Correction: 10 June 2015

www.sciencemag.org/content/348/6232/333/suppl/DC1

Supplementary Materials for

Oxytocin-gaze positive loop and the coevolution of human-dog bonds
Miho Nagasawa, Shouhei Mitsui, Shiori En, Nobuyo Ohtani, Mitsuaki Ohta, Yasuo
Sakuma, Tatsushi Onaka, Kazutaka Mogi, Takefumi Kikusui*
*Corresponding author. E-mail: kikusui@azabu-u.ac.jp

Published 17 April 2015, Science 348, 333 (2015)

DOI: 10.1126/science.1261022

This PDF file includes:

Materials and Methods

Figs. S1 to S5
Tables S1 to S4
Data Tables S1 and S2
Captions for Movies S1 to S3
References (2830)

Other Supplementary Materials for this manuscript include the following:

(available at www.sciencemag.org/content/348/6232/333/suppl/DC1)

Movies S1 to S3

Correction: Some numbers of dogs were corrected in the results of experiment 1 (cluster
analysis). Fig. S1 and its legend, Fig. S2, and Table S1A were replaced with corrected
versions.
 Submitted Manuscript: Confidential 7th September 2014

Supplementary Materials:
Materials and Methods

[Experiment 1]
Nagasawa et al. (19) found that dog owners could be divided into 2 groups based on the
duration of the dogs gaze at its owner and the degree of communication owners reported
having with their dogs on a questionnaire. Owners who received a longer duration of gaze
from their dogs and reported a higher degree of communication with their dogs had higher
urinary oxytocin concentrations than owners who received a shorter-duration gaze and
reported a lower relationship. Therefore, we examined changes in urinary oxytocin
concentrations in dogs, as well as owners, using the same procedure as in the previous study
(19). We conducted this experiment in wolves in order to demonstrate the evolutionary
differences between dogs and wolves regarding social bonding to owners. In addition, we
examined in the dogs and wolves the association between our results and early-life
experience with humans in order to test the possibility that the observed changes in oxytocin
release and behavior were influenced by such experiences.

Experimental Procedures
Participants
Participants included 30 volunteers (6 males, 24 females; 36.6 13.7 years old) and their
dogs (2 gonadally intact male, 13 castrated, 4 gonadally intact female, 11 spayed; 4.7 2.7
years old; 5 Golden Retrievers, 3 Labrador Retrievers, 3 Miniature Schnauzers, 3 Standard
Poodles, 2 Miniature Dachshunds, 2 Shiba-inu, 2 Toy Poodles, Border Collie, Boxer, Flat-
Coated Retriever, German Shepherd, Jack Russell Terrier, Miniature Bull Terrier, Papillon,
Shetland Sheepdog, 2 mongrels). Recruitment was conducted in dog training classes,
veterinary clinics, and dog parks. Six pure wolves and 5 wolves with little possibility of
genetic contamination with dogs (< 1%) were tested to the same experimental procedures.
The wolves we used were not kept as pets. Their owners were animal management
professionals that were authorized by the local government to keep wolves. Their facilities
satisfied all requirements for keeping wolves. They acquired wolf puppies officially from
abroad, and also bred and reared wolves independently according to the standard methods
(27). They were hand-raised in houses by individuals and had daily contact with their owners,
who fed them, played with them, and slept in the same room. Participants were aware of the
procedure of the experiment but blind to its purpose. All participants were healthy and not
taking prescription medication.

Procedures
Experiments for dogs were conducted in an experimental room (5 x 4.5 m) at Azabu
University which was entirely unfamiliar to dogs. The procedure primarily followed that of
Nagasawa et al. (19). Dog owners were asked to avoid eating and drinking for 2 h before the
experiment. An hour before the experiment, the owners urinated and rested in the waiting
room. Their dogs urinated 2 h before and rested in the same room without interference of
each other until urine sampling just before interaction. During this resting period, stress signs
of dog were not recognized. Most of them were relaxed and laid down on the floor, some of
them even started sleeping. Immediately before the 30-min interaction between the owner
and his/her dog, the owner was asked to provide a urine sample. An experimenter also
collected a urine sample from the dog. The experimenter took the dogs outside and waited
until they urinated spontaneously. When dogs urinated, the experimenter caught their urine
with a tray and stored urine samples in a -80 C deep freezer as soon as possible. During the
30-min interaction, the owner was instructed to remain sitting in the chair, while the dog was
allowed to move freely in the room. The owner was forbidden to give the dog food or toys.
With the exception of these instructions, the owner was able to interact freely with the dog.
After the 30-min interaction, the owner and dog rested for 30 min. A post-experiment urine
sample was collected from both the owner and dog right after the rest. All dogs urinated
within 30 min-1 hour after the interaction. The experiments with wolves were conducted in a
space that was entirely unfamiliar to the animals (novel room or pen, Movie S2). One wolfs
urine could not be collected until 2 hours after interaction with his owner because he fell
asleep during the rest time. For wolves, urine was collected using a piece of sterilized
absorbent cotton (5 cm x 5 cm) attached to a 20-cm stick. We did not use trays with wolves,
as they may have been afraid of the tray. Wolf urine was stored in a cool box with dry ice
until it was transferred to a -80 C deep freezer. The other procedures were conducted in the
same manner as the experiment with dogs.

