ArticlePDF Available

Validation of the child models of the Radboud Faces Database by children

Authors:

Abstract and Figures

To draw valid and reliable conclusions from child studies involving facial expressions, well-controlled and validated (child) facial stimuli are necessary. The current study is the first to validate the facial emotional expressions of child models in school-aged children. In this study, we validated the Radboud Faces Database child models in a large sample of children ( N = 547; 256 boys) aged between 8 and 12. In addition, associated validation measures such as valence, clarity, and model attractiveness were examined. Overall, the results indicated that children were able to accurately identify the emotional expressions on the child faces in approximately 70% of the cases. The highest accuracy rates were found for “happiness,” whereas “contempt” received the lowest accuracy scores. Children confused the emotions “fear” and “surprise,” and the emotions “contempt” and “neutral” with one another. Ratings of all facial stimuli are available (https://osf.io/7srgw/) and can be used to select appropriate stimuli to investigate the processing of children’s facial emotional expressions.
Content may be subject to copyright.
Methods & Measures
Validation of the child models of the Radboud
Faces Database by children
Geraly Bijsterbosch,
1
Lynn Mobach,
1,2
Iris A. M. Verpaalen,
1
Gijsbert Bijlstra,
1
Jennifer L. Hudson,
2
Mike Rinck,
1
and Anke M. Klein
1
Abstract
To draw valid and reliable conclusions from child studies involving facial expressions, well-controlled and validated (child) facial stimuli are
necessary. The current study is the first to validate the facial emotional expressions of child models in school-aged children. In this study,
we validated the Radboud Faces Database child models in a large sample of children (N¼547; 256 boys) aged between 8 and 12. In
addition, associated validation measures such as valence, clarity, and model attractiveness were examined. Overall, the results indicated
that children were able to accurately identify the emotional expressions on the child faces in approximately 70% of the cases. The highest
accuracy rates were found for “happiness,” whereas “contempt” received the lowest accuracy scores. Children confused the emotions
“fear” and “surprise,” and the emotions “contempt” and “neutral” with one another. Ratings of all facial stimuli are available (https://osf.io/
7srgw/) and can be used to select appropriate stimuli to investigate the processing of children’s facial emotional expressions.
Keywords
Facial emotions, child expressions, facial emotion recognition, Radboud Faces Database (RaFD), validation
Facial emotion recognition (FER) plays an important role in day-to-
day social interactions: Facial expressions convey crucial informa-
tion on the thoughts and feelings of an individual, which is neces-
sary for successful communication between humans (Leppa¨nen &
Hietanen, 2001). The ability to differentiate between facial expres-
sions develops between infancy and early adulthood (Lawrence
et al., 2015). Child development researchers, therefore, highlight
the importance of FER in the emotional development of children,
and they often employ facial stimuli in their studies (e.g., Chronaki
et al., 2014; Lawrence et al., 2015). However, almost all studies
have included adult facial expressions when examining FER in
children. This may make sense since children could be more com-
petent in evaluating adult faces, based on their dependency on
adults for their basics needs. However, solely including adult faces
is problematic because of the repeated finding that children per-
ceive and recognize emotional expressions from adults differently
compared to emotional expressions from same-aged peers (e.g.,
Hills & Lewis, 2011; Rhodes & Anastasi, 2012). Moreover, from
the age of six, most children will spend more time with peers
outside their family (Rubin et al., 2011). During early adolescence,
the importance of the peer group increases even further (Lansford
et al., 2009). Therefore, using adult stimuli in child studies might
lead to inaccurate representations of FER in children. Fortunately,
there are a few studies that included child facial stimuli to study
emotional development in children; however, these stimuli were
mostly only validated by adults (e.g., Dalrymple et al., 2014;
Marusak et al., 2013). To draw valid and reliable conclusions from
studies involving facial expression paradigms in children, research-
ers need access to well-controlled and validated child facial stimuli
that were validated by children. Therefore, the aim of the present
study was to validate a child face database by children.
Currently, the authors are aware of the existence of five child face
databases, including the NIMH Children Emotional Faces Picture Set
(Egger et al., 2011), the Dartmouth Database of Children’s Faces
(Dalrymple et al., 2013), the Child Affective Facial Expression
(CAFE) set (LoBue & Thrasher, 2015), the Child Emotion Picture
Set (Romani-Sponchiado et al., 2015), and the Radboud Faces Data-
base (RaFD; Langner et al., 2010). Of these databases, the only child
face database that has been validated by both adults (e.g., LoBue &
Trasher, 2015) and (preschool aged) children is the CAFEset (LoBue
et al., 2018). They found that adults differ in recognition accuracy
from preschoolers, with preschoolers still being able to recognize
facial expressions above chance level. Thereby, this study suggested
that the faces of the CAFE set could be applied to child research.
However, this database was only validated in children aged between
3 years and 4 years, covering only a small target group in child
research. This is problematic because child FER research is largely
focused on school-aged children who are thought to have better FER
abilities than preschool children (e.g., Chronaki et al., 2014). There-
fore, the current study focuses on validating a child faces database,
namely the RaFD, in 8- to 12-year-old children.
The RaFD (Langner et al., 2010) is a freely available database in
which specific image characteristics are systematically varied,
1
Radboud University, The Netherlands
2
Macquarie University, Australia
Corresponding author:
Geraly Bijsterbosch, Department of Clinical Psychology, Behavioural
Science Institute, Radboud University, Montessorilaan 3, 6525 HR
Nijmegen, The Netherlands.
Email: g.bijsterbosch@psych.ru.nl
International Journal of
Behavioral Development
1–7
ªThe Author(s) 2020
Article reuse guidelines:
sagepub.com/journals-permissions
DOI: 10.1177/0165025420935631
journals.sagepub.com/home/jbd
including eight different facial expressions. It consists of an adult
part, with Caucasian and Moroccan models, and a child part. The
Radboud Faces Database child models (RaFD-C) consists of 1,200
pictures of 4 boys and 6 girls, all of them Caucasian (10 models 8
emotions 3 gaze directions 5 camera angles). The models were
trained to express the emotions “happiness,” “sadness,” “fear,”
“surprise,” “disgust,” “anger,” “contempt,” and a “neutral” facial
expression, according to the Facial Action Coding System (Ekman
et al., 2002, see Langner et al. (2010) for the procedure). The RaFD
extends other databases, like FACES and the Karolinska Directed
Emotional Faces database (KDEF), by including the emotion con-
tempt. Although findings with regard to contempt are less clear and
received less attention in research so far, this emotion continues to
raise scientific interest in (child FER) research (e.g., Chaplin &
Aldao, 2013; Fischer & Giner-Sorolla, 2016). Moreover, the pic-
tures are highly standardized, in that models do not wear glasses or
makeup, or have facial hair. Langner et al. (2010) concluded that
the RaFD can be considered an adequate tool for research using
facial stimuli. Specifically, adult participants could accurately iden-
tify the expressed emotions in 82%of the cases, for both child faces
and adult faces. For the child faces, it was found that adult partici-
pants best recognized the emotion “happiness,” followed by
“surprise,” whereas “contempt” was recognized worst. Addition-
ally, recent research showed that school-aged children could accu-
rately recognize 72%of the adult expressions (Verpaalen et al.,
2019), but the child expressions have not been evaluated by chil-
dren yet. The highly standardized features and qualities of the
RaFD, together with its popularity (more than 1,100 citations since
its availability; Rothermund & Koole, 2018) emphasize the value of
choosing the RaFD-C as a nominee for validation by children as
well.
The aim of the present study was to validate the RaFD-C in a
large sample of school-aged children (aged 8–12). As the primary
validation measure, we assessed the degree of agreement between
the intended and chosen emotions. Overall, we expected the chil-
dren to accurately recognize the expressed emotions, such that
the agreement rate between the intended expressed emotion and
the emotion chosen by the children would be above chance level.
