You are on page 1of 10

Anthrozoös

A multidisciplinary journal of the interactions of people and animals

ISSN: 0892-7936 (Print) 1753-0377 (Online) Journal homepage: www.tandfonline.com/journals/rfan20

Dogs Look Like Their Owners: Replications with


Racially Homogenous Owner Portraits

Sadahiko Nakajima, Mariko Yamamoto & Natsumi Yoshimoto

To cite this article: Sadahiko Nakajima, Mariko Yamamoto & Natsumi Yoshimoto (2009) Dogs
Look Like Their Owners: Replications with Racially Homogenous Owner Portraits, Anthrozoös,
22:2, 173-181, DOI: 10.2752/175303709X434194

To link to this article: https://doi.org/10.2752/175303709X434194

Published online: 28 Apr 2015.

Submit your article to this journal

Article views: 411

View related articles

Citing articles: 4 View citing articles

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=rfan20
AZ VOL. 22(2).qxp:Layout 1 3/31/09 10:43 PM Page 173

ANTHROZOÖS VOLUME 22, ISSUE 2 REPRINTS AVAILABLE PHOTOCOPYING © ISAZ 2009


PP. 173–181 DIRECTLY FROM PERMITTED PRINTED IN THE UK
THE PUBLISHERS BY LICENSE ONLY

Dogs Look Like Their Owners:


Replications with Racially
Homogenous Owner Portraits
Sadahiko Nakajima, Mariko Yamamoto and
Natsumi Yoshimoto
Department of Psychological Science, Kwansei Gakuin University, Japan
Address for correspondence: ABSTRACT Previous studies conducted in the United States and Venezuela have
Dr Sadahiko Nakajima,
Department of Psychological
demonstrated that people can correctly match portraits of unknown purebred
Science, dogs with their owners at statistically significant levels, suggesting that the popu-
Kwansei Gakuin University, lar belief in dog–owner physical resemblance is empirically valid. We investigated
Nishinomiya, Hyogo
662-8501, Japan. the perceived physical resemblance of dog–owner pairs in Japan, where the own-
E-mail: ers are racially more homogenous than in the countries in which the previous
nakajima@kwansei.ac.jp studies were conducted. In experiment 1, the matching performance by naïve
judges was statistically significant, and a supplementary test suggested that per-
ceived dog–owner physical resemblance plays a critical role in this. Experiment 2
presented a new procedure to test the perceived dog–owner physical resem-
blance and demonstrated that two-thirds of the judges selected a set of 20 real
dog–owner pairs over a set of 20 fake dog–owner pairs, irrespective of whether
it was an ownership-guessing task or a resemblance-based choice task. The
ability to match correct dog–owner pairs (experiment 1) and the sensitivity to dif-
ferentiate between real and fake dog–owner pairs (experiment 2) were not unique

Anthrozoös DOI: 10.2752/175303709X434194


to the characteristics of the judges (e.g., whether or not they were dog lovers) be-
cause the task performance was independent of any measured judge-related
factor. These results, taken together, provide another piece of positive evidence
for the popular belief that there is a physical resemblance between dogs and their
owners. Furthermore, the demonstration of dog–owner physical resemblance
with racially homogeneous owner samples supports the generality and robust-
ness of this phenomenon. The mechanism underlying this phenomenon (i.e.,
owners’ selection of dogs that look like themselves or the convergence of
appearance over time), however, remains to be elucidated by future studies.

Keywords: dog–owner resemblance, face, physical appearance


Disney’s “One Hundred and One Dalmatians” starts by briefly
portraying the physical resemblance between dogs and their
owners walking together down a street. In order to examine the
173

validity of this belief in the physical resemblance between dogs and their
AZ VOL. 22(2).qxp:Layout 1 3/31/09 10:43 PM Page 174

