Promoting social behavior with oxytocin in high-
functioning autism spectrum disorders
Elissar Andaria, Jean-René Duhamela, Tiziana Zallab, Evelyn Herbrechtb, Marion Leboyerb, and Angela Sirigua,1
aCentre de Neuroscience Cognitive, Unité Mixte de Recherche 5229, Centre National de la Recherche Scientifique, 69675 Bron, France; andbInstitut National
de la Santé et de la Recherche Médicale U 841, Department of Psychiatry, Hôpital Chenevier-Mondor, 94000 Créteil, France
Edited by Leslie G. Ungerleider, National Institute of Mental Health, Bethesda, MD, and approved January 7, 2010 (received for review September 8, 2009)
Social adaptation requires specific cognitive and emotional compe-
tences. Individuals with high-functioning autism or with Asperger
syndrome cannot understand or engage in social situations despite
preserved intellectual abilities. Recently, it has been suggested that
oxytocin, a hormone known to promote mother-infant bonds, may
be implicated in the social deficit of autism. We investigated the
behavioral effects of oxytocin in 13 subjects with autism. In a
simulated ball game where participants interacted with fictitious
partners, we found that after oxytocin inhalation, patients
exhibited stronger interactions with the most socially cooperative
partner and reported enhanced feelings of trust and preference.
Also, during free viewing of pictures of faces, oxytocin selectively
increased patients’ gazing time on the socially informative region of
the face, namely the eyes. Thus, under oxytocin, patients respond
more strongly to others and exhibit more appropriate social behav-
ior and affect, suggesting a therapeutic potential of oxytocin
through its action on a core dimension of autism.
simple social signals such as eye contact, as well as to infer
from more complex behaviors intrinsically social qualities of
other people such as fairness or cooperation. Individuals suf-
fering from high-functioning autism spectrum disorders (HF-
ASD), a neurodevelopmental disorder, are impaired in under-
standing social cues and in responding to them. These patients
generally have normal language or general intellectual abilities,
yet in everyday life they avoid eye contact (1–3) and do not
spontaneously interact with people (4). On formal tests of social
cognitive skill, they show specific impairments in understanding
the intentions of others (1, 5) and lack of fast intuitive judgments
about social contexts (4).
The pathogenesis of autism is unclear, although mutations in
genes implicated in synaptogenesis have been identified (6, 7)
and different neurochemical, neurophysiological, and neuro-
pathological abnormalities have been demonstrated in these
patients (8). An interesting current hypothesis has implicated
oxytocin in the etiology of autism, and in particular in the social
disorders that are the hallmark of HF-ASD (8–10).
Oxytocin is a hormone synthesized in the hypothalamus. Best
known for its facilitatory role in parturition and lactation, oxy-
tocin is also involved in the regulation of emotions and has
receptors distributed in various brain regions including the limbic
system and amygdala (11, 12). In mammals, it has been asso-
ciated specifically with the development of prosocial behavior
such as mother-infant attachment, grooming, approach behavior,
sexual activity, and stress regulation (13, 14). Oxytocin anomalies
have been reported in children with autism. They have sig-
nificantly lower plasma oxytocin levels compared to control
subjects (15) and fail to show the normal developmental increase
in oxytocin blood levels. Moreover, plasma samples are asso-
ciated with higher oxytocin precursor levels, suggesting that
autism may be related to anomalies in the way this hormone is
ur brain is endowed with the ability to detect and respond to
Experimental manipulation of brain oxytocin levels in healthy
human subjects confirms its involvement in the expression of
human affiliative social behavior (17). In a simulated economic
investment game, subjects who received an intranasal spray of
oxytocin were more inclined, as compared to a placebo control
group, to trust another player by sending him money with no
guarantee of reciprocation, suggesting that oxytocin acts on brain
circuits that promote social proximity and affiliation with peers
(17). Recently, it has been shown that oxytocin facilitates recog-
nition of memorized faces and strengthens the encoding of social
stimuli (18, 19). Moreover, oxytocin has been reported to
increase the time spent looking at socially important cues, such as
the eyes, when viewing pictures of human faces (20). In the light
of the above findings, a key question regarding both the role of
oxytocin in the nervous system and the pathophysiology of social
disorders in autism is whether administration of oxytocin can
influence social interaction behavior in individuals with autism.
We investigated the effects of intranasal oxytocin on the social
behavior of 13 patients suffering from HF-ASD and compared
these effects to a placebo condition and to the behavior of
matched healthy subjects. Two different behavioral measures
were used: (i) decision making and affect in a social interaction
game, and (ii) eye movement recordings during a face perception
task. We also measured plasma oxytocin levels in patients before
and after nasal spray intake, to establish whether patients dis-
played physiological abnormalities in oxytocin and to verify the
effectiveness of the nasal administration procedure in enhancing
plasma oxytocin levels.
