Friend or foe: The effect of implicit trustworthiness judgments
in social decision-making
M. van ’t Wout*, A.G. Sanfey
Neural Decision Sciences Laboratory, Department of Psychology, University of Arizona, 1503 E University Boulevard, P.O. Box 210068, Tucson, AZ 85721, USA
a r t i c l e i n f o
Received 30 June 2007
Revised 8 July 2008
Accepted 14 July 2008
a b s t r a c t
The human face appears to play a key role in signaling social intentions and usually people
form reliable and strong impressions on the basis of someone’s facial appearance. There-
fore, facial signals could have a substantial influence on how people evaluate and behave
towards another person in a social interaction, such as an interactive risky decision-making
game. Indeed, there is a growing body of evidence that demonstrates that social behavior
plays a crucial role in human decision-making. Although previous research has demon-
strated that explicit social information about one’s partner can influence decision-making
behavior, such as knowledge about the partner’s moral status, much less is known about
how implicit facial social cues affect strategic decision-making. One particular social cue
that may be especially important in assessing how to interact with a partner is facial trust-
worthiness, a rapid, implicit assessment of the likelihood that the partner will reciprocate a
generous gesture. In this experiment, we tested the hypothesis that implicit processing of
trustworthiness is related to the degree to which participants cooperate with previously
unknown partners. Participants played a Trust Game with 79 hypothetical partners who
were previously rated on subjective trustworthiness. In each game, participants made a
decision about how much to trust their partner, as measured by how much money they
invested with that partner, with no guarantee of return. As predicted, people invested more
money in partners who were subjectively rated as more trustworthy, despite no objective
relationship between these factors. Moreover, the relationship between the amount of
money offered seemed to be stronger for trustworthy faces as compared to untrustworthy
faces. Overall, these data indicate that the perceived trustworthiness is a strong and impor-
tant social cue that influences decision-making.
? 2008 Elsevier B.V. All rights reserved.
Research that investigates social behavior and human
interaction often focuses on the processing of facial
features. Indeed, the face is an excellent source of informa-
tion and in particular appears to play a key role in signaling
social intentions from which people infer meaning, such as
personality traits and complex social characteristics (Frith
& Frith, 1999). Generally, people form reliable and strong
impressions on the basis of facial appearance. In addition,
facial cues related to threat are generally processed quickly
and automatically, with only 40 ms exposure enough to
generate an opinion about someone’s intentions (Bar, Neta,
& Linz, 2006). Moreover, there is ample evidence that the
human brain has specialized areas for the processing of
faces (Kanwisher, McDermott, & Chun, 1997). Therefore,
facial signals could have a substantial influence on how
people evaluate and behave towards another person in a
social interaction, such as an interactive risky decision-
A particularly useful task in examining interactive
(Berg, Dickhaut, & McCabe, 1995). In a standard Trust
well-studied Trust Game
0010-0277/$ - see front matter ? 2008 Elsevier B.V. All rights reserved.
* Corresponding author. Tel.: +1 520 6268597; fax: +1 520 6219306.
E-mail address: email@example.com (M. van ’t Wout).
Cognition 108 (2008) 796–803
Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/locate/COGNIT
Game two anonymous persons, an investor and partner,
interact with one another. The investor is endowed with
a certain amount of money and informed that she has
endowment to her partner. She can transfer no money,
the entire endowment, or any amount in between. The
money transferred by the investor is then multiplied by
the experimenter (usually by a factor of 3 or 4) and given
to the partner. The partner, with whom the investor is
not allowed to communicate during the game, then has
the opportunity to transfer back some of the multiplied
amount to the investor. In essence, the partner has a choice
between honoring and abusing trust. If the partner honors
trust and returns more money than transferred by the
investor, both players end up with a higher monetary
payoff than was originally obtained. However, if the
partner abuses trust, the investor ends up with less money
than the initial endowment, whereas the partner ends up
with a large profit.
In a standard Trust Game, the investor and partner
interact only once and therefore sharing the initial
endowment is risky for the partner. From a traditional
Game Theoretic standpoint, a rational (and selfish) partner
would never honor trust given by the investor, and would
keep the entire transferred amount for themselves. The
investor, herself being rational, naturally realizes this,
and so no amount would ever be sent over to the partner.
Thus, by this theoretical view, trust is neither given nor
repaid during this game (Berg et al., 1995; Camerer,
2003). However, in most experimental studies of the Trust
Game, investors are usually willing to send some amount
of money to the partner, and generally this trust is repaid
to some extent. For instance in the original study by Berg
et al. (1995), 30 out of 32 people did send some money
to their partners, with on average half of the initial
endowment being invested. In return, this trust was repaid
with approximately one third of the multiplied amount,
although there was a wide distribution with almost half
of people sending back very little or nothing. Recently, a
growing body of research acknowledges the role of social
behavior in why people deviate from traditional Game
important questions as to exactly how social interaction
can influence decision-making.