Dogs Look Like Their Owners: Replications with Racially Homogenous Owner Portraits

owners, Roy and Christenfeld (2004) conducted a simple portrait-matching experiment. The
headshots of 45 dogs (25 purebreds and 20 mongrels) and their owners were used to create
triads consisting of one owner, her/his dog, and the dog of another owner. The results re-
vealed that a majority of the naïve, undergraduate student judges (i.e., students who were un-
familiar with the tested dogs and owners) correctly matched 16 of the purebreds to their
respective owners, whereas the remaining 9 were incorrectly matched to other owners. Since
a ratio of 16:9 is statistically different from the chance level, the researchers concluded that
there was a resemblance between purebreds and their owners. However, in the case of the
mongrels, no such significant matching outcome was obtained.
In response to Levine’s (2005) criticism of the experimental procedure used by Roy and
Christenfeld (2004), Roy and Christenfeld (2005) conducted another experiment in which a dif-
ferent group of judges was requested to rank the photos of 6 owners on the basis of the like-
lihood that they were the owners of a target dog. The results showed that when the target dog
was purebred, the real owner of the dog had an average rank of 3.0, which was statistically
better than the chance level of 3.5. Accordingly, Roy and Christenfeld reiterated their stance
that purebreds resemble their owners.
One shortcoming of Roy and Christenfeld’s (2004, 2005) studies is that the measures used
were indirect—the judges’ performance on the dog–owner matching tasks were not reported
in a conventional manner (e.g., number or percentage of correct choices), thereby suggest-
ing an insignificant effect in these measures. Another drawback of their research is that they
used the term “matchability” as a synonym for the term “resemblance.” Semantically, correct
dog–owner matching does not necessarily mean that the matching was done on the basis of
similarity; it might have been accomplished with the help of other cues such as complemen-
tarity (i.e., a thin, female owner might have a stout dog as her guard).
Payne and Jaffe (2005) conducted an independent study on the same subject, which was
devoid of the shortcomings of Roy and Christenfeld’s studies. Six sets of 6 dog–owner pairs
were created using the headshots of 36 purebred dogs and their owners, and the portraits
within each set were randomly shuffled before the naïve judges were asked to match the dogs
with their owners. The number of pairs that the individual judges matched correctly had a dis-
tribution that was different from that of the Monte Carlo random simulation data, and was
skewed in the direction of better performance. The judges had been requested to use re-
semblance cues to solve the task, although this point was not clearly stated in the article (we
obtained this information by personal communication with Jaffe on February 8, 2008).
In the present research, our aims were threefold. First, we attempted to test dog–owner
resemblance by using Japanese samples. Although Roy and Christenfeld (2004, 2005) did
not provide details of the photographed dog owners, they were most likely residents of San
Diego or neighboring areas and consisted of racially diverse individuals. The dog owners in
the experiment conducted by Payne and Jaffe (2005) were Venezuelans, who are typically
a mix of African, Caucasian, and American Indian blood. Accordingly, one may argue that
dog–owner resemblance is not observed in the case of racially homogeneous (i.e., percep-
Anthrozoös

tually more similar) owner portraits. Thus, the replication of the above-chance performance
in Japanese samples would strengthen the claim of dog–owner resemblance. In our
research, not only the owners but also the judges were Japanese. This was the case mainly
in order to make it easier to recruit judges, and also to avoid the possible difficulty in the
recognition of individuals of other races (e.g., Malpass and Kravitz 1969; Goldstein and
174

Chance 1979).
AZ VOL. 22(2).qxp:Layout 1 3/31/09 10:43 PM Page 175

Nakajima et al.

The second aim of this research was to present another method for demonstrating dog–
owner resemblance. After obtaining the statistically reliable, but weak, evidence for dog–owner
resemblance in experiment 1, which was designed in a similar manner to Payne and Jaffe’s
(2005) experiment, we decided to conduct a quick and simple two-choice test with a large
number of judges, the data of which are documented in experiment 2.
Finally, we aimed to test an intuitively valid but semantically arguable assumption that
judges who are requested to match dogs to their real owners resort to dog–owner resem-
blance as a cue to solving the task. This was investigated not only by conducting post-session
interviews of the judges but also by comparing the perceived similarities of the real dog–owner
pairs that were the most and least correctly matched by the judges (experiment 1), and
by directly comparing the performance on an ownership-guessing task with that of a
resemblance-based choice task (experiment 2).