Social Ball Tossing Game. We used a social interaction task
inspired by the Cyberball game (21) in which the participant
engages in a multiround ball-toss game over a computer network
with three fictitious partners (Fig. 1A Inset). In our variant game,
we manipulated the amount of reciprocation exhibited by the
three fictitious players. The critical task manipulation was the
probability that each of the three fictitious players would throw
the ball to the participant, which allowed us to create different
cooperative behavior profiles (good, bad, and neutral) (Materials
and Methods and SI Materials and Methods).
Oxytocin Effect on Social Decision. The behavioral decision variable
of interest in this task is the participant’s ball-toss choices. Under
placebo treatment, patients showed little evidence that they
discriminated the three players’ cooperative profiles. Whereas
healthy subjects sent significantly more balls to the good than to
the bad (Wilcoxon test, z = 3, P < 0.003) or neutral player (z =
Author contributions: E.A. and A.S. designed research; E.A., J.-R.D., T.Z., E.H., M.L., and A.
S. performed research; E.A., J.-R.D., and A.S. analyzed data; and E.A., J.-R.D., and A.S.
wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
1To whom correspondence should be addressed. E-mail: firstname.lastname@example.org.
This article contains supporting information online at www.pnas.org/cgi/content/full/
www.pnas.org/cgi/doi/10.1073/pnas.0910249107 PNAS Early Edition
| 1 of 6
2.76, P < 0.005) (Fig. 1B Left), patients under placebo responded
in the same manner to all players (z = 0.36, P = 0.72; z = 0.2,
P = 0.84) (Fig. 1B Right). In striking contrast, oxytocin intake led
patients to engage more often with the good player and to send
significantly more balls to this player as compared to the bad one
(z = 2.04, P < 0.041; two-tailed) (Fig. 1B Middle). When com-
paring directly the effects of placebo and oxytocin, we also found
a significantly larger difference in the number of balls sent to the
good versus the bad player in the oxytocin condition (z = 1.99,
P < 0.047; two-tailed). Finally, the difference in performance
(number of balls sent to the good versus the bad player) between
the control subjects and patients, which was significant under
placebo (Mann–Whitney U test: z = 3.1, P < 0.0021), disap-
peared when the comparison was made with the oxytocin treat-
ment condition (z = 1.62, P = 0.11).
A finer-grained image of the patients’ decision making was
obtained by examining the distribution of ball tosses over time.
Data were binned with respect to intervals defined by player A’s
turns, as it was through the observation of A’s behavior that the
participant could learn to cooperate with him more than with the
other two players. The first six tosses by A were unbiased; hence
the probability of the participant receiving the ball p[A→P] =
0.33. On the next four tosses, A increased the proportion of balls
sent to the participant up to p[A→P] = 0.50. From the 11th toss
on, A sent all of its balls to P, that is, p[A→P] = 1.0. Over the
same period, player C started to exclude P, revealing himself as
the bad player. The effect of these biases on the behavior of the
participants is illustrated in Fig. 2. Because preliminary analyses
in both healthy subjects and patients failed to reveal any differ-
ences in behavior toward the neutral versus the bad player, we
focus on good and bad players only. Both healthy subjects and
patients under oxytocin begin to cooperate preferentially with
the good player at about the same time, with the cumulative
number of balls sent to the good and bad players diverging sig-
nificantly in the 15–17 interval (Fig. 2 A and B; first of two
consecutive significant bins for the difference between good and
bad, healthy subjects: z = 2.7, P < 0.007; oxytocin: z = 2.04, P <
0.041; two-tailed, Wilcoxon test). By contrast, under placebo, the
patients’ cumulative ball-toss curves never diverged significantly
Oxytocin Effect on Emotions. The emotional response of the
patients to the fictitious players’ personality was assessed after
completion of the task using a seven-point rating scale. These
emotional self-ratings were consistent with their decision
behavior under both treatment conditions. Whereas feelings
(trust and preference) expressed toward the three fictitious
players did not differ in the placebo condition (Friedman’s
ANOVA, respectively: χ2= 2.39, P = 0.3; χ2= 1.19, P = 0.55),
patients reported that they trusted more and showed stronger
preference for the good than the bad player after playing under
oxytocin (Friedman’s ANOVA, respectively: χ2= 17.89, P <
0.0002; χ2= 13.63, P < 0.001; posthoc pairwise comparisons P <
0.05; Fig. 3). No significant differences were found between
feelings toward the neutral and the other two players.
One question which could be raised about the effect of oxy-
tocin on ball-toss choices is whether it mainly acted on social
engagement or on the perception of monetary rewards. To
address this issue, we tested a new group of seven HF-ASD
patients on the same ball-toss game but modified the contextual
framing of the task to eliminate any reference to monetary
incentives. The task conditions and oxytocin administration
procedures were exactly the same as in the original version
except that subjects were instructed that the goal of the task was
to play a friendly ball-toss game with other players, but no
monetary reward was promised and the participant did not
receive any feedback about the number of balls he/she received.
