Article
Reference
Oxytocin increases trust in humans
KOSFELD, Michael, et al.
Abstract
Trust pervades human societies1,2. Trust is indispensable in friendship, love, families and
organizations, and plays a key role in economic exchange and politics3. In the absence of
trust among trading partners, market transactions break down. In the absence of trust in a
country's institutions and leaders, political legitimacy breaks down. Much recent evidence
indicates that trust contributes to economic, political and social success4,5. Little is known,
however, about the biological basis of trust among humans. Here we show that intranasal
administration of oxytocin, a neuropeptide that plays a key role in social attachment and
affiliation in non-human mammals6,7,8, causes a substantial increase in trust among humans,
thereby greatly increasing the benefits from social interactions. We also show that the effect of
oxytocin on trust is not due to a general increase in the readiness to bear risks. On the
contrary, oxytocin specifically affects an individual's willingness to accept social risks arising
through interpersonal interactions. These results concur with animal research suggesting an
essential role for oxytocin as [...]
KOSFELD, Michael, et al. Oxytocin increases trust in humans. Nature, 2005, vol. 435, no.
7042, p. 673-676
DOI : 10.1038/nature03701
Available at:
http://archive-ouverte.unige.ch/unige:101739
Disclaimer: layout of this document may differ from the published version.
1 / 1
Seediscussions,stats,andauthorprofilesforthispublicationat:https://www.researchgate.net/publication/7811372
OxytocinIncreasesTrustinHumans
ArticleinNature·July2005
DOI:10.1038/nature03701·Source:PubMed
CITATIONS
2,124
READS
1,104
5authors,including:
Someoftheauthorsofthispublicationarealsoworkingontheserelatedprojects:
InequalityAversion,Efficiency,andMaximinPreferencesinSimpleDistributionExperiments:Comment
Viewproject
PrejudiceandTruthAbouttheEffectofTestosteroneonHumanBargainingBehaviourViewproject
MichaelKosfeld
Goethe-UniversitätFrankfurtamMain
47PUBLICATIONS3,769CITATIONS
SEEPROFILE
UrsFischbacher
UniversitätKonstanz
130PUBLICATIONS25,111CITATIONS
SEEPROFILE
ErnstFehr
UniversityofZurich
269PUBLICATIONS47,276CITATIONS
SEEPROFILE
AllcontentfollowingthispagewasuploadedbyErnstFehron01December2015.
Theuserhasrequestedenhancementofthedownloadedfile.
Oxytocin increases trust in humans
Michael Kosfeld
1
*, Markus Heinrichs
2
*, Paul J. Zak
3
, Urs Fischbacher
1
& Ernst Fehr
1,4
Trust pervades human societies
1,2
. Trust is indispensable in friend-
ship, love, families and organizations, and plays a key role in
economic exchange and politics
3
. In the absence of trust among
trading partners, market transactions break down. In the absence
of trust in a country’s institutions and leaders, political legitimacy
breaks down. Much recent evidence indicates that trust contrib-
utes to economic, political and social success
4,5
. Little is known,
however, about the biological basis of trust among humans. Here
we show that intranasal administration of oxytocin, a neuro-
peptide that plays a key role in social attachment and affiliation
in non-human mammals
6–8
, causes a substantial increase in trust
among humans, thereby greatly increasing the benefits from social
interactions. We also show that the effect of oxytocin on trust is
not due to a general increase in the readiness to bear risks. On the
contrary, oxytocin specifically affects an individual’s willingness
to accept social risks arising through interpersonal interactions.
These results concur with animal research suggesting an essential
role for oxytocin as a biological basis of prosocial approach
behaviour.
In non-human mammals, the neuropeptide oxytocin has a central
role in general behavioural regulation, particularly in positive social
interactions. Aside from its well-known physiological functions in
milk letdown and during labour, oxytocin receptors are distributed
in various brain regions associated with behaviour
9,10
, including pair
bonding, maternal care, sexual behaviour, and the ability to form
normal social attachments
6–8,11–15
. Thus, oxytocin seems to permit
animals to overcome their natural avoidance of proximity and
thereby facilitates approach behaviour. Given that oxytocin is
believed to promote social attachment and affiliation in non-
human mammals, we hypothesized that oxytocin might also pro-
mote prosocial approach behaviours
—
such as trust
—
in humans.