Assessment
Behaviors of owners and their dogs/wolves
Two video cameras were placed in the corners and a video camera was placed on the
ceiling of the experimental room to record behavior (Gigashot R60 GSC-R60, Toshiba,
Japan; Gopro HERO3, Gopro inc., US). We measured the duration that the dog/wolf gazed at
the owner (dog/wolf-to-owner gaze), and that the owner talked to the dog/wolf (dog/wolf-
talking), touched the dog/wolf (dog/wolf-touching). Dog/wolf-to-owner gaze was defined as
an orientation of the dog/wolfs nose toward the humans face, which was characterized by
lifting of the head; this definition was modified from that of Gcsi et al. (29). Dog/wolf-
touching was defined as the owners voluntary touching of the dog/wolf. Incidental physical
contact was excluded.
Because the majority of active interactions between owners and dogs/wolves tended to
occur in the first 5 min of a 30-min interaction in most cases, the behaviors recorded during
the first 5 min were analyzed by two persons who were blind to details of this study. The
scores by the two observers were highly correlated (rs = 0.95-1.00, p < 0.01). Most of dogs
did not bark or touch their owner; therefore, we excluded these behaviors from the analysis.

Measurement of urinary oxytocin in owners and dogs/wolves

In this study, the urine collection procedures were different for dogs and wolves. For
wolves, urine was collected on a cotton swab, whereas for dogs, the urine was collected
directly in a tray. Another difference was in the duration until post-urine collection: urine was
collected from all the dogs and wolves within 1 hour after the interaction, expect for one wolf
which urinated and was collected 2 hours after the interaction with his owners. To assess
whether the collection by tray or by cotton swab affected the measurement of oxytocin and
creatinine, we collected urine from 28 dogs using both procedures simultaneously. There
were no significant differences (oxytocin: U = 432, p = 0.512; creatinine: U = 386, p = 0.921)
and very high correlations (oxytocin: rs = 0.953, p < 0.001; creatinine: rs = 0.974, p < 0.001)
between the tray urine and cotton urine. The intra-assay coefficients of variation (CV) of the
oxytocin and creatinine assays were 3.57% and 5.99%, respectively. In addition, we
reanalyzed the raw data for urinary oxytocin from our previous study, in which plasma
oxytocin concentration reached a maximum 15 min after injection and the peak urinary
oxytocin concentration occurred at Time 60 and returned to baseline levels slowly (18). We
analyzed the correlation between cumulative urinary oxytocin concentration from injection to
90 min after and that from injection to 150 min, and found significantly high correlation (rs =
0.970, p = 0.001). Therefore, we were able to exclude the possibility that these different urine

1
collection procedures influenced the measurement, and we proceeded to analyze the wolves
and dogs urine together.
Immediately after collection, urine samples were frozen at -80C until the assay was
performed. After thawing urine samples were centrifuged at 4C in a refrigerated centrifuge,
and urinary oxytocin concentrations were measured by radioimmunoassay (18). Urinary
oxytocin concentrations are expressed as the oxytocin-to-creatinine ratio.

Early-life experience with humans in dogs/wolves

We asked dog owners where they had acquired their dogs and the dogs ages (in weeks)
upon acquisition. Some dogs came from facilities such as experimental laboratories and
consequently had minimal early-life interaction with humans (lower socialization group).
These dogs were compared with dogs that received direct care from humans from an early
developmental period and lived in houses (normal socialization group). We excluded from
this analysis the dogs of which early-life experience data were missing (Table S1-A). We also
obtained the following information from the wolf owners: 1) the age at which the wolf
puppies were pulled from their mothers to live with humans, 2) how much time they
interacted/interact with humans (the developmental period and adulthood), 3) the age at
which the puppies left their owners house (Table S1-B).

Statistical analysis
A hierarchical cluster analysis with Wards method was used to divide participants. A
linear mixed model (LMM) was conducted and if significant differences were shown, a post
hoc analysis with a Bonferroni correction was performed. We also conducted a power
analysis and used resampling tests for the comparison of wolves with other groups because
the sample size for wolves was small (Table S4). We only presented results that were
significant in the resampling tests. The relationships between the variables were examined
using a Kruskal Wallis test, Spearmans rank correlation coefficient (two-tailed), and
multiple linear regression analysis. After a Kruskal Wallis test, Mann-Whitney U tests were
conducted with Bonferroni correction. Data from dogs and wolves were separately analyzed
using correlation and regression. The results are expressed as mean SE (parametric) or
median quartile (non-parametric) (SPSS v.19.0).

Results
Cluster analysis of duration of dogs gaze and owners touch
In accordance with our hypothesis of a positive loop from which attachment and nursing
behaviors mutually arise, we divided participants with a hierarchical cluster analysis based on
two behavioral variables: the duration of the dog-to-owner gaze and the duration of dog-
touching. Visual examination of the dendrogram showed that the participants could be
divided into two separate groups as we previously described (Fig. S1) (19). The number of
participants in each group was as follows: Cluster 1, n = 22 (dog: 2 gonadally intact male, 9
castrated, 4 gonadally intact female, 7 spayed, 4.1 0.6 years old; owner: 5 males, 17
females, 36.6 3.3 years old), and Cluster 2, n = 8 (dogs: 4 castrated, 4 spayed, 5.9 0.7
years old; owners: 1 male, 7 females, 36.6 4.0 years old). When these two behavioral
variables were compared between the two clusters using a Mann-Whitneys U-test,
significant differences were found in both behavioral variables (duration of the dog-to-owner
gaze, U = 0, p < 0.001; and the duration of dog-touching, U = 29, p = 0.008). A longer
duration of both variables was observed in Cluster 2, which was named the Long Gaze group
(LG), relative to that of Cluster 1, which was named the Short Gaze group (SG). There were
no significant differences between the two groups in breed and sex of the dogs, or age of
owners or dogs.