They were only told that whenever they tossed the ball to
someone, that player could either send it back or toss it to
another player. Following completion of the task, the partic-
ipants again estimated their feelings of “trust” and “preference”
with respect to the fictitious players. Despite the smaller size of
the patient sample, we again found a significant, positive effect of
oxytocin on the participant’s capacities to discriminate between
the two extreme player profiles (Fig. S1). Comparing directly the
effects of placebo and oxytocin, we found a significantly larger
difference in number of balls sent to the good versus the bad
player in the oxytocin condition (z = 1.99, P < 0.047; two-tailed)
(Fig. S1). Consistent with this behavior, patients also reported
that they felt more trust toward the good than the bad player
under oxytocin (z = 2.11, P < 0.035; two-tailed), whereas under
placebo, there was no such difference (z = 0.59, P > 0.58). In the
oxytocin condition, a similar trend was found in the feelings of
preference but it did not reach significance (z = 1.57, P = 0.11).
Visual Scanning of Faces. To strengthen our observations of the
effects of oxytocin on the processing of socially relevant infor-
mation, we investigated how patients looked at a fundamental
social stimulus, such as human faces. Participants examined
partners. (A) Schematic representation of the modified Cyberball game. On
successive trials, the role of the participant P alternates between turns as
observer of ball exchanges between two of the other players, as ball recip-
ient, and as ball sender. The behavior of players A, B, and C is computer-
generated so as to define three different profiles from P’s standpoint: an
includer or good profile, a neutral profile, and an excluder or bad profile.
The length of the gray arrows is proportional to the number of balls sent by
a given player to each of the other players. The profiles represent the
average behavior over the entire game but, rather than being fixed from
the beginning, they were tuned progressively using an algorithm described
in SI Text. Black arrows represent the behavior of a representative healthy
subject. (B) Ball-toss distributions for healthy subjects and for patients with
HF-ASD treated with oxytocin or placebo. Under oxytocin, there was a nearly
significant trend in the number of balls sent toward the good as compared
to the neutral player (significant trend, z = 1.82, P = 0.06; two-tailed) (mid-
dle) (mean and SEM; * indicates significant difference at P < 0.05 or better
on posthoc pairwise comparisons).
Cyberball game and ball-toss distributions toward each of the three
2 of 6
| www.pnas.org/cgi/doi/10.1073/pnas.0910249107 Andari et al.
pictures of faces presented one at a time on a computer monitor
while their eye movements were being recorded. The partic-
ipants’ task was to report either the gender (male/female) or the
gaze direction (direct/averted) of the depicted face (Fig. S2).
Offline, we computed the total fixation time inside each of six
regions of interest (eyes, nose, mouth, forehead, cheeks, and
outside of facial contour) and the number of saccades (rapid
displacements of the line of gaze) elicited by the face stimuli.
Normal subjects directed their gaze preferentially within the
contours of the faces [gender identification (GI): 80% ± 2.09;
gaze direction (GD): 84% ± 1.74; Tables S1 and S2]. By contrast,
patients under placebo spent significantly less time looking
directly at the faces as compared to healthy subjects (53% ± 7.2
in GI: Mann–Whitney U test, z = 2.67, P < 0.0077; 41% ± 8.04 in
GD: z = 4.08, P < 0.00005). More detailed analysis of scanning
pattern by regions of interest shows that they specifically avoided
the eye region (z = 2.88, P < 0.004; z = 3.48, P < 0.0005 for GI
and GD, respectively). Interestingly, during gaze direction
judgments, patients also produced more saccades than healthy
subjects (GD: z = 2.45, P < 0.015). This increase in saccade rate
was only present during epochs when patients looked at faces
directly and not when they explored the rest of the display (GD:
z = 0.71, P > 0.47). Such underexploration of the face and eyes,
in association with high saccade frequency, implies that patients
explored these images hastily by means of multiple brief fix-
ations, probably with high levels of anxiety and discomfort.
Saccade frequency was not increased during gender decisions
(GI: z = 1.36, P > 0.17), possibly because, in contrast to gaze
direction judgment, it does not depend critically upon attending
to the eye region of the faces.