Recent research has shown that neuropeptides cross the blood-brain
barrier after intranasal administration
16
, providing a useful method
for studying the central nervous system effects of oxytocin in
humans
17,18
. We used a double-blind study design to compare
trusting behaviour in a group of subjects that received a single dose
of intranasal oxytocin with that of subjects in a control group that
received placebo.
We analysed the effect of exogenously administered oxytocin on
individuals’ decisions in a trust game with real monetary stakes
19–22
.
In this trust game, two subjects interacting anonymously play either
the role of an investor or a trustee (Fig. 1). First, the investor has the
option of choosing a costly trusting action by giving money to the
trustee. If the investor transfers money, the total amount available for
distribution between the two players increases but, initially, the
trustee reaps the whole increase. The trustee is then informed
about the investor’s transfer and can honour the investor’s trust by
sharing the monetary increase generated by the investor’s transfer.
Thus, if the investor gives money to the trustee and the latter shares
the proceeds of the transfer, both players end up with a higher
monetary payoff. However, the trustee also has the option of
violating the investor’s trust. As sharing the proceeds is costly for
the trustee, a selfish trustee will never honour the investor’s trust
because the investor and the trustee interact only once during the
experiment.
The investor is therefore caught in a dilemma: if he trusts and the
trustee shares, the investor increases his payoff, but he is also subject
to the risk that the trustee will abuse this trust. In the latter case, the
investor is worse off than if he had not trusted at all and, adding insult
to injury, the trustee has an unfair payoff advantage relative to the
investor. Substantial evidence exists to show that humans are averse
to such risks
22–24
. Moreover, the aversion of investors to abuse of trust
seems to have an important role across different human cultures and
social groups in the context of our game
22,25
. The investors have to
overcome their aversion against these risks in order to trust, allowing
us to address the question of whether oxytocin modulates this
trusting behaviour in humans.
LETTERS
Figure 1 | The trust game. Both subjects receive an initial endowment of 12
monetary units (MU). The investor can send 0, 4, 8 or 12 MU to the trustee.
The experimenter triples each MU the investor transfers. After the investor’s
decision is made, the trustee is informed about the investor’s transfer. Then
the trustee has the option of sending any amount between zero and his total
amount available back to the investor. For example, if the investor has sent
12 MU, the trustee possesses 48 MU (12 MU own endowment þ 36 MU
tripled transfer) and can, therefore choose any back transfer from 0 to
48 MUs. The experimenter does not triple the back transfer. The investor’s
final payoff corresponds to the initial endowment minus the transfer to the
trustee, plus the back transfer from the trustee. The trustee’s final payoff is
given by his initial endowment plus the tripled transfer of the investor,
minus the back transfer to the investor. At the end of the experiment, the
earned MU are exchanged into real money according to a publicly
announced exchange rate (see Methods). Each subject made four decisions
in the same player role while paired with four different, randomly selected
interaction partners.
1
University of Zurich, Institute for Empirical Research in Economics, Blumlisalpstrasse 10, CH-8006 Zurich, Switzerland.
2
University of Zurich, Department of Clinical Psychology
and Psychotherapy, Zurichbergstrasse 43, CH-8044 Zurich, Switzerland.
3
Center for Neuroeconomics Studies, Claremont Graduate University, Claremont, California 91711-6165,
USA.
4
Collegium Helveticum, Schmelzbergstrasse 25, CH-8092 Zurich, Switzerland.
*These authors contributed equally to this work.
Vol 435|2 June 2005|doi:10.1038/nature03701
673
© 2005 Nature Publishing Group
Our hypothesis that oxytocin increases the trusting behaviour of
investors implies that the investors in the oxytocin group (n ¼ 29)
will show higher money transfers than those in the placebo group
(n ¼ 29). In fact, our data show that oxytocin increases investors’
trust considerably. Out of the 29 subjects, 13 (45%) in the oxytocin
group showed the maximal trust level, whereas only 6 of the 29
subjects (21%) in the placebo group showed maximal trust (Fig. 2a).