Experiment 1
This experiment is a replication of Payne and Jaffe’s study (2005). We used portraits of Japan-
ese owners and their dogs to study dog–owner resemblance in a racially homogenous owner
group. The test procedure was essentially identical to that followed by Payne and Jaffe (2005)
(i.e., the dog portraits had to be matched with the corresponding owners), except that, here,
eight sets of five pairs, rather than six sets of six pairs, were employed, because of portrait
availability. Another difference between the present experiment and Payne and Jaffe’s (2005)
study was that in this experiment, we did not explicitly instruct our judges to use resemblance
cues for solving the task.
Methods
Judges: Seventy students (undergraduate and graduate; 48 females and 22 males, between
19 and 33 years of age) from the authors’ university participated in the experiment as volun-
tary judges. They were all native Japanese and were not familiar with the dogs and their owners
in the photos. Twenty-three students had kept dogs as pets before, while the remaining 47 had
never done so. A post-session interview revealed that 52 of the students generally liked dogs,
5 disliked dogs, and the remaining 13 were neutral towards them.

Materials: The photos of the dogs and their owners were taken at an annual festival of a
dog-lovers’ society, held at a public park in mid-May, 2006. In order to prevent any selec-
tion bias, the camera crew approached all the owners that they saw until the number of
owners who let their pictures be taken reached 50. Fortunately, none of the owners ap-
proached refused to be photographed with their dogs. For the photos, the owners were
asked to smile slightly at the camera. The head-and-shoulders digital photos, taken
separately for the dogs and their owners, were edited in the laboratory using Adobe
Photoshop 7.0, to remove the background.
The photographs were then individually color-printed on 8.9  12.7 cm (height  width)
Anthrozoös

cards, and the faces of the individuals were placed in the centre. The distance between the
vertex and an individual’s chin was 5.5–6.0 cm. Seven pairs of photos were excluded be-
cause the backgrounds were either difficult to remove or at least one member of the pair was
not facing forward, thereby leaving us with 43 photo pairs for the experiment. Forty pairs of
purebred dogs and their owners were employed for the experimental trials (see Figure 1, for
175

examples), while three pairs of mongrels and their owners were employed for the practice
AZ VOL. 22(2).qxp:Layout 1 3/31/09 10:43 PM Page 176

Dogs Look Like Their Owners: Replications with Racially Homogenous Owner Portraits

trial—incidentally, an overwhelming majority of the members of the dog-lovers’ society who


participated in this study were owners of purebred dogs. Each portrait was placed in a clear,
plastic card case, to protect it from damage while handling.

Figure 1. Examples of the headshots used in the study (dogs and their owners).
Printed with written permission from the owners.

According to the interviews conducted with the owners before their photographs were
taken, the purebred dogs1 consisted of 24 females and 16 males, and their ages ranged from
6 months to 10.7 years, with a mean age of 3.6 years. Eighteen dogs in the photos had col-
lars and one dog had a ribbon on its head, but the colors and designs of the dogs’ acces-
sories were different from those of the clothes or accessories of the owners. The ages of the
40 owners (24 females and 16 males) of the purebred dogs ranged between the 20s and the
60s, with the mode being 30s—we asked the owners to mention their age in decade age
ranges, rather than provide their exact age, so as to protect their privacy. The owners were
dressed casually, and five males and five females wore caps or hats. Only one male owner
wore sunglasses, but they were placed on top of his head. The number of years for which the
owners had kept their dogs ranged from 1 to 10 (with a mean duration of 3.5 years), and the
number of hours per day that they had contact with their dogs ranged from 0 to 12 hours on
weekdays and from 1 to 24 hours on weekends. Incidentally, 28 owners (i.e., 70%) had chosen
their dogs by themselves.
The 40 purebred dog–owner pairs were semi-randomly divided into eight sets of five pairs
each, ensuring that each set included 3 female owners and 2 male owners, and that the
average size of the dogs was equivalent across the sets. In addition, no set contained dogs
of the same breed.

Procedure: The experimenter placed a set of portraits, which were arranged in the form of a
2  5 matrix, on a round table—the portraits of the dogs were in the top row and the portraits
of the owners were in the bottom row. A student (judge), sitting at the table, was then requested
to make correct dog–owner pairs within 2 minutes. The experimenter was made to face away
from the judge, to eliminate the possibility of the experimenter giving the judge cues, until she/he
Anthrozoös

(the judge) made the final decision. After this, the experimenter recorded the judge’s decision
in a notebook without giving her/him any feedback. The same procedure was repeated for all
the remaining sets of photos, without any recess. The sequence of sets was changed from
judge to judge, and the portraits in each set were shuffled before each session. The eight sets
of experimental trials were preceded by a practice trial, where three pairs of photos of mongrels
176

and their owners were employed, to instruct the judge on how to match the portraits.
AZ VOL. 22(2).qxp:Layout 1 3/31/09 10:43 PM Page 177

Nakajima et al.