The chance level, for this specific choice task with nine (answer)
options, is 11%. Hereafter, this agreement will be referred to as
accuracy, to indicate how accurately children identified the dis-
played facial emotion expressions. Particularly, we expected that
the emotion “happiness” would be recognized best compared to all
other emotions, and “contempt” worst (Langner et al., 2010; Lawr-
ence et al., 2015; LoBue et al., 2018). Finally, in order to assess the
quality of the database and to provide researchers with overall
ratings of the models, associated validation measures (model attrac-
tiveness, clarity, and valence of the expressed emotions) were
examined (following Langner et al., 2010). Since clarity reflects
whether an emotion is expressed well, and valence best represents
the emotional experience of the emotion (Shuman, Sander, &
Scherer, 2013), these stimulus features will support researchers in
their stimulus selection.
Methods
Recruitment and Participants
A total of 547 children (256 boys) aged between 8 years and
12 years (M¼9.9, SD ¼1.2) participated in the study. Children
were recruited via nine elementary schools, representing both urban
and rural regions of The Netherlands.
1
Schools were recruited via
the personal networks of the involved authors and were selected
based on their availability to participate in the study. The schools
received an information letter about the study and received a phone
call afterward to provide the principal and participating teachers the
opportunity to ask questions about the study. If the school agreed to
participate, an informed consent form was signed. The researchers
supplied the schools with information letters and informed consent
forms for parents and the children. Children were asked to indicate
at the beginning of the session if they wanted to participate. Both
parents and children of at least 12 years old were asked to sign
active informed consent. The study was approved by the Ethics
Committee of the Faculty of Social Sciences of the Radboud Uni-
versity, the Netherlands.
Materials
RaFD-C validation task. The RaFD-C (Langner et al., 2010)
includes pictures of 10 different Caucasian (i.e., “white-colored
skin”) Dutch children (4 male and 6 female models) who vary in
hair and eye color. Each child displays eight different emotions
(anger, contempt, disgust, fear, happiness, neutrality, sadness, sur-
prise; see also Langner et al., 2010). The included child models are
aged between 7 years and 12 years, with the age of one child model
being unknown. For the current study, 80 straight-gaze, frontal
(90) camera, pictures of the children were used. To not burden the
participants too much, each child rated only a subset of the pictures.
In total, 10 versions of a slideshow were created, containing a
subset of the RaFD-C pictures with the constraint that each child
model was included in four different slideshow versions. The first
block of each slideshow started with 10 pictures of child models
displaying a “neutral” expression. In order to rate the attractiveness
of the models and to familiarize the children with the procedure,
they were asked to indicate for each of the 10 “neutral” faces: “How
attractive do you find this child?” (5-point Likert scale [1 ¼not at
all to 5 ¼very attractive]).
During the second block, a subset of 32 pictures was presented,
in which four different models expressed each of the eight emotions
in a randomized order. The expression rating was forced-choice
based. Children were instructed to answer three questions per pic-
ture: (1) “Which emotion does this person express?” (happiness,
anger, fear, sadness, surprise, disgust, neutral, contempt, other); (2)
“How clear do you find this emotion?” (5-point Likert scale [1 ¼
not at all to 5 ¼very clear]); and (3) “How positive do you find this
emotion?” (5-point Likert scale [1 ¼not at all to 5 ¼very
positive]).
Procedure
The study was conducted at the participating schools during school
hours in the children’s regular classroom environment. The task and
questionnaires were completed using pen and paper under close
supervision. Before receiving instructions, the children were
reminded about their right to stop at any time and/or ask questions.
Researchers provided the children with the definitions of the emo-
tions, as well as the specific questions and answering scales that were
used. If necessary, the explanations were repeated, also during the
tasks. Children were randomly assigned to one of the 10 slideshow
versions, based on the date of testing. All children in one classroom
received the same version. Children were seated so that they were
2International Journal of Behavioral Development
able to clearly see the RaFD-C pictures, presented on a big screen in
front of the class, and could not see each other’s answers. The pic-
tures were presented one by one, with four blocks of eight pictures.
Each picture was presented until the last child finished all three
corresponding questions in the paper booklet. The order of the choice
options for the expressed emotions was kept constant.
The current study was part of a larger project on childhood
anxiety. As a result, children also completed other questionnaires
and tasks (see Baartmans et al., 2019; Klein et al., 2018). The
completion of the RaFD-C validation task took approximately
30–45 min and the total testing time was approximately 60 min.
Data Preparation and Analyses
Calculations of accuracy rates, unbiased hit rates, and arcsin-
transformed scores. First, raw hit rates were calculated separately
for each emotion category. Raw hit rates are mean percentages of
correct responses and indicate to which degree the chosen emotion
is in agreement with the intended emotional expression. A formula
developed by Wagner (1993) was used to create unbiased hit rates
for tasks that use multiple choice answer formats in FER tasks (see
also Elfenbein & Ambady, 2002). This formula corrects for possi-
ble answer habits, for example, if participants answer with one
specific emotion category for all expressions. In line with Langner
et al. (2010), the unbiased hit rates were calculated in two steps:
First, a choice matrix was created with chosen emotion expressions
as columns and intended emotional expressions as rows. Second,
the number of ratings in each cell was squared and divided by the
product of the corresponding row and column marginal values
(Wagner, 1993). Finally, to correct for skewed variances of decimal
fractions obtained from counts (Fernandez, 1992), unbiased hit
rates were calculated and arcsin-transformed (recommended by
Bromiley & Tacker, 2002; Bishara & Hittner, 2012; Winer,
1971). Arcsin-transformations are commonly used for proportional
data (Bishara & Hittner, 2012), and they are a commonly applied
“variance stabilizing transformers” for binominal distributions,
especially with a large sample size (Bromiley & Tacker, 2002, p. 4).
Unusual response patterns. Next, the data were checked for
unusual response patterns. Specifically, we checked if participants
answered with a single emotion category, which cannot be cor-
rected for by the used formula to correct for possible answer habits.
None of the participants showed such an unusual response pattern.
Missing data. In total, 54 children did not report their age and/or
gender. In the validation task, at the item level, 148 data points were
missing (0.8%of the data points), 127 (0.7%) clarity ratings were
missing, as were 190 (1.1%) data points in the positivity ratings.
Only one subject was missing entirely. The answers of the partici-
pants were scored as correct (1) or incorrect (0). In total, each
participant rated 80 pictures; every emotion (expressed by a differ-
ent model) was shown 4 times. If the participant rated all four
expressions of a specific emotion correctly, the unbiased score was
1.00 (4/4; based on the formula developed described in the data
preparation section and recommended by Bromiley & Tacker,
2002; Bishara & Hittner, 2012; Wagner, 1993; Winer, 1971). If
there was a missing value, and the participant, therefore, rated a
specific emotion only 3 times, this was taken into account while
computing the unbiased score for that emotion. In this case, the
maximum score for the participant was 0.75 (if all three were cor-
rect), that is, the missing data point was treated as an incorrect
response. The data used for the final analyses (i.e., arcsine trans-
formed scores) did not have missing data points.
Analyses. To examine the overall accuracy across the different
emotions, repeated-measures analyses of variance (ANOVAs) were
computed in SPSS 21 (IBM Corp, 2012). Separate ANOVAs were
computed for clarity and valence with emotion (happiness, surprise,
neutral, sadness, anger, disgust, fear, and contempt) as a within-
subject factor. Intraclass correlation coefficients (ICCs) were cal-
culated to determine the interrater reliability between the children’s
accuracy rates on valence, clarity, and model attractiveness.