In contrast, only 21% of the subjects in the oxytocin group had a trust
level below 8 monetary units (MU), but 45% of the subjects in the
control group showed such low levels of trust. These differences in
the distribution of trust result in higher average and median trust
levels for subjects given oxytocin (Table 1). The investors’ average
transfer is 17% higher in the oxytocin group (Mann-Whitney U-test;
z ¼ 21.897, P ¼ 0.029, one-sided), and the median transfer in the
oxytocin group is 10 MU, compared to a median of only 8 MU for
subjects in the placebo group.
In the trust game, the risk on the part of the investor’s is due to the
uncertainty of the trustee’s behaviour
—
that is, a social interaction
with a specific trustee constitutes the risk. This raises the question of
whether oxytocin helps humans to overcome a general aversion
against risks or whether oxytocin specifically affects trusting beha-
viour in social interactions. In order to answer this question, we
conducted a risk experiment in which the investor faced the same
choices as in the trust game but in which a random mechanism, not
the trustee’s decision, determined the investor’s risk. The random
mechanism in the risk experiment replicated the trustees’ decisions in
the trust experiment. Therefore, the investors faced exactly the same
risk as in the trust experiment (see Methods); however, their transfer
decisions were not embedded in a social interaction because there
were no trustees in the risk experiment.
In this risk experiment, the investors’ behaviour does not differ
between the oxytocin and the placebo groups (Table 1 and Fig. 2b).
The median transfer is 8 MU and the average transfer is 7.5 MU in
both groups (Mann-Whitney U-test; z ¼ 0.022, P ¼ 0.983; two-
sided test, n ¼ 31 in oxytocin group, n ¼ 30 in placebo group).
Moreover, there is no significant difference in a comparison of the
placebo group in the trust experiment with the oxytocin group and
the placebo group in the risk experiment (Kruskal-Wallis test;
x
2
¼ 0.533, d.f. ¼ 2, P ¼ 0.766), with identical median transfers
across groups (Table 1). However, if we add the oxytocin group in
the trust experiment to these three samples, significant differences
are observed (Kruskal-Wallis test;
x
2
¼ 8.610, d.f. ¼ 3, P ¼ 0.035),
indicating that only the investors in the oxytocin group of the trust
experiment behave differently. Thus, oxytocin increases the investors’
transfer levels in the trust experiment but not in the risk experiment.
This finding is illustrated by a comparison of Figs 2a and b, which
show that only 10% of the subjects with oxytocin choose the maximal
transfer level in the risk experiment, whereas 45% choose the
maximal level in the trust experiment. Therefore, the differences
between the oxytocin group in the trust experiment and the oxytocin
group in the risk experiment are highly significant (Mann-Whitney
U-test; z ¼ 22.563, P ¼ 0.010, two-sided), suggesting that oxytocin
specifically affects trust in interpersonal interactions.
The risk experiment constitutes a powerful control for the effects
of oxytocin on trusting behaviour because everything is kept con-
stant relative to the trust experiment, except that the investors’ risk in
the risk experiment is not generated through a social interaction.
Specifically, all the indirect effects of oxytocin on the state of a
subject, such as possible effects on mood or calmness, would be
present in both the trust and the risk experiment. Therefore, these
potential indirect effects of oxytocin cannot be responsible for the
effect of oxytocin on trusting behaviour. Moreover, in order to
provide an additional control for non-specific effects that might be
associated with oxytocin administration, we explicitly measured
mood and calmness before substance administration and 50 min
after administration (but before subjects played the trust or the risk
game). We used a questionnaire suitable for repeated measures
within short periods of time, one that is widely used in neuropharma-
cological studies in humans
26
and correlates with physiological
measures
17
. There were no statistical differences in the levels of
mood and calmness before and after the administration of oxytocin
in either the trust or the risk experiment. (Trust experiment:
z ¼ 21.541, P ¼ 0.123 for calmness; z ¼ 1.452, P ¼ 0.146 for
mood; n ¼ 29. Risk experiment: z ¼ 0.620, P ¼ 0.535 for calmness;
z ¼ 20.841, P ¼ 0.400 for mood; n ¼ 31; two-sided Wilcoxon
signed rank tests.) This provides further support for our conclusion
Figure 2 | Transfers in the trust and the risk experiment. Each observation
represents the average transfer amount (in MU) over four transfer decisions
per investor. a, Relative frequency of investors’ average transfers in oxytocin
(filled bars) and placebo (open bars) groups in the trust experiment
(n ¼ 58). Subjects given oxytocin show significantly higher transfer levels.