The session ended with a brief interview, to determine the self-confidence of the judges in
their task performance. The judges had to rate their level of self-confidence on a 7-point scale,
ranging from “absolutely not confident” to “highly confident.” They were also asked about the
matching strategies they used, their previous experience of keeping dogs as pets, and their
affection towards dogs.
Results and Discussion
All the statistical decisions in this study were based on an alpha level set at p < 0.05, two-tailed.
Since each set of photos had 5 dog–owner pairs, the number of possible combinations was
120 (i.e., 5  4  3  2  1), consisting of 44 cases of 0 correct pairs, 45 cases of 1 correct
pair, 20 cases of 2 correct pairs, 10 cases of 3 correct pairs, and 1 case of 5 correct pairs (the
4-correct match inevitably leads to a 5-correct match by exclusion). Consequently, the theo-
retically expected chance value of the number of correct pairs per set was 1.00. Compared
with this value, the overall mean number of observed correct matches (1.11) was significantly
high (t(69) = 2.63, p = 0.0105), when the number of correct pairs averaged over the 8 sets for
the individual judges was employed as the sample data for the statistical tests.
An analysis of variance (ANOVA), with the factors of the judges’ gender and experience of
keeping dogs at home, yielded no significant outcomes (Fs(1,66) < 2.80). A separate analysis
of the judges’ affection for dogs, by comparing the dog lovers (n = 52) and the others (n = 18),
failed to reveal any influence of this factor on the task performance (t < 1). In addition, the
judges’ self-confidence was not correlated with their task performance.
None of the dogs’ and owners’ attributes, which we recorded from the interview with the
owners before their photos were taken, had any effect on how the given portrait was correctly
matched. In addition, the matching results were not affected by the dogs’ breed categories,
which were classified according to the standards of the Japan Kennel Club (almost equivalent
to those of the American Kennel Club) or by lineage, according to the standards of the
Fédération Cynologique Internationale.
If the judges had randomly paired the portraits, the expected number of judges who
matched a dog to one of the five owners in each set would have been 14 (70 judges divided
by the 5 possible options). However, there was a bias in each of the 8 sets employed (²s(16) >
56.86, ps < 0.0001), suggesting some sort of dog–owner compatibility stereotype in the judges.
Coren (1999) has reported that women with long hair tend to prefer droop-eared dogs
and that short-haired women tend to prefer prick-eared dogs. Accordingly, we classified the
female owners in our study into either a long- or short-haired group and examined the ears of
their dogs. The results were in line with Coren’s claim, although the association ( = 0.43) was
statistically marginal (Fisher exact test, p = 0.0893). Specifically, 10 out of 15 long-haired
women owned droop-eared dogs, while 7 out of 9 short-haired women owned prick-eared
dogs. Thus, our judges might have used this cue in their matching task, despite the fact that
no judge mentioned this strategy. However, even if this were the case, its role must have been
limited because the association between the length of owner’s hair and the shape of dog‘s ear
Anthrozoös

was far from perfect. Notably, Roy and Christenfeld (2004) failed to find any such association
in their study.
Only 31 out of the 70 judges (44%) voluntarily stated that they used the dog–owner phys-
ical resemblance to match the portraits, but this does not imply that the other judges did not
use resemblance as a cue to solve the task. This may be due to the fact that they might have
177

considered it obvious to use resemblance in the task. In a supplementary test, we selected


AZ VOL. 22(2).qxp:Layout 1 3/31/09 10:43 PM Page 178

Dogs Look Like Their Owners: Replications with Racially Homogenous Owner Portraits

dog–owner pairs that had been the most and least correctly matched for each male and fe-
male owner and asked 16 new judges which of the 2 pairs was more similar than the other
pair in each owner–gender set. In 24 out of 32 instances, the pairs that had been the most
correctly matched in experiment 1 were chosen (Fisher exact test, p = 0.0035), suggesting that
the perceived similarity between the dogs and their owners played a role in the matching of
the dogs and the owners in experiment 1.

