A Neuromarketing Study of Consumer Satisfaction

Abstract

The interest of marketing science in using neuroscience techniques to understand the consumer’s thought processes, dates back to the 1970s, when EEG data were recorded while subjects were watching TV commercials. Recently, fMRI was used to study the neural correlates of culturally based brands and neural predictors of purchases. These studies have discovered important properties of the neural circuits that are associated with consumer decision-making process and satisfaction. Here, EEG brain mapping was used to study the dynamics of the brain activity associated with these processes. The present study validated the EEG technology as an adequate neuromarketing tool and shows that consumer’s satisfaction evaluation with the aesthetical dermatological treatment involved the activation of neural circuits involved with facial beauty evaluation.

Full text

A Neuromarketing Study of Consumer Satisfaction
Armando F. Rocha
1
, Fábio Theoto Rocha
1
and Lucia Helena Fávaro Arruda
2
1
RANI – Research on Natural an Artificial Intelligence
2
Department of Dermatology, Catholic University of Campinas
http://papers.ssrn.com/abstract=2321787

Summary: The interest of marketing science in using neuroscience techniques to
understand the consumer’s thought processes, dates back to the 1970s, when EEG data
were recorded while subjects were watching TV commercials. Recently, fMRI was used to
study the neural correlates of culturally based brands and neural predictors of purchases.
These studies have discovered important properties of the neural circuits that are associated
with consumer decision-making process and satisfaction. Here, EEG brain mapping was
used to study the dynamics of the brain activity associated with these processes. The
present study validated the EEG technology as an adequate neuromarketing tool and shows
that consumer’s satisfaction evaluation with the aesthetical dermatological treatment
involved the activation of neural circuits involved with facial beauty evaluation.
Keywords: neuromarketing, consumer satisfaction, EEG mapping, decision-making
modeling, aesthetic treatment

1 Introduction
The interest of marketing science in using neuroscience techniques to understand
the consumer’s thought processes, dates back to the 1970s, when EEG data were recorded
while subjects were watching TV commercials. (Young, 2002). Recently, fMRI was used to
study the neural correlates of culturally based brands (Ambler et al., 2000; McClure et al.,
2004; Schaefer et al., 2006 and Yoon et al., 2006) and neural predictors of purchases
(Knutson et al., 2007).
The microeconomic theory states that purchases are driven by a combination of the
consumer’s preference and price. Using event-related fMRI, Knutson et al. (2007) showed
that activation of the nucleus accumbens correlated with the consumer’s preference,
whereas excessive prices activated the insula and deactivated the medial prefrontal cortex
prior to the purchase decision. Coke® and Pepsi® are nearly identical in chemical
composition; however, humans routinely display strong subjective preferences for one or
the other. McClure et al. (2004) showed that the anonymous delivery of Coke® or Pepsi®
activated the ventromedial cortex, but, when knowledge about the brand was available,
only Coke® and not Pepsi® activated the hippocampus, dorsolateral prefrontal cortex and
the midbrain. They concluded that the consumer’s preference was a complex construct that
involved judgments based on sensory information and the history of the relationship
between the individual and the brand. Consumer preferences were, therefore, based on a
history of customer satisfaction.

Although neuroscience has extensively examined the reward systems (Aktsuki et al.,
2003; Bretier et al., 2001; Plat and Padoa-Schioppa, 2009; Polezzi et al., 2010; Roger et al.,
2004; Tobler, Fiorillo and Schultz, 2005) involved in satisfaction assessment and
preference encoding, the customer’s brain activity associated with the evaluation of service
satisfaction has not been elucidated. Therefore, we examined the brain activity of women
by evaluating their satisfaction in using an aesthetic dermatological filler treatment (Arruda,
Rocha and Rocha, 2008).
1.2 A Satisfaction study
Competition makes people worry about physical appearance, which occurs
primarily with respect to aging of the face and skin. Studies analyzing attitudes toward
aging and the elderly have found that older women are judged more negatively than older
men because modern urbanized societies allow two standards of male beauty (Berman,
O’Nan and Floyd, 1981; Deutsch, Zalenski and Clark, 1986; Sontag, 1972
)
, those of the
boy and of the man, but only one standard of female beauty, that of the girl. Thus, women
are more prone to enroll in cosmetic dermatology procedures.
1.3 Facial recognition
Facial attractiveness is an important Darwinian factor in human reproduction and is
an important social factor of motivated behavior (Aharon et al., 2001). In addition, facial
recognition is important in human evolution because facial expressions are external signals
of the internal experienced emotions (Britton et al., 2006b), and the emotional information

exchange is fundamental to social relations. Because of its importance for human behavior,
facial recognition is supported by a specific and diverse neural circuit involving a) regions
of the extrastriate cortex that process the visual identification of individuals; b) the superior
temporal sulcus, where gaze directions and speech related to movements are processed; c)
the amygdala and insula, where facial emotional expression is processed; d) the fusiform
face areas and superior temporal sulcus, where attractiveness, gender and age are identified;
and e) regions in the prefrontal cortex and in the reward circuitry, such as the nucleus
accumbens and orbitofrontal cortex, where the assessment of beauty is processed (Aharon
et al., 2001; Britton et al., 2006a; Brady, Campbell and Flaherty, 2004; Ishai, 2007; Ishai,
Schmidt and Boesiger, 2005; Kircher et al., 2001; Singer et al., 2004).
Figure 1 – Magnetic resonance images showing several neural structures involved in face
processing.
Quiroga et al. (2005) have shown that specific neurons specialize in recognizing
specific faces, which demonstrated the specificity of the facial recognition neural circuits