Results
Accuracy Rates of the Expressed Emotions
Table 1 presents an overview of the means and standard deviations
of all accuracy measures (i.e., raw hit rates, unbiased hit rates,
arcsin-transformed hit rates) separately for each expression.
2
For
an overview of all choice rates and significant differences from
chance level, see Table 2. To assess accuracy, a repeated-
measures ANOVA was computed on the arcsin-transformed
unbiased hit rates with emotion (happiness, surprise, neutral, sad-
ness, anger, disgust, fear, and contempt) as a within-subject factor.
As the assumption of sphericity was violated, w
2
(27, N¼547) ¼
186.62, p< .001, the Huynh–Feldt correction was applied.
3
The
repeated-measures ANOVA yielded a significant within-subject
effect of emotion, F(6.40, 3,496.23) ¼234.10, p< .001, h2
p¼
.30, suggesting that particular emotions were significantly better
recognized than others.
4
To test which specific emotions were bet-
ter recognized than others, we calculated associated deviation
Table 1. Mean Hit Rates per Intended Emotion Category, Analyzed per Picture.
Intended emotion
Range
Type of hit rate
Happiness Surprise Anger Neutral Sadness Disgust Fear Contempt
M(SD)M(SD)M(SD)M(SD)M(SD)M(SD)M(SD)M(SD) Min–max
Raw hit rates (%) 92.34 (6) 64.36 (23) 77.31 (6) 80.54 (8) 75.35 (20) 68.34 (12) 62.89 (15) 39.30 (14) 0.00–100.00
Unbiased hit rates 0.88 (0.2) 0.56 (0.3) 0.64 (0.3) 0.64 (0.3) 0.62 (0.3) 0.56 (0.3) 0.49 (0.3) 0.32 (0.3) 0.00–1.00
Arcsin-transformed unbiased hit rates 1.29 (0.4) 0.67 (0.4) 0.79 (0.4) 0.81 (0.5) 0.78 (0.5) 0.68 (0.5) 0.57 (0.4) 0.38 (0.5) 0.00–1.57
Note. Mean hit rates are presented for all participants (N¼547). The variable range for all different intended emotions and different rates is included in the table and is
the same for all emotions per type of hit rate.
Bijsterbosch et al. 3
contrasts from the grand mean accuracy rate (M
grand
¼.75). Taking
the high number of ratings per emotion into account, only contrasts
with an effect size of h2
p.10 are reported (in line with Langner
et al., 2010). “Happiness” was better recognized (M¼1.29) than
the grand mean accuracy rate between all intended and chosen
emotions F(1,546) ¼1,258.73, p< .001, h2
p¼.70. On the contrary,
“contempt” was significantly worse recognized (M¼.38) than the
grand mean accuracy rate, F(1,546) ¼73.24, p¼.482, h2
p¼.48.
To compare the different emotions with each other, we com-
puted Bonferroni-corrected pairwise comparisons. These compar-
isons indicated that “happiness” was significantly better recognized
than all other emotions (p< .001). Next, the emotions “neutral,”
“sadness,” and “anger” received the second-highest overall accu-
racy rates and differed significantly from the emotions with lower
accuracy rates (p< .001). Accuracy for “disgust” and “surprise”
was significantly higher than for “fear” and “contempt” (ps < .001).
Finally, “fear” was found to have a significantly higher accuracy
rate than “contempt” (p< .001).
Interestingly, the results indicated a pattern of “emotion con-
fusions” for specific emotions, which means that children often
misidentified some of the intended emotions with a specific other
emotion above the chance level of 11%(in this forced-choice task
with nine answer options). The rationale for these patterns was that
it might be possible that participants may choose particular emo-
tions more frequently than other emotions when they are not sure
of their response. One-sample t-tests showed three significant
emotion confusions: Children significantly misidentified the
indented expression of “fear” as “surprise” (p¼.023), and they
significantly misidentified “contempt” as “neutral” (p¼.008) or
“other” (p¼.002).
Associated Validation Measures
Clarity and valence ratings of the pictures. Next, mean clarity and
valence ratings were computed (see Table 3). The ANOVA for
clarity revealed a significant main effect of emotion for clarity
ratings, F(7,72) ¼13.94, p< .001, h2
p¼.58. This indicates that
some emotions were rated as significantly clearer than other emo-
tions. Bonferroni-corrected pairwise comparisons revealed that
“happiness” was rated as significantly more clear than all other
emotions (all ps<.01)and“contempt”wasratedaslessclear
compared to all other emotions (all ps <.001). Moreover, “sadness”
was rated as significantly less clear as “surprised” (p¼.046). All
other emotions did not differ significantly from each other. The
second ANOVA, with regard to the emotional valence of the pic-
tures, revealed that some emotions were rated as significantly more
positive than other emotions, F(7,72) ¼177.17, p< .001, h2
p¼.95.
Bonferroni-corrected pairwise comparisons showed that
“happiness” was rated as significantly more positive than all other
emotions (all ps < .001). Moreover, “neutral” and “surprise” were
rated as significantly more positive than “contempt,” “fear,”
“disgust,” “sadness,” and “anger” (for both “neutral” and “surprise”
all ps < .001). Next, the valence rates for “fear” were significantly
higher than for “sadness” (p¼.001) and “anger” (p< .001), but not
for “disgust.” However, “disgust” was found to have a significantly
higher valence rate than “anger” (p< .001).
Interrater reliability. To objectively select the appropriate ICCs, the
model fit (deviance information criterion) was calculated for all
possible ICC models. The model with the best fit to our data was
the ICC 2, applying a two-way cross-classified multilevel model.
5
The corresponding ICCs (2, x) for valence, clarity, and model
attractiveness are reported in Table 4. Generally, the calculated
ICCs confirm the expected high agreement of the raters for clarity,
valence, and model attractiveness.
Discussion
The goal of the present study was to validate the RaFD-C by chil-
dren. As the primary validation measure, we investigated to which
Table 2. Choice Percentages per Intended and Chosen Emotion Category.
Chosen emotion
Intended emotion Happiness Anger Neutral Surprise Sadness Disgust Fear Contempt Other
Happiness 92.34 0.42 2.54 0.59 0.09 0.72 0.57 1.37 1.36
Surprise 0.48 77.31 1.53 0.46 0.20 0.94 14.79 0.89 3.40
Neutral 1.00 3.28 80.54 1.03 1.75 1.86 1.11 3.34 6.23
Sadness 0.31 6.14 2.53 64.36 2.35 7.16 9.64 3.49 4.13
Anger 0.31 5.65 1.29 1.70 75.35 4.42 1.83 4.58 4.70
Disgust 0.43 6.03 1.11 0.68 11.53 68.35 0.77 5.73 5.47
Fear 1.01 16.66* 0.60 1.36 4.84 7.26 62.89 1.07 4.29
Contempt 2.73 8.29 22.48** 1.08 0.92 3.93 0.96 39.30 20.17**
Note. Choice percentages are presented for all participants (N¼547). The cells contain raw hit percentages, hits are marked in light gray. Contoured cells, noted with
asterisks, contain percentages of emotion confusions above chance level (*p< .05, **p< .01, ***p< .001).
Table 3. Mean Clarity and Valence Ratings of the Pictures.