Oxytocin modified how patients responded to pictures of
faces, as compared to the placebo condition. Total gaze time
over the face increased significantly under both task conditions
(GI: z = 2.27, P < 0.023; GD: z = 2.19, P < 0.029; two-tailed
Wilcoxon test; Fig. 4 A and B Left). Broken down by region of
interest, the effects of oxytocin are found to be largely accounted
for by an increased fixation time over the eye region (GI: z =
2.12, P < 0.04; nearly significant trend for GD: z = 1.88, P =
0.059; Fig. 4 C and D). No effects of oxytocin were observed over
the other regions of interest (mouth + nose, GI: z = 1.18, P >
0.23; GD: z = 1.41, P > 0.15; forehead + cheeks, GI: z = 39, P >
0.69; GD: z = 0.86, P > 0.38) (Fig. 4 C and D Right). Finally,
oxytocin significantly reduced the abnormally high saccade fre-
quency observed under placebo during gaze direction judgments
(z = 2.12, P < 0.03; two-tailed; Table S2).
Although oxytocin significantly enhanced patients’ visual
scanning of faces, as compared to the placebo condition, their
gaze time on the face and eye region remained significantly lower
than that of healthy subjects for all comparisons with the
exception of whole-face scanning in the gender identification
condition (Mann–Whitney U test; face: GD: z = 3.21, P < 0.002;
GI: z = 1.96, P > 0.05; eye: GI: z = 2.77, P < 0.006; GD: z = 2.99,
P < 0.003). The fact that oxytocin did not fully restore a normal
visual exploration pattern in patients is discussed below in rela-
tion to the magnitude of the changes in blood oxytocin.
In summary, under oxytocin, patients with HF-ASD spent
more time looking at the face pictures and, specifically, at the eye
region. The accompanying decrease in saccade frequency, that is,
the increase in the average duration of individual fixations,
suggests that oxytocin may reduce the fear or anxiety induced by
face stimuli in these patients.
We tested for a possible effect of treatment order by com-
paring patients’ performance between the two visits in the pla-
cebo and the oxytocin condition. No differences were found
between the two visits on any of the dependent variables meas-
ured in the ball-tossing and face-scanning tasks (Mann–Whitney
U test; z = 0, P = 1). We also found that oxytocin’s effect on
patients’ performance during both tasks was not related to a
simple mood effect (see SI Results).
Plasma Oxytocin Levels. Baseline plasma oxytocin concentration in
patients (1.08 pg/mL ± 1.04) was significantly below the values
observed in a normative group of healthy subjects (7.28 pg/mL ±
spaced bins, for healthy subjects and for patients with HF-ASD under oxytocin and under placebo. Each data point falls in a bin defined by an interval
between player A’s tosses n and n+2. The participant had, on average, three ball-toss opportunities in each interval. The transition in player A’s behavior from
unbiased (equal probability of throwing the ball to each player) to positively biased toward P (100% probability of throwing the ball to P) was progressive
(mean and SEM; * indicates significant difference at P < 0.05 or better for the first of two significant consecutive bins).
Time course of ball tosses during the Cyberball game. Cumulative number of balls sent by the participant P to players A (good) and C (bad) in regularly
treatment. Rating (1–7) of subjective feeling states toward the three players
in the placebo and oxytocin-treated patients with HF-ASD (mean and SEM;
* indicates significant difference at P < 0.05 or better on posthoc pairwise
Subjective postexperimental rating under oxytocin and placebo
Andari et al. PNAS Early Edition
| 3 of 6
4.49) (Mann–Whitney U test; z = 4.69, P < 0.0001). A second
measurement made 10 min after nasal administration of a dose
of 24 IU of Syntocinon spray showed a significant increase in
plasma oxytocin concentration (2.66 pg/mL ± 2.2) (Wilcoxon
test; z = 1.88, P < 0.02; Fig. S3), indicating successful assim-
ilation of the substance (SI Materials and Methods).
Individual Variability in Response to Oxytocin. Although the social
symptoms of HF-ASD can be diagnosed on the basis of well-
established, reliable criteria, it is unclear whether these symp-
toms are related to a common etiological process. A degree of
heterogeneity in responsiveness to oxytocin treatment can thus
be expected in such a patient group. Inspection of individual
performances revealed that some patients responded strongly to
oxytocin, others more weakly, and some not at all (Table S3).
Furthermore, oxytocin effects on the social game and on the face
perception tasks were only weakly correlated (GI: r = 0.23; GD:
r = 0.54, P > 0.05), indicating that the two tasks are sensitive to
different aspects of social information processing. Indeed,
although explicit social engagement is required in the ball game,
visual inspection of facial stimuli may involve more implicit,
automatic mechanisms. Also, looking directly at large face
stimuli may be more threatening to some patients than inter-
acting with other people via a computer network, whereas for
others, the dynamical aspect of the social interaction may be
more difficult to apprehend than the simple perceptual decision
required by the face task. Possible relationships between oxy-
tocin effects and general clinical data on patients were inves-
tigated. We found no significant correlation between patients’
performance with the Autism Diagnostic Interview-Revised
(ADI-R) (ball game: r = −0.31; GI: r = 0.03; GD: r = 0.35; all P
> 0.05), IQ (ball game: r = 0.43; GI: r = −0.43; GD: r = −0.24;
all P > 0.05), or age (ball game: r = 0.25; GI: r = −0.2; GD: r =
−0.09; all P > 0.05). Different authors have suggested that
patients with autism may display different social interaction
styles (15, 22). According to these authors, one can distinguish
between “aloof” individuals who avoid physical proximity with
others and actively reject social contact, “passive” individuals
who do not reject approaches but neither engage in social rela-
tions, and “active-but-odd” individuals who display approach
behavior but in a somewhat inappropriate or one-sided manner.