(FRNCs). All of these studies highlight the engagement of the FRNCs in deliberative and
implicit social judgments. Figure 1 shows the location of some of the structures involved in
these judgments.
Neuropsychological and functional neuroimaging studies frequently use facial
expressions to probe brain regions involved in affect recognition such as the amygdala, the
insula and the orbitofrontal cortex (O´Doherty et al., 2003). One feature of a face that can
elicit a strong affective response is its attractiveness or beauty. Attractiveness impacts not
only mating success, but also kinship opportunities, the evaluation of personality,
performance and employment prospects (Cellerino et al., 2007; Ishai, 2007; Kranz and
Ishai, 2006; Werheid, Schacht and Sommer, 2007). Functional magnetic resonance imaging
(fMRI) studies have shown that a complex network involving different regions of the
orbitofrontal cortex, the medial prefrontal cortex, the paracingulate cortices, the insula, the
amydgala and the superior temporal cortex are involved in processing attractiveness (Ishai,
2007; Kranz and Ishai, 2006; Winston et al., 2007). More specifically, the orbitofrontal
cortex, the amygdala and the insula are involved in sensing the value of social
attractiveness (Winston et al., 2007).
The electroencephalogram (EEG) has also been used to study the temporal
characteristics by which facial attractiveness is appraised. These studies have shown that
facial analysis involves distinct steps, with early events correlating with the recognition of
physical characteristics and later components being associated with emotional, gender and
social information carried by facial expression (Cellerino et al., 2007; Werheid, Schacht
and Sommer, 2007).

1.4 Investigating customer satisfaction with aesthetic dermatological treatments
Many people worry about physical appearances, mostly with respect to aging of the
face and skin. This phenomenon has motivated new developments in cosmetic dermatology
and the need for evaluating patients’ levels of satisfaction with the new proposed treatments.
Poll questionnaires have been used for such evaluations, and the analysis of the
electroencephalogram mapping, which is obtained while the patient answers the satisfaction
questionnaire, may cause the results to be less subjective (Arruda, Rocha and Rocha, 2008).
Because patients should be allowed free time for answering pool questions, the EEG
analysis has to examine the brain activity preceding the moment the patient provides the
answer. In contrast, previous studies involving EEG and decision-making focused on the
analysis of brain activity following the decision-making or stimulus presentation (Chen et
al., 2009; Polezzi et al., 2010; Utku et al., 2002). Rocha et al. (2010) examined vote
decision-making and used an EEG brain mapping technique, which considered the brain to
be a distributed processing system (Foz et al., 2001; Rocha, Massad and Pereira Jr., 2004;
Rocha et al., 2005; Rocha et al., 2010 and Rocha, Rocha and Massad, 2011). Using this
approach, Rocha and colleagues (2010), for instance, were able to disclose different
patterns of brain activity associated with different voting decisions. In a similar vein, we
used the same EEG mapping technique to investigate the brain activity associated with
customer satisfaction.
The purpose of the present paper is to extend the analysis on consumer satisfaction
reported by Arruda et al (2008) by studying brain mappings associated with level of
satisfaction with treatment and self-evaluation of face components before and after
treatment.

2. Methods
2.1 The EEG brain mapping of a cosmetic dermatology treatment
Hyaluronic acid (HA) was used to correct nasolabial folds and in lip augmentation
in 33 women aged 30 to 55 years with a mean age of 44 years. At the initial evaluation,
patients were inspected for nasolabial fold depth and lip volume loss. Informed consent was
obtained from all patients before treatment, and the experimental protocol was approved by
the Ethics Research Committee of the Catholic University of Campinas. Treatment
consisted of an injection of 1.0 ml of HA in each nasolabial fold or in the upper and lower
lip. This procedure was performed under local anesthesia or infraorbital nerve blockage.
Patients were reevaluated at 48 hrs and 1, 2 and 3 months after the initial procedure. The
reevaluation detected the side effects and assessed the treatment durability.
2.2 The experiment
At the third-month evaluation, two networked personal computers were used to record
the EEG and present the patients with a questionnaire about:

1) the self-evaluation of face components: hair, forehead, eyebrows, eyes, nose, chin,
facial contours, cheeks and neck, which were classified as superb, great, regular,
bad or very bad (F).
2) the reasons for deciding to undergo the treatment. Patients selected from one or
more of the following options: because it was a free treatment; because she was
dissatisfied with her appearance; because she had already planned to submit
herself to an aesthetic treatment; because it was recommended by a friend; none of
these.
3) the level of satisfaction with the results of the treatment by comparing before and
after photos and declaring the patient very satisfied, satisfied, unsatisfied, very
unsatisfied, or none of these (A) .
4) the self-evaluation of appearance after the treatment: very much improved, improved,
did not change, worsened or badly worsened (R).
5) how the patients’ family, friends and coworkers evaluated the results of the
treatment: excellent, good, bad, very bad, no opinion.
6) the decision to repeat the treatment: definitely yes, yes, no, definitely no, undecided.
7) the decision to recommend the treatment to other people: definitely yes, yes, no,
definitely no, undecided.
Regression analysis was used to disclose possible associations between the answers
to different items on the questionnaire. The correlation entropy h(c
i
) of the EEG activity
recorded while answering the questionnaire was calculated, as described by Rocha et al.
(2004, 2005, 2010, 2011), for each electrode e
i
of the 10/20 system and for each item of the
questionnaire (see appendix 1 for a full description of the EEG methodology). Regression