Intended emotion
Clarity Valence
M(SD) Min–max M(SD) Min–max
Happiness 4.14 (0.5) 3.25–4.50 4.05 (0.2) 3.61–4.27
Sadness 3.38 (0.3) 2.91–3.78 2.37 (0.1) 2.23–2.68
Surprise 3.69 (0.1) 3.57–3.85 2.94 (0.1) 2.77–3.11
Neutral 3.41 (0.3) 2.85–3.85 3.03 (0.1) 2.79–3.19
Anger 3.42 (0.5) 2.52–3.91 2.26 (0.2) 2.10–2.52
Disgust 3.39 (0.2) 3.08–3.81 2.47 (0.0) 2.36–2.52
Fear 3.54 (0.4) 2.75–3.89 2.57 (0.1) 2.45–2.65
Contempt 2.72 (0.2) 2.38–2.99 2.71 (0.1) 2.46–3.00
Note. Mean clarity and valence ratings are presented from all participants
(N¼547) for the different pictures (N¼80). Each mean and standard deviation,
for the separate emotions, is based on clarity and the valence ratings of
10 pictures (n¼10). The variables ranged between 1 and 5.
4International Journal of Behavioral Development
degree children aged 8–12 years labeled the emotions in agreement
with the intended emotions. Finally, clarity and valence of the
emotions, and model attractiveness were investigated. This study
establishes childhood norms for the recognition of seven emotions
and a neutral expression using the RaFD-C.
In the current study, we observed that children were able to
correctly identify, on average, 70%of the emotions. Overall, this
means that the children were relatively accurate in recognizing the
emotions from the child models. The agreement rate in adults rating
the RaFD-C was approximately 82%(Langner et al., 2010), and
72%in children who rated pictures of RaFD adult expressions
(Verpaalen et al., 2019). In line with our hypotheses and consistent
with other studies, we found that “happiness” was recognized best,
whereas “contempt” was least accurately identified. Specific dif-
ferences between emotions were found in the following rank order:
“Happiness” accuracy rates were higher than all other emotions,
followed by “neutral,” “sadness,” and “anger,” next by “disgust”
and “surprise,” then by “fear,” and finally “contempt” accuracy
rates were lower than all others.
These results roughly follow the pattern found in the RaFD
validation study by adults (Langner et al., 2010), suggesting con-
tinuity in the development of FER. They also concur with the
results of Verpaalen et al. (2019) and the FER-study in 6- to
15-year-olds by Lawrence et al. (2015), who investigated the devel-
opment of emotion recognition in adult faces through childhood
and adolescence. Also, the child faces validation study in preschool
children by LoBue et al. (2018) approximately found the same
emotion accuracy patterns as in the current study. For a comparison
of the current results with the agreement rates of the validation
study in adults (Langner et al., 2010) and the validation study in
preschool children (LoBue et al., 2018), see Table 5. Moreover, our
results replicate previous findings that “happiness” is easiest to
recognize for children (e.g., Lawrence et al., 2015). The finding
that the emotions “surprise,” “disgust,” and “fear” were found to be
disproportionally difficult to recognize resonates with earlier stud-
ies as well (Gagnon et al., 2014; Rozin et al., 2005). It is likely that
these complex emotions are generally more difficult to interpret,
more easily confused with other emotions, and mature significantly
through the course of late childhood into adolescence (Gagnon
et al., 2010; Lawrence et al., 2015). The similarity with other
research using child participants suggests that the observed differ-
ences between the expressions are not likely to be a specific char-
acteristic of the RaFD-C.
The current study had several strengths. First, this study is the
first of its kind to have a child faces database validated by school-
aged children (8–12 years), including an assessment of all six basic
emotions, plus the emotion “contempt” and a “neutral” expression.
This provides researchers with a validated child faces database,
allowing them to draw valid and reliable conclusions from studies
involving facial expression paradigms in children. The current find-
ings can be generalized to other same-aged Caucasian child faces.
However, limitations of the current study need to be mentioned as
well. First, conducting the current study in a class-wise manner
implied some practical constraints. Fast-responding children had
to wait before the next picture was presented. This might have led
to disinterest and consequently, interrupted attention patterns in
these children (Malkovsky et al., 2012). Besides, children who
needed more time to answer the questions might have felt pressured
or hurried because other children had to wait for them to finish.
Both situations might have negatively biased the FER accuracy
rates of the current study; either lacking attention or time pressure
may have led children to select other answers than they would have
chosen if they had followed their own pace. Using a forced-choice
task could possibly rely on the development of executive function-
ing, meaning that a process of elimination could have been used to
deduce which emotion was presented. A review on developmental
studies regarding deductive reasoning stated that that deductive
reasoning is a characteristic of the concrete operational period,
between 7 years and 11 years (Jansson, 1974). However, unbiased
hit rates and the option other in the answering format were used to
minimize this bias possibility. Adding other to the answering for-
mat counteracts the bias in forced-choice FER paradigms (Frank &
Stennet, 2001). Next to this, it could be that the used method to
handle the missing data inherently also has led to biased results.
However, the number of missing data points is very small, so is
unlikely to have led to biased estimates, and we are being conser-
vative using this method. Finally, it would have been valuable to
collect more detailed information about the race/ethnicity and
social economic status of our participants to be able to explore
possible differences in FER between subgroups of our sample.
Moreover, it was not registered why children did not participate
in the current study. This would be valuable to determine the repre-
sentativeness of the sample but also to control for possible
Table 4. ICCs for Clarity and Valence of the Emotions, and Model
Attractiveness.
Dimension ICC (2,1) ICC (2, k¼547)
ICC (2, kadjusted
for missings)
Clarity 0.14 0.99 0.97
Valence 0.31 0.96 0.99
Attractiveness 0.01 0.85 0.85
Note. ICCs are presented for all participants (N¼547). ICC (2, 1) ¼intraclass
correlation coefficient, each subject is measured by each rater with reliability
estimated from a single measurement, ICC (2, k)¼as before, but here reliability
is calculated by imputing the average of the kraters’ measurements.
Table 5. Raw Hit Rates for the RaFD-C Child and Adult Validations, and
the CAFE Database Validated by Children.
RaFD CAFE set
Emotion
category
Child validation
(current study)
M(SD)
Adult validation
(Langner et al., 2010)
M(SD)
Child validation
(LoBue et al., 2018)
M(SD)
Happiness 92 (6) 97 (6) 78 (42)
Surprise 64 (23) 91 (25) 61 (49)
Anger 77 (6) 89 (8) 65 (48)
Neutral 81 (8) 84 (16) 44 (50)
Sadness 75 (20) 75 (14) 49 (50)
Disgust 68 (12) 83 (18) 47 (50)
Fear 63 (15) 79 (13) 26 (44)
Contempt 40 (14) 59 (19) NA
Note. Raw hit rates for the current validation study are presented for all parti-
cipants (N¼547) in the column on the left, and the raw hit rates for the adult
validation study by Langner et al. (2010) in the middle (N¼276), whereas the raw
hit rates for the child validation study by LoBue et al. (2018) is presented in the
column on the right (N¼58). The variable range for the raw hit rates for all three
studies is between 0 and 100. The included agreement rates are the raw hit rates
and standard deviations per emotion, the agreement rates from the child valida-
tion of the CAFE set are the scores from the full set at Time 1. RaFD ¼Radboud
Faces Database; CAFE ¼the Child Affective Facial Expression Set.
Bijsterbosch et al. 5
confounds related to participation refusal. The generalizability of
the results is an important point to be addressed in future research
because we currently cannot generalize to emotion recognition in
non-Caucasian models. It would be beneficial for future studies to
validate a child faces database representing a variety of ethnicities,
together with an ethnically diverse sample of participants.
The validation data per picture are provided freely online (see
https://osf.io/7srgw/), enabling researchers to include specific pic-
tures as stimuli and select the most appropriate items for their
research. Based on this study, some recommendations about select-
ing stimuli can be made. Specifically, we recommend to be cautious
in combining (pictures with) the emotions “fear” and “surprise,”
and the emotions “contempt” and “neutral” together in one study
since these were found to be confused with each other. Finally,
researchers who are interested in pictures that are easily recogniz-
able and do not include ambiguity of the different emotional
expressions can use the available validation data to select those
pictures that were identified with at least 60%accuracy by the
children. Although this will lead to a more limited choice in terms
of suitable pictures, this subset of 60 pictures will still enable
researchers to select pictures based on criteria of their choice.