For instance, research has demonstrated that the
willingness of the investor to trust can be altered by
several experimental factors. Specifically, labeling the
investor’s partner with terms that imply cooperation or
competition can affect decision-making, such as when
the term ‘opponent’ is used instead of ‘partner’. Use of
the former resulted in half as much trusting behavior as
the latter (Burnham, McCabe, & Smith, 2000). Information
about the partner, such as how they treated someone else
in the past (King-Casas et al., 2005), can also change the
behavior of the investor. Studies have also examined
how explicit information about the social and moral
status of the partner can influence the investor’s willing-
ness to place trust (Delgado, Frank, & Phelps, 2005;Singer,
Kiebel, Winston, Dolan, & Frith, 2004). Delgado et al.
(2005) showed that investors were more likely to transfer
greater amounts of money to partners described in
vignettes as having high moral character as opposed to
those demonstrating more neutral or more venal behavior.
In this study, participants also gave higher trustworthiness
ratings to the morally ‘good’ partners.
However, even in Trust Games where investors know
performance of their partner but are simply paired with
an anonymous partner, there may well be substantial dif-
ferences in terms of how much money is transferred to
certain partners. Investors do make a (usually very rapid)
decision regarding the trustworthiness of a previously
unencountered partner and in the absence of any other
information, it seems likely that automatically processed
cues that signal one’s social intentions, attitudes, or
behavior could be vital in determining the likelihood of
cooperation in the Trust Game. Since the face is particu-
larly important in determining someone’s intentions as
outlined above, facial signals could have a substantial
influence on people’s behavior in the Trust Game. For
instance, emotional expressions in general (Eckel& Wilson,
1999), and smiling in particular (Scharlemann, Eckel,
Kacelnik, & Wilson, 2001), are social-emotional cues that
could influence the decision to trust in economic games.
Despite this, relatively little is known about how non-emo-
tional, purely social, facial signals, such as competence or
trustworthiness, influence strategic decision-making.
Trustworthiness in particular would appear to be a
social facial signal of special significance, as it can
potentially provide information about whether another
individual is someone to approach or avoid, trust or dis-
trust. Moreover, the trustworthiness of someone is read-
ily judged from appearance with high reliability across
people (Berry & Brownlow, 1989; Berry & McArthur,
1986; Brownlow, 1992). Indeed, Willis and Todorov
(2006) have shown that people are able to judge the
trustworthiness of faces very quickly (within 100 ms)
and that this judgment is quite invariant even when
more time is provided, suggesting that people are partic-
ularly efficient in judging the trustworthiness of faces.
Though ‘trustworthiness’ is a rather ambiguous term, it
has been argued that the detection of trustworthiness
is essential for human survival (Cosmides
2000), and there is good evidence for a set of brain
regions which may be involved in the perception of
trustworthiness. An fMRI study by
and O’Doherty Dolan (2002) showed that facial trustwor-
thiness evaluation is associated with activation of the
amygdala and insula, specifically that these areas are
increases. This is particularly interesting, as these brain
areas have been implicated in detecting potentially
dangerous stimuli (Amaral, 2002), encoding feelings of
disgust (Phillips et al., 2004) and in other studies of
(social) strategic decision-making
Aronson, Nystrom, & Cohen, 2003).
The relationship between these highly automatic, rapid,
judgment processes and strategic decision-making is
clearly relevant in many real-world interactions. Though
it seems likely that perception of an individuals’ trustwor-
thiness is related to decisions made about that person in
strategic interactions, this question has surprisingly not
character or previous
M. van ’t Wout, A.G. Sanfey/Cognition 108 (2008) 796–803
been studied in the laboratory to date. The present study
sought to remedy this, and examined the role of perceived
trustworthiness on risky social decision-making. Partici-
pants played the role of investor in a one-shot, two-person
Trust Game with partners who had been previously inde-
pendently rated on trustworthiness. We hypothesized that
investors who are confronted with partners perceived as
less trustworthy would invest less money than when con-
fronted with more trustworthy-looking partners. After the
Trust Game rounds participants saw their partners again
and rated their trustworthiness. In addition, we were
interested in whether the relationship between judged
trustworthiness and trust behavior might be is stronger
for faces that are perceived as either trustworthy or
untrustworthy. This question is motivated by the fMRI
study of Winston etal. (2002) who reported amygdala acti-
vation particularly in response to untrustworthy faces,
suggesting that untrustworthiness might be a stronger
facial cue than trustworthiness. On the other hand, in an
fMRI study by Singer et al. (2004) participants showed
brain activation in areas associated with social cognition
particularly in responses for partners that had shown to
be cooperative in a Prisoner’s Dilemma Game, which is
similar in structure to the Trust Game. Moreover, this
study design also allowed a preliminary analysis whether
people recognize particular faces better than others. More
specifically, Singer et al. (2004) showed that memory
performance was better for faces that were associated with
cooperation than with defection. Thus, we hypothesized
that recognition would be better for faces of partners that
repaid trust as compared to those that abused trust, and
additionally that this might be mediated by perceived
trustworthiness, i.e., recognition memory is best for
subjectively trustworthy faces that also acted trustworthy.