analysis was used to study the correlation between the type of answer for each
questionnaire item and the associated h(c
i
). The values of the angular coefficients for the
calculated linear regression were color-coded to build the regression EEG mappings
(Figure 2) associated with each questionnaire item.
3 Results
3.1 The poll data
Table 1 shows what patients liked most of their facial components (questionnaire
item 1 - EEG mappings F in Figures 1 and 2). Eyes and hair were the preferred elements,
and the forehead, eyebrows, nose and neck had the highest rate of disapproval. We recoded
the facial component evaluation according to the following rule: Superb (S) = 5, Good (G)
= 4, Regular (R) = 3, Bad (B) = 2 and Very Bad (VB) = 1, and we calculated a general
appearance index for each patient as the mean of her evaluation of all facial components.
The mean general appearance index calculated for all patients was 4,07. Hence, the
majority of the patients considered themselves to be attractive.
Table I – Face components evaluation
hair forehead
eyebrows
eyes
nose
cheeks
ears
lips chin
neck
contour
S 15 8 12 8 0 4 0 15 0 0 12
G 65 50 54 71 54 50 73
58 65 35 46
R 0 0 0 9 8 12 15
12 19 8 19
B 20 12 8 8 35 34 8 15 12 46 23
VB
0 30 26 4 3 0 4 0 4 11 0

S: Superb; G: Great; R: Regular; B: Bad and VB: Very bad. Data are presented as a
percentage.
Patients decided to undergo the treatment (questionnaire item 2) because they were
already considering it (54%) or dissatisfied with their lips or nasolabial folding (52%). The
fact that the treatment was free of charge did not influenced patient’s decision
Patients were very satisfied or satisfied with the results of the treatment
(questionnaire item 3 – EEG mappings R in Figures 1 and 2) and with their facial
attractiveness after the treatment (questionnaire item 4 – EEG mappings A in Figures 1 and
2). No patient claimed to be unsatisfied with both the immediate and later treatment results.
In addition, patients declared that family and friends made highly positive comments about
their new appearances (questionnaire item 5). Therefore, patients were firmly determined
(60%) or determined (32%) to repeat the treatment (questionnaire item 6) and to
recommend it (questionnaire item 7) to family (70%), friends (60%) and others (30%).
3.2 The Brain Mappings
The mean entropy mapping values calculated for each of the studied EEG epoch are
shown in Figure 2 with the corresponding Z score mappings (Appendix 1). The minimum Z
score obtained for all mappings was 1.979, which invalidated the null hypothesis that the
entropy values could be due to chance.
The minimum values for the calculated mean entropies were obtained for the
occipital electrodes O
1
, O
2
and O
z
for all of the studied EEG epochs (Table 2). The
maximum values of the calculated mean entropies were obtained for the electrodes F4, F8,
CZ, C4 and T4 for all of the studied EEG epochs (Table 2). The minimum Pearson’s

correlation coefficient obtained for comparing the mean entropy mapping values was 0.93,
confirming that all 3 mappings were very similar.
Figure 2 – Mean entropy brain mapping (E) associated with facial self-evaluation (F),
satisfaction evaluation of the results (R) and attractiveness self-evaluation (A) and the Z
score mapping (Z).
The regression analysis results showed that Holm-Bonferroni and FWER
procedures were equivalent in selecting the most significant statistical inferences, as shown
in Table 2. Back-wise regression was found to be more conservative than forward-wise

regression, and it may enhance type III error frequency. Therefore, the forward-wise
inferences found to be significant according to the FWER procedure (marked in red in
Table 3) were used to build the brain mapping shown in Figure 3.
Table 2 – Entropy statistics
F R A
Mean Std.Dev. Mean Std.Dev. Mean Std.Dev.
C3 6.55 2.53 7.15 2.10 6.58 2.53
C4 6.90 2.36 7.20 1.75 6.98 2.19
CZ 6.83 2.49 7.37 1.85 6.86 2.36
F3 6.42 2.19 6.79 1.99 6.30 2.35
F4 6.75 2.45 7.49 1.66 6.89 2.35
F7 5.97 2.39 6.60 1.98 6.10 2.28
F8 6.73 2.66 7.50 1.84 6.93 2.52
FP1 6.35 2.15 6.90 1.72 6.44 2.16
FP2 6.24 2.29 6.76 1.83 6.37 2.23
FZ 6.75 2.23 7.05 1.80 6.77 2.18
O1 5.72 3.48 5.61 3.40 5.73 3.46
O2 5.44 3.28 5.44 3.24 5.44 3.25
OZ 5.53 3.33 5.45 3.30 5.54 3.33
P3 6.39 2.54 6.84 1.86 6.45 2.47
P4 6.50 2.22 6.75 1.70 6.48 2.12
PZ 6.45 2.27 6.68 1.76 6.53 2.16
T3 6.42 2.46 6.91 2.03 6.52 2.40
T4 6.67 2.55 7.11 2.04 6.72 2.41
T5 6.53 2.53 7.00 2.07 6.50 2.55
T6 6.51 2.29 6.66 2.71 6.65 2.26
The regression brain mapping values associated with the facial component self-
evaluation, treatment results and the self-evaluation of appearance after the treatment are
shown in Figure 3. The h(c
i
) calculated for the central (FZ, CZ and PZ) and right (FP2, T4
and P4) electrodes (green to blue electrodes in Figure 2F) was positively correlated with the
facial component of self-evaluation, such that a high h(c
i
) at these electrodes was
associated with a very positive self-evaluation (Max = 5). In contrast, the h(c
i
) calculated
for the left (F3, F7, C3, P3 and T5) and right frontal (F4 and F8) electrodes (rose to dark
red electrodes in Figure 2F) was negatively correlated with the facial component of self-