In conclusion, the current study indicated that 8- to 12-year-old
children were able to accurately identify approximately 70%of the
RaFD-C pictures. The differences in the child FER rates compared
to the FER rates of adults emphasize the importance of a validation
study of child faces stimuli from a child’s perspective. Addition-
ally, the results from the current study indicate that subtle emotion
differences should be considered when studying FER in children, as
reflected in our recommendations above. As this is the first study
that validated the RaFD-C in children, we expect that researchers
focusing on FER in childhood will benefit from this study for
further and better understanding of emotion recognition in children.
Acknowledgments
We are grateful to the schools and the children for their participa-
tion in the study. We thank Pierre Souren for calculating the inter-
rater reliability indices and Giovanni ten Brink for his help with the
data preparation. One of the authors was supported by a Niels
Stensen Fellowship.
Funding
The author(s) received no financial support for the research, author-
ship, and/or publication of this article.
ORCID iD
Geraly Bijsterbosch https://orcid.org/0000-0003-4390-9794
Gijsbert Bijlstra https://orcid.org/0000-0002-0827-7376
Notes
1. Student populations of Dutch nonprivate elementary schools
have mixed socioeconomic characteristics and a Caucasian eth-
nic majority (Centraal Bureau voor de Statistiek, 2019). There-
fore, we can assume that the included schools in the current
study represented varied socioeconomic characteristics similar
to other Dutch elementary schools. However, not all children
from a class participated in the study. Generally, the reason for
refusal to participate was decided by the researcher on site as
this was mostly due to parents forgetting to hand in the informed
consent form within due time. Overall, approximately 121 (22.
2%) of the children were in the third grade, 145 (26.7%) in the
fourth grade, 128 (23.5%) in the fifth grade, and 150 (27.6%)in
the sixth grade.
2. Due to the importance of age for facial emotion recognition
(FER) performance (Field & Lester, 2010; Martin & Ruble,
2013) and given that researchers may want to select the best
fitting stimuli for different age groups, the current study also
investigated age group differences in FER performance. For an
overview of the means and standard deviations of the accuracy
for each emotion, separately for age, see Table S1 in the online
supplementary materials (https://osf.io/7srgw/).
3. If the data violate the sphericity assumption, corrections should
be made to produce a valid F-ratio. The used correction is based
on the estimates of sphericity (). When > .75, it is advised to
use the Huynh–Feldt correction (Field, 1998; Girden, 1992). In
the current study, it was found that ¼.915, resulting in the
valid usage of the Huynh–Feldt correction to correct for spheri-
city violations.
4. As an additional check to justify the use of arcsin-transformed
scores in the current study, the exact same repeated-measures
ANOVA was conducted with the raw hit rates as dependent
variable. This analysis yielded similar results.
5. Cross-classified multilevel Markov Chain Monte Carlo intra-
class correlation coefficients (ICCs) were calculated. This type
of ICC is uncommon, however, it was decided to use this type in
this study since this type fitted the data best.
References
Baartmans, J. M. D., Rinck, M., Hudson, J. L., Lansu, T. A. M., van
Niekerk, R. E., Bo¨gels, S. M., & Klein, A. M. (2019). Are socially
anxious children really less liked, or do they only think so? Cogni-
tive Therapy and Research,43, 1043–1050. https://doi.org/10.1007/
s10608-019-10028-9
Bishara, A. J., & Hittner, J. B. (2012). Testing the significance of a
correlation with nonnormal data: Comparison of Pearson, Spear-
man, transformation, and resampling approaches. Psychological
Methods,17(3), 399–417. https://doi.org/10.1037/a0028087
Bromiley, P. A., & Thacker, N. A. (2002). The effects of an arcsin
square root transform on a binomial distributed quantity (TINA
memo no. 2002-007). University of Manchester.
Centraal Bureau voor de Statistiek. (2019). Leerlingen in (speciaal)ba-
sisonderwijs; migratieachtergrond, woonregio. https://opendata.
cbs.nl/statline/#/CBS/nl/dataset/83295NED/table?
ts¼1572185466637
Chaplin, T. M., & Aldao, A. (2013). Gender differences in emotion
expression in children: A meta-analytic review. Psychological
Bulletin,139(4), 735–765. https://doi.org/10.1037/a0030737
Chronaki, G., Hadwin, J. A., Garner, M., Maurage, P., & Sonuga--
Barke, E. J. (2014). The development of emotion recognition from
facial expressions and non-linguistic vocalizations during child-
hood. British Journal of Developmental Psychology,33(2),
218–236. https://doi.org/10.1111/bjdp.12075
Dalrymple, K. A., Gomez, J., & Duchaine, B. (2013). The Dartmouth
database of children’s faces: Acquisition and validation of a new
face stimulus set. PLoS ONE,8(11), e79131. https://doi.org/10.13
71/journal.pone.0079131
Dalrymple, K., Garrido, L., & Duchaine, B. (2014). A dissociation
between face perception and face memory in adults, but not chil-
dren, with developmental prosopagnosia. Journal of Vision,14(10),
1434–1434. https://doi.org/10.1167/14.10.1434
6International Journal of Behavioral Development
Egger, H. L., Pine, D. S., Nelson, E., Leibenluft, E., Ernst, M., Towbin,
K. E., & Angold, A. (2011). The NIMH child emotional faces
picture set (NIMH-ChEFS): A new set of children’s facial emotion
stimuli. International Journal of Methods in Psychiatric Research,
20(3), 145–156. https://doi.org/10.1002/mpr.343
Ekman, P., Friesen, W. V., & Hager, J. C. (2002). Facial action coding
system: The manual. Research Nexus.
Elfenbein, H. A., & Ambady, N. (2002). Is there an in-group advantage
in emotion recognition? Psychological Bulletin,128(2), 243–249.
https://doi.org/10.1037/0033-2909.128.2.243
Fernandez, G. C. (1992). Residual analysis and data transformations:
Important tools in statistical analysis. HortScience,27(4), 297–300.
Field, A. P. (1998). A bluffer’s guide to ...sphericity. British Psycho-
logical Society-MSC Newsletter,6(1), 13–24. http://www.discover
ingstatistics.com/
Field, A. P., & Lester, K. J. (2010). Is there room for ‘development’ in
developmental models of information processing biases to threat in
children and adolescents? Clinical Child and Family Psychology
Review,13(4), 315–332. https://doi.org/10.1007/s10567-010-0078-8
Fischer, A., & Giner-Sorolla, R. (2016). Contempt: Derogating others
while keeping calm. Emotion Review,8(4), 346–357. https://doi.
org/10.1177/1754073915610439
Frank, M. G., & Stennett, J. (2001). The forced-choice paradigm and
the perception of facial expressions of emotion. Journal of Person-
ality and Social Psychology,80(1), 75–85. https://doi.org/10.1037/
0022-3514.80.1.75
Gagnon, M., Gosselin, P., Hudon-ven der Buhs, I., Larocque, K., &
Milliard, K. (2010). Children’s recognition and discrimination of
fear and disgust facial expressions. Journal of Nonverbal Behavior,
34(1), 27–42. https://doi.org/10.1007/s10919-009-0076-z
Gagnon, M., Gosselin, P., & Maassarani, R. (2014). Children’s ability
to recognize emotions from partial and complete facial expressions.
The Journal of Genetic Psychology,175(5), 416–430. https://doi.
org/10.1080/00221325.2014.941322
Girden, E. R. (1992). ANOVA: Repeated measures (pp. 7–84). Sage
University Press Series on Quantitative Applications in the Social
Sciences.
