Psychology
2013. Vol.4, No.10, 753-759
Published Online October 2013 in SciRes (http://www.scirp.org/journal/psych) http://dx.doi.org/10.4236/psych.2013.410107
How Emotional Context Influences Facial Preferences and
Impressions
Satomi Hara1,2, Kentaro Katahira3,4, Kazuo Okanoya2,3,4,5*
1College of Arts and Sciences, The University of Tokyo, Tokyo, Japan
2ERATO, Okanoya Emotional Information Project, Japan Science Technology Agency, Saitama, Japan
3Center for Evolutionary Cognitive Sciences, The University of Tokyo, Tokyo, Japan
4Emotional Information Joint Research Laboratory, RIKEN Brain Science Institute, Saitama, Japan
5Graduate School of Arts and Sciences, The University of Tokyo, Tokyo, Japan
Received July 9th, 2013; revised August 12th, 2013; accepted September 16th, 2013
Copyright © 2013 Satomi Hara et al. This is an open access article distributed under the Creative Commons At-
tribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the
original work is properly cited.
Individuals change their evaluations of human faces depending on the valence of the stimuli presented
with the faces. The present study investigated whether repeatedly presenting picture stimuli in pairs would
evoke various emotions that would influence the subjects’ preferences for and impressions of facial stim-
uli. The subjects’ preferences for the facial stimuli did not differ consistently between before and after the
facial stimuli were presented in tandem with emotional pictures. The direction of the change differed de-
pending on the sex of the participant and the sex of face stimulus, even when the face was paired with a
picture of the same valence. The emotional pictures appeared to have an effect on the emotions experi-
enced in response to the face stimuli: the male participants, who were likely to feel pleasant emotions to-
ward the pleasant pictures, were also likely to feel positive emotions toward the face stimuli paired with
those pictures. Moreover, the female participants, who were likely to feel unpleasant emotions toward the
unpleasant pictures, were also likely to feel afraid of the male faces paired with those pictures. These re-
sults suggest that the ability of an emotional stimulus to affect our preferences for and impressions of a
face stimulus, as well as the degree of this effect, are highly sensitive to factors such as the sex of the par-
ticipant and the sex of facial stimulus.
Keywords: Facial Preference; Facial Impression; Emotional Pictures; Evaluative Conditioning
Introduction
Preference is a critical factor that determines human behavior
in various daily situations. Preference has a great influence on
many aspects of our everyday decisions, such as what to eat,
with whom to spend time, and what to buy. Thus, knowing how
preferences are formed and manipulated is closely related to
understanding human behavior and decision-making. To ad-
dress this, many researchers have been studying a phenomenon
called evaluative conditioning (EC). EC is a specific form of
classical conditioning that refers to a change (either positive or
negative) in one’s evaluation of an initially neutral stimulus (a
conditioned stimulus: CS) due to the valence of a stimulus pre-
sented in tandem with the CS (an unconditioned stimulus: US).
Since EC was first observed by researchers such as Razran
(1954) and Staats & Staats (1957), a wide range of areas of
psychology, including social psychology (e.g., Olson & Fazio,
2001), learning psychology (e.g., Baeyens, Eelen, Crombez, &
Van den Bergh, 1992), and consumer psychology (e.g., Bierley,
McSweeney, & Vannieuwkerk, 1985), have been focused on
this phenomenon.
Researchers have used stimuli of several domains, such as
visual, auditory, and gustatory, to show that diverse types of
stimuli can function as CS or US (see De Houwer, Thomas, &
Baeyens, 2001 for a review). Many of these studies have
adopted pictures of human faces as CS. Baeyens et al. (1992)
showed that faces initially rated as neutral became preferred
after they were presented in pairs with highly preferred faces,
and less preferred when presented with faces with low prefer-
ence scores. Additionally, Hermans, Vansteenwegen, Crombez,
Baeyens, & Eelen (2002) used electric shocks and adjectives
that described either positive or negative personalities as US
and found that the preference scores for the faces paired with
the negative adjectives or with the electric shocks decreased,
whereas the preference scores for the faces paired with positive
adjectives increased.