We examined whether such qualitative differences in social
interaction profiles might account for the variability in the
response to oxytocin treatment. Patients in our study were
assigned to one of those three categories based on clinical
records and parent interviews (6 patients were classified as aloof,
7 as active-but-odd, none as passive). Interestingly, 6/8 patients
who showed positive changes on the ball game under oxytocin
had been labeled as active-but-odd, whereas 4/5 who showed no
positive change were of the aloof type.
In this study, we investigated whether oxytocin could modify how
high-functioning autistic patients process social signals and social
feedback. Oxytocin was shown to enhance visual scanning of
faces and, in particular, of the eye region, as compared to a
placebo condition. Eye contact between individuals can be con-
sidered a basic form of social aptitude. Previous studies in nor-
mal individuals indicate that oxytocin enhances processing of
facial stimuli (20) and the ability to infer others’ mental states
from the eye region (23). Here we further demonstrate that
oxytocin promotes a first level of prosocial approach by over-
turning what constitutes a core deficit of patients with HF-ASD,
namely the lack of eye contact. How does oxytocin facilitate
patients’ prosocial behavior? The present data provide some
suggestions of the neural mechanisms mediating these effects. It
has been proposed that oxytocin enhances affiliation partly by
reducing fear of social unfaithfulness and by suppressing avoid-
ance behavior (14) and that it reduces the activity of the amyg-
dala, resulting in a decrease of fear responses (14, 24, 25). It is
possible that patients with autism possess latent social skills and
that oxytocin may thus favor social engagement behavior by
suppressing fear and mistrust.
The results from the ball-tossing task suggest the possibility of
other mechanisms underlying the effects of oxytocin on social
interaction. In this simple game simulating social exchanges,
the face and outside the face region under oxytocin and placebo when patients had to identify the face’s gender (male/female) and face’s gaze direction
(direct/averted), respectively. (C and D) Gaze time spent on main regions of interest: the eyes, nose and mouth, and other regions such as forehead and cheeks
during gender identification and gaze direction detection, respectively. * indicates significant difference at P < 0.05 or better on Wilcoxon test.
Mean gaze time spent on different regions of interest for patients with HF-ASD under placebo and oxytocin treatment. (A and B) Gaze time spent on
4 of 6
| www.pnas.org/cgi/doi/10.1073/pnas.0910249107 Andari et al.
patients tested under placebo conditions did not take into
account the behavior of other players and showed no differential
emotional responses to the different players. By contrast, under
oxytocin, these patients engaged more often in exchanges with
the player who reciprocated strongly, less often with the player
who reciprocated weakly, and they exhibited emotional respon-
ses congruent with this behavior. Thus, oxytocin enhanced
patients’ ability to process socially relevant cues and acquire their
meaning in an interactive context. A study conducted in normal
subjects showed that oxytocin increases trust of others in the
absence of any certainty of reciprocation (17), hinting at the
possibility that oxytocin may promote indiscriminate prosocial
behavior and “blind” trust. The task used here to study the
effects of oxytocin on autistic social difficulties was different in
that it involved multiple iterations in which the participant was
presented with successive feedback from partners that were
endowed with different reciprocating tendencies. Therefore, the
behavior exhibited by the participants could evolve over time
through a learning process which can be interpreted within a
social reinforcement learning framework, with social inclusion
acting as reinforcer. The fact that oxytocin allowed recognition
of the partner who was willing to reciprocate the most cannot be
explained only in terms of prosocial attitudes such as reduced
fear or increased approach and trust. The patients’ ability to
discriminate between the good and bad partners shows that
oxytocin facilitated learning, which may in turn result from an
increased drive for social affiliation or from an enhancement of
reinforcers satisfying this drive. This hypothesis is consistent with
data from animal and human studies. Animal studies show that
oxytocin promotes social bonding behavior. In rodents, oxytocin
has been reported to enhance social recognition, as indicated by
a decrease of exploration behavior toward a conspecific during a
second encounter (13). Moreover, in oxytocin knockout mice,
social memory is impaired but recovered after a single shot of the
hormone before initial social encountering takes place. Finally,
in humans, different studies have shown that oxytocin improves
recognition memory of social relevant cues (19) (i.e., faces) and
memory of positive social information (i.e., happy faces) (18).