2
Differences in behavioral variables among LG dogs, SG dogs, and wolves
We excluded the data from two owner-dog dyads due to outlier in urinary oxytocin
concentrations in the owners (LG: n = 8, SG: n = 20). We conducted a LMM using the
factors of dog/wolfs sex and group (LG, SG, and wolves) for each behavioral variable.
We observed the following significant main effects and interaction. For dog/wolf-to-owner
gaze, there was a significant main effect of group [F(2,22) = 289.367, p < 0.001]. For
dog/wolf-touching, there was a significant main effect of group [F(2,22) = 5.621, p = 0.005]
and a significant interaction between dog/wolfs sex and group [F(2,22) = 6.194, p = 0.003].
For dog/wolf-talking, there was a significant main effect of group [F(2,22) = 3.317, p =
0.042]. A post hoc test with a Bonferroni correction showed the duration of dog-to-owner
gaze in the LG group was the longest among the three groups (SG: p < 0.001, wolves: p <
0.001), and that wolves gazed less than did the other groups (SG: p < 0.001). LG owners
touched and talked to their dogs more than did SG owners (touch: p = 0.018, talk: p = 0.036).
The owners of male wolf touched their wolves more than did the owners of female wolf (p =
0.005). These results were shown in Fig. 1-A and Fig. S2 (Movie S1, S2).

Differences in oxytocin concentrations between the three groups

We were able to successfully collect urine from 5 out of 11 wolves, and these samples
were subsequently used for oxytocin measurement. A LMM analysis using dog/wolfs sex
and time of urine collection (pre and post) as factors was performed for owners and dogs.
No significant effect of sex was observed in owners or dogs/wolves; therefore, we conducted
the following analysis without regard to dog/wolfs sex.
To compare urinary oxytocin concentrations, a LMM was conducted using the factors
group (LG, SG, and wolves) and time of urine collection (pre and post). In the owners,
there was a significant main effect of time of urine collection [F(1,25) = 4.391, p = 0.043]
and a significant interaction between group and time of urine collection [F(2,25) = 4.016, p =
0.027]. In dogs/wolves, there was a significant main effect of group [F(2,30) = 12.005, p <
0.001]. A post hoc test with a Bonferroni correction showed that oxytocin concentration after
the 30-min interaction significantly increased in LG owners (p = 0.004, Fig. 1-B). Oxytocin
concentrations in wolves were significantly higher than those of SG dogs (p < 0.001) and LG
dogs (p = 0.001, Fig. 1-D). We also compared the oxytocin change ratio between LG, SG,
and wolves, and found significant differences [owners: 2(2) = 13.937, p = 0.001, dogs: 2(2)
=7.268, p = 0.026]. The oxytocin change ratio was significantly higher in LG than in SG
owners (U = 9, p < 0.01, Fig. 1-C) and dogs (U = 28, p < 0.05, Fig. 1-E). The oxytocin
change ratio was also higher in LG owners than in wolf owners (U = 10, p < 0.05). The intra-
assay coefficient of variation (CV) of the oxytocin assay was 5.23%. Raw data of wolf
owners and wolves are shown in Fig. S3.

Association between oxytocin concentrations and behavior variables in dogs/wolves and

owners
A Spearmans rank correlation analysis showed that the oxytocin change ratio in response
to the 30-min interaction in owners significantly correlated with that of their dogs (rs = 0.847,
p < 0.001), the duration of dog-touching (rs = 0. 455, p = 0.015), and the duration of dog-to-
owner gaze (rs = 0.534, p = 0.003). The oxytocin change ratio in dogs also correlated with
the duration of dog-touching (rs = 0.421, p = 0.026) and the duration of dog-to-owner gaze
(rs = 0.422, p = 0.025). The duration of dogs gaze was significantly correlated with that of
dog-touching (rs = 0.330, p = 0.013) and dog-talking (rs = 0.288, p = 0.031) (Table S2-A). In
wolves, a significant correlation was only found between wolf-talking and the oxytocin
change ratio of wolf owners, and wolf-talking and wolf-to-owner gaze (oxytocin change

3
ratio: rs = 0.670, p = 0.024, gaze: rs = 0.612, p = 0.002). These results are shown in Table
S2-B.

Multiple regression on owner and dog/wolf oxytocin concentrations

A multiple linear regression analysis was conducted to analyze the effect of owner/dog
behavior on oxytocin concentrations in owners and dogs. We attempted to predict the
oxytocin change ratio by using the duration of dog-talking, dog-touching, and dog-to-owner
gaze. Results indicated that the oxytocin change ratio in both owners and dogs showed
significant equalities [owners: R2 = 0.306, F(3,24) = 4.959, p = 0.008; dogs: R2 = 0.247,
F(3,24) = 3.957, p = 0.020]. In both cases, the duration of dog-to-owner gaze was
significantly related to the oxytocin change ratio (owner: = 0.458, p = 0.015; dog: = 0.388,
p = 0.045). These results are shown in Table 1-A. For wolves, the duration of wolf-talking
was excluded from analysis because multicollinearity was found between wolf-to-owner gaze
and wolf-talking. No significant equalities were found in wolves and owners (Table S3).

Influence of early-life experience with humans

We examined the correlation between dogs age when acquired by their owners and the
oxytocin change ratio and behavioral variables. We also compared these variables between
the lower and normal socialization groups. The results did not indicate any significant
differences. For wolves, we examined the association between the indexes of wolves early-
life experience (13 in Early-life experience with humans in dogs/wolves, Assessment) and
the oxytocin change ratio and behavioral variables, and did not find any significant
correlation or differences (Table S1-C).

