stranger,
overall they spent more time oriented towards their owner. Frequency of
looking at the owner and stranger was probably in uenced by a number of
factors and may have served different functions. Dogs are known to engage
250
PRATO-PREVIDE, CUSTANCE, SPIEZIO & SABATINI
in mutual eye contact with their owners and spend considerable time observing their movements, an activity which according to Serpell (1986) plays a
central role in building and maintaining the relationship between people and
their pets. Prior to separation (ep. 1 and 2), most dogs looked at their owner
only occasionally. In episode 2, their attention was clearly captured by the
stranger: this appeared to be due to a variable combination of curiosity, exploration, wariness and sociability. All dogs looked at the entering stranger,
but not all approached, sniffed or greeted her. A number of subjects observed
the stranger from a distance remaining close to their owner’s chair without
interacting or playing with her. This Ž nding is consistent with what reported
in both human infants and chimpanzees. Visual orientation toward the owner
increased signiŽ cantly during reunion episodes and occurred frequently during physical contact. In contrast, visual orientation toward the stranger (ep.
3 and 6) occurred mainly during play or in response to the stranger’s attempts to interact. These differences in the use of visual orientation suggest
that looking at the stranger was related to socialisation toward humans in
general, whereas looking at the owner was related to affection and comfort
seeking.
The behaviour of the dogs during the last episode, when they remained
alone in the presence of the owner’s and stranger’s objects, suggests that at
least a number of the dogs used the owner’s objects as a source of comfort.
The dogs showed a clear preference for their owner’s objects: there were
signiŽ cantly more contacts with the owners’ objects and prolonged physical
contact was only ever directed to the owner’s objects. Behaviour in episode 8
differed from that of episode 5 in a number of ways. First, barking, whining
and scratching at the door decreased signiŽ cantly. Second, the time spent oriented to the door decreased while the time spent oriented to the owner’s chair
increased signiŽ cantly. Last but not least, the amount of time spent by the
dogs in the owner’s chair area where the owner’s objects were increased signiŽ cantly. In fact, during episode 8 a number of dogs jumped on the owner’s
chair and sat on the clothes while others remained next to their owner’s chair
sitting or lying down. Although potentially confounding order effects cannot be excluded, the signiŽ cant increase in the amount of time spent close
to the owner’s chair in episode 8 suggests that the owner’s clothing could
have represented a source of comfort. As dogs strongly rely on olfaction and
are reported to prefer familiar odours (Millott, 1994) it is possible that they
DOG–HUMAN ATTACHMENT
251
used the owner’s clothing as a source of comfort because it carried familiar odours. It is also possible that previous experience during the procedure
could have decreased distress (i.e. by promoting the expectation in the dogs
of their owner’s return) and this could have lead to a decrease in searching
and proximity seeking behaviours and a corresponding increase in passive
behaviours.
In summary, we found that adult dogs’ behaviour in the Strange Situation
test resembled that of human and chimpanzee infants to a remarkable degree.
Our results clearly indicate that the dog–human relationship is an affectional
bond, but our evidence that it conforms to an attachment is not entirely
conclusive. The order effects inherent in Ainsworth’s procedure meant that
we could Ž nd only one positive indicator of the secure base effect, namely,
the fact that the dogs would play with the stranger in the presence of the
owner, but not in his or her absence. This suggests that the conclusions drawn
in the previous study by Topal et al. were premature and that further research
is needed to gain a better understanding of the nature of the dog–human
bond and to test speciŽ cally for secure base effects. ModiŽ ed versions of
the procedure could be devised to test speciŽ c hypotheses avoiding the order
effects typical of Ainsworth’s original design.
Nevertheless, the Strange Situation procedure and possible modiŽ ed versions aimed at counteracting order effects would appear to be an appropriate
and useful tool to test hypotheses about the bond between dogs and their
owners and to evaluate the quality of the relationship. In fact, in terms of
the dog–human affectional bond and attachment, there are very many areas
of possible future research. With a larger sample size than that used here,
one could compare different breeds of dogs and the effect of gender with
respect to the dogs and their owners. Even with our relatively small sample,
our data show males dogs spent more time exploring the new environment
than females during the Ž rst episode, whereas female were signiŽ cantly more
passive.
One could also investigate the effect of early socialisation with humans
and previous experiences on the dog–human bond. A rather dramatic experience that can have a profound in uence on dog–human relationship is abandonment, especially if it takes place when the affectional bond is already
well established. Thirteen of the dogs in our sample had been abandoned
at some point in their life before being adopted by their present owner. We
252
PRATO-PREVIDE, CUSTANCE, SPIEZIO & SABATINI
found several trends (although most were statistically non-signiŽ cant) suggesting that these dogs were more anxious and perhaps less securely bonded:
they explored and played less in the presence of the owner, were less passive
when left alone and withdrew from the stranger more than those dogs that
had lived in the same household since puppyhood. A particularly interesting
Ž nding, that deserves further attention, is the signiŽ cant difference in social
play with the stranger between abandoned and non-abandoned dogs.
Physiological measures could be used to support the behavioural data with
respect to anxiety, distress and comfort seeking (Palestrini et al., 2001). None
of the owners in our study reported that their dogs suffered any serious behavioural problems, and almost all reported that their dogs could stay alone at
home without problems, however during the test a number of dogs showed
behaviours that could be related to separation anxiety (Voith & Borchelt,
1985). The Strange Situation could be used to investigate the behaviour of
dogs that exhibit ‘separation anxiety’ and to make comparisons with dogs
not showing this behaviour problem.