However, the emotions and impressions that people experi-
enced in their everyday lives are much more complicated than
simple terms such as “like”, “dislike”, “good” or “bad” or the
disgust and avoidance produced by temporary punishments. In
various situations, we become happy, angry, and sad and feel
much more diverse emotions. Therefore, to profoundly under-
stand how preferences are formed and manipulated in daily and
social contexts, one must consider the influence of these types
of complex and various emotions on the formation of prefer-
ence. Nevertheless, no research has yet shown that such emo-
tional stimuli can change individual preferences for CS.
*Corresponding author.

S. HARA ET AL.
Therefore, in this study, we investigated whether US that in-
duce complex and various types of emotions could change sub-
jects’ preferences for paired CS. We used the International
Affective Picture System (IAPS) developed by Lang, Bradley
& Cuthbert (2008) as US. The IAPS is a database of pictures
depicting a wide range of human experiences, which are scored
using three criteria: valence, arousal, and dominance. Although
this database was originally standardized using Western sam-
ples, many studies of Japanese samples (e.g., Katahira, Fuji-
mura, Okanoya, & Okada, 2011; Fujimura, Katahira, & Oka-
noya, 2013) have confirmed that the images evoke similar emo-
tions in individuals from both cultures. For the CS, we used
pictures of human faces, which are socially important. We in-
vestigated whether the subjects’ preferences for these faces
would change according to the valence of IAPS pictures that
were repeatedly paired with the faces. Although it has been
confirmed that pleasant IAPS pictures can serve as appetitive
stimuli and unpleasant IAPS pictures can be aversive stimuli in
an instrumental conditioning paradigm (Katahira, Fujimura,
Okanoya, & Okada, 2011), it is not yet known whether these
pictures can function as US in the EC paradigm.
There were four between-subject conditions that involved
pairing the emotional pictures and the face stimuli: three ex-
perimental conditions and one control condition. The experi-
mental conditions involved pairing a face stimulus with either a
pleasant emotional picture (pleasant condition), an unpleasant
emotional picture (unpleasant condition), or with a pleasant
picture in half of the repetitions and with an unpleasant picture
in the other half (pleasant-unpleasant condition). The control
condition involved pairing a face stimulus with a picture with
no particular emotional valence (neutral condition). Thus, we
investigated how the subjects’ preferences for the faces in each
condition differed before and after the face stimulus and the
emotional picture were presented in tandem. We hypothesized
that the emotional pictures would have the same effect as the
other stimuli in previous studies: the faces in the pleasant con-
dition were expected to be preferred, those in the unpleasant
condition to be less preferred, and those in the neutral condition
to remain neutral. The pleasant-unpleasant condition is a novel
condition created for the current study, and we expected it to
affect preferences in three possible ways. First, faces in this
condition might be greatly preferred due to the enhancement of
the effect of the pleasant pictures by the unpleasant pictures.
Second, the opposite pattern might occur: faces in this condi-
tion might be much less preferred due to the enhancement of
the effect of the unpleasant pictures by the pleasant pictures.
Third, the preference might not change due to a neutralizing
effect between the unpleasant and pleasant pictures. We asked
the subjects to what degree they felt a particular type of emo-
tion, such as sadness or happiness, in response to the face stim-
uli after the conditioning. The responses to this question were
used to assess the relationship among the emotions induced by
the pictures, the changes in the preference rating, and the emo-
tions evoked by the face stimuli.
Methods
Participants
A total of 53 Japanese university students (26 male and 27
female) participated in the study. The average age was 19.2
years (SD = .76, range = 18 - 20). All the experimental proce-
dures were approved by the Ethics Committee of the University
of Tokyo.