[Experiment 2]
Experiment 1 showed a positive relationship between duration of dog-to-owner gaze and
urinary oxytocin concentration in owners and dogs during a 30-min normal interaction. In
contrast, there was no such increase of oxytocin in wolf-owner dyads, suggesting that dogs
are able to establish a biological bond with their owners that is facilitated and maintained
through mutual gaze. Given this, in the next step, oxytocin was nasally administered to dogs
to directly examine whether oxytocin enhanced dog-to-owner gazing behavior, and whether
an increase in this behavior affected urinary oxytocin concentrations in owners. Participants
were additionally instructed to refrain from interacting with the dogs to reveal the causal
direction of the human-dog bond by exogenous manipulation of oxytocin on the oxytocin-
mediated positive loop using mutual gazing between humans and dogs, ensuring that urinary
oxytocin concentrations in dogs were not affected by extraneous stimuli.

Experimental Procedures
Preparatory experiment
To evaluate the effectiveness of oxytocin nasal administration in dogs, we measured heart
rate variability, plasma or urinary concentrations of oxytocin in dogs that received oxytocin
(40 IU oxytocin/100 l) or saline treatment. Results showed that following administration of
both oxytocin and saline, the average sympathetic nerve (LF/HF) index increased rapidly, and
rapidly decreased 10 min after oxytocin administration; thus, the availability of nasally
administered oxytocin to the central nervous system in dogs was suggested. Dogs treated with
oxytocin also displayed a significant increase in plasma oxytocin relative to saline post-spray,
which might indicate an activation of central oxytocin neurons. Plasma oxytocin
concentrations after oxytocin administration were significantly elevated at 15 min relative to
both baseline and saline conditions. Oxytocin concentrations at the end of the experimental

4
session were also higher after oxytocin administration than after saline administration. This
variation in plasma oxytocin after spray intake was also reflected in the oxytocin
concentrations of the urine samples collected before and after the experimental sessions. We
found a significant positive correlation between the accumulative value of plasma oxytocin
and the increase in urinary oxytocin concentrations. Furthermore, posttest urinary oxytocin
concentrations in dogs were significantly elevated following intranasal oxytocin
administration compared to pretest oxytocin concentrations of the same experimental session
and posttest oxytocin concentrations after saline spray. These results were published in
Romero et al. in 2014 (29).
Similarly, we also collected plasma and urine samples from humans that received oxytocin
or saline treatment (male: n = 8, age: 29.6 3.2 years) in order to examine whether urinary
oxytocin concentrations were reflecting plasma oxytocin concentrations. Plasma and urine
samples were collected just before nasal administration (baseline) and after oxytocin (24 IU
oxytocin/100 l) or saline administration. Following this, plasma and urine were collected
every 15 and 30 min, respectively. The experimental session was a within-subjects design,
and subjects were blind to their treatment condition. The generalized linear models were
conducted using administration (oxytocin and saline) and time of collection (every 15
min) as factors [plasma: 2(13) = 58.634, p < 0.001; administration: 2(1) = 23.922, p <
0.001; time of collection: 2(6) = 20.539, p = 0.002; administration and time of collection:
2(6) = 16.686, p = 0.011; urine: 2(9) = 47.734, p < 0.001; time of collection: 2(4) = 59.058,
p < 0.001; administration and time of collection: 2(4) = 10.030, p = 0.040]. When subjects
were treated with oxytocin, plasma oxytocin concentrations tended to increase 15 min after
administration relative to baseline (p = 0.074, Fig. S5-A). Meanwhile, urinary oxytocin
concentrations began to increase 60 min after oxytocin administration (p = 0.086, Fig. S5-B).
This result was consistent with Amicos previous study, which showed that urinary oxytocin
concentrations increased approximately 30 min after intravenous administration of oxytocin
(30). Moreover, there was a significant positive correlation between the accumulative plasma
oxytocin concentrations from 15 to 90 min and urinary oxytocin concentrations from 30 to
120 min (rs = 0.521, p = 0.046, Fig. S5-C). Therefore, as the urine sampling procedure in
Experiment 1 was appropriate for the evaluation of oxytocin activity, this method was
adopted in Experiment 2.

Participants
This experiment involved 30 volunteers (male: n = 7; female: n = 23; age: 39.6 16.4
years) and their dogs (2 gonadally intact male, 13 castrated, 2 gonadally intact female, 13
spayed; age: 5.65 0.8 years; 7 Standard Poodles, 6 Golden Retrievers, 6 Labrador
Retrievers, 2 Beagles, 2 Border Collies, 2 Schnauzers, Bernese Mountain Dog, Boxer,
German Shepherd Dog, Shetland Sheepdog, Siberian Husky). Participants were recruited in
dog training classes, veterinary clinics, dog runs, and through the Internet. We also recruited
students and staff from the university who were similar to owners in sex and appearance, but
unfamiliar to the dogs, to serve as unfamiliar people during the 30-min interaction. Our
reasons for including two unfamiliar people in this experiment were to examine whether
oxytocin-induced behavioral changes in the dog were specific to its owner. Owners and the
unfamiliar people were aware of the procedure of the experiment, but blind to its purpose and
treatment.