Experiment 2
To further support the notion that naïve judges correctly match purebred dogs and their own-
ers, this experiment presents a new test procedure involving a large number of judges. Two
sets of 20 dog–owner pairs were prepared. One set comprised of real pairs, while the other
was a fake set (i.e., each dog was paired with another dog’s owner). The judges were required
to choose one of these sets either by guessing the real relationship or by using the dog–owner
resemblance to make their choice. With this simple test we sought to achieve more sizable
results than those obtained in experiment 1 and the previous studies (Payne and Jaffe 2005;
Roy and Christenfeld 2004, 2005) and to examine the role of task instructions, by comparing
the choice performance of judges in the ownership-guessing task with that of judges in the
resemblance-based choice task.
Methods
Judges: The participants in the experiment were 373 Japanese undergraduate students (144
females, 229 males; 18–34 years old) from the authors’ university who were attending an
introductory psychology class in a large lecture hall. None of them were familiar with the dogs
or owners in the photos used in the experiment. The number of judges who had kept dogs as
pets before was 129, while the remaining 244 had never kept dogs as pets. There were 250
dog lovers, 25 dog haters, and 98 who were neutral to dogs.

Materials: The 40 dog–owner pairs employed in experiment 1 were divided into two sets,
each consisting of 20 pairs and containing the same number of female (12 per set) and male
(8 per set) owner portraits. We tried as far as possible to ensure that these two sets were
equivalent with regard to the variety in the dogs’ appearance (breeds and colors) and the
matching outcome of the pairs in experiment 1. One of these sets was directly employed as
the “matching set,” while the pairs of the other sets were recombined by swapping owners’
photos within each gender (e.g., a female owner’s dog was paired with another female owner)
to create the “mismatching set.” Each set was arranged in a 5  4 (row  column) matrix of
dog–owner pairs, and both sets were color-printed on a sheet of paper 29.7  42.0 cm,
(height  width), with the size of each photo being 1.5  1.9 cm. The face of a dog or owner
in each photo was 1.1–1.2 cm (from the vertex to the chin). Each pair (left: owner, right: dog)
was surrounded by a 2.3  4.8 cm rectangle with a black contour for emphasis. In order to
prevent any unique features of the pairs’ positions in the matrices from affecting the results,
Anthrozoös

we prepared 10 different patterns of printed sheets, ensuring that the allotted positions for
female-owner photos (and consequently, those for male-owner photos, too) were equivalent
between the matching and mismatching sets of each sheet. Five of the ten patterns had the
matching set on the left and the mismatching set on the right, while the other five had these
sets in the opposite locations. In all the sheets, the left set was labeled “A” and the right set
178

was labeled “B.”


AZ VOL. 22(2).qxp:Layout 1 3/31/09 10:43 PM Page 179

Nakajima et al.

Each sheet also served as a questionnaire consisting of task instructions (see below) and
questions regarding the judge’s gender and age, confidence in her/his decisions (using a
7-point scale, ranging from “absolutely not confident” to “highly confident”), strategy used to
reach a final decision (open-response question), experience with regard to keeping a dog as
a pet (yes or no), the number of years for which a dog had been kept, the breed(s) of the
dog(s), general affection for dogs (like, dislike, or neutral), and the reasons for this attitude
(open-response question).

Procedure: The survey was administered simultaneously to all the participants at the begin-
ning of a class as a voluntary task. A student who returned the sheet without filling it out was
excluded from our judge sample. Each student received a photo/questionnaire sheet and was
required to read the task instructions, which were either “choose the set of real dog–owner
pairs, Set A or Set B” (the ownership task) or “choose the set of dog–owner pairs that phys-
ically resemble each other, Set A or Set B” (the resemblance task). The aim of the research was
described simply with a neutral statement “This is a survey on dog–owner relationships.”
Discussions with other students were prohibited. Furthermore, we randomly distributed 20
patterns of sheets (10 patterns for each task), as a precaution against sneak discussions and
peeking at neighbors’ sheets. The session took about 20 minutes, including the time for the
distribution and collection of all the sheets.
Results and Discussion
Since the number of collected sheets did not vary significantly among the 20 patterns (range,
16–20 sheets per pattern) and the patterns did not affect the task performance, this factor
was collapsed for the following analyses. For the ownership task, 115 out of 186 judges
(62%) correctly chose the matching set (Fisher exact test, p = 0.0015), against the chance
level of 50%. Similarly, 124 out of 187 judges (66%) who performed the resemblance task
chose the matching set (Fisher exact test, p < 0.0001). In the other words, our students
chose the set consisting of matched dog–owner pairs rather than the set consisting of mis-
matched dog–owner pairs, irrespective of whether the task instructions were to choose the
real-pair set or the similar-pair set. Furthermore, because these percentages did not differ
significantly (Fisher exact test, p = 0.3891), we can safely argue that the type of instructions
had no effect on the judges’ choice, although only 65 out of 186 judges (35%) who per-
formed the ownership task voluntarily mentioned that they had used dog–owner resem-
blance to solve the task. In total, 239 out of 373 judges (64%) correctly chose the real
dog–owner pair set (Fisher exact test, p < 0.0001).
The task performance was independent of the gender of the judges, their reported confi-
dence in their decision, their experience of keeping dogs as pets, the number of years for
which they had kept dogs as pets, the breeds of the dogs, and their affection for dogs. The
factor of the judges’ ages was not assessed, because almost all the judges were within a nar-
row range (18–20 years old). Interestingly, the judges who reported that their decision was
Anthrozoös