One question that can be raised is whether oxytocin mainly
acted by enhancing sensitivity to social rewards (i.e., to being
sent the ball, a social engagement gesture) or by enhancing
sensitivity to the accompanying nonsocial reward (the monetary
value of the received ball). Motivation for money could not have
been the main factor determining their choices. Both healthy
subjects and patients were biased toward adopting a prosocial
attitude because they sent fewer balls to the good player and
more balls to the bad player than would be predicted by an
optimal reward-seeking strategy such as the matching law (26).
More direct evidence that, in this task, oxytocin is acting on
social motivation comes from a second ball-toss game that was
performed by an independent group of HF-ASD patients, in
which ball exchange was not associated with monetary reward.
Also in this case, oxytocin enhanced the propensity to interact
with the reciprocating partner, as compared to placebo. This is in
keeping with results obtained in a similar social task showing that
normal subjects preferred to avoid being excluded from the game
even when, as a consequence, they ended up losing money (27).
Although previous studies have shown that oxytocin can
reduce repetitive behavior in subjects with autism (28) and
enhance the comprehension of affective speech (29), here we
demonstrated that oxytocin can promote social approach and
social comprehension in patients with autism. Individual varia-
bility was observed in the effects of oxytocin on the social tasks
used in this study. Nevertheless, the results from the ball-toss
game were statistically robust and found in two independent
groups of HF-ASD patients. More work will be needed to
understand the relationship between changes in social behavior
induced by oxytocin administration in individuals with autism
and local changes in brain oxytocin metabolism. This could be
accomplished using functional imaging techniques. Finally, our
results highlight the therapeutic potential of oxytocin through its
action on core deficits of patients with HF-ASD such as affili-
ation and cooperative behavior. Although the effect we meas-
ured here is certainly transient, it serves to show that these
patients are quite able to engage in social relationships. Future
research is necessary to investigate whether a long-term intake of
social functioning of
Materials and Methods
Participants. A group of 13 adults (11 men and 2 women, mean age = 26,
range = 17–39) with a clinical diagnosis of Asperger syndrome (AS) (n = 10) or
high-functioning autism (HFA) (n = 3) according to Diagnostic and Statistical
Manual-Revision 4 (DSM-IV R) (American Psychiatric Association, 2000) and
ASDI (Asperger Syndrome Diagnostic Interview) (30) were recruited from the
expert centers (Foundation FondaMental), Chenevier-Mondor Hospital in
Créteil, France. Interviews with parents or caregivers using the ADI-R (Autism
Diagnostic Interview-Revised) (31) (Table S4) confirmed the diagnoses. As
part of the checking process, the French translation of A-TAC (autism, tics,
AD-HD, and other comorbidities) (32) was completed by the parents.
Patients received verbal and performance IQ tests (WAIS-III) and all showed
average to above average estimates of intelligence (Table S4). Patients were
medication-free for at least 2 weeks before and throughout the study
(SI Materials and Methods). A second group of patients was recruited to test
oxytocin effects on a social ball-toss game involving no monetary incentives.
It included seven new HF-ASD patients (7 men, mean age = 28, range =
18–38; verbal IQ: 96 ± 15.85, performance IQ: 87 ± 20.57, total IQ: 92 ± 17.47;
ADI-R: social interaction 12.6 ± 7.21, communication 6.7 ± 3.73, repetitive
behaviors 3.1 ± 2.03) with a clinical diagnosis of AS (n = 4) or HFA (n = 1) or
pervasive developmental disorder-not otherwise specified (PDD-NOS) (n = 2).
The study also included a control group of 13 healthy subjects matched for
chronological age and sex to the patients (11 men and 2 women, mean
age = 26, range = 18–40). The study was approved by the Local Ethical
Committee (Centre Léon Berard, Lyon IV). The French Agency (Agence
Française de Sécurité Sanitaire des Produits de Santé) competent for clinical
trials on a medicinal product for human use also gave its approval.
Behavioral Experiments. Social ball-tossing game. During this variant version of
the Cyberball game, three players depicted by cartoon characters and their
corresponding photographs were presented on a touch-sensitive computer
display. The participant (player P) was featured by an additional cartoon
representing a pair of animated hands assuming a first-person perspective.