Procedures
Experiments were conducted in an experimental area (4.5 4.5 m) that was divided by
partitions in an experimental room at Azabu University, or unfamiliar and similar size of
space for subjects. Three chairs were set in a circle in the experimental area, and marks were

5
placed at a 1 m radius from each chair with vinyl tape. This experiment consisted of three
phases: resting before the interaction, a 30-min interaction between the participants and dog,
and resting after the interaction. Two video cameras were placed in the corners and two video
cameras were placed on the ceiling of the experimental room to record the behaviors of the
dog during the 30-min interaction (Gigashot R60 GSC-R60, Toshiba, Japan; PC-355micro,
Sun-Mechatronics, Japan) (Fig. S4). Oxytocin and saline treatments were conducted on
separate days in each participant, at least one month apart. The owners were asked to avoid
eating and drinking for 2 h before the experiment. The owners urinated 1 h before the 30-min
interaction, and rested on the chair with the dog without interference of each other until urine
sampling just before 30-min interaction. Just before the interaction, a urine sample from the
dog was collected. While the owner was providing a urine sample, two unfamiliar people
entered the experimental area and sat on the chairs. We adopted a three-way choice test,
which included the owner and two unfamiliar people, to increase the sensitivity of preference
and avoid incorrect interpretation of results. The order of seating was randomized by using
the randomize function in Excel (Microsoft).
After the owner was seated, 100 l of oxytocin (40 IU) or 100 l of saline was
administered to the dog using a hand-compressing air spray bottle. The order of
administration was counterbalanced. Following oxytocin or saline administration, the dog
entered the experimental area. All participants were instructed to remain seated in chairs,
while the dog was allowed to move freely in the room. In order to prompt the dogs
movement during the interaction, participants were instructed to change their seats every 10
min. They were forbidden to talk to each other or to touch the dog voluntarily; however, if
the dog touched participants, the participants were allowed to pet the dog in return. After the
30-min interaction, the unfamiliar people left the experimental area. The owner and the dog
stayed and rested, and a second urine sample was collected from both 30 min after the
interaction.

Assessment
Dog behavior
The dogs behaviors were recorded by video cameras during the 30-min interaction. We
measured the total amount of time that the dog touched participants (touch), were in close
proximity (dogs head or body was within a 1-m radius from a participant) to participants
(proximity), and that the dogs nose oriented toward the participants face, as in Experiment 1
(gaze). These videos were analyzed by two people who were blind to the details of this study.
The scores by the two observers were highly correlated (rs = 0.95-1.00, p < 0.01). The
participants measured with stopwatches the total amount of time that they met the dogs eyes
(mutual gaze, which was defined as the orientation of dogs nose toward the participants
face or the dogs eyelids and eyebrows lifted to see the participants face), and this
significantly correlated with gaze duration (rs = 0.892, p < 0.001). The average of each
behavioral variable towards the two unfamiliar people was used for analysis.

Measurement of urinary oxytocin in owners and dogs

Urine was collected from both owners and dogs just before and 30 min after the 30-min
interaction. The methods of preservation and analysis were the same as those in Experiment 1.

Early-life experience with humans in dogs/wolves

We also asked dog owners the same questions about their dogs early-life experience with
humans as were asked in Experiment 1 (Table S1-A).

Statistical analysis

6
 As with Experiment 1, a LMM was conducted to compare behaviors and oxytocin
concentrations, and if significant differences were shown, a post hoc analysis with a
Bonferroni correction was conducted. A Wilcoxon signed ranks test was conducted to
compare the oxytocin change ratio. The relationships between the variables were examined
using a Spearmans rank correlation analysis (two-tailed) and multiple linear regression
analysis. Results were expressed as mean SE (parametric) or median quartile (non-
parametric) (SPSS v.19.0).

Results
Differences in behavioral variables between the two treatments
We excluded the data from three ownerdog dyads because of outliers (defined as values
more than 3 standard deviations from the mean) in urinary oxytocin concentrations in the
owners or dogs. We conducted a LLM using the factors dogs sex, participant (owner
and unfamiliar people), and administration (oxytocin and saline) for each behavioral
variable. We observed the following significant main effects and interactions. For gaze, there
were significant main effects of participant [F(1,75) = 22.285, p < 0.001], and significant
interactions between dogs sex and administration [F(1,75) = 3.981, p = 0.050] and dogs sex,
participant, and administration [F(1,75) = 4.267, p = 0.042]. For touch, there were no
significant effects. Finally, for proximity, there was a significant main effect of participant
[F(1,75) = 68.006, p < 0.001]. A post hoc test with a Bonferroni correction showed the sex
difference was evident in gazing behavior following oxytocin treatment (p = 0.05). Dogs
exhibited gazing and proximity for a significantly greater amount of time towards their
owners than towards unfamiliar people (gaze: p < 0.001, proximity: p < 0.001). Female dogs
gazed at their owners significantly longer than they did at unfamiliar people in oxytocin
administration (p < 0.001), while male dogs gazed at their owners significantly longer than
unfamiliar people in saline administrations (p = 0.006). Female dogs also gazed at their
owners significantly longer following oxytocin administration than following saline
administration (p = 0.006, Fig. 2-A).

Differences in oxytocin concentrations between the two treatments

To compare urinary oxytocin concentration in owners and dogs, LLMs were conducted
using dogs sex, administration to dogs (oxytocin and saline), and time of urine
collection (pre and post) as factors. In the owners, results showed significant main effects of
dogs sex [F(1,25) = 6.071, p = 0.018], administration [F(1,75) = 4.909, p = 0.032], and time
of urine collection [F(1,75) = 7.950, p = 0.007]. Analysis of owners showed significant
interactions between dogs sex and time of urine collection [F(1,75) = 4.545, p = 0.038];
administration and time of urine collection [F(1,75) = 9.558, p = 0.003]; and dogs sex,
administration, and time of urine collection [F(1,75) = 4.233, p = 0.045]. A post hoc test with
a Bonferroni correction indicated a significant increase in oxytocin concentrations in owners
of female dogs in the oxytocin condition (p < 0.001, Fig. 2-D). Further, post-interaction
oxytocin concentrations in these owners were significantly higher in the oxytocin condition
relative to the saline condition (p = 0.001, Fig. 2-D), and post-interaction oxytocin
concentrations were significantly higher in the owners of female dogs than in the owners of
male dogs in the oxytocin treatment group (p = 0.001, Fig. 2-D). We also compared the
change ratio of oxytocin between oxytocin treatment and saline in each sex; we found that the
oxytocin change ratio in the owners of female dogs following oxytocin administration was
significantly higher than that following saline administration (Z = 2.353, p = 0.019, Fig 2-E).
No significant difference in dogs was observed (Fig. 2-F, 2-G). The intra-assay coefficient of
variation (CV) of the oxytocin assay was 6.12%.