Sixty undergraduate students (33 women, 27 men) par-
ticipated in this study. Two individuals were excluded
from the analyses as they did not follow the rules of game
properly, i.e., offering more to the partner than was al-
lowed or providing no trustworthiness ratings. This re-
sulted in a group of 32 women and 26 men. The number
of men and women that participated in the experiment
did not differ significantly from each other, Chi2= 0.62,
p = .43. Mean age was 18.7 years (SD = 1.4). The study
was conducted in compliance of the Declaration of Helsinki
and local ethics committee approval, and all participants
provided written informed consent.
2.2. Trust Game
Participants were instructed as to the nature and rules
of the Trust Game. Participants acted only in the role of
investor in the Trust Game. In the task instructions it was
emphasized that the participant’s partners in the game
would play the game independently of each other, with
no collusion. Participants were also told that in addition
to course credit they would be paid based on their choices
in the game. After finishing the game, they received, in
cash, a small percentage of their overall game earnings
(average of $4).
Participants played against 79 different partners, and
began each trial by being endowed with $10. On each trial,
participants could decide to transfer over any amount of
this money, in one dollar increments. That is, they decided
to send over some amount between $0 (placing no trust)
and $10 (placing full trust). If they decided to trust the
partner, the amount of money selected was multiplied by
four and transferred to the partner.
Partners responded in pre-programmed fashion. When
a non-zero amount was transferred by the investor (that
is, some trust was placed), 50% of the time this trust was
repaid and 50% of the time this trust was abused. When
trust was repaid, the partner sent back half of the trans-
ferred, multiplied, amount to the investor. For example, if
$4 was transferred initially, this was multiplied by 4 for a
total of $16, and then half of this (i.e., $8) was sent back.
The investor therefore earned this $8, plus her original $6
(the $10 endowed minus the transferred $4), for a total
of $14 on this trial.
When trust was abused the exact outcome depended on
the amount of money transferred to the partner. When the
participant transferred less than $5, nothing was returned
back. If the transferred amount was more than $5, half the
amount money the participant initially invested (12.5% of
the total multiplied amount) was sent back. In both of
these cases, the investor ultimately earned less than her
originally endowed $10.
presented with a picture of their partner for that trial.
Pictures were 79 grayscale frontal photographic images of
emotionally neutral faces that were selected from a larger
set of 150 images; 100 are from the set of Adolphs, Tranel,
and Damasio (1998), supplemented with 50 images from
ment of Stirling University (http://pics.psych.stir.ac.uk/).
Faces were selected on the basis of trustworthiness
ratings given by 36 healthy subjects in a separate pilot
study (9 men, 27 women; mean age 21.6, SD = 3.3). All
faces were randomly presented in the game.
Following presentation of the photograph, participants
then decided how much of their endowment to offer to
their partner on that trial and pressed the appropriate
key on the computer keyboard to send the money over.
After a short delay, the participant was informed of their
partner’s decision and the total amount earned in the trial
was shown (see Fig. 1 for a trial timeline).
2.3. Subjective trust rating and recognition memory
After playing the Trust Game, participants completed an
unrelated task for approximately 20 min. Following that,
participants were presented serially with 150 faces, com-
prised of the 79 faces that were presented in the Trust
Game and 71 decoy faces. All 150 faces were randomly
presented. They were asked to rate how trustworthy they
thought each face was on a scale of ‘‘1” (highly untrustwor-
thy) to ‘‘7” (highly trustworthy), with ‘‘4” being neutral,
and also to rate whether they had seen that face before,
M. van ’t Wout, A.G. Sanfey/Cognition 108 (2008) 796–803
on a scale from ‘‘1” (definitely not seen) to ‘‘5” (definitely
seen). Due to technical problems, six people failed to
complete this part of the experiment.
3.1. Trust behavior and trustworthiness
Overall, participants placed some amount of trust in
85% of trials. Nine participants placed some trust in every
partner, i.e., sent over some amount of the $10 on every
round, whereas the remaining 49 participants sometimes
sent nothing to their partner. Although there was variation
in what was sent to a partner by different participants,
most participants used the full range from $0 to $10, and
only nine participants never sent more than $5 (half their
endowment) to their partners. The mean amount of money
offered when trust was placed was $3.52 (SD 2.65, range
$0-$10). On average, less money was sent over to a partner
at the end of the game (last 20 rounds) as compared to the
beginning of the game first (20 rounds), t(57) = 3.39,
p = .001. Mean subjective trustworthiness rating of faces
of partners after the game was 4.1 (SD = 1.65, range 1–7).