References
Ainsworth, M.D.S. (1964). Patterns of attachment behavior shown by the infant in interaction
with his mother. — Merrill-Palmer Quart. 10, p. 51-58.
— — (1967). Infancy in Uganda: Infant care and the growth of attachment. — Johns Hopkins
University Press, Baltimore.
— — (1969). Objects relations, dependency and attachment: A theoretical review of the
infant-mother relationship. — Child Devel. 40, p. 969-1025.
— — (1989). Attachments beyond infancy. — Am. Psychol. 44, p. 709-716.
— — & Bell, S.M. (1970). Attachment, exploration, and separation: illustrated by the behavior of one-year-olds in a strange situation. — Child Devel. 41, p. 49-67.
— —, Blehar, M.C., Waters, E. & Wall, S. (1978). Patterns of attachment: A psychological
study of the strange situation. — Lawrence Erlbaum Associates, Publishers, Hillsdale,
New Jersey.
Askew, H.R. (1996). Treatments of behaviour problems in dog and cat. A guide for the small
animal veterinarian. — Blackwell Science, Oxford.
Bard, K. (1991). Distribution of attachment classiŽ cations in nursery chimpanzees. — Am. J.
Primatol. 24, p. 88.
Bowlby, J. (1958). The nature of the child’s tie to his mother. — Int. J. Psycho-Anal. 39, p.
350-373.
— — (1969/1982). Attachment and loss: Vol. 1. Attachment. — Basic Books, New York.
Bretherton, I. (1992). The origins of attachment theory: John Bowlby and Mary Ainsworth.
— Devel. Psychol. 28, p. 759-775.
DOG–HUMAN ATTACHMENT
253
Cassidy, J. (1999). The nature of the child’s ties. — In: Handbook of attachment: theory,
research, and clinical applications (J. Cassidy & P.R. Shaver, eds). Guilford Press, New
York, p. 3-20.
Collis, J.M. & McNicholas, J. (1998). A theoretical basis for health beneŽ ts of pet ownership.
— In: Companion animals in human health (C.C. Wilson & D.C. Turner, eds). Sage
Publications, London, p. 105-122.
Clutton-Brock, J. (1999). A natural history of domesticated mammals. — Cambridge University Press, Cambridge.
Endenburg, N. (1995). The attachment of people to companion animals. — Anthrozoös 8, p.
83-89.
Glantz, S.A. (1997). Primer of biostatistics. — McGraw-Hill, New York.
Hart, L.A. (1995). Dogs as human companions: a review of the relationship. — In: The
domestic dog: its evolution, behavior, and interactions with people (J. Serpell, ed.).
Cambridge University Press, Cambridge, p. 162-178.
Jagoe, A. & Serpell, J. (1996). Owner characteristics and interaction and the prevalence of
canine behaviour problems. — Appl. Anim. Behav. Sci. 47, p. 31-42.
Kretchmer, K.R. & Fox, M.W. (1975). Effects of domestication on animal behavior. —
Veterin. Rec. 96, p. 102-108.
Menanche, S. (1998). Dogs and Human Beings: A Story of Friendship. — Soc. Anim. 6, p.
67-86.
Millott, J.L. (1994). Olfactory and visual cues in the interaction systems between dogs and
children. — Behav. Proc. 33, p. 177-188.
O’Farrell, V. (1995). Effects of owner personality and attitudes on dog behaviour. — In: The
domestic dog: its evolution, behaviour, and interactions with people (J. Serpell, ed.).
Cambridge University Press, Cambridge, p. 154-158.
Palestrini, C., Prato-Previde, E., Custance, D.M., Spiezio, C., Sabatini, F. & Verga, M. (2001).
Heart rate and behavioural responses of dogs (Canis familiaris) in the Ainsworth’s
Strange Situation Test: a pilot study. — Proc. 3rd Int. Congr. Vet. Behav. Med., Vancouver, Canada, p. 89-91.
Rowan, A.N. & Beck, A.M. (1994). The health beneŽ ts of human-animal interactions. —
Anthrozoös 7, p. 85-89.
Scott, J.P. (1992). The phenomenon of attachment in human-nonhuman relationships. — In:
The inevitable bond. Examining scientist-animal interaction (H. Davis & D. Balfour,
eds). Cambridge University Press, Cambridge, p. 72-91.
Serpell, J. (1986). In the company of animals: a study of human-animal relationships. —
Cambridge University Press, Cambridge.
— — (1995). The domestic dog: its evolution, behaviour and interactions with people. —
Cambridge University Press, Cambridge.
Topal, J., Miklosi, A., Csanyi, V. & Doka, A. (1998). Attachment behavior in dogs (Canis
familiaris): A new application of Ainsworth’s (1969) strange situation test. — J. Comp.
Psychol. 112, p. 219-229.
Voith, V.L. (1985). Attachment of people to companion animals. — Vet. Clin. North Am.:
Small Anim. Pract. 15, p. 289-296.
— — & Borchelt, P.L. (1985). Separation anxiety in dogs. Compendium on continuing
education for the Practicing. — Veterinarian 7, p. 42-52.
— —, Wright, J.C. & Danneman, P.J. (1992). Is there a relationshipbetween canine behavior
problems and spoiling activities, anthropomorphism and obedience training? — Appl.
Anim. Behav. Sci. 34, p. 263-272.
254
PRATO-PREVIDE, CUSTANCE, SPIEZIO & SABATINI
Wilson, C.C. & Turner, D.C. (1998). Companion animals in human health (C.C. Wilson &
D.C. Turner, eds). — Sage Publications, London.