evaluation, such that a high h(c
i
) at these electrodes was associated with a positive self-
evaluation (Min = 4).
Table 3 – Angular coefficients ( _
b
) and their statistical significance (
p
) for the
regressions used to obtain the EEG mapping values in Figure 2. The statistically
significant
p
based on the FWER procedure is shown in red.
F R A
_
b
p
_
b
p
_
b
p
C3 0.3363 0.0166 -0.0230
0.6689
0.4598 0.0081
C4 -0.0344 0.7093 -0.2452
0.0322
-0.3050 0.0316
CZ -0.2165 0.0060 -0.3622
0.0007
-0.3214 0.1868
F3 0.3610 0.0000 -0.2620
0.0003
-0.4456 0.0095
F4 0.4165 0.0000 -0.8284
0.0000
-0.1021 0.6973
F7 0.2857 0.0000 -0.0870
0.0540
0.1553 0.1055
F8 0.5271 0.0000 0.4088 0.0032
0.5323 0.0362
FP1 -0.0534 0.4255 0.3037 0.0018
-0.4322 0.0456
FP2 -0.3182 0.0000 0.0844 0.3773
0.4278 0.0316
FZ -0.4143 0.0000 -0.0072
0.9348
0.0699 0.6613
O1 -0.1035 0.0002 0.0590 0.1369
0.1294 0.1719
O2 0.2495 0.0000 -0.0197
0.5931
0.1855 0.0392
OZ -0.0810 0.0024 -0.2029
0.0000
-0.0782 0.4126
P3 0.1875 0.0012 -0.3408
0.0000
-0.0012 0.9943
P4 -0.1616 0.0308 0.3843 0.0001
-0.3195 0.0390
PZ -0.2618 0.0000 0.6696 0.0000
0.6001 0.0001
T3 -0.4515 0.0000 -0.1887
0.0277
0.2269 0.2449
T4 -0.5192 0.0000 0.1818 0.0215
-0.3742 0.0376
T5 0.4243 0.0000 0.1757 0.0675
0.0974 0.6441
T6 -0.2110 0.0000 0.1509 0.0001
0.0762 0.3896
R
2
44% 20% 57%
F/R F/A R/A
Pc 0,1 0 0,46

Figure 3 – Regression mapping values )(...)(
202011
chbchbaD +++= calculated for all
volunteers, such that negative values of b
i
were color-coded from rose to dark red, positive
values of b
i
were color-coded from yellow to dark blue, and those b
i
= 0 were color-coded
as orange. The correlation entropy h(c
i
) calculated for the electrodes e
i-
associated to
negative b
i
contributed to make D Min, whereas h(c
i
) calculated for the electrodes e
i+
associated to positive b
i
contributed D Max. For the facial element component of self-
evaluation (F), Max = 5 (for Superb) and Min = 1 (for Very bad). For the satisfaction with
the treatment results (R), Max = 5 (for Very satisfied) and Min = 4 (for Satisfied), and for
the attractiveness self-evaluation after treatment (A), Max = 5 (for Superb) and Min = 3
(for regular). The computed values of
i
b
, their statistical significance p and the computed
R
2
for each regression are shown in Table 3. The similarity between the mapping values, as
evaluated by the Pearson’s coefficient (Pc), is shown in Table 3.

Before and after photo comparison was positively associated with the h(c
i
) that was
calculated from the F3, P3, FZ, F4 and F8 electrodes (green to blue electrodes in Figure
2R), implying that a high h(c
i
) was associated with a very positive evaluation (Max = 5). In
contrast, the h(c
i
) calculated for the electrodes FP1, P4, CZ, C4 and PZ was negatively
associated (rose to dark red electrodes in Figure 2R) with the before and after decision,
where a high h(c
i
) was associated with a positive evaluation (Min = 4).
Finally, the patients’ self-evaluation of their appearance was positively correlated
with the h(c
i
) calculated for the electrodes FP1, F3, CZ, P4 and T4 (green to blue electrodes
in Figure 2A), such that the high values of h(c
i
) were associated with a highly positive
appearance (Max = 5). In contrast, h(c
i
) calculated for the electrodes FP2, F8, C3 and PZ
(rose to dark red electrodes in Figure 2A) was negatively correlated with the patient’ self-
evaluation, such that the high values of h(c
i
) were associated with a regular appearance
(Min = 3).
4 Discussion
Beauty is strongly influential in human reproduction and socially motivated
behaviors (Aharon et al., 2001), and it is more important for women than for men. Also,
women are judged more critically than men concerning aging because modern urbanized
societies allow only one standard of female beauty, that of the girl (Berman, O’Nan and
Floyd, 1981; Deutsch, Zalenski and Clark, 1986; Sontag, 1972
)
. Finally, beauty is the result
of both self-evaluation and social recognition. The female’s sense of her own beauty is
determined by the feeling she has about herself and the cumulative opinions of her partner,
family and friends.