Subjective trust ratings given by participants correlated
extremely highly with trust ratings of the norm group,
Spearman r = .89, p < .0001.
There was a significant correlation between the normed
trustworthiness rating of the partner and the amount of
money that was transferred by the participant, Spearman
r = .75, p < .0001 (see Fig. 2). In addition, there was a
significant correlation between subjective trustworthiness
rating of the partner and the amount of money transferred
by the participant, Spearman r = .82, p < .0001. This result
strongly suggests that participants were indeed influenced
by the perceived subjective trustworthiness of the partner
in their judgment of whether to trust, and how much trust
to place. Of additional interest was whether the amount
transferred to a partner is also influenced by the outcome
experience by the participant (trust repaid or abused) in
the previous round. Therefore we performed a mixed linear
model with subjects as random factors to test whether the
amount of money sent over to a partner was influenced by
normed trustworthiness ratings, the outcome on the previ-
ous round (trust repaid or abused) and their interaction.
This resulted in a significant effect of normed trustworthi-
F(1,4702) = 81.15,
F(2,4702) = 6.66, p = .001); but no significant interaction
between normed trustworthiness ratings and outcome,
F(2,4702) = 0.61, p = .55.
It is possible that other factors, besides trustworthiness,
may play a role in the decision as to how much money to
p < .0001; outcome,
send over to the partner. Although we tried to eliminate
possible factors such as facial emotional expression, which
are known to influence decision-making, participants
could have picked up other facial features, such as attrac-
tiveness. Therefore we collected attractiveness ratings of
the set of faces, rated by an independent sample of under-
graduate students who did not play the game. Results
showed that although attractiveness correlated with trust-
worthiness, Spearman r = .55, p = .0002, a linear mixed
model analysis showed that attractiveness was not corre-
lated with the amount of money sent over to that partner,
F(1,2266) = 0.26, p = .61, whereas trustworthiness ratings
were F(1,2266) = 9.78, p = .002.
In their fMRI study, Winston et al. (2002) reported that
untrustworthy faces in particular appeared to activate
emotional brain areas, suggesting that the perceptual
system may be especially sensitive to untrustworthy faces.
To explore this question in the context of decisions made
in the Trust Game, we split the group of faces on the basis
of scores given by participants into untrustworthy (score
below neutral, ‘‘4”) and trustworthy (score above neutral).
Interestingly, the correlation between subjective trustwor-
thiness and amount of money transferred was stronger for
partners rated in general as trustworthy, Spearman r = .65,
p < .0001 (n = 45), as opposed to those rated as untrustwor-
thy, Spearman r = 0.29, p = .10 (n = 34). To test whether
these two correlations differ significantly from one another
we performed a Fischer z transformation for independent
samples. Using the participant trust ratings, the correla-
tions were significantly different (p < .05). Similar results
were found when we used normed trustworthiness ratings
(correlation with amount of money transferred was Spear-
man r = 0.45, p = .001 and Spearman r = 0.19, p = .29 for
trustworthy and untrustworthy faces, respectively) and
this difference between the correlation reached trend level
significant (p < .08). It is important to note here that the
number of faces rated as untrustworthy is smaller than
the number of faces rated as trustworthy. However, the
range of scores for trustworthy and untrustworthy faces
is comparable, i.e., 1.20 (variance = 0.092) and 1.31 (vari-
ance = 0.086), respectively. This suggests that the lack of
correlation between untrustworthy faces and amount of
money transferred is probably not due to a smaller sample
size or a restricted range in ratings.
Because we measured subjective trustworthiness after
playing the Trust Game, playing the Trust Game itself
could influence subjective trustworthiness ratings. For
instance, merely seeing a given face twice (once in the
game and once in the ratings test) may increase subjective
trustworthiness ratings due to familiarity. Indeed, subjec-
tive trust ratings were significantly higher for faces that
Fig. 1. Single trial in the Trust Game.
M. van ’t Wout, A.G. Sanfey/Cognition 108 (2008) 796–803
were seen previously than for the new decoy faces,
t(51) = ?9.67, p < .0001). Additionally, partners in whom
trust was placed, but who did not reciprocate this trust,
might be rated as less trustworthy afterwards. Although,
subjective trustworthiness ratings were not significantly
different for partners that repaid trust, as compared to
abused trust, t(57) = 0.12, p = .91, we performed a linear
mixed model analysis in which participants’ trustworthi-
ness ratings were regressed on normed trustworthiness
ratings, the partner’s return (calculated as the percentage
of the amount of money sent back as a function of what
was sent over), and their interaction, to specifically test
whether experience with a partner influences subjective
trustworthiness above and beyond normed trustworthi-
ness ratings. This resulted in a significant relationship only
between participant and normed trust ratings, F(1,4132) =
448.05, p < .0001. Neither the relationship between partner
return, F(3,4132) = 0.96, p = .41, or the interaction between
normed trust ratings and partner return were significant,
F(3,4132) = 0.43, p = .73. This suggests that post-game
participant trust ratings were specifically predicted by
normed trust ratings.