Stimuli
The face stimuli (8 males and 8 females) were selected from
a database that consisted of 49 male and 51 female faces, ac-
cording to the data generated by a pilot study. The database was
developed by the JST ERATO Okanoya Emotional Information
Project (Fujimura and Okanoya, unpublished). In the pilot
study, 11 participants aged 16 - 25 years (6 male and 5 female,
average age 21.4 years) were shown each of the 100 face pic-
tures in the database and rated how much they liked each face
one a score of 1 to 9 (1: do not like it at all; 5: neutral; 9: like it
very much). Among the pictures of comparatively young faces,
we chose those that constantly received ratings close to 5. The
emotional pictures (30 pleasant, 30 unpleasant, and 20 neutral)
were selected from the IAPS database using the valence ratings
and arousal data in Lang et al. (2008).
Procedures
The experimental program was created with Inquisit ver.
3.0.6.0 by Millisecond Software and was presented on an LCD
monitor. The participants were required to perform the neces-
sary tasks by pressing a keyboard.
The experiment consisted of 4 parts: a pre-conditioning pref-
erence rating phase, a conditioning phase, a post-conditioning
preference rating phase, and an emotion rating phase.
In the pre-conditioning preference rating phase, we measured
the baseline preference of each participant for each face stimu-
lus using a method similar to those of Field (2006) and Johns-
rude, Owen, Zhao & White (1999). We presented 2 of the 8
male or female face stimuli in turn, and we asked the partici-
pants to state which stimulus was preferred. Each of the possi-
ble pairs was presented twice: one on the left and the other on
the right the first time, and the former on the right and the latter
on the left the second time. Thus, each participant performed
112 trials in total. The preferred stimulus in each trial received
1 point, and the other received 0; thus, each face stimulus re-
ceived a total of 0 to 14 points. The order between and within
the male and female stimuli was randomized for each partici-
pant. This binary rating method was considered superior be-
cause when the participants were asked to rate the same stimuli
again in the post-conditioning phase, they would be less likely
to be influenced by memories of the ratings they had given to
the stimuli in the pre-conditioning phase.
During the conditioning phase, we repeatedly presented each
face stimulus paired with an emotional picture. The combina-
tions used to create pairs of these two stimuli consisted of 3
experimental conditions and 1 control condition. The experi-
mental conditions involved pairing a face stimulus with either a
pleasant emotional picture (pleasant condition), an unpleasant
emotional picture (unpleasant condition), or a pleasant picture
in half of the repetitions and an unpleasant picture in the other
half (pleasant-unpleasant condition). The control involved
pairing a face stimulus with a picture with no particular emo-
tional valence (neutral condition). Eight male and female face
stimuli were divided into the 4 conditions so that each condi-
tion contains one male face and one female face. The assign-
ment of face stimuli was counterbalanced between the partici-
pants.
The flow of one trial is shown in Figure 1. First, a face
stimulus appeared on either the right or left side of the screen.
754

S. HARA ET AL.
1 s 2.5 s3 s
Figure 1.
The flow of the screen in the conditioning phase. The sample face stimulus and emotional picture were
taken from other source materials to avoid showing materials of the databases used in the present study.
Results
One second later, an emotional picture appeared on the opposite
side of the screen. After another 2.5 seconds, these two stimuli
disappeared simultaneously, and the screen remained empty for
3 seconds. During this interval, the participant rated the subjec-
tive level of pleasantness or unpleasantness of the emotional
picture that had been presented on a scale of 1 to 9 (1: very
unpleasant; 5: neutral; 9: very pleasant). After the interval, the
next trial began.