Each trial consisted of a single ball exchange depicted by a short animation
showing one player handling the ball, and a few seconds later another player
catching the ball. If a trial ended with the participant as recipient, he/she
became the next trial’s sender and had to address the ball to player A, B, or C
bytouchingthecorresponding photograph.Atgame start,probabilities were
homogeneous for all players, that is, the participant had a probability P = 1/3
of receiving the ball from any of the three players. After a predetermined
number of rounds, player profiles diverged such that player A (the “good”
profile) sent, on average, 70% of its played balls to the participant (P), player
B (the “neutral” profile) sent 30% of its played balls to P, and player C (the
“bad” profile) sent 10% of its played balls to P. These proportions are rep-
resented by the length of the gray arrows in Fig. 1A. The game included a
monetaryincentivetoenhance theparticipant’scognitive engagement inthe
task. Any player receiving the ball earned 2€ (see SI Materials and Methods
for details about the algorithm used to dynamically set the three player
profiles). To optimize cognitive engagement in the task, the participant was
told that each ball received was worth 2 euros, and that when returning the
ball two outcomes were possible: either the recipient would toss the ball back
to the participant, generating further income, or toss it to another player,
earning that player 2 euros. The participant’s cumulative gains were dis-
played on the screen and he/she was led to expect a percentage of the gains
at the end of the game. The participant was instructed that the game ended
after a total of 80 tosses. The main dependent variable in this experiment was
the distribution of the participant’s toss choices between players A, B, and C
(illustrated for a representative healthy subject by the black arrows in Fig.
1A). Also, following completion ofthe task, theparticipantestimated, using a
subjective seven-point rating scale, their sentiments of “trust” and “prefer-
ence” with respect to the fictitious players.
Face perception tasks. Pictures of faces were presented on the 17-inch video
display of a Tobii 1750 eye tracker and patients’ gaze movements were
Andari et al.PNAS Early Edition
| 5 of 6
analyzed offline using ClearView software. Five regions of interest (ROI) on Download full-text
the face were defined: the two eyes, nose, mouth region, forehead, and the
two cheeks. A sixth ROI consisted of the portion of the image outside the
contour of the face. For each picture, gaze fixation time (in milliseconds) was
computed for each of the six ROIs. The number of saccadic eye movements
(rapid changes in gaze direction) made for two ROIs, inside and outside the
facial contours, was computed (a finer-grained parcellation of the face yields
too few eye movement samples) and converted into saccade frequency =
(number of saccades)/(total fixation time in ROI) (SI Materials and Methods).
Procedure. The study used a randomized, placebo-controlled, double-blind
within-subject experimental design. Patients received oxytocin and placebo
ball-tossing game and on face perception tasks and completed a number of
rating scales following the ball game on each visit (Table S5). Healthy subjects
were tested on a single visit. General affect was measured after oxytocin and
after placebo intake for each participant using the PANAS scale to assess the
possible mood-altering effects of oxytocin (SI Materials and Methods).
Statistics. Statistical analyses were conducted on the behavioral data in the
social ball-tossing game, eye movement measurements, and plasma levels of
oxytocin. Nonparametric tests were used because the distribution of the data
(number of balls, gaze time, oxytocin levels) was non-Gaussian (SI Materials
and Methods). Comparisons between groups and drug treatment conditions
were made using nonparametric ANOVA (Friedman) as well as Wilcoxon
signed-rank and Mann–Whitney rank-sum tests. The sequence effect of
treatment was tested using Mann–Whitney rank tests. All tests were eval-
uated against two-tailed probabilities.
Physiological Measures. Oxytocin administration. Each subject served as his or
her own control and received oxytocin and placebo with a 1-week interval, in
a balanced within-subject design. Participants received a single intranasal
dose of 24 IU oxytocin (Syntocinon Spray; Novartis; three puffs per nostril,
(see SI Materials and Methods for the details on the measurement
ACKNOWLEDGMENTS. We thank Paul El Khoury and Philippe Vindras for
methodological assistance and Dr. Nicolas Georgieff for assistance during
approval of the Ethical Committee protocol. We are particularly grateful to
Dr. Carmine Mottolese for providing clinical facilities during blood tests and
oxytocin administration. This research is sponsored by CNRS and supported
by a Human Frontier Science Program Research Grant (RGP0056/2005-C) and
by Fondation de France (A.S.). E.A. is supported by a French Ministry of
Research fellowship. The authors have declared no conflict of interest. M.L. is
supported by Inserm, Fondation Orange, and Fondation FondaMental.
1. Baron-Cohen S, Wheelwright S, Hill J, Raste Y, Plumb I (2001) The “Reading the Mind
in the Eyes” Test revised version: A study with normal adults, and adults with
Asperger syndrome or high-functioning autism. J Child Psychol Psychiatry 42:241–251.
2. Dalton KM, et al. (2005) Gaze fixation and the neural circuitry of face processing in
autism. Nat Neurosci 8:519–526.
3. Klin A, Jones W, Schultz R, Volkmar F, Cohen D (2002) Visual fixation patterns during
viewing of naturalistic social situations as predictors of social competence in
individuals with autism. Arch Gen Psychiatry 59:809–816.
4. Volkmar FR, Klin A (2000) Asperger’s syndrome, eds Klin A, Volkmar F, Sparrow SS
(Guilford, New York), pp 25–71.