In this study, the volunteers were satisfied or very satisfied with the components of
their faces, and the EEG mapping values showed that this evaluation was supported by a
diverse set of neurons, where their activity was recorded by a large number of electrodes
(Figure 2).
The present results were consistent with those of previous studies showing that
facial recognition was supported by a specific and diverse neural circuit (Aharon et al.,
2001; Britton et al., 2006a; Brady, Campbell and Flaherty, 2004; Ishai, 2007; Ishai,
Schmidt and Boesiger, 2005; Kircher et al., 2001; Singer et al., 2004). The h(c
i
) values
calculated for the left and right anterior frontal electrodes were inversely correlated with
this self-evaluation, and the values obtained for the right posterior electrodes were directly
correlated with a very positive classification of patients’ facial elements. Previous studies
have shown that the left hemisphere was focused on self-evaluation of the body and the
right hemisphere with perception of other people’s bodies (Alisson, Puce and McCarthy,
2001; Brady, Campbell and Flaherty, 2004; Ishai, Schmidt and Boesiger, 2005; Kircher et
al., 2001; Stone and Valentine, 2005).
Because female beauty is determined by the feelings that females have about
themselves and socially collected opinions, we would propose that the left brain contributes
to the self component and the right brain encodes the social component of the volunteers’
evaluation of their beauty. If this is true, then the present results suggest that the self
component reduced a more positive social evaluation of the volunteer’s beauty. In addition,
the patients were dissatisfied with the age effects on their looking and motivated to undergo
the aesthetic treatment. Therefore, it may be concluded that aging creates a necessity to
remedy decaying beauty, which, in turn, motivates the search for an aesthetic treatment.

The volunteers were asked to compare their pre- and post-treatment photos and to
decide if their attractiveness was improved or worsened. They were unanimous in deciding
that the treatment improved or very much improved their appearance. The very much
improved decision was supported by the increase in the h(c
i
) calculated from electrodes F8,
F4, F3, and P3 (green to blue electrodes in Figure 2R), whereas the high values of h(c
i
) at
P4, C4, CZ and PZ (rose to dark red electrodes in Figure 2R) reduced their enthusiasm with
the treatment results. A comparison of Figures 2F and 2R showed a reversion of the
correlation between h(c
i
) for the left electrodes and attractiveness self-evaluation. Before
the treatment, the left hemisphere contributed to a less positive decision about the
attractiveness of the components of the facial elements, whereas the high values of h(c
i
) at
the left electrodes F3 and P3 were associated with the very much improved decision after
the treatment. The only difference in the right hemisphere was the reversion of the
correlation between the h(c
i
) and decision-making, where the frontal electrodes became
associated with the more positive evaluations and the posterior electrodes correlated more
to the less positive evaluations made after the treatment than before the treatment.
If the left hemisphere was more concerned with self-evaluation, then these results
showed that the volunteers were more positive about their appearance after the treatment,
and the social evaluation remained unchanged. Therefore, the analysis of the brain activity
during these different decisions supports the “subjective” poll opinion.
The very positive self-evaluation of patients’ appearances after the treatment
(questionnaire item 4) was associated with the high values of h(c
i
) calculated from the
anterior left and posterior right electrodes (figure 2A). This supports the hypothesis that the
positive evaluation from the treatment was due both to personal and social factors. This

finding is in agreement with the fact that the left hemisphere contributed more to a more
positive appearance self-evaluation after than before the treatment and with the high
satisfaction of family and friends with the volunteer’s new appearance. Because the patients
were very decided about repeating the treatment and to promote it among family and
friends, we concluded that their satisfaction with the results of the aesthetic intervention
translated into trust. Whether trust was solely correlated to the treatment technique, or it is
inputted to the physician’s abilities remains to be investigated.

Appendix 1 – The EEG mapping technology
Two networked personal computers were used to record the EEG and another for
sequentially displaying the questionnaire items (Figure 3). The volunteers were allowed to
take as much time as needed to make a decision. The times at which the questionnaire item
was displayed (t
q
) and at which the decision was made (t
d
) were recorded. The EEG was
visually inspected for artifacts before its processing, and the EEG epochs associated with a
bad EEG were discarded (e.g., when eye movements could compromise the results of
regression analysis). The linear correlation coefficients
ji
r
,
for the activity at each
recording site
i
e that referred to the activity for each of the other 19 electrodes
j
e
were
calculated for the EEG epoch (
qd
tt −) of each decision, which was used to select the EEG
sequence to calculate the correlation entropy )(
i
ch (Foz et al., 2001; Rocha, Massad and
Pereira Jr., 2004; Rocha et al., 2005; Rocha et al., 2010 and Rocha, Rocha and Massad,
2011) as:
)1(log)1(log)(
,2,,2,, jijijijiji
rrrrrh −−−−=
(1)
19
19
1,
∑
=
ji
i
r
r (2)
)1(log)1(log)(
22 iiiii
rrrrrh
−−−−=
(3)

∑
−=
19
1,
)()()(
jiii
rhrhch
(4)
It may be stressed that we did not intend to assign any physiological meaning to the
entropy )(
i
ch . The correlation entropy )(
i
ch was assumed to be a measure of the uncertainty
about the existence of a correlation between the activity recorded by pairs of
electrodes
ji
ee ,. The entropy )(
,ji
rh is equal to 1 when 5.0
,
=
ji
r and equal to 0 when
0
,
=
ji
r or 1
,
=
ji
r. Thus, )(
,ji
rh measures the uncertainty of the correlation between the
EEG activity recorded by
ji
ee ,. The entropy )(
i
rh of the mean correlation ŕ
i
provides more
information about the covariance of the correlation between the activity recorded by
i
eand
all other
j
e. If 5.0
,
=
ji
rfor all
j
e, then 5.0=
i
r and 1)( =
i
rh . Also, if 0
,
→
ji
rfor some
j
e,
1
,
→
ji
r
for some other
j
e
and
5.0
,
→
ji
r
for the remaining
j
e
, then 5.0=
i
r and 1)( =
i
rh .
However, if 1
,
→
ji
r( 0
,
→
ji
r) for most of the
j
e, then 1=
i
r ( 0=
i
r) and 0)( =
i
rh . All
other conditions imply 0)( →
i
ch . Therefore, the actual value of )(
i
ch is a measure of how
much the EEG activity recorded by the electrode
i
e
may be associated with the task being
processed by the brain.