Finally, there may be a general effect whereby trust
increases or decreases across the whole experiment, inde-
pendent of the actual faces and partner behavior. We did
not find evidence of this, with no significant difference
observed in subjective trust ratings for the first 20 partners
as compared to last 20 partners faced in the Trust Game,
i.e., trust ratings for faces at the end of the game are not
higher than at the beginning, t(52) = ?0.98, p = .33).
Instead, only partners in whom no trust was placed were
rated afterwards as significantly less trustworthy, t(48) =
?4.60, p < .0001).
3.2. Recognition memory
With respect to memory for faces, faces that were
presented in the Trust Game were given higher recognition
ratings than faces that were new, t(51) = 10.76, p < .0001).
As mentioned previously, we hypothesized that recogni-
tion would be better for faces that repaid trust compared
to those that abused trust, that subjectively trustworthy
faces would be recognized better than untrustworthy
faces, and a potential interaction, namely that faces rated
as trustworthy that subsequently repaid trust would have
the highest recognition scores.
Recognition was indeed better for faces that were rated
as trustworthy by participants as compared to faces that
were rated as untrustworthy, F(1,50) = 35.59, p < .0001).
However, recognition ratings for faces that abused trust
did not differ significantly as compared to faces that repaid
trust, F(1,50) = 0.01, p = .92), nor was there a significant
interaction between trustworthiness of the face and deci-
sion of the partner, F(1,50) = 1.14, p = .29) (see Fig. 3). This
suggests that a partner who is a priori thought to be trust-
worthy is remembered better than a partner who is a priori
thought to be untrustworthy, regardless of their actual
behavior in the game. However, it is possible that subjec-
tively trustworthy faces are perceived as more familiar
because they are perhaps more average or attractive
Fig. 2. Scatter plot for normed trustworthiness ratings of a particular face (horizontal axis) and the average offer amount send to that partner (vertical axis).
M. van ’t Wout, A.G. Sanfey/Cognition 108 (2008) 796–803
compared to the untrustworthy faces. Therefore, we used
signal detection methods to calculate the false alarm rate
(saying a face is remembered whether it was absent in
the task), the hit rate (correct recognition of a face), false
negative rate (saying a face is new when it was present
in the task), correct rejection, and d’ (sensitivity of recogni-
tion memory). Results showed that there is no difference
between d’ for trustworthy (0.76) and untrustworthy faces
(0.79), t(49) = 0.49, p = .62, suggesting there is no differ-
ence between sets of faces in recognition sensitivity. How-
ever, it should be noted that the overall the recognition
sensitivity is quite low with a hit rate of 67%, a correct rec-
ognition rate of 60%, a false negative rate of 33% and a false
alarm rate of 39% across all faces.
In this study individuals played a standard, one-shot,
Trust Game with 79 hypothetical partners, each of whom
had been previously rated by an independent group on
their subjective trustworthiness. Results demonstrated
that implicit processing of social cues, i.e., trustworthiness,
had a reliable effect on decisions made in the Trust Game,
namely that participants chose to invest more money with
those partners who had higher trustworthiness ratings.
These results suggest that an initial impression of trust-
worthiness is a prevailing cue that is automatically, and
rapidly, processed (Willis & Todorov, 2006; Winston et
al., 2002), and can extend a strong influence on decision-
making. Interestingly, despite playing the game with each
partner and thereby obtaining direct evidence of the trust-
worthiness of the person, the post-game trustworthiness
ratings of the partners were not influenced by whether
the partners had in fact abused or repaid trust. Instead, it
appears that the rapid, intuitive judgment of trustworthi-
ness based on facial appearance exerts a rather dominant
influence. Moreover, it is important to realize that the
amount of money that was transferred to a partner corre-
lated highly with both ratings of trustworthiness given by
an independent sample of subjects who did not play the
game, as well as the subjects that did play the game.
Despite the trustworthiness ratings of partners, the
outcome on a previous round (whether trust was repaid
or abused) also had an effect of how much money as send
over. Thus after trust was abused people send over less
money in the next round. This is in line with our finding
that participants send over less money at the end of the
game compared to the beginning of the game. This thus
may have biased participants towards making less use of
the facial features of trustworthiness (and untrustworthi-
ness) on each trial. Yet, we observed a strong significant
influence in the amount of money transferred to partners
based on their facial features of trustworthiness. Addition-
ally, the effect of outcome on the amount of money send
over did not interact with trustworthiness ratings, which
suggests that although participants changed their strategy
throughout the game, the effect of perceived trustworthi-
ness was robust.