Hills, P. J., & Lewis, M. B. (2011). Rapid communication: The own-age
face recognition bias in children and adults. Quarterly Journal of
Experimental Psychology,64(1), 17–23. https://doi.org/10.1080/
17470218.2010.537926
IBM Corp. (2012). IBM Corp statistics for windows (Version 21.0).
Jansson, L. C. (1974). The development of deductive reasoning: A
review of the literature. Preliminary version.
Klein, A. M., Bakens, R., van Niekerk, R. E., Ouwens, M. A., Rapee, R.
M., B, S. M., o¨ , E. S., gels, M., Becker, O., & Rinck, R. (2018).
Children with generalized anxiety disorder symptoms show a
content-specific interpretation bias using auditory stimuli. Journal
of Behavior Therapy and Experimental Psychiatry,61, 121–127.
https://doi.org/10.1016/j.jbtep.2018.06.011
Langner, G., Dotsch, D. H. J., Bijlstra, S. T., Wigboldus, A., Hawk, J. E.,
& van Knippenberg, L. A. (2010). Presentation and validation of the
Radboud Faces Database. Cognition & Emotion,24(8), 1377–1388.
https://doi.org/10.1080/02699930903485076
Lansford, S., Killeya-Jones, P. R., Miller, K., & Costanzo, R. (2009).
Early adolescents’ social standing in peer groups: Behavioral cor-
relates of stability and change. Journal of Youth and Adolescence,
38(8), 1084–1095. https://doi.org/10.1007/s10964-009-9410-3
Lawrence, D., Campbell, J. M., & Skuse, J. K. (2015). Age, gender, and
puberty influence the development of facial emotion recognition.
Frontiers in Psychology,6(761). https://doi.org/10.3389/fpsyg.
2015.00761
Leppa¨nen, J. M., & Hietanen, M. (2001). Emotion recognition and
social adjustment in school–aged girls and boys. Scandinavian
JournalofPsychology,42(5), 429–435. https://doi.org/10.1111/
1467-9450.00255
LoBue, C., Baker, V., & Thrasher, C. (2018). Through the eyes of a
child: Preschoolers’ identification of emotional expressions from
the child affective facial expression (CAFE) set. Cognition and
Emotion,32(5), 1122–1130. https://doi.org/10.1080/02699931.
2017.1365046
LoBue, E., & Thrasher, C. (2015). The child affective facial expression
(CAFE) set: Validity and reliability from untrained adults. Frontiers
in Psychology,5, 1532. https://doi.org/10.3389/fpsyg.2014.01532
Malkovsky, Y., Merrifield, J., Goldberg, C. L., & Danckert, D. N.
(2012). Exploring the relationship between boredom and sustained
attention. Experimental Brain Research,221(1), 59–67. https://doi.
org/10.1007/s00221-012-3147-z
Martin, H. A., & Ruble, J. M. (2013). Patterns of gender development.
Annual Review of Psychology,61, 353–381. https://doi.org/10.
1146/annurev.psych.093008.100511
Marusak, M. E., Carr´e, M. G., & Thomason, J. S. (2013). The stimuli
drive the response: An fMRI study of youth processing adult or child
emotional face stimuli. NeuroImage,83, 679–689. https://doi.org/10.
1016/j.neuroimage.2013.07.002
Rhodes, A., & Anastasi, B. (2012). The own-age bias in face recogni-
tion: A meta-analytic and theoretical review. Psychological Bulle-
tin,138(1), 146–174. https://doi.org/10.1037/a0025750
Romani-Sponchiado, C., Sanvicente-Vieira, D., Mottin, A.,
Hertzog-Fonini, K., & Arteche, S. L. (2015). Child emotions
picture set (CEPS): Development of a database of children’s
emotional expressions. Psychology & Neuroscience,8(4),
467–478. https://doi.org/10.1037/h0101430
Rothermund, P., & Koole, C. (2018). Three decades of cognition &
emotion: A brief review of past highlights and future prospects.
Cognition and Emotion,32(1), 1–12. https://doi.org/10.1080/
02699931.2018.1418197
Rozin, L., Taylor, G., Ross, A., Bennett, K. H., & Hejmadi, W. M.
(2005). General and specific abilities to recognise negative emo-
tions, especially disgust, as portrayed in the face and the body.
Cognition & Emotion,19(3), 397–412. https://doi.org/10.1080/
02699930441000166
Rubin, B., Bukowski, I. A. M., & Laursen, G. (2011). Handbook of peer
interactions, relationships, and groups (1st ed.). Guilford Press.
Shuman, V., Sander, D., & Scherer, K. L. (2013). Levels of valence.
Frontiers in Psychology,4. https://doi.org/10.3389/fpsyg.2013.
00261
Verpaalen, L., Bijsterbosch, G., Mobach, M., Bijlstra, A., Rinck, H. L.,
& Klein, B. (2019). Validating the Radboud Faces Database from a
child’s perspective. Cognition and Emotion. https://doi.org/10.
1080/02699931.2019.1577220
Wagner (1993). On measuring performance in category judgment stud-
ies of nonverbal behavior. Journal of Nonverbal Behavior,17(1),
3–28. https://doi.org/10.1007/BF00987006
Winer (1971). Statistical principles in experimental design (2nd ed.).
McGraw-Hill.
Bijsterbosch et al. 7
... Validation was determined by the 'matching accuracy' variable, which is the matching percentage of the number of photos that were correctly recognized by the instrument for the predicted emotional expressions of prototypical faces for each of the 20 images [31][32][33]. Therefore, matching accuracy can be regarded as the percentage for correct matching. ...
... The 20 most relevant studies with specific sources of data (either from figure, words, or tables) were documented for data extraction. The criteria when screening for usage was adapted as the mean of 20 studies (Table 1) [16,[30][31][32][33][34][35][36][37][38][39][40][41][42][43][44][45][46][47][48]. The specific process is shown in Fig 2. Only expressions that passed the criteria were used for further assessment. ...
... The matching accuracy in these three cases were around 30%, which was much lower than the recognition of other expressions. In addition, two cases of accuracies were achieved through artificial rating [31,48], and only one case published a matching accuracy that was given by an instrument, E-Prime [41]. Therefore, further improvement is needed for FireFACE to recognize the contempt expression because scores were too low for accurate determination. ...
Article
Full-text available
An outcome of building sustainable urban forests is that people's well-being is improved when they are exposed to trees. Facial expressions directly represents one's inner emotions , and can be used to assess real-time perception. The emergence and change in the facial expressions of forest visitors are an implicit process. As such, the reserved character of Asians requires an instrument rating to accurately recognize expressions. In this study, a dataset was established with 2,886 randomly photographed faces from visitors at a constructed urban forest park and at a promenade during summertime in Shenyang City, Northeast China. Six experts were invited to choose 160 photos in total with 20 images representing one of eight typical expressions: angry, contempt, disgusted, happy, neutral, sad, scared, and surprised. The FireFACE ver. 3.0 software was used to test hit-ratio validation as an accuracy measurement (ac.) to match machine-recognized photos with those identified by experts. According to the Kruskal-Wallis test on the difference from averaged scores in 20 recently published papers, contempt (ac. = 0.40%, P = 0.0038) and scared (ac. = 25.23%, P = 0.0018) expressions do not pass the validation test. Both happy and sad expression scores were higher in forests than in promenades, but there were no difference in net positive response (happy minus sad) between locations. Men had a higher happy score but lower disgusted score in forests than in promenades. Men also had a higher angry score in forests. We conclude that FireFACE can be used for analyzing facial expressions in Asian people within urban forests. Women are encouraged to visit urban forests rather than promenades to elicit more positive emotions.