7
Association between oxytocin concentration and behavioral variables
A Spearmans rank correlation analysis was conducted to examine the correlation between
oxytocin concentrations and dog behavior. No significant correlations were found in any of
the measured variables.

Multiple regression of oxytocin concentrations in owners and dogs

To analyze the effect of owners and dogs behaviors on oxytocin concentrations in owners
and dogs, a multiple linear regression analysis was conducted. We attempted to predict the
oxytocin change ratio in owners by using the following factors: dogs sex, oxytocin
administration to dogs, duration of touch, duration of proximity, and duration of gaze. Results
indicated that the oxytocin change ratio in owners showed a significant regression equation
[owners: R2 = 0.248, F(5,48) = 4.031, p = 0.005]. The duration of the dogs gaze was
significantly related to the oxytocin change ratio ( = 0.458, p = 0.006). In dogs, no
significant equation was observed [R2 = -0.046, F(5,48) = 0.619, p = 0.686]. These results are
shown in Table 1-B.

Influence of early-life experience with humans

We examined the correlation between dogs age when acquired and the oxytocin change
ratio and behavioral variables, and compared these variables between the lower and normal
socialization groups. As the results, there was neither a significant correlation nor difference
(Table S1-C).

8
Fig. S1. Cluster analysis in Experiment 1
The pairs of dogs and owners were divided into two clusters (Short Gaze group: n = 22, Long
Gaze group: n = 8) based on the total amount of time the dog gazed at and touched the owner.

9
Fig. S2. Histogram of numbers of dog/wolf-to-owner gaze.
Bars show the number of dogs/wolves that gazed at their owners for each duration during a 5-
min interaction. LG, Long Gaze dogs (black bars); SG, Short Gaze dogs (white bars); Wolf,
wolves (gray bar).

10
Fig. S3. Individual changes in urinary oxytocin concentrations after a 30-min interaction in
wolf owners (A) and wolves (B). The number of the explanatory note corresponds to wolves
and theirs owners ID in Table S1-B.

11
Fig. S4. Experimental setup.

12
Fig. S5. The change ratio of humans plasma oxytocin (A) and urinary oxytocin (B) at
baseline (0 min) and every 15 and 30 min after oxytocin or saline nasal administration. C is
the association between the accumulative plasma oxytocin concentrations from 15 to 90 min
and urinary oxytocin concentrations from 30 to 120 min. Solid and dash lines indicate
respectively oxytocin and saline treatment conditions. A. a-b: p = 0.074, b-c: p = 0.066, b-d:
p = 0.059, oxytocin-saline: * p = 0.025. B. a-b: p = 0.086, a-d: p < 0.001, b-d: p = 0.061, x-y:
p = 0.026, x-z: p = 0.009.

13
Table S1. Information of interaction with human.
A. Dogs information
Acquired age (week) Socialization level a (n)
Mean SE Range Normal Lower NA
Experiment 1
LG (n=8) 56.63 25.82 5 - 201 4 3 1
SG (n=22) 65.73 17.42 5 - 226 10 5 7
Total 63.30 14.31 5 - 226
Experiment 2
Male (n=15) 74.23 28.45 7 - 336 9 4 2
Female (n=15) 17.60 8.76 6 - 140 15 0 0
Total 43.89 14.76 6 - 336

B. Wolves individual information

How the Age of Duration of interaction with
wolves Age at which the
separation humans in a day (hour) Urine
ID Sex were puppies left their
from mother In first collection
acquired In adulthood owners house (month)
(day) month
1 Male Bred b 13 24 2 12 pre, post
2 Female Bred 13 24 2 12 pre
3 Male Bred 7 24 2 24
4 Female Bred 15 10 2 18
5 Female Bred 15 10 2 18
6 Female Bred 15 10 2 18
7 Female Bred 7 24 6 60 pre, post
8 Male Bred 21 24 4 36 pre, post d
9 Male Captured c 21 24 4 36 pre
10 Male Captured 9 24 4 4 pre, post
11 Female Bred 9 24 4 4 pre, post
a. The dogs in Normal socialization group were received direct care from humans from an
early developmental period and lived in houses. The dogs in Lower socialization group came
from facilities such as experimental laboratories and consequently had minimal early-life
interaction with humans.
b. Bred by humans.
c. Captured from the wild as puppies.
d. This post urine was collected 2 hours after the interaction with the owners.

14
 C. Influence of early-life experience with humans
Age at
Duration of interaction with which the
Age at human in a day (hour) puppies left Acquired
separation Socializa-
their age
from mother In first In tion level
owner's (week)
(day) month adulthood house
(month)
rs p rs p rs p rs p rs p U p
Experiment 1
Oxytocin change
Dogs -0.24 0.23 31 0.31
ratio (n = 28)
Dog-to-owner gaze
0.17 0.39 31 0.55
(n = 28)
Oxytocin change
Wolves -0.30 0.37 0.84 0.65 0.43 0.41
ratio (n = 5)
Wolf-to-owner
-0.41 0.19 0.35 0.45 0.34 0.60 0.35 0.33
gaze (n = 11)

Experiment 2
Oxytocin change ratio (n = 27) -0.28 0.22 26 0.96
Gaze (n = 27) -0.01 0.95 23 0.21

Socialization level: comparison between normal and lower socialization group.