The robustness of this influence is further demonstrated
by the strong correlation between normed trustworthiness
ratings (from an independent sample of raters) and trust-
worthiness ratings obtained from our participants. The
strong relationship between subjective trustworthiness
and amount of money transferred to a partner is in accor-
dance with recent studies that show that the moral and so-
cial status of a hypothetical partner can influence behavior
in economic decision-making using similar tasks (Delgado
et al., 2005; Singer et al., 2004). However, our findings ex-
tend these studies in demonstrating that even non-explicit,
automatically processed, social information signaled by fa-
cial features appears to be a strong cue in influencing deci-
sion-making in an interactive social context. In addition,
our results are in line with, and indeed extend, previous re-
search showing that facial cues, such as smiling, influences
strategic decision-making (Eckel & Wilson, 1999; Scharle-
mann et al., 2001). These results could have real-life impli-
cations to what extent facial features guide us in real-life
decision-making, such as negotiation, and how much con-
trol we have over these influences. A previous study inves-
tigating the strength of character traits derived from facial
features showed that although impressions are particularly
strong other sources of information could change these
impressions (Hassin & Trope, 2000).
Since it has been demonstrated that strong and reliable
impressions, such as trustworthiness, are derived from the
face (Bond, Berry, & Omar, 1994; Hassin & Trope, 2000;
Willis & Todorov, 2006), it is interesting to know whether
people are more sensitive to positive (trustworthiness) or
negative (untrustworthiness) facial cues. Here, the influ-
ence of subjective trustworthiness on economic decision-
making appeared to be strongest in the most trustworthy
faces. This finding was not due to a restricted range in rat-
ings of untrustworthy faces. These results are in line with
the interpretation given by Singer et al. (2004) who found
that cooperators in particular elicited more activation in
social-emotional brain areas, such as the amygdala and in-
sula, suggesting that trustworthiness might be a more
important social cue than untrustworthiness. Neverthe-
less, the converse, i.e., a stronger correlation between
investment and untrustworthy faces, could also been
hypothesized. For instance, Winston et al. (2002) found
more activation of social brain areas when people judged
Fig. 3. Recognition ratings (SE) for faces when trust was abused or repaid
for trustworthy and untrustworthy faces.
M. van ’t Wout, A.G. Sanfey/Cognition 108 (2008) 796–803
subjectively untrustworthy faces as compared to trustwor-
thy faces. An interpretation of this finding is that the hu-
man brain reacts in particular to untrustworthiness, as it
is a more salient threat-related social cue that might there-
fore influence decision-making in a social interaction to a
stronger degree than trustworthiness. This leads to the
question of whether people are more sensitive to positive
or negative feedback, a question also addressed by both
Singer et al. (2004) and Rilling et al. (2002). Some proposed
models of facial processing in humans suggest that this
skill evolved because it allows individuals to keep track
of others who had proved to be cooperative in interactions
(Cosmides & Toobey, 2000). According to this account,
trustworthy partners in particular should be remembered
better and offered the most money.
With respect to memory, our data shows that although
recognition memory ratings for subjectively trustworthy
partners were higher than for subjectively untrustworthy
partners, the signal detection analysis shows that there
was no difference between participants’ ability in recog-
nizing trustworthy and untrustworthy faces. However, it
must be noted that the hit rate and correct rejection rate
of all faces was 67% and 60%, respectively, which shows
that participants were relatively poor at recognizing their
partners. In addition, we found no influence of experience
in the game on the ability to remember partners. This is
contrary to our hypothesis that participants would recog-
nize partners who repaid trust better than those who
abused trust (based on Singer et al., 2004) who showed
that memory for cooperative partners was better than for
defectors. The present data are also not in line with Mea-
ley, Daood, and Krage (1996) or Yamagishi, Tanida, Mashi-
ma, Shimoma, and Kanazawa (2003), who reported that
people who showed to be untrustworthy (by cheating or
not cooperating) are remembered better, even when they
do not know who are cheaters and cooperators. On the
other hand, in a series of well controlled experiments,
Mehl and Buchner, (2008) recently reported that there
was no difference in recognition memory between faces
associated with a history of cheating and faces associated
with a history of trustworthiness. Conversely, our results
suggest that the perception of subjective trustworthiness
may in fact be stronger than the actual game experience.
However, interpretation of these conclusions should be
made with caution, as the present study was not designed
to exclusively focus on memory. For example, the large
amount of faces the participants were exposed to (n = 79)
might have made it too difficult to keep track of the expe-
rience with each partner, and therefore may have diluted
the effect of experience on recognition memory. Indeed,
as mentioned earlier we found low correct recognition
rates across all faces. In addition, we did not ask
participants specifically whether they remembered if each
partner either abused or repaid their trust. Taken together,
we believe it is still unclear in what situation people are
more likely to attend to defectors or to cooperators. Future
research should investigate this interesting question in
more detail. For instance, these results most likely change
drastically when people play a repeated-round game with
partners as in such a situation memory for previous
outcomes, i.e., reputation is important, whereas in the
present single-shot version we highlight the fact that after
a decision is made by a partner the game is over. For these
reasons future research is needed to investigate the
interaction between first impressions based on rapid visual
presentation and the actual outcome experienced with
that partner on memory and behavior in (sequential)
strategic decision-making. Further, the issue of whether
cooperators or defectors are better encoded and remem-
bered is likely to be related to personality factors (some
people learn better from negative feedback and others
Zuckerman, Joireman, Kraft, & Kuhlman, 1999) and
perhaps even the specific task used.