Ratings for the Emotional Pictures
A 2 (sex of participant: male vs. female) × 3 (type of emo-
tional picture: pleasant, unpleasant, or neutral) ANOVA was
used to analyze the ratings for the emotional pictures in the
conditioning phase. The results revealed a significant main
effect of the type of emotional picture (F(2,102) = 733.59, p
< .001) and a significant sex × type interaction (F(2,102) =
15.49, p < .001). The post-hoc comparison of the sex × type
interaction (Figure 2) revealed that both the male and female
participants gave significantly higher ratings (more pleasant) to
the neutral pictures than to the unpleasant ones (male: F(1,51 =
16.55, p < .001; female: F(1,51) = 18.29, p < .001). Similarly,
both sexes gave significantly higher ratings to the pleasant pic-
tures than to the neutral ones (male: F(1,51) = 14.09, p < .001,
female: F(1,51) = 21.82, p < .001) or to the unpleasant ones
(male: F(1,51) = 17.43, p < .001, female: F(1,51) = 22.93, p
< .001). The female participants gave significantly higher rat-
ings to the pleasant pictures and significantly lower ratings to
the unpleasant ones compared with the male participants
(pleasant: F(1,51) = 17.77, p < .001; unpleasant: F(1,51) =
10.11, p < .01). In addition, a one-sample t-test performed to
compare the ratings given to the neutral pictures by the male
and female participants revealed no significant difference from
the most neutral rating, 5 (male: t(25) = −.87, p = .39; female:
t(26) = 1.55, p = .13). These results suggest that all of the pleas-
ant, unpleasant and neutral pictures had the expected valence
for both the male and female participants. A comparison of the
male and female participants, however, revealed that the fe-
males rated the pleasant pictures more positively and the un-
pleasant pictures more negatively than the males, suggesting
that the emotional pictures may have produced a somewhat
greater effect on the females than on the males.
Each face stimulus was presented in tandem with an emo-
tional picture 10 times, the number that was expected to pro-
duce the most effective conditioning according to Baeyens et al.
(1992). The face stimuli in the pleasant, unpleasant, and neutral
conditions were paired with 10 different pleasant, unpleasant,
and neutral pictures, respectively. The face stimuli in the
pleasant-unpleasant condition were paired with 5 pleasant pic-
tures and 5 unpleasant pictures. Each of the 30 pleasant, 30
unpleasant, and 20 neutral pictures was presented twice, once
with a male stimulus and once with a female stimulus. Thus,
each participant performed 160 trials in total. All the trials were
divided into 5 sessions, for a total of 32 trials, and between the
sessions, the participants could take a break of the desired
length. Because the order in which the stimuli were presented
and the timing of the rating were somewhat complicated, there
was a practice session of 13 trials before the 5 experimental
sessions. The practice session employed different face stimuli
and emotional pictures from those used in the 5 experimental
sessions.
In the post-conditioning preference rating phase, we meas-
ured the preference (after conditioning) for each face stimulus
following the same method as in the pre-conditioning phase.
In the emotion rating phase, we investigated whether and to
what degree the participants felt emotions while observing the
face stimuli. The face stimuli were presented singly beneath
one of 4 words: “sad”, “scared”, “happy”, or “delighted”, and
the participants were asked to what degree they experienced the
feeling of the depicted word while observing the face (1: not at
all; 5: very much). Each of the 16 stimuli (8 male and 8 female)
was presented with each of the 4 emotional words, for a total of
64 trials. The order in which the face stimuli and the words
within each face were presented was randomized for each par-
ticipant.
Preference Ratings for the Face Stimuli
A 2 (sex of participant: male vs. female) × 2 (sex of stimulus:
male vs. female) × 4 (condition: pleasant, unpleasant, pleasant-
unpleasant, or neutral) × 2 (time: pre-conditioning vs. post-
conditioning) ANOVA was performed to analyze the prefer-
ence ratings for the face stimuli. The results showed a signifi-
cant sex of stimulus × condition × time interaction (F(3, 312) =
12.77, p < .05), although none of the 8 conditions (sex of
stimulus [2] × condition [4]) differed significantly between the
pre-conditioning and the post-conditioning ratings, according to
the post-hoc comparison. Figure 3 shows the average prefer-
ence scores calculated for the male and female participants,
Statistical Analyses
R and anovakun ver. 4.3.1, an ANOVA function that runs on
R software, was used for the statistical analyses. The signifi-
cance level was set at α = .05 for the ANOVA, and post-hoc
comparisons were conducted using a modified Bonferroni
method.

S. HARA ET AL.
Figure 2.