... The validation was made with the detection of the variable of 'matching accuracy', which is the matching percentage of the number of photos that were correctly recognized by instrument for the most prototypical expression for each of the 20 images with the predicted emotional expressions [28][29][30]. Therefore, the validation of matching accuracy can be regarded as the percent value for correct matching. ...
... The top 20 relevant studies with specific source of data (eight from figure, words, or tables) were documented for data extraction. The standard for employment screening was adapted as the mean of 20 studies (Table 1) [14,[28][29][30][31][32][33][34][35][36][37][38][39][40][41][42][43][44][45][46]. The specific process of this part of study is shown in Figure 3. ...
... The matching accuracy in these three cases were around 30%, which was much lower than the simultaneous recognition of other expressions. In addition, two cases of accuracies were achieved by artificial rating [28,46], and only one case published the matching accuracy given by an instrument of E-Prime [39]. Therefore, further improvement is needed of contempt expression for use by FireFACE because the scores were too low for accurate determination. ...
Preprint
Full-text available
The construction of sustainable urban forests follows the principle that well-being in people is promoted when exposed to tree population. Facial expression is the direct representation of inner emotion that can be used to assess real-time perception in urban forests. The emergence and change of facial expressions for forest visitors are in an implicit process. The reserved character of oriental races strengthens the requirement for the accuracy to recognize expressions through instrument rating. In this study, a dataset was established with 2,886 randomly photographed faces from visitors in a constructed urban forest park and a promenade at summertime in Shenyang City at Northeast China. Six experts were invited to choose 160 photos in total with 20 images representing one of eight typical expressions as angry, contempt, disgusted, happy, neutral, sad, scared, and surprised emotions. The FireFACE ver. 3.0 software was used to test hit-ratio validation as the accuracy (ac.) to match machine-recognized photos with those identified by experts. According to Kruskal-Wallis test on the difference from averaged scores in 20 recently published papers, contempt (ac.=0.40%, P=0.0038) and scared (ac.=25.23%, P=0.0018) expressions cannot pass the validation test. Therefore, the FireFACE can be used as an instrument to analyze facial expression from oriental people in urban forests but contempt and scared expressions cannot be identified.
... To this end, children with SAD (SAD group) were compared to healthy control children (NONANX group) and to children with other, nonsocial anxiety disorders (OTHANX group). For all children, the accuracy of identifying emotions and the responses to ambiguous, validated (Bijsterbosch et al., 2021), emotional child faces of varying intensities (10%, 30%, 50%, 70%) was studied for three relevant emotions (happy, angry, fearful). ...
... Leeuw, 2015). Stimuli were derived from the validated Radboud Face Database-Children (RAFD-C; Bijsterbosch et al., 2021;Langner et al., 2010). In total, 8 (Caucasian) child models were chosen based on highest agreement rates. ...
Article
Objective: The current study examined whether children with a social anxiety disorder (SAD) demonstrate divergent facial emotion processing and a disorder-specific negative interpretation bias in the processing of facial emotional expressions. This study aimed to overcome previous study limitations by including both a nonsocially anxious control group and a healthy control group to examine whether childhood SAD is characterized by a general emotion labeling deficit, and/or by a negative interpretation bias, indicated by systematic misclassifications, or a lower threshold for recognizing threatening emotions. Method: Participants were 132 children aged 7-12 years (Mage = 9.15; 45.5% female). Children with SAD (n = 42) were compared to children with other, nonsocial, anxiety disorders (n = 40) and healthy control children (n = 50) on a novel facial emotion recognition task. Children judged ambiguous happy/neutral, angry/neutral and fear/neutral faces that were morphed at different intensities (10%, 30%, 50%, 70%). Results: Children with SAD did not differ from other groups in their accuracy of identifying emotions. They did not show systematic misclassifications or a heightened sensitivity to negative, threatening faces either. Rather, children with nonsocial anxiety disorders showed a generally heightened sensitivity to emotional faces. Conclusions: The current study does not provide evidence for a general deficit in labeling of emotional faces in childhood SAD. Childhood SAD was not characterized by an interpretation bias in processing emotional faces. Children with nonsocial anxiety disorders may benefit from assistance in accurately interpreting the degree of emotionality in interpersonal situations.
... As emotional face stimuli, we used images from the Radboud Faces Database (RaFD; Langner et al., 2010). These stimuli are well-validated (Bijsterbosch et al., 2021;Langner et al., 2010;Mishra et al., 2018;Verpaalen et al., 2019). Moreover, the actors were instructed to express the emotions using the "Facial Action Coding System" (FACS; Ekman et al., 2002), which is based on the contraction and relaxation of emotion-specific facial muscles. ...
Article
Full-text available
Face masks are now worn frequently to reduce the spreading of the SARS-CoV-2 virus. Their health benefits are undisputable, but covering the lower half of one's face also makes it harder for others to recognize facial expressions of emotions. Three experiments were conducted to determine how strongly the recognition of different facial expressions is impaired by masks, and which emotions are confused with each other. In each experiment, participants had to recognize facial expressions of happiness, sadness, anger, surprise, fear, and disgust, as well as a neutral expression, displayed by male and female actors of the Radboud Faces Database. On half of the 168 trials, the lower part of the face was covered by a face mask. In all experiments, facial emotion recognition (FER) was about 20% worse for masked faces than for unmasked ones (68% correct vs. 88%). The impairment was largest for disgust, followed by fear, surprise, sadness, and happiness. It was not significant for anger and the neutral expression. As predicted, participants frequently confused emotions that share activation of the visible muscles in the upper half of the face. In addition, they displayed response biases in these confusions: They frequently misinterpreted disgust as anger, fear as surprise, and sadness as neutral, whereas the opposite confusions were less frequent. We conclude that face masks do indeed cause a marked impairment of FER and that a person perceived as angry, surprised, or neutral may actually be disgusted, fearful, or sad, respectively. This may lead to misunderstandings, confusions, and inadequate reactions by the perceivers.
Article
Full-text available
Studies have demonstrated that a forest experience can improve the mental state of visitors. The questionnaire methodology accounts for most of current findings, but subjective bias in self-reported scores cannot be eliminated. In this study, FireFACE V3.0 software was used to evaluate facial expressions of visitors unconsciously entering (urban experiencer) or leaving an urban forest park (forest experiencer) from 9:00 am to 5:00 pm. A total of 458 photos were tested for the hit-rate of machine-recognized facial expression scores to those identified by experts. Neutral, happy, sad, and scared expressions passed the validation test and were used for further analysis. The positive response index (PRI) was calculated as the difference between happy and sad scores. All five variables were analyzed by a mixed-model analysis of variance of walking direction (into the park vs. out of the park), gender (male vs. female), and age (senior, middle-aged, and young) on repeated records every 30 min in a daytime except noon. In the morning, pedestrians walking out of the forest park did not show more positive expressions than those leaving the park. Young visitors walking through the exit averaged a higher sadness score but a lower PRI than middle-aged visitors. In the afternoon, middle-aged visitors walking through the entrance averaged a higher PRI relative to young visitors leaving the park in the morning. Compared to urban experi-encers, forest experiencers averaged lower scores for neutral and scared expressions, but forest experiencers unlikely revealed more positive expressions. Therefore, a forest experience does not always evoke positive expression on unconscious faces which was a responsive variable to go with time. Middle-aged people are encouraged to walk in forests with a longer time than youths in the morning. Male visitors, especially those in senior ages, are suggested to spend more time in forests than females.