based on some facial cues outperformed the other judges—121 out of 168 judges (72%) who
mentioned the facial cues were correct, while only 118 out of the remaining 205 judges (58%)
were correct (Fisher exact test, p = 0.0047). However, further analyses concerning the reported
details (e.g., facial parts and contours) were unsuccessful in identifying the critical factor
underlying this result.
179
AZ VOL. 22(2).qxp:Layout 1 3/31/09 10:43 PM Page 180

Dogs Look Like Their Owners: Replications with Racially Homogenous Owner Portraits

General Discussion
This research shows that the facial resemblance of purebred dogs with their owners (Roy and
Christenfeld 2004, 2005; Payne and Jaffe 2005) is also confirmed in Japan, where the owners
are more racially homogenous than those that have been studied in the previous studies. The
replication of an above-chance matching performance with even less diversity in owners’ faces
strengthens the claim of dog–owner resemblance. Furthermore, the same conclusions from
three independent laboratories using different sets of photographs reduces the possibility that
the obtained positive results are peculiar to the portraits employed. The reliability of dog–owner
resemblance is also supported by the fact that it has been demonstrated with a variety of test
procedures employed by these studies.
The ability to match correct dog–owner pairs (experiment 1) and the sensitivity to
differentiate between real and fake dog–owner pairs (experiment 2) were not limited to spe-
cial groups of people (e.g., dog lovers), because the task performance was independent of any
measured judge-related factor including gender, reported confidence in the decision,
experience with respect to keeping dogs as pets, the total number of years for which dogs
had been kept as pets, breeds of the dogs kept as pets, and affection for dogs. Incidentally,
the previous studies (Roy and Christenfeld 2004, 2005; Payne and Jaffe 2005) did not mention
how the judge-related factors affected task performance.
Next, we discuss the roles of the attributes of the dogs, owners, and their relationship
between the two in perceived dog–owner resemblance. One study reports a dog–owner
compatibility stereotype of matching a cat or a small dog to a female owner, and a large dog
to a male owner (Budge et al. 1997). However, Payne and Jaffe (2005) reported that an
owner’s age and gender had no direct effect on the results of dog–owner matching.
Experiment 1 of the present research replicated their findings—no examined attributes of the
dogs, owners, and their relationship had any significant effect on how the given portraits
were correctly matched.
Two non-significant effects of the factors examined in experiment 1 require special men-
tion. First, the matching accuracies of the real dog–owner pairs were not correlated with the
years of their companionship. Having obtained the same kind of results, Roy and
Christenfeld (2004) argued that dog–owner resemblance is due to the selection of dogs by
their owners rather than the convergence of appearance over time. However, such a con-
clusion might be premature, or at least tentative, because the general performance was not
good enough to allow analysis of the individual factors in Roy and Christenfeld’s (2005)
study, as well as ours.
Second, one may question why the portraits of the owners who had chosen their dogs by
themselves did not yield a better matching outcome, than did the portraits of the owners who
had played no critical role in choosing their dogs. If people seek dogs that look like them, the
matching outcome should have been better if they had chosen their dogs by themselves than
if they had not. A plausible answer to this question is that the key person in each of the cases
where the dog was not chosen by the owner was a member of the owner’s family. In other
Anthrozoös

words, the owner and the key person were genetically similar or were alike in terms of the
“self seeks like” mechanism (Spuhler 1968) or through the convergence of appearance (Za-
jonc et al. 1987). Consequently, the owner should be similar to her/his dog by way of the key
person, even if she/he played no role in dog selection. Of course, such an “indirect” resem-
blance would be lesser than the direct dog–owner resemblance, although a more powerful test
180

would be necessary to detect the difference.