Figure 4 – Outline of the experiment. Two networked microcomputers were used to record
the EEG activity (10/20 system) while the volunteer decided about a questionnaire item.
The beginning of the questionnaire item was displayed, and the moment a decision was
made, data were saved in the database together with the type of decision-making and time
required to achieve the decision. The linear correlation coefficients r
i,j
for the recorded
activity at each recording electrode e
i
referred to the recorded activity for each of the other
19 recording sites e
j
that were calculated for each questionnaire item and volunteer. These
r
i,j
were used to calculated the correlation entropy h(r
i
) for each recording electrode e
i
.
Therefore, h(r
i
) was calculated for all 20 recoding electrodes. The corresponding values of
h(r
i
) constituted the entropy data base. Regression analysis concerning the decision about
each questionnaire item and the h(r
i
) was used to create the cognitive brain mapping values.
Each mapping image shows the contribution of the product β
i
*h
m
(r
i
) of each electrode e
i
to
the decision made. The h
m
(r
i
) is the average of the h(r
i
) calculated for all volunteers. The
location of each 10/20 system electrode is displayed at the left brain drawings.

A null hypothesis was constructed by randomly reordering the EEG channels, i.e.,
the mean entropy was recomputed from a hypothetical brain obtained by randomly
shuffling the EEG recorded activity from the individuals. Next, the Z scores between the
mean entropy and the null hypotheses were computed for each of the EEG epochs (facial
evaluation – F, results satisfaction evaluation – R and attractiveness self-evaluation – A).
The mean entropy calculated for each channel was used to generate the mean entropy brain
mapping values shown in Figure 2, and the corresponding Z scores were used to generate
the Z score mapping values. The minimum Z core for all of these calculations was 1.979,
showing that the null hypotheses were rejected for all EEG epochs.
Linear regression analysis was used to study the correlation between the )(
i
ch and
the response time
qs
ttST −= , and, the logistic regression analysis was used to study the
correlation between the )(
i
ch and
2
P
(adequate) or
2
−
P (not adequate) decision. The
normalized values of the )(
ii
chb were used to build the color-coded brain mapping images
to display the results of the regression analysis. The color-coding routine used commercial
software. Statistically positive betas were coded from green (normalized )(
ii
chb tending to
0) to dark blue (normalized )(
ii
chb tending to 1). Statistically negative )(
ii
chb are
displayed from rose (normalized )(
ii
chb tending to 0) to dark red (normalized )(
ii
chb
tending to -1), and statistically non-significant )(
ii
chb are shown in orange. Brain contours
are used as references for the spatial location of the 10/20 system electrodes.
Let
p
be the probability that the null hypothesis for a statistical inference is true. A
statistical inference is assumed to be true if the value of
p
is less than a given significance

threshold
α
, which is considered the maximum permissible error in making the inference.
The threshold
05.0
=
α
is commonly used for single inferences because the inference is not
due to chance alone at least 95 times out of 100. The risk of declaring significant what is
not is, at most, 5 times out of 100.
The analysis becomes complicated for multivariate inferences because the total
possible error
_
α
is dependent on the number
n
of inferences (Benjamin et al., 2001;
Blakesley et al., 2009; Genovese et al., 2002; Huizenga et al., 2007; Marroquin et al., 2011;
Nandy and Cordes, 2007; Vechiato et al., 2010). For statistical inferences supported by
independent variables, the possible error is multiplied by
n
because, for each inference, the
risk of its being wrong is 5%. The total possible error is, therefore,
αα
n=
_
, which implies
that the certainty of making a wrong inference increases as the number of inferences
increases. The solution is to decrease the value of
α
to minimize the total possible error.
Bonferroni proposed
α
to be
n/1
of the total admissible error
_
α
. Because the Bonferroni
procedure may be too conservative and may increase the type II error, the Holm-Bonferroni
method has been commonly used instead. However, when the independence hypothesis
can be removed,
αα
n<
_
. In this condition,
α
may be greater than
n/1
of the total
admissible error
_
α
. As the correlation between the variables increases, the FWER
calculation as
n
)1(1
_
αα
−−≤ maintains the level of possible total error at around
_
α
.
The entropies calculated for the 10/20 electrodes are dependent measurements. This
condition is because both the Holm-Bonferroni and FWER methods were used to calculate

the significance of the statistical inferences about the volunteer’s evaluations and the EEG
activity, as measured by h(c
i
).
Stepwise regression includes regression models in which the selection of predictive
variables is performed by an automatic procedure. Commonly, a sequence of F-tests is used.
Stepwise regression is another tool available to find the best relevant statistical inferences.
Forward selection (forward-wise regression), which involves starting with no variables in
the model and testing the variables individually and including them if they are 'statistically
significant. Backward elimination (back-wise regression), which involves starting with all
candidate variables and testing them individually for statistical significance and deleting
any that are not significant. Stepwise regression is used for selecting the most useful
statistical inferences in neurosciences (Antonakis and Dietz, 2011; Mueller et al., 2011;
Song et al., 2008; Stadler et al., 2007), and it will be used here for the same purpose.
We combined stepwise regression, the Holm-Bonferroni and the FWER methods to
select the most relevant statistical inferences between the EEG activity and the volunteers’
poll evaluations.
The Pearson’s correlation coefficient was calculated to provide an index of
similarity between both the mean entropy and the regression brain mapping values obtained
for each of the studied EEG epochs.