The average valence rating of the emotional pictures. The error bars
represent the s.e.m. **p < .01, ***p < .001.
Figure 3.
The difference between the pre- and post-conditioning preference
scores for the face stimuli. The error bars represent the s.e.m.
the male and female face stimuli, the conditions of pairing, and
the pre-conditioning and post-conditioning ratings. These re-
sults indicate that the preference scores did not change in the
same direction within each condition. Especially in the pleas-
ant-unpleasant condition, the female faces received higher
scores and the male face received lower scores in the post-
conditioning rating compared with the pre-conditioning rating
(although these differences did not reach statistical signifi-
cance).
In addition, to examine the possibility that the pre-condi-
tioning score (minimum: 0, maximum: 14) affected the change
in the preference score (post-conditioning score minus pre-
conditioning score), we calculated the Pearson product-moment
correlation coefficient between the pre-conditioning scores and
the averages of the change in the preference scores; the result
showed a strongly negative correlation (n = 15, r = −.92, p
< .001; Figure 4). This result indicates that the face stimuli that
received lower scores in the pre-conditioning rating tended to
be rated more highly in the post-conditioning phase. The oppo-
site was also true: the face stimuli that received higher scores in
the pre-conditioning rating tended to receive lower scores in the
post-conditioning rating.
Emotional Ratings of the Face Stimuli
A 2 (sex of participant: male vs. female) × 2 (sex of stimulus:
male vs. female) × 4 (condition: pleasant, unpleasant, pleas-
ant-unpleasant, or neutral) × 4 (rated emotion: sad, scared,
happy, or delighted) ANOVA was used to analyze the emo-
tional ratings for the face stimuli. The results showed a signifi-
cant main effect of the rated emotion (F(3,312) = 4.14, p < .01)
and significant interactions between the sex of the participant ×
emotion (F(3,312) = 2.65, p < .05), the sex of the stimulus ×
emotion (F(3,312) = 5.08, p < .01) and the sex of the partici-
pant × the sex of the stimulus × emotion (F(3,312) = 7.47, p
< .001). The post-hoc comparisons of the cubic interaction
revealed different patterns between the male and female par-
ticipants (Figure 5). The female participants evaluated the male
and female stimuli differently (Figure 5(a)): for “sad” (F(1,53)
= 6.44, p < .05) and “scared” (F(1,53) = 11.76, p < .01), they
gave higher ratings to the male stimuli than to the female stim-
uli; oppositely, for “happy” (F(1,53) = 15.02, p < .001) and
“delighted” (F(1,53) = 10.21, p < .01), the female participants
gave higher ratings to the females than to the males. The ratings
of the male participants, in contrast, did not differ between the
male and female stimuli (Figure 5(b)). This finding indicates
that the female participants were more likely to experience
positive emotions toward the female stimuli and negative emo-
tions toward the male stimuli, whereas the male participants felt
the same way toward both the male and female stimuli.
The lack of any significant effects of the paired picture con-
dition (pleasant, unpleasant, pleasant-unpleasant, or neutral)
indicates that there was no significant difference in the emo-
tional ratings among the conditions, although the emotional
rating was conducted after the conditioning process. However,
Figure 4.
The correlation between the pre-conditioning scores and the
average change in the preference scores (from post-condi-
tioning score to pre-conditioning score).
Figure 5.
The difference between males and females in the emotions experienced
toward the face stimuli. The female participants felt strongly positive
emotions in response to female stimuli and negative emotions in re-
sponse to male stimuli. In contrast, the male participants felt the same
way toward both the male and female stimuli. The error bars represent
the s.e.m. *p < .05, **p < .01, ***p < .001.
756

S. HARA ET AL.
the degree to which a participant experiences an emotional
picture as pleasant or unpleasant varies between individuals.