b, Relative frequency of investors’ average transfers in oxytocin (filled bars)
and placebo (open bars) groups in the risk experiment (n ¼ 61). Subjects in
the oxytocin and the placebo group show statistically identical transfer
levels.
Table 1 | Median and average transfer behaviour of investors
Trust experiment Risk experiment
Oxytocin group Placebo group Oxytocin group Placebo group
Mean average transfer (MU) 9.6 8.1 7.5 7.5
Median average transfer (MU) 10 8 8 8
Standard deviation of transfers (MU) 2.8 3.1 3.3 3.4
Number of observations 29 29 31 30
LETTERS NATURE|Vol 435|2 June 2005
674
© 2005 Nature Publishing Group
that the effect of oxytocin on human trust is not caused by non-
specific, psychotropic effects of oxytocin.
What mechanisms might be involved in generating the effect of
oxytocin on trusting behaviour? One possibility is that oxytocin
causes a general increase in prosocial inclinations. This implies that
oxytocin should affect not only the prosocial behaviour of the
investors but also that of the trustees. We would therefore predict
that those trustees who are given oxytocin should make higher back
transfers at any given level than the trustees who received placebo.
However, trustees given oxytocin do not show more trustworthy
behaviour (Fig. 3). At every positive transfer level (4, 8 or 12 MU),
their back transfers are statistically indistinguishable from those of
placebo trustees (Mann Whitney U-tests; P . 0.243, two-sided tests
for each positive transfer level). Thus, oxytocin does not increase
the general inclination to behave prosocially. Rather, oxytocin
specifically affects the trusting behaviour of investors.
We hypothesize that the differing effect of oxytocin on the
behaviour of investors and trustees is related to the fact that investors
and trustees face rather different situations. Specifically, investors
have to make the first step; they have to ‘approach’ the trustee by
transferring money. In contrast, the trustees can condition their
behaviour on the basis of the investors’ actions. Thus, the psychology
of trust is important for investors, whereas the psychology of strong
reciprocity
27
is relevant for trustees. The fact that oxytocin affects
subjects’ approach or trust behaviour, but not their degree of
reciprocity, is in agreement with animal studies. There is substantial
evidence that oxytocin promotes prosocial approach behaviour by
inhibiting defensive behaviours
6,13
, but there is no evidence that
oxytocin affects reciprocity in animals.
A second mechanism behind the effect of oxytocin on trust could
be based on subjects’ beliefs. Oxytocin might render subjects more
optimistic about the likelihood of a good outcome. In order to
address this question, we measured the investor’s subjective expec-
tation about the trustee’s back transfer after every transfer decision.
A Mann-Whitney U-test indicates that these expectations do not
differ significantly between oxytocin and placebo groups at every
feasible positive transfer level (P . 0.357, two-sided tests at transfer
levels of 4, 8 or 12 MU). Thus, the investors given oxytocin show
more trusting behaviour but do not hold significantly different
beliefs about the trustworthiness of others. Moreover, oxytocin
does not affect investors’ beliefs about the likelihood of a good
outcome in the risk experiment (P . 0.128, two-sided Mann
Whitney U-tests for transfer levels of 4, 8 or 12 MU).