References
Aharon, I., N. Etcoff, D. Ariely, C.F. Chabris, E. O'Connor and H.C. Breiter (2001)
Beautiful Faces Have Variable Reward Value: fMRI and Behavioral Evidence.
Neuron, 32: 537-551.
Aktsuki Y., M. Sigiura, J. Watanabe, K. Yamashita, Y. Sassa, S. Awata, H. Matwuoka, Y.
Maeda, Y. Matuse, H. Fukuda and R. Kawashima (2003) Context-dependent cortical
activation in response to financial reward and penalty: and event-related fMRI study.
NeuroImage, 19:1674-1685.
Alisson, T., A. Puce and G. McCarthy (2001) Social perception from visual cues: role of
the STS region. Trends in Cognitive Sciences, 11:267-279.
Ambler T., A. Ioannides and S. Rose (2000) Brands on the Brain: Neuro-Images of
advertsing. Business Strategy Rev, 11:17-30.
Antonakis J. and J. Dietz (2011) Looking for validity or testing it? The perils of stepwise
regression, extreme-scores analysis, heteroscedasticity, and measurement error. Personality
and Individual Differences, 50(3):409-415.
Arruda L.H., F.T. Rocha and A.F. Rocha, (2008) Studying the satisfaction of patients on
the outcome of an aesthetic dermatological filler treatment. Journal of Cosmetic
Dermatology, 7(4):246-250.
Benjamin Y., D. Drai, G. Elmer, N. Kafkafi and I. Golani (2001) Controlling the false
discovery rate in behavior genetics research. Behavioral Brain Research, 125(1-2): 279-284.
Berman P.W., B.A. O’Nan and W. Floyd (1981) The double standard of aging and the
social situation: judgments of attractiveness of the middle-aged woman. Sex Roles, 7:87–96.

Blakesley R.E., S. Mazumdar, M.A. Dew, P.R. Houck, G. Tang, C.F. Reynolds III and
M.A. Butters (2009) Comparisons of Methods for Multiple Hypothesis Testing in
Neuropsychological Research. Neuropsychology, 23(2):255-264.
Brady N., M. Campbell and M. Flaherty (2004) My left brain and me: a dissociation in the
perception of self and others. Neuropsychologia, 42:1156-1161.
Breiter H.C., I. Aharon, D. Kahneman, A. Dale and P. Shizgal (2001) Functional imaging
of neural responses to expectancy and experience of monetary gains and losses. Neuron,
30:619-639.
Britton J.C., K.L. Phan, S.F. Taylor, R.C. Welsh, K.C. Berridge and I. Liberszon (2006a)
Neural correlates of social and nonsocial emotions: An fMRI study. NeuroImage, 31:906-
919.
Britton J.C., S.F. Taylor, K.D. Sudheimer and I. Liberzon (2006b) Facial expressions and
complex IAPS pictures: Common and differential networks. NeuroImage, 31:906-919.
Cellerino, A., D. Borghetti, D.R. Valenzano, G. Tartarelli, A. Mennucci, L. Murri and F.
Sartucci (2007) Neurophysiological correlates for the perception of facial sexual
dimorphism. Brain Res. Bulletin 71:515-522.
Chen. P., J. Qiu, H. Li, and Q. Zang (2009) Spatiotemporal cortical activation underlying
dilemma decision-making: An event-related potential study. Biological Psychology,
82:111-115.
Deutsch F.M., C.M. Zalenski and M.E. Clark (1986) Is there a double standard of aging? J
Appl Soc Psychol, 16:771–785.

Foz, F.B., F.L.P. Lucchini, S.Palmieri, A.F. Rocha, E.C. Rodella, A.G. Rondó, M.B.
Cardoso, P.B. Ramazzini, C.C. Leite (2002) Language plasticity revealed by
electroencephalogram mapping. Pediatric Neurology, 26(2):106-115.
Genovese C.R., N.A. Lazar, and T. Nichols (2002) Thresholding of Statistical Maps in
Functional Neuroimaging Using the False Discovery Rate1. NeuroImage 15:870-878.
Huizenga H.M., H. Smeding, R.P.P.P. Grasman and B. Schmand (2007) Multivariate
normative comparisons. Neuropsychologia, 45(11):2534-2542.
Ishai A., C.F. Schmidt and P. Boesiger (2005) Face perception is mediated by a distributed
cortical network. Brain Res. Bulletin, 67(13):87-93.
Ishai, A. (2007) Sex, beauty and the orbitofrontal cortex. Int. J. Psychophysiology, 63:181-
185.
Kircher T.T.J., C. Senior, M.L. Phillips, S. Rabe-Hesketh, P.J. Benson, E.T. Bullmores, M.
Brammer, A. Simmons, M. Bartels and A.S. David (2001) Recognizing one´s own face.
Cognition, 78:b1-b15.
Knutson B., S. Rick, G.E. Wimmer, D. Prelec and G. Loewentein (2007) Neural predictors
of purchases. Neuron, 53:147-156.
Kranz F., and A. Ishai (2006) Face perception is modulated by sexual preference. Current
Biol, 16:63-68.
Marroquin J.L., R.J. Biscay, S. Ruiz-Correa, A. Alba, R. Ramirez and J.L. Armony (2011)
Morphology-based hypothesis testing in discrete random fields: A non-parametric method
to address the multiple-comparison problem in neuroimaging. NeuroImage, In Press,
Corrected Proof, Available online 8 April 2011.