There was a noteworthy correlation between the pleasant-un-
pleasant ratings for the emotional pictures and the emotional
ratings for the face stimuli. The correlation was calculated
separately for the sex of the participant and the stimulus and
between the personal averages of the emotional ratings for the 4
kinds of emotion and those of the ratings for the emotional
pictures that were paired with each face stimulus in the condi-
tioning phase. As a result, the male participants produced par-
tially consistent correlations irrespective of the sex of the face
stimulus (Figures 6(a) and (b)): with regard to the pleasant
condition face stimuli, the emotional ratings for “happy” and
“delighted” showed positive correlations with the ratings for the
pleasant pictures paired with those face stimuli (male stimulus,
“happy”; n = 26, r = .40, p < .05: male stimulus, “delighted”; n
= 26, r = .26, p = .20: female stimulus, “happy”; n = 26, r = .42,
p < .05: female stimulus, “delighted”; n = 26, r = .43, p < .05:
Figure 6(a)). With regard to the pleasant-unpleasant condition
face stimuli, the emotional ratings for “happy” and “delighted”
were also positively correlated with the ratings for the pleasant
pictures paired with those face stimuli (male stimulus, “happy”;
n = 26, r = .31, p = .13: male stimulus, “delighted”; n = 26, r
= .52, p < .01: female stimulus, “happy”; n = 26, r = .52, p
< .01: female stimulus, “delighted”; n = 26, r = .48, p < .05:
Figure 6(b)). That is, in both conditions in which the face
stimuli were paired with pleasant pictures at least 5 times, the
more that participants experienced pleasant emotions in re-
sponse to the pleasant pictures, the more they experienced posi-
tive emotions in response to the face stimuli paired with those
pictures. In contrast, the female participants’ responses to the
male stimuli of the unpleasant and pleasant-unpleasant condi-
tions exhibited negative correlations between the emotional
ratings for “scared” and the ratings for the unpleasant pictures
(a smaller value indicates a more unpleasant response) that
were paired with them (unpleasant condition; n = 27, r = −.38,
p < .05: pleasant-unpleasant condition; n = 27, r = −.38, p < .05:
Figure 6(c)). In both conditions in which the male stimuli were
paired with unpleasant pictures at least 5 times, the more that
the female participants experienced unpleasant emotions in
response to the unpleasant pictures, the more they felt “scared”
of the male stimuli that were paired with those pictures.
Discussion
The purpose of the current study was to consider the influ-
ence of emotional pictures on the preference for face stimuli
presented in tandem with the emotional pictures and the emo-
tions experienced toward such stimuli. Based on the hypothesis
that the emotional pictures would produce the same effects as
those of stimuli used in previous studies, the face stimuli that
were paired with pleasant pictures (pleasant condition) were
predicted to be preferred, and those paired with unpleasant
pictures (unpleasant condition) were predicted to be less pre-
ferred. One of the purposes of this study was to investigate the
effects of face stimuli paired with both pleasant and unpleasant
pictures.
However, the preference scores for the face stimuli in each
condition did not differ significantly before and after the condi-
tioning process. Therefore, we would primarily like to discuss
Figure 6.
The correlation between subjective emotional reactions to emotional pictures and to face
stimuli after the conditioning process. Panels a and b show the responses of the male par-
ticipants to both the male and female face stimuli; panel c shows the response of the fe-
male participants to the male stimuli alone. This figure indicates that the more that the
male participants experienced pleasant emotions in response to the pleasant pictures, the
more they experienced positive emotions in response to the face stimuli paired with those
pictures. The more that the female participants experienced unpleasant emotions in re-
sponse to the unpleasant pictures, the more they felt “scared” of the male stimuli paired
with those pictures.

S. HARA ET AL.
the reasons for the results differing from our expectations. One
possible reason is the considerable variance in the pre-condi-
tioning scores. In previous studies that adopted human face
stimuli as CS (Baeyens et al., 1992; Hermans et al., 2002), the
participants were asked to evaluate many more face stimuli
before the conditioning process than in this study; the purpose
was to enable the researchers to choose as the CS the face with
the most neutral score for each participant. This method was
likely chosen because preferences for human faces are more
susceptible to individual differences than are other stimuli.