15
Table S2. Correlations between the oxytocin change ratio and behavioral variables in owners
and dogs (A) and owners and wolves (B) in Experiment 1. ** p < 0.01, * p < 0.05.

A. Dogs
Owners Dogs
Oxytocin Dog-to- M SE
Dog-talking Dog-touching
change ratio owner gaze
Owners Oxytocin change ratio 0.076 0.455 * 0.847 *** 0.534 ** 1.787 0.442
Dog-talking - 0.156 -0.100 0.288 * 28.933 4.151
Dog-touching - 0.421 * 0.330 * 14.994 3.499
Dogs Oxytocin change ratio - 0.422 * 1.387 0.405
Dog-to-owner gaze - 56.413 6.520

B. Wolves
Owners Wolves
Wolf- Oxytocin Wolf-to-owner M SE
Wolf-talking
touching change ratio gaze
Owners Oxytocin change ratio 0.670 * -0.145 -0.600 0.167 1.089 0.136
Wolf-talking - 0.360 -0.722 0.612 ** 28.727 6.874
Wolf-touching - -0.018 0.407 59.825 19.468
Wolves Oxytocin change ratio - -0.035 0.658 0.201
Wolf-to-owner gaze - 0.317 0.113

Table S3. Results of multiple linear regression analysis using the oxytocin change ratio and
behavioral variables in wolf owners and wolves. ** p < 0.05, * p < 0.05.

Oxytocin change ratio

Owners Wolves
Owner talking to wolf - -
Owner touching wolf -0.255 -0.019
Wolf-to-owner gaze 0.271 -0.036
R 0.287 0.040
2
Adjusted R -0.147 -0.997
P 0.709 0.998

16
Table S4. The results of power analysis and resampling in comparison of wolves with other
groups. We conducted a power analysis for the comparison of wolves with other groups
because of the small number of wolf samples. Power was not sufficiently high for some tests;
therefore, we also analyzed using a resampling test.

Power Resampling
Effect size d
(1- err prob) test
LG 13.619 1.000 0.000
Dog-to-owner gaze
SG 2.904 1.000 0.000
LG 0.230 0.075 0.351
Behavior Dog-talking
SG 0.523 0.271 0.338
LG 0.573 0.214 0.238
Dog-touching
SG 0.179 0.075 0.611
LG 1.205 0.818 0.000
Total
SG 1.223 0.923 0.000
LG 1.310 0.554 0.004
pre
SG 1.226 0.651 0.003
Dog
LG 0.975 0.345 0.065
post
SG 1.102 0.560 0.004
LG 0.801 0.250 0.222
Change ratio
SG 0.282 0.084 0.408
LG 0.137 0.069 0.122
Oxytocin Total
SG 0.085 0.061 0.035
LG 0.793 0.363 0.109
pre
SG 0.026 0.051 0.359
Owner
LG 0.459 0.154 0.545
post
SG 0.140 0.065 0.060
LG 1.080 0.592 0.004
Change ratio
SG 0.200 0.081 0.145
Wolf pre-post 0.216 0.067 0.438
Wolf's owner pre-post 0.124 0.066 0.755