It is worth noting that although we report a strong
relationship between subjective trustworthiness and trust
behavior in the Trust Game, we did not specifically inves-
tigate other social facial cues that could affect behavior
in the Trust Game. For instance, Willis and Todorov,
(2006) showed that besides trustworthiness, facial features
such as attractiveness, liking, competence, and aggressive-
ness are processed fast and automatically. Although the
present set of faces was controlled for emotional expres-
sion, the faces were initially not rated on these other facial
features. However, since previous studies showed that
more attractive people are more trusted (Mulford, Orbell,
Shatto, & Stockard, 1998; Solnick & Schweitzer, 1999;
Wilson & Eckel, 2006), we asked an independent group of
undergraduate students rate the faces on attractiveness.
Although attractiveness correlated with trustworthiness,
the attractiveness ratings did not correlate with the
amount of money send over to that partner. Though we
cannot rule out that other factors could have influenced
decisions in this game, it seems plausible that many of
these additional features could influence decision-making
via the perception of trustworthiness. Future studies could
investigate the influence of these automatically processed
social signals in decision-making. Additional questions
that need investigation refer to whether people are
actually accurate in their assessment of trustworthiness,
i.e., are partners that are perceived as more trustworthy
actually in fact more trustworthy, are there certain
markers for trustworthiness, such as the distance between
particular facial features (e.g., eyes, eyebrows, mouth), or
can trustworthiness be mimicked? Although it is not with-
in the scope of present study to tackle the question of
whether trustworthiness is a useful and accurate signal,
studies that address the question of whether facial signals
can be correctly mimicked showed that people are rather
accurate in distinguishing a fake smile from a genuine
one (Ekman & Friesen, 1982). Moreover, in a study on
authentic smilingand customer
showed both higher perceived friendliness and increased
customer satisfaction when servers’ smiles were genuine
(Grandey et al.,2005). Future studies should elucidate the
accuracy of one’s judgments as people might treat trust-
worthy-looking people consistently different, which can
elicit the expected behavior of being more trustworthy.
In summary, we observed a strong relationship between
the perceived trustworthiness of a partner and the amount
of money that was transferred to this partner by our
participants. These data suggest that even when people
M. van ’t Wout, A.G. Sanfey/Cognition 108 (2008) 796–803
are not explicitly asked about subjective facial trustworthi- Download full-text
ness of a partner they seem to evaluate this social cue
implicitly and act upon it. This underlines the fact that
automatically processed facial features of partners influ-
ence strategic decision-making, and that these processes
should be integrated into models of cooperation and reci-
procal behavior in interactive decision-making.
The authors thank Professor Ralph Adolphs for his
generous permission to use the set of faces used by his
research group and Luke Chang for his helpful comments.
This study was supported by funding from the Netherlands
Organisation for Scientific Research (NWO), Rubicon Grant
Adolphs, R., Tranel, D., & Damasio, A. R. (1998). The human amygdala in
social judgment. Nature, 393, 470–474.
Amaral, D. G. (2002). The primate amygdala and the neurobiology of
social behavior: Implications for understanding social anxiety.
Biological Psychiatry, 51, 11–17.
Bar, M., Neta, M., & Linz, H. (2006). Very first impressions. Emotion, 6,
Berg, J., Dickhaut, J., & McCabe, K. (1995). Trust, reciprocity, and social
history. Games and Economic Behaviour, 10, 122–142.
Berry, D. S., & Brownlow, S. (1989). Were the physiognomists right?
Personality correlates of facial babyishness. Personality and Social
Psychology Bulletin, 15, 266–279.
Berry, D. S., & McArthur, L. Z. (1986). Perceiving character in faces: The
impact of age-related craniofacial changes on social perception.
Psychological Bulletin, 100, 3–18.
Bond, C. F., Berry, D. S., & Omar, A. (1994). The kernel of truth in
judgments of deceptiveness. Basic and Applied Social Psychology, 15,
credibility, and attitude change. Journal of Nonverbal Behavior,
Burnham, T. C., McCabe, K., & Smith, V. L. (2000). Friend-or-foe
intentionality priming in an extensive form trust game. Journal of
Economic Behavior & Organization, 43, 57–73.
Camerer, C. F. (2003). Dictator, ultimatum, and trust games. In Behavioral
game theory (pp. 43–116). New York: Russell Saga Foundation.