5. Frith U (2004) Emanuel Miller lecture: Confusions and controversies about Asperger
syndrome. J Child Psychol Psychiatry 45:672–686.
6. Durand CM, et al. (2007) Mutations in the gene encoding the synaptic scaffolding
protein SHANK3 are associated with autism spectrum disorders. Nat Genet 39:25–27.
7. Jamain S, et al.; Paris Autism Research International Sibpair Study (2003) Mutations of
the X-linked genes encoding neuroligins NLGN3 and NLGN4 are associated with
autism. Nat Genet 34:27–29.
8. WaterhouseL,FeinD,ModahlC(1996) Neurofunctional mechanismsinautism.Psychol
9. Bartz JA, Hollander E (2006) The neuroscience of affiliation: Forging links between
basic and clinical research on neuropeptides and social behavior. Horm Behav 50:
10. Insel TR (1997) A neurobiological basis of social attachment. Am J Psychiatry 154:
11. Huber D, Veinante P, Stoop R (2005) Vasopressin and oxytocin excite distinct neuronal
populations in the central amygdala. Science 308:245–248.
12. Landgraf R, Neumann ID (2004) Vasopressin and oxytocin release within the brain: A
dynamic concept of multiple and variable modes of neuropeptide communication.
Front Neuroendocrinol 25:150–176.
13. Ferguson JN, Young LJ, Insel TR (2002) The neuroendocrine basis of social recognition.
Front Neuroendocrinol 23:200–224.
14. Lim MM, Young LJ (2006) Neuropeptidergic regulation of affiliative behavior and
social bonding in animals. Horm Behav 50:506–517.
15. Modahl C, et al. (1998) Plasma oxytocin levels in autistic children. Biol Psychiatry 43:
16. Green L, et al. (2001) Oxytocin and autistic disorder: Alterations in peptide forms. Biol
17. Kosfeld M, Heinrichs M, Zak PJ, Fischbacher U, Fehr E (2005) Oxytocin increases trust in
humans. Nature 435:673–676.
18. Guastella AJ, Mitchell PB, Mathews F (2008) Oxytocin enhances the encoding of
positive social memories in humans. Biol Psychiatry 64:256–258.
19. Savaskan E, Ehrhardt R, Schulz A, Walter M, Schächinger H (2008) Post-learning intranasal
oxytocin modulates human memory for facial identity. Psychoneuroendocrinology 33:
20. Guastella AJ, Mitchell PB, Dadds MR (2008) Oxytocin increases gaze to the eye region
of human faces. Biol Psychiatry 63:3–5.
21. Williams KD, Cheung CK, Choi W (2000) Cyberostracism: Effects of being ignored over
the Internet. J Pers Soc Psychol 79:748–762.
22. Wing L, Gould J (1979) Severe impairments of social interaction and associated ab-
normalities in children: Epidemiology and classification. J Autism Dev Disord 9:11–23.
23. Domes G, Heinrichs M, Michel A, Berger C, Herpertz SC (2007) Oxytocin improves
“mind-reading” in humans. Biol Psychiatry 61:731–733.
24. Kirsch P, et al. (2005) Oxytocin modulates neural circuitry for social cognition and fear
in humans. J Neurosci 25:11489–11493.
25. Petrovic P, Kalisch R, Singer T, Dolan RJ (2008) Oxytocin attenuates affective evalu-
ations of conditioned faces and amygdala activity. J Neurosci 28:6607–6615.
26. Herrnstein, R.J. (1961) Relative and absolute strength of response as a function of
frequency of reinforcement. J Exp Anal Behav 4:267–272.
27. van Beest I, Williams KD (2006) When inclusion costs and ostracism pays, ostracism still
hurts. J Pers Soc Psychol 91:918–928.
28. Hollander E, et al. (2003) Oxytocin infusion reduces repetitive behaviors in adults with
autistic and Asperger’s disorders. Neuropsychopharmacology 28:193–198.
29. Hollander E, et al. (2007) Oxytocin increases retention of social cognition in autism.
Biol Psychiatry 61:498–503.
30. Gillberg C, Gillberg C, Råstam M, Wentz E (2001) The Asperger Syndrome (and high-
functioning autism) Diagnostic Interview (ASDI): A preliminary study of a new
structured clinical interview. Autism 5:57–66.
31. Lord C, Rutter M, Le Couteur A (1994) Autism Diagnostic Interview-Revised: A revised
version of a diagnostic interview for caregivers of individuals with possible pervasive
developmental disorders. J Autism Dev Disord 24:659–685.
32. Hansson SL, et al. (2005) Psychiatric telephone interview with parents for screening of
comorbidities (A-TAC): Preliminary reliability and validity. Br J Psychol 187:262–267.
hyperactivitydisorder and other
6 of 6
| www.pnas.org/cgi/doi/10.1073/pnas.0910249107 Andari et al.