Article
Full-text available
The purpose of this study was to examine the relation of social anxiety with self-perceived and peer-reported likability, while controlling for the possible influence of depression. In total, 586 children (7 to 13 years) completed questionnaires to measure social anxiety, self-assessed likability, and depression. Peer-reported likability was derived from sociometric data on likability. As expected, children with higher self-reported social anxiety perceived themselves as less liked by classroom peers than children with lower self-reported social anxiety. In reality, children with higher levels of social anxiety were more liked by peers than children with lower levels of social anxiety. Multilinear regression analyses indicated no confounding effect of depression: Social anxiety, but not depression, was a significant predictor of biased perceived likability. Correcting the discrepancy between objective versus subjective likability may be a crucial target in the prevention and treatment of social anxiety disorders in children.
Article
Full-text available
Facial expressions play a central role in diverse areas of psychology. However, facial stimuli are often only validated by adults, and there are no face databases validated by school-aged children. Validation by children is important because children still develop emotion recognition skills and may have different perceptions than adults. Therefore, in this study, we validated the adult Caucasian faces of the Radboud Faces Database (RaFD) in 8- to 12-year-old children (N = 652). Additionally, children rated valence, clarity, and model attractiveness. Emotion recognition rates were relatively high (72%; compared to 82% in the original validation by adults). Recognition accuracy was highest for happiness, below average for fear and disgust, and lowest for contempt. Children showed roughly the same emotion recognition pattern as adults, but were less accurate in distinguishing similar emotions. As expected, in general, 10- to 12-year-old children had a higher emotion recognition accuracy than 8- and 9-year-olds. Overall, girls slightly outperformed boys. More nuanced differences in these gender and age effects on recognition rates were visible per emotion. The current study provides researchers with recommendation on how to use the RaFD adult pictures in child studies. Researchers can select appropriate stimuli for their research using the online available validation data.
Article
Full-text available
Researchers have been interested in the perception of human emotional expressions for decades. Importantly, most empirical work in this domain has relied on controlled stimulus sets of adults posing for various emotional expressions. Recently, the Child Affective Facial Expression (CAFE) set was introduced to the scientific community, featuring a large validated set of photographs of preschool aged children posing for seven different emotional expressions. Although the CAFE set was extensively validated using adult participants, the set was designed for use with children. It is therefore necessary to verify that adult validation applies to child performance. In the current study, we examined 3- to 4-year-olds’ identification of a subset of children’s faces in the CAFE set, and compared it to adult ratings cited in previous research. Our results demonstrate an exceptionally strong relationship between adult ratings of the CAFE photos and children’s ratings, suggesting that the adult validation of the set can be applied to preschool-aged participants. The results are discussed in terms of methodological implications for the use of the CAFE set with children, and theoretical implications for using the set to study the development of emotion perception in early childhood.
Article
Full-text available
While philosophers have discussed the emotion of contempt from antiquity to the present day, contempt has received less attention in psychological research. We review the defining features of contempt, both as a short-term emotion and as a more long-lasting sentiment. Contempt is similar to anger in that it may occur after (repeated) social or moral transgressions, but it differs from anger in its appraisals, actions, and emotivational goals. Unlike anger, contempt arises when a person’s or group’s character is appraised as bad and unresponsive to change, leading to attempts to socially exclude the target. We discuss associative, self-regulatory, and social distancing functions of contempt and present a dynamic social model of contempt versus anger.
Article
Full-text available
The recognition of facial emotions plays an important role in establishing social interactions through nonverbal communication. This is particularly true during childhood because language skills are not yet fully developed. Nevertheless, few studies have focused on facial emotion processing in children. Moreover, studies that have addressed this issue commonly have the major limitation of using only adult facial stimuli, and stimulus sets with children's faces are rare. The present study sought to fill this gap by developing the Child Emotions Picture Set (CEPS). The CEPS is a database of the facial emotions of 17 children who posed or naturally expressed 6 basic emotions. A panel of expert judges rated each picture. The set is split into 3 age groups (6 –7 years old, 8 –9 years old, and 10 –11 years old) and includes 225 photos of boys and girls who display facial expressions of happiness, sadness, anger, fear, disgust, and surprise at 3 different intensities (low, medium, and high) plus neutrality. This work contributes to the scientific field by constructing a set of child facial stimuli that is intended for use in further research on emotional processing in children. Additionally, the database may allow future studies on the mechanisms that are involved in processing facial emotions across childhood and investigating effects associated with psychiatric disorders.
Article
Full-text available
Our ability to differentiate between simple facial expressions of emotion develops between infancy and early adulthood, yet few studies have explored the developmental trajectory of emotion recognition using a single methodology across a wide age-range. We investigated the development of emotion recognition abilities through childhood and adolescence, testing the hypothesis that children's ability to recognize simple emotions is modulated by chronological age, pubertal stage and gender. In order to establish norms, we assessed 478 children aged 6–16 years, using the Ekman-Friesen Pictures of Facial Affect. We then modeled these cross-sectional data in terms of competence in accurate recognition of the six emotions studied, when the positive correlation between emotion recognition and IQ was controlled. Significant linear trends were seen in children's ability to recognize facial expressions of happiness, surprise, fear, and disgust; there was improvement with increasing age. In contrast, for sad and angry expressions there is little or no change in accuracy over the age range 6–16 years; near-adult levels of competence are established by middle-childhood. In a sampled subset, pubertal status influenced the ability to recognize facial expressions of disgust and anger; there was an increase in competence from mid to late puberty, which occurred independently of age. A small female advantage was found in the recognition of some facial expressions. The normative data provided in this study will aid clinicians and researchers in assessing the emotion recognition abilities of children and will facilitate the identification of abnormalities in a skill that is often impaired in neurodevelopmental disorders. If emotion recognition abilities are a good model with which to understand adolescent development, then these results could have implications for the education, mental health provision and legal treatment of teenagers.
Article
The view that certain facial expressions of emotion are universally agreed on has been challenged by studies showing that the forced-choice paradigm may have artificially forced agreement. This article addressed this methodological criticism by offering participants the opportunity to select a none of these terms are correct option from a list of emotion labels in a modified forced-choice paradigm. The results show that agreement on the emotion label for particular facial expressions is still greater than chance, that artifactual agreement on incorrect emotion labels is obviated, that participants select the none option when asked to judge a novel expression, and that adding 4 more emotion labels does not change the pattern of agreement reported in universality studies. Although the original forced-choice format may have been prone to artifactual agreement, the modified forced-choice format appears to remedy that problem.
Article
Over the past three decades, Cognition & Emotion has been one of the world’s leading outlets for emotion research. In this article, we review past highlights of and future prospects for the journal. Our tour of history covers three periods: The first period, from 1987 to 1999, was a pioneering era in which cognitive theories began to be applied to the scientific analysis of emotion. The second period, from 2000 to 2007, was characterised by a sharp increase in the number of empirical research papers, a lot of which were concerned with automatic processing biases and their implications for clinical psychology. During the third period, from 2008 to 2017, a new focus emerged on self-regulatory processes and their implications for emotion. We then turn to the present profile of Cognition & Emotion and introduce our new editorial team. Finally, we consider how the journal’s future success can be continued and increased by a) providing authors with fast and high-quality feedback; b) offering attractive publication formats, including the newly introduced Registered Reports for pre-registered studies; and c) consolidating key methodological paradigms with reproducible findings.
Book
The definitive handbook on peer relations has now been significantly revised with 55% new material. Bringing together leading authorities, this volume presents cutting-edge research on the dynamics of peer interactions, their impact on multiple aspects of social development, and the causes and consequences of peer difficulties. From friendships and romance to social withdrawal, aggression, and victimization, all aspects of children's and adolescents' relationships are explored. The book examines how individual characteristics interact with family, group, and contextual factors across development to shape social behavior. The importance of peer relationships to emotional competence, psychological well-being, and achievement is analyzed, and peer-based interventions for those who are struggling are reviewed. Each chapter includes an introductory overview and addresses theoretical considerations, measures and methods, research findings and their implications, and future directions.