McClure, S.M., J. Li, D. Tomlin, K.S. Cypert, L.M. Montague and R.M. Montague (2004)
Neural correlates of behavioral preference for culturally familiar drinks. Neuron 44:379-
387.
Mueller S.G., L.L. Chao, B. Berman and M.W. Weiner (2011) Evidence for functional
specialization of hippocampal subfields detected by MR subfield volumetry on high
resolution images at 4 T. NeuroImage, 56(3):851-857.
Nandy R. and D. Cordes (2007) A semi-parametric approach to estimate the family-wise
error rate in fMRI using resting-state data. NeuroImage, 34(4):1562-1576.
O'Doherty J., J. Winston, H. Critchley, D. Perrett, D.M. Burt and R.J. Dolan (2003) Beauty
in a smile: the role of medial orbitofrontal cortex in facial attractiveness. Neuropsychologia,
41:147–155.
Plat M. and C. Padoa-Schioppa (2009) Neuronal representations of value. In
Neuroeconomics, eds, Glimcher, Camerer, Fehr and Poldrack. Academic Press, 441-462.
Polezzi D., G. Sartori, R. Rumiati, G. Vidotto and I. Daum I. (2010) Brain correlates of
risky decision-making. NeuroImage, 49:1886–1894.
Quiroga R.Q., L. Reddy, G. Kreiman, C. Kcoh and I. Fried (2005) Invariant visual
representation by single neurons in the human brain. Nature, 435:1102-1107.
Rocha, A.F., E. Massad and A. Pereira Jr (2004) The Brain: From Fuzzy Arithmetic to
Quantum Computing. Springer Verlag, Heildelberg.
Rocha, F.T., A.F. Rocha, E. Massad and R. Menezes (2005) Brain mappings of the
arithmetic processing in children and adults. Cognitive Brain Research, 22(3):359-372.

Rocha A.F., F.T. Rocha, E. Massad and M.N. Burattini (2010) Neurodynamics of an
election. Brain Research, 1351:198-211.
Rocha, A.F., F.T. Rocha and E. Massad (2011) The Brain as a Distributed Intelligent
Processing System: An EEG Study. Plos One, 6:e17355.
Roger R.D., N. Ramnani, C. Mackay, J.L. Wiilson, P. Jezzard, C.S. Carter and S.M. Smith
(2004) Distinct portions of Anterior Cingulate Cortex and Medial Prefrontal Cortex are
activated by reward processing in separable phases of decision-making cognition. Biol.
Psychiatry, 55:594-602.
Schaefer M., H. Berens, H.J. Heinze and M. Rotte (2006) Neural correlates of culturally
familiar brands of car manufacturers. NeuroImage, 31:861-865
Singer T., S.J. Kiebel, J.S. Winston, R.J. Dolan and C.D. Frith (2004) Brain Responses to
the Acquired Moral Status of Faces. Neuron, 41:653-662.
Song M., Y. Zhou, J. Li, Y. Liu, L. Tian, C. Yu and T. Jiang (2008) Brain spontaneous
functional connectivity and intelligence. NeuroImage, 41(3):1168-1176.
Sontag S. (1972) The double standard of aging. Saturday Review of Literature, 39:29–38.
Stadler C., P. Sterzer, K. Schmeck, A. Krebs, A. Kleinschmidt and F. Poustka (2007)
Reduced anterior cingulate activation in aggressive children and adolescents during
affective stimulation: Association with temperament traits. Journal of Psychiatric Research,
41(5):410-417.
Stone A. and T. Valentine (2005) Strength of visual percept generated by famous faces
perceived awareness: Effects of affective valence, response latency and visual field.
Consciousness and Cognition, 14:548-564.

Tobler P.N., C.D. Fiorillo and W. Schultz (2005) Adaptative coding of reward by dopamine
neurons. Science, 307:1642-1645.
Utku H., Ö.U. Erzengin, E.D. Cakmak and S. Karaka (2002) Discrimination of brain's
neuroelectric responses by a decision-making function. Journal of Neuroscience Methods,
114:25-31.
Vecchiato G., F. De Vico Fallani, L. Astolfi, J. Toppi, F. Cincotti, D. Mattia, S. Salinari
and F. Babiloni (2010) The issue of multiple univariate comparisons in the context of
neuroelectric brain mapping: An application in a neuromarketing experiment. Journal of
Neuroscience Methods, 191(2):283-289.
Werheid K., A. Schacht and W. Sommer (2007) Facial attractiveness modulates early and
late event-related brain potentials. Biol. Psychol, 76:100-108.
Winston J.S., J. O´Doherty, J.M. Kilner, D.I. Perret and R.J. Dolan (2007) Brain systems
for assessing facial attractiveness. Neuropyschologia, 45:195-206.
Yoon C., A.H. Gutchess, F. Feinberg and T.A. Polk (2006) A functional magnetic
resonance imaging study of neural dissociations between brand and person judgments. J.
Consumer Res., 33:31-40.
Young, C. (2002) Brain Waves, Picture Sorts® and Branding Moments. Journal of
Advertising Research, 42(4):42-53.