Carver, C. S., & White, T. L. (1994). Behavioral inhibition, behavioral
activation, and affective responses to impending reward and
punishment: The BIS/BAS scales. Journal of Personality and Social
Psychology, 67, 319–333.
Cosmides, L., & Toobey, J. (2000). The cognitive neuroscience of social
reasoning. In M. S. Gazzaniga (Ed.), The new cognitive neuroscience
(pp. 1259–1276). Cambridge; Massachusetts: MIT Press.
Delgado, M. R., Frank, R. H., & Phelps, E. A. (2005). Perceptions of moral
character modulate the neural systems of reward during the trust
game. Nature Neuroscience, 8, 1611–1618.
believing: Facial appearance,
Eckel, C. C., & Wilson, R. K. (1999). Reciprocal fairness and social signaling:
Experiments with limited reputations. New York, NY: American
Ekman, P., & Friesen, W. V. (1982). Felt, false, and miserable smiles.
Journal of Nonverbal Behavior, 6, 238–252.
Frith, C. D., & Frith, U. (1999). Interacting minds – A biological basis.
Science, 286, 1692–1695.
Grandey, A. A., Fisk, G. M., Mattila, A. S., Jansen, K. J., & Sideman, L. A.
(2005). Is ‘‘service with a smile” enough? Authenticity of positive
displays during service encounters. Organizational Behavior and
Human Decision Processes, 96, 38–55.
Hassin, R., & Trope, Y. (2000). Facing faces: Studies on the cognitive
aspects of physiognomy. Journal of Personality and Social Psychology,
Kanwisher, N., McDermott, J., & Chun, M. M. (1997). The fusiform face
area: A module in human extrastriate cortex specialized for face
perception. The Journal of Neuroscience, 17, 4302–4311.
King-Casas, B., Tomlin, D., Anen, C., Camerer, C. F., Quartz, S. R., &
Montague, P. R. (2005). Getting to know you: Reputation and trust in
a two-person economic exchange. Science, 308, 78–83.
Mealey, L., Daood, C., & Krage, M. (1996). Enhanced memory for faces of
cheaters. Ethology and Sociobiology, 17, 119–128.
Mehl, B., & Buchner, B. (2008). No enhanced memory for faces of cheaters.
Evolution and Human Behavior, 29, 35–41.
Mulford, M., Orbell, J., Shatto, C., & Stockard, J. (1998). Physical
attractiveness, opportunity, and success in everyday exchange. The
American Journal of Sociology, 103, 1565–1592.
Phillips, M. L., Williams, L. M., Heining, M., Herba, C. M., Russell, T.,
Andrew, C., et al. (2004). Differential neural responses to overt and
covert presentations of facial expressions of fear and disgust.
NeuroImage, 21, 1484–1496.
Rilling, J. K., Gutman, D. A., Zeh, T. R., Pagnoni, G., Berns, G. S., & Kilts,
C. D. (2002). A neural basis for social cooperation. Neuron, 35,
Sanfey, A. G., Rilling, J. K., Aronson, J. A., Nystrom, L. E., & Cohen, J. D.
(2003). The neural basis of economic decision-making in the
ultimatum game. Science, 300, 1755–1758.
Scharlemann, J. P. W., Eckel, C. C., Kacelnik, A., & Wilson, R. K. (2001). The
value of a smile: Game theory with a human face. Journal of Economic
Psychology, 22, 617–640.
Singer, T., Kiebel, S. J., Winston, J. S., Dolan, R. J., & Frith, C. D. (2004).
Brain responses to the acquired moral status of faces. Neuron, 41,
Solnick, S. J., & Schweitzer, M. E. (1999). The influence of physical
Organizational Behavior and Human Decision Processes, 79, 199–215.
Willis, J., & Todorov, A. (2006). First impressions: Making up your mind
after a 100-ms exposure to a face. Psychological Science, 17, 592–598.
Wilson, R. K., & Eckel, C. C. (2006). Judging a book by its cover: Beauty and
expectations in the trust game. Political Research Quarterly, 59,
Winston, J. S., Strange, B. A., & O’Doherty Dolan, R. J. (2002). Automatic
and intentional brain responses during evaluation of trustworthiness
of faces. Nature Neuroscience, 5, 277–283.
Yamagishi, T., Tanida, S., Mashima, R., Shimoma, E., & Kanazawa, S. (2003).
You can judge a book by its cover: Evidence that cheaters may look
different from cooperators. Evolution and Human Behavior, 24,
Zuckerman, M., Joireman, J., Kraft, M., & Kuhlman, M. D. (1999). Where do
motivational and emotional traits fit within three factor models of
personality? Personality and Individual Differences, 26, 487–504.
on ultimatumgame decisions.
M. van ’t Wout, A.G. Sanfey/Cognition 108 (2008) 796–803