Finally, there is the possibility that oxytocin helps subjects to
overcome their betrayal aversion in social interactions. This expla-
nation is consistent with the differing effects of oxytocin across the
trust and the risk experiments, and is further supported by the fact
that investors faced a considerable betrayal risk. An increase in the
transfer level from 4 or 8 MU to 12 MU decreased the investor’s
average payoff slightly, whereas it increased the objective risk of very
low back transfers by the trustee. However, betrayal aversion alone
cannot explain why investors given oxytocin make higher transfers in
the trust experiment compared with the risk experiment, because
betrayal is impossible in the risk experiment. The higher transfers in
the trust experiment can be reconciled with betrayal aversion if one
acknowledges that investors’ behaviour in the trust experiment is also
likely to be driven by the motive to increase the available amount for
distribution between the two players
28
. As this motive cannot operate
in the risk experiment, it can only increase transfers levels in the trust
experiment. Our interpretation of oxytocin’s effect on trust in terms
of betrayal aversion may be seen in the light of animal studies
indicating that increased availability of oxytocin in the central
nervous system facilitates approach behaviour, by linking the over-
coming of social avoidance with the activation of brain circuits
implicated in reward (for example, the nucleus accumbens)
12,15
.
The ubiquity of trusting behaviour is perhaps one of the dis-
tinguishing features of the human species. An element of trust
characterizes almost all human social interactions. Here we have
sought to examine the effect of oxytocin on trust in humans.
Research in non-human mammals suggests that oxytocin has a key
role in social attachment and affiliation. We find that intranasal
administration of oxytocin causes a substantial increase in trusting
behaviour. Subjects given oxytocin seem better able to overcome
trust obstacles such as betrayal aversion. Of course, this finding could
be misused to induce trusting behaviours that selfish actors sub-
sequently exploit. However, our findings may also have positive
clinical implications for patients with mental disorders that are
associated with social dysfunctions (for example, social phobia or
autism). In particular, social phobia ranks as the third most common
mental health disorder and is characterized by marked social deficits,
including persistent fear and avoidance of social interactions. Thus,
our results might lead to fertile research on the role of oxytocin in
several mental health disorders with major public health significance.
METHODS
Subjects. A total of 194 healthy male students (mean age ^ s.d., 22.0 ^ 3.4 yr)
from different universities in Zurich participated in the study. The trust
experiment had 128 participants, and 66 subjects participated in the risk
experiment. Exclusion criteria for participation were significant medical or
psychiatric illness, medication, smoking more than 15 cigarettes per day, and
drug or alcohol abuse. Subjects were instructed to abstain from food and drink
(other than water) for 2 h before the experiment, and from alcohol, smoking and
caffeine for 24 h before the experiment. Participants were informed at the time of
recruitment that the experiment would evaluate the effects of a hormone on
decision making. In total, 16 individuals out of the original sample of 194 were
excluded because of incorrect substance administration (7 in the trust experi-
ment, 5 in the risk experiment) or their stated disbelief that the opponent in the
trust game was actually a human being (4 participants). The study protocol was
approved by the ethics committee of the University of Zurich. All subjects gave
written, informed consent before participation.
Substance administration. Subjects received a single intranasal dose of 24 IU
oxytocin (Syntocinon-Spray, Novartis; 3 puffs per nostril, each with 4 IU
oxytocin) or placebo 50 min before the start of the trust or the risk experiment.
Subjects were randomly assigned to the oxytocin or placebo group (double-
blind, placebo-controlled study design). In order to avoid any subjective
substance effects (for example, olfactory effects) other than those caused by
oxytocin, the placebo contained all inactive ingredients except for the
neuropeptide.
Behavioural experiment and questionnaires. After substance administration,
Figure 3 | Average back transfer of trustees to their investors. The graph
shows the average back transfer by trustees for different levels of investor
transfer in the oxytocin and placebo groups. The dotted line shows the level
of the back transfer necessary to achieve payoff equality between the investor
and the trustee. The dashed line shows a level of back transfer equal to the
investor’s transfer to the trustee. The trustees’ back transfers are on average
slightly higher than the amount sent by the investor. Trustees in both
treatment groups make higher back transfers in response to higher original
investor transfer levels. However, there is no statistically significant
difference in back transfers between subjects in the oxytocin and the placebo
groups.
NATURE|Vol 435|2 June 2005 LETTERS
675
© 2005 Nature Publishing Group