Previous studies thus employed a truly neutral face stimulus for
each participant; therefore, there was little difference between
the preference scores before conditioning. In contrast, we se-
lected in advance face stimuli with stably neutral ratings by
performing a pilot study and used them for all of the partici-
pants in the main experiment. Therefore, the preference scores
for the face stimuli prior to conditioning varied considerably,
depending on the participant and the stimulus. In addition, even
the average scores within each of the 4 conditions were not
equal: the average score in the pleasant condition was slightly
higher than 7 (the middle score), and the average score in the
unpleasant condition was slightly lower than 7.
The strongly negative correlation between the pre-condi-
tioning scores and the variation in the preference scores provide
evidence that the above factor contributed to the difference
between the results and the expected patterns. There are several
possible causes of the correlation. A bias may have been intro-
duced by the phenomenon that when the minimum score is
limited, those who receive a low score in the first measurement
will be more likely to receive a higher score in the second
measurement, and vice versa. It is also possible that initially
high or low evaluations of faces are difficult to manipulate or
are likely to change in a definite direction (first high, then
negative and first low, then positive), irrespective of the va-
lence of the paired US (pleasant or unpleasant). Regardless of
the cause, it is likely that the variance in the pre-conditioning
ratings and the differences among the averages in each condi-
tion influenced the variations, as this correlation indicates, and
thus made the direction of the preference change in each condi-
tion unclear.
Nonetheless, it is quite suspicious that we observed such a
simple effect as expected although the influence stated above
was excluded, considering that there were no constant tenden-
cies at all within the conditions such that the preference scores
of the pleasant condition face stimuli increased as those of the
unpleasant condition face stimuli decreased. This finding sug-
gests that the influence of emotional stimuli on preferences for
faces is not necessarily identical to the influence of simple
positive or negative stimuli, which were used in the previous
studies. Emotional pictures may produce different effects, de-
pending on the sex of the participant and the stimulus.
The results of the emotional ratings, on the other hand, sug-
gest that the emotional pictures influenced the emotions ex-
perienced in response to the face stimuli. The male participants
who were likely to feel pleasant emotions toward the pleasant
pictures were also likely to feel positive emotions, such as
“happy” and “delighted”, toward the face stimuli in the pleasant
and pleasant-unpleasant conditions, in which the face stimuli
were paired with pleasant pictures at least 5 times. The female
participants who were likely to experience unpleasant emotions
toward the unpleasant pictures were also likely to feel “scared”
of the male face stimuli in the unpleasant and pleasant-un-
pleasant conditions, in which the face stimuli were paired with
unpleasant pictures at least 5 times. This difference between the
sexes suggests that the emotions that are readily attached to
faces by conditioning may vary between males and females:
males may be susceptible to conditioning between pleasant
emotions and face stimuli, whereas females may be susceptible
to conditioning between unpleasant emotions and male faces in
particular, which must be avoided for safety under certain cir-
cumstances.
Moreover, the female participants experienced positive emo-
tions, such as “happy” and “delighted”, more strongly in re-
sponse to the female face stimuli than to the male face stimuli.
The females also experienced negative emotions, such as “sad”
and “scared”, more strongly in response to the male stimuli
than to the female stimuli. The male participants, however,
exhibited almost identical responses to the male and female
stimuli. This result can be interpreted in several possible ways:
it may reflect how males and females initially feel in response
to unknown face stimuli or how they feel after unknown face
stimuli are presented in tandem with emotional pictures, re-
gardless of their valence. However, both interpretations indicate
that this difference between the sexes reflects differences be-
tween males and females in the conditioning effect of emo-
tional pictures on faces.
In this study, the emotional pictures produced no significant
changes in the preference for the face stimuli presented in pairs
with themselves. The influence of the emotional pictures on this
preference may not have been as powerful and simple as the
effects of the stimuli used in the previous studies, such as the
liked and the disliked stimuli and the electric shocks, which
changed the preference for CS in a constant direction irrespec-
tive of the sex of the participant or the stimulus. Concerning the
emotions felt in response to the face stimuli, the emotional
pictures did have an effect, the strength of which may have
depended on the sex of the participant and the stimulus.