17
REFERENCES
1. B. Hare, M. Tomasello, Human-like social skills in dogs? Trends Cogn. Sci. 9, 439444
(2005). Medline doi:10.1016/j.tics.2005.07.003
2. A. Miklsi, E. Kubinyi, J. Topl, M. Gcsi, Z. Virnyi, V. Csnyi, A simple reason for a
big difference: Wolves do not look back at humans, but dogs do. Curr. Biol. 13, 763
766 (2003). Medline doi:10.1016/S0960-9822(03)00263-X
3. A. P. Melis, B. Hare, M. Tomasello, Chimpanzees recruit the best collaborators. Science
311, 12971300 (2006). Medline doi:10.1126/science.1123007
4. R. Coppinger, J. Glendinning, E. Torop, C. Matthay, M. Sutherland, C. Smith, Degree of
behavioral neoteny differentiates canid polymorphs. Ethology 75, 89108 (1987).
doi:10.1111/j.1439-0310.1987.tb00645.x
5. M. Somel, H. Franz, Z. Yan, A. Lorenc, S. Guo, T. Giger, J. Kelso, B. Nickel, M.
Dannemann, S. Bahn, M. J. Webster, C. S. Weickert, M. Lachmann, S. Pbo, P.
Khaitovich, Transcriptional neoteny in the human brain. Proc. Natl. Acad. Sci. U.S.A.
106, 57435748 (2009). Medline doi:10.1073/pnas.0900544106
6. S. Dickstein, R. A. Thompson, D. Estes, C. Malkin, M. E. Lamb, Social referencing and
the security of attachment. Infant Behav. Dev. 7, 507516 (1984). doi:10.1016/S0163-
6383(84)80009-0
7. S. Kim, P. Fonagy, O. Koos, K. Dorsett, L. Strathearn, Maternal oxytocin response
predicts mother-to-infant gaze. Brain Res. 1580, 133142 (2014). Medline
doi:10.1016/j.brainres.2013.10.050
8. L. J. Young, Z. Wang, The neurobiology of pair bonding. Nat. Neurosci. 7, 10481054
(2004). Medline doi:10.1038/nn1327
9. H. E. Ross, L. J. Young, Oxytocin and the neural mechanisms regulating social cognition
and affiliative behavior. Front. Neuroendocrinol. 30, 534547 (2009). Medline
doi:10.1016/j.yfrne.2009.05.004
10. G. Dlen, A. Darvishzadeh, K. W. Huang, R. C. Malenka, Social reward requires
coordinated activity of nucleus accumbens oxytocin and serotonin. Nature 501, 179
184 (2013). Medline doi:10.1038/nature12518
11. I. D. Neumann, Involvement of the brain oxytocin system in stress coping: Interactions
with the hypothalamo-pituitary-adrenal axis. Prog. Brain Res. 139, 147162 (2002).
Medline doi:10.1016/S0079-6123(02)39014-9
12. M. Nagasawa, S. Okabe, K. Mogi, T. Kikusui, Oxytocin and mutual communication in
mother-infant bonding. Front. Hum. Neurosci. 6, 31 (2012). Medline
doi:10.3389/fnhum.2012.00031
13. J. K. Rilling, L. J. Young, The biology of mammalian parenting and its effect on
offspring social development. Science 345, 771776 (2014). Medline
doi:10.1126/science.1252723
14. J. Topl, M. Gcsi, . Miklsi, Z. Virnyi, E. Kubinyi, V. Csnyi, Attachment to humans:
A comparative study on hand-reared wolves and differently socialized dog puppies.
Anim. Behav. 70, 13671375 (2005). doi:10.1016/j.anbehav.2005.03.025
15. M. Nagasawa, K. Mogi, T. Kikusui, Attachment between humans and dogs. Jpn. Psychol.
Res. 51, 209221 (2009). doi:10.1111/j.1468-5884.2009.00402.x
16. D. S. Tuber, S. Sanders, M. B. Hennessy, J. A. Miller, Behavioral and glucocorticoid
responses of adult domestic dogs (Canis familiaris) to companionship and social
separation. J. Comp. Psychol. 110, 103108 (1996). Medline doi:10.1037/0735-
7036.110.1.103
17. J. S. Odendaal, R. A. Meintjes, Neurophysiological correlates of affiliative behaviour
between humans and dogs. Vet. J. 165, 296301 (2003). Medline doi:10.1016/S1090-
0233(02)00237-X
18. S. Mitsui, M. Yamamoto, M. Nagasawa, K. Mogi, T. Kikusui, N. Ohtani, M. Ohta,
Urinary oxytocin as a noninvasive biomarker of positive emotion in dogs. Horm.
Behav. 60, 239243 (2011). Medline doi:10.1016/j.yhbeh.2011.05.012
19. M. Nagasawa, T. Kikusui, T. Onaka, M. Ohta, Dogs gaze at its owner increases owners
urinary oxytocin during social interaction. Horm. Behav. 55, 434441 (2009).
Medline doi:10.1016/j.yhbeh.2008.12.002
20. M. W. Fox, The Soul of the Wolf (Burford Books, New York, 1997).
21. M. Gcsi, J. Vas, J. Topl, . Miklsi, Wolves do not join the dance: Sophisticated
aggression control by adjusting to human social signals in dogs. Appl. Anim. Behav.
Sci. 145, 109122 (2013). doi:10.1016/j.applanim.2013.02.007
22. J. K. Rilling, A. C. Demarco, P. D. Hackett, X. Chen, P. Gautam, S. Stair, E. Haroon, R.
Thompson, B. Ditzen, R. Patel, G. Pagnoni, Sex differences in the neural and
behavioral response to intranasal oxytocin and vasopressin during human social
interaction. Psychoneuroendocrinology 39, 237248 (2014). Medline
doi:10.1016/j.psyneuen.2013.09.022
23. I. D. Neumann, Brain oxytocin: A key regulator of emotional and social behaviours in
both females and males. J. Neuroendocrinol. 20, 858865 (2008). Medline
doi:10.1111/j.1365-2826.2008.01726.x
24. C. K. De Dreu, L. L. Greer, M. J. Handgraaf, S. Shalvi, G. A. Van Kleef, M. Baas, F. S.
Ten Velden, E. Van Dijk, S. W. Feith, The neuropeptide oxytocin regulates parochial
altruism in intergroup conflict among humans. Science 328, 14081411 (2010).
Medline doi:10.1126/science.1189047
25. E. Meins, Security of Attachment and the Social Development of Cognition (Psychology
Press, Philadelphia, 1997).
26. J. Shinozaki, T. Hanakawa, H. Fukuyama, Heterospecific and conspecific social cognition
in the anterior cingulate cortex. Neuroreport 18, 993997 (2007). Medline
doi:10.1097/WNR.0b013e3281ac2161
27. E. Klinghammer, P. A. Goodmann, in Man and Wolf: Advances, Issues, and Problems in
Captive Wolf Research, H. Frank, Ed. (Dr W. Junk Publishers, Dordrecht, 1987),
chap. 2.
28. M. Gcsi, B. Gyori, A. Miklsi, Z. Virnyi, E. Kubinyi, J. Topl, V. Csnyi, Species-
specific differences and similarities in the behavior of hand-raised dog and wolf pups
in social situations with humans. Dev. Psychobiol. 47, 111122 (2005). Medline
doi:10.1002/dev.20082
29. T. Romero, M. Nagasawa, K. Mogi, T. Hasegawa, T. Kikusui, Oxytocin promotes social
bonding in dogs. Proc. Natl. Acad. Sci. U.S.A. 111, 90859090 (2014). Medline
doi:10.1073/pnas.1322868111

2
30. J. A. Amico, J. S. Ulbrecht, A. G. Robinson, Clearance studies of oxytocin in humans
using radioimmunoassay measurements of the hormone in plasma and urine. J. Clin.
Endocrinol. Metab. 64, 340345 (1987). Medline doi:10.1210/jcem-64-2-340