AZ VOL. 22(2).qxp:Layout 1 3/31/09 10:43 PM Page 181

Nakajima et al.

As mentioned previously, Roy and Christenfeld (2004) claimed that owners choose dogs
that look like themselves, but they do not provide the reason for this preference. Payne and
Jaffe (2005), however, claim that the assortative mating system for selecting a conspecific
mating partner is also at play when owners choose dogs as their companions. In contrast to
such a nativist account, Coren (1999) has argued that we prefer ourselves due to a mere pre-
exposure effect (Zajonc 1968) (because we see our face in the mirror many times) and
suggests that the same preference may affect our choice of dog. Unfortunately, the present
study cannot answer the question of which of these accounts is valid to explain the mechanism
of dog–owner resemblance.

Acknowledgements
We are grateful to the dogs and their owners for being photo models, and thank the KNOTS
for granting the camera crew, consisting of the second author and her friends (Hisae Mo-
tokawa, Kanako Tatsumi, and Mariko Mori), the permission to take photos at their annual
event. We also express our thanks to Tsuneo Shimazaki, Takahisa Masaki, Yumiko Ogawa,
Takatoshi Nagaishi, and Gentaro Shishimi for their technical support and for their comments
on this study. This research was financially supported by a grant from MEXT to the Center for
Applied Psychological Science, Kwansei Gakuin University.

Note
1. The 40 purebred dog photos employed in this research were as follows: 3 Toy Poodles, 2 Pugs, 2 Papillons,
1 Japanese Terrier, 1 Yorkshire Terrier, 1 Miniature Pinscher, 1 Pomeranian, 2 Shiba Inus, 1 Kishu, 1 Siberian
Husky, 2 Pembroke Welsh Corgis, 2 Border Collies, 1 Belgian Tervuren, 3 Miniature Dachshunds, 1 Beagle,
2 Labrador Retrievers, 1 Flat-Coated Retriever, 1 Golden Retriever, 1 Irish Setter, 1 English Cocker Spaniel,
1 Weimaraner, 2 Jack Russell Terriers, 2 Wirehaired Fox Terriers, 1 Airedale Terrier, 1 Boston Terrier, 1 Dal-
matian, 1 French Bulldog, and 1 Standard Poodle.

References
Budge, R. C., Spicer, J., St. George, R. and Jones, B. R. 1997. Compatibility stereotypes of people and pets:
A photograph matching study. Anthrozoös 10: 37–46.
Coren, S. 1999. Do people look like their dogs? Anthrozoös 12: 111–114.
Goldstein, A. G. and Chance, J. 1979. Do “foreign” faces really look alike? Bulletin of the Psychonomic Society
13: 111–113.
Levine, D. W. 2005. Do dogs resemble their owners?: A reanalysis of Roy and Christenfeld (2004). Psycholog-
ical Science 16: 83–84.
Malpass, R. S. and Kravitz, J. 1969. Recognition for faces of own and other race. Journal of Personality and
Social Psychology 13: 330–334.
Payne, C. and Jaffe, K. 2005. Self seeks like: Many humans choose their dog pets following rules used for
assortative mating. Journal of Ethology 23: 15–18.
Roy, M. M. and Christenfeld, N. J. S. 2004. Do dogs resemble their owners? Psychological Science 15: 361–363.
Roy, M. M. and Christenfeld, N. J. S. 2005. Dogs still resemble their owners. Psychological Science 16: 743–744.
Spuhler, J. N. 1968. Assortative mating with respect to physical characteristics. Eugenics Quarterly 15: 128–140.
Zajonc, R. B. 1968. Attitudinal effects of mere exposure. Journal of Personality and Social Psychology,
Anthrozoös

Monograph Supplement 9(2): 1–27.


Zajonc, R. B., Adelmann, P. K., Murphy, S. T. and Niedenthal, P. M. 1987. Convergence in the physical
appearance of spouses. Motivation and Emotion 11: 335–346.
181

You might also like