Therefore, to investigate how emotion-evoking stimuli affect
our preferences and impressions of other stimuli, we must per-
form detailed analyses that consider such conditions. Addition-
ally, larger samples than that of the present study are necessary
to examine the difference between the sexes. In future research,
more sophisticated experimental procedures and detailed analy-
ses will be needed to clarify the relationship between stimuli
that evoke complicated emotions and our preferences and emo-
tions in response to other people and things.
Acknowledgements
We thank Maiko Fujimori for providing us with the database
of facial stimuli. This work was supported in part by funding
from the Japan Science and Technology Agency, Exploratory
Research for Advanced Technology, Okanoya Emotional In-
formation Project, and Grants-in-Aid for Scientific Research,
no. 24700238.
REFERENCES
Baeyens, F., Eelen, P., Crombez, G., & Van den Bergh, O. (1992).
Human evaluative conditioning: Acquisition trials, presentation sche-
dule, evaluative style and contingency awareness. Behaviour Re-
search and Therapy, 30, 133-142.
http://dx.doi.org/10.1016/0005-7967(92)90136-5
Bierley, C., McSweeney, F. K., & Vannieuwkerk, R. (1985). Classical
conditioning of preferences for stimuli. Journal of Consumer Re-
758

S. HARA ET AL.
search, 12, 316-323. http://dx.doi.org/10.1086/208518
De Houwer, J., Thomas, S., & Baeyens, F. (2001). Associative learning
of likes and dislikes: A review of 25 years of research on human
evaluative conditioning. Psychological Bulletin, 127, 853-869.
http://dx.doi.org/10.1037/0033-2909.127.6.853
Field, A. P., (2006). I don’t like it because it eats sprouts: Conditioning
preferences in children. Behaviour Research and Therapy, 44, 439-
455. http://dx.doi.org/10.1016/j.brat.2005.03.006
Fujimura, T., Katahira, K., & Okanoya, K. (2013). Contextual modula-
tion of physiological and psychological responses triggered by emo-
tional stimuli. Frontiers in Psychology, 4, 212.
http://dx.doi.org/10.3389/fpsyg.2013.00212
Hermans, D., Vansteenwegen, D., Crombez, G., Baeyens, F., & Eelen,
P. (2002). Expectancy-learning and evaluative learning in human
classical conditioning: Affective priming as an indirect and unobtru-
sive measure of conditioned stimulus valence. Behaviour Research
and Therapy, 40, 217-234.
http://dx.doi.org/10.1016/S0005-7967(01)00006-7
Johnsrude, I. S., Owen, A. M., Zhao W. V., & White, N. M. (1999).
Conditioned preference in humans: A novel experimental approach.
Learning and Motivation, 30, 250-264.
http://dx.doi.org/10.1006/lmot.1999.1031
Katahira, K., Fujimura, T., Okanoya, K., & Okada, M. (2011). Deci-
sion-making based on emotional images. Frontiers in Psychology, 2,
311. http://dx.doi.org/10.3389/fpsyg.2011.00311
Lang, P. J., Bradley, M. M., & Cuthbert, B. N. (2008). International
affective picture system (IAPS): Affective ratings of pictures and in-
struction manual. Technical Report A-8, Gainesville, FL: University
of Florida.
Olson, M. A., & Fazio, R. H. (2001). Implicit attitude formation
through classical conditioning. Psychological S c i e n c e, 12, 413-417.
http://dx.doi.org/10.1111/1467-9280.00376
Razran, G. (1954). The conditioned evocation of attitudes (cognitive
conditioning?). Jou rn a l o f Experimental Psychology, 48 , 278-282.
http://dx.doi.org/10.1037/h0058778
Staats, A. W., & Staats, C. K. (1957). Meaning established by classical
conditioning. Journal of Experimental Psych olo gy, 54, 74-80.
http://dx.doi.org/10.1037/h0047716