Universal Patterns in Color-Emotion Associations Are Further Shaped by Linguistic and Geographic Proximity

Abstract

Many of us “see red,” “feel blue,” or “turn green with envy.” Are such color-emotion associations fundamental to our shared cognitive architecture, or are they cultural creations learned through our languages and traditions? To answer these questions, we tested emotional associations of colors in 4,598 participants from 30 nations speaking 22 native languages. Participants associated 20 emotion concepts with 12 color terms. Pattern-similarity analyses revealed universal color-emotion associations (average similarity coefficient r = .88). However, local differences were also apparent. A machine-learning algorithm revealed that nation predicted color-emotion associations above and beyond those observed universally. Similarity was greater when nations were linguistically or geographically close. This study highlights robust universal color-emotion associations, further modulated by linguistic and geographic factors. These results pose further theoretical and empirical questions about the affective properties of color and may inform practice in applied domains, such as well-being and design.

Full text

https://doi.org/10.1177/0956797620948810
Psychological Science
2020, Vol. 31(10) 1245 –1260
© The Author(s) 2020
Article reuse guidelines:
sagepub.com/journals-permissions
DOI: 10.1177/0956797620948810
www.psychologicalscience.org/PS
ASSOCIATION FOR
PSYCHOLOGICAL SCIENCE
Research Article
948810PSSXXX10.1177/0956797620948810Jonauskaite et al.Universal Color-Emotion Associations in 30 Nations
research-article2020
Corresponding Author:
Domicele Jonauskaite, University of Lausanne, Institute of Psychology, Quartier Mouline, Bâtiment Géopolis, CH-1015, Lausanne, Switzerland
E-mail: domicele.jonauskaite@unil.ch
Universal Patterns in Color-Emotion
Associations Are Further Shaped by
Linguistic and Geographic Proximity
Domicele Jonauskaite1, Ahmad Abu-Akel1, Nele Dael1,2,
Daniel Oberfeld3, Ahmed M. Abdel-Khalek4,
Abdulrahman S. Al-Rasheed5, Jean-Philippe Antonietti1,
Victoria Bogushevskaya6, Amer Chamseddine7, Eka Chkonia8,
Violeta Corona9,10, Eduardo Fonseca-Pedrero11, Yulia A. Griber12,
Gina Grimshaw13, Aya Ahmed Hasan4, Jelena Havelka14,
Marco Hirnstein15, Bodil S. A. Karlsson16, Eric Laurent17,18,
Marjaana Lindeman19, Lynn Marquardt15, Philip Mefoh20,
Marietta Papadatou-Pastou21,22, Alicia Pérez-Albéniz11,
Niloufar Pouyan1, Maya Roinishvili23, Lyudmyla Romanyuk24,25,26,
Alejandro Salgado Montejo27,28,29, Yann Schrag1, Aygun Sultanova30,
Mari Uusküla31, Suvi Vainio32, Graz˙yna Wa˛sowicz33,
Suncˇica Zdravkovic´34,35, Meng Zhang36, and Christine Mohr1
1Institute of Psychology, University of Lausanne; 2Department of Organizational Behavior, University of Lausanne;
3Institute of Psychology, Johannes Gutenberg-Universität Mainz; 4Department of Psychology, Faculty of Arts,
Alexandria University; 5Department of Psychology, King Saud University; 6Department of Linguistic Sciences and
Foreign Literatures, Catholic University of the Sacred Heart; 7School of Computer and Communication Sciences,
Swiss Federal Institute of Technology Lausanne; 8Department of Psychiatry, Tbilisi State Medical University;
9Escuela de Ciencias Económicas y Empresariales, Universidad Panamericana; 10Business Management Department,
Universitat Politècnica de València; 11Department of Educational Sciences, University of La Rioja; 12Department of
Sociology and Philosophy, Smolensk State University; 13School of Psychology, Victoria University of Wellington;
14School of Psychology, University of Leeds; 15Department of Biological and Medical Psychology, University of
Bergen; 16Division of Built Environment, Research Institutes of Sweden AB, Gothenburg, Sweden; 17Laboratory
of Psychology, University Bourgogne Franche–Comté; 18Maison des Sciences de l’Homme et de l’Environnement,
Centre National de la Recherche Scientifique (CNRS) and University of Franche-Comté; 19Department of
Psychology and Logopedics, University of Helsinki; 20Department of Psychology, University of Nigeria; 21School of
Education, National and Kapodistrian University of Athens; 22Biomedical Research Foundation (BRFaa), Academy
of Athens, Athens, Greece; 23Laboratory of Vision Physiology, I. Beritashvili Center of Experimental Biomedicine,
T'bilisi, Georgia; 24Faculty of Psychology, Taras Shevchenko National University of Kyiv; 25Department of
Psychology, V. I. Vernadsky Taurida National University; 26Department of Psychology, Kyiv National University
of Culture and Arts; 27Escuela Internacional de Ciencias Económicas y Administrativas, Universidad de La Sabana;
28Center for Multisensory Marketing, BI Norwegian Business School; 29Neurosketch, Bogotá, Colombia; 30National
Mental Health Centre, Ministry of Health, Baku, Azerbaijan; 31School of Humanities, Tallinn University; 32Faculty
of Social Sciences, University of Helsinki; 33Department of Economic Psychology, Kozminski University;
34Department of Psychology, University of Novi Sad; 35Laboratory for Experimental Psychology, University of
Belgrade; and 36Department of Psychology and Behavioral Sciences, Zhejiang University

1246 Jonauskaite et al.
Abstract
Many of us “see red,” “feel blue,” or “turn green with envy.” Are such color-emotion associations fundamental to
our shared cognitive architecture, or are they cultural creations learned through our languages and traditions? To
answer these questions, we tested emotional associations of colors in 4,598 participants from 30 nations speaking 22
native languages. Participants associated 20 emotion concepts with 12 color terms. Pattern-similarity analyses revealed
universal color-emotion associations (average similarity coefficient r = .88). However, local differences were also
apparent. A machine-learning algorithm revealed that nation predicted color-emotion associations above and beyond
those observed universally. Similarity was greater when nations were linguistically or geographically close. This study
highlights robust universal color-emotion associations, further modulated by linguistic and geographic factors. These
results pose further theoretical and empirical questions about the affective properties of color and may inform practice
in applied domains, such as well-being and design.
Keywords
affect, color perception, cross-cultural, universality, cultural relativity, pattern analysis, open data, open materials
Received 10/24/19; Revision accepted 5/7/20
Statement of Relevance
Why do we “see red,” “feel blue,” or “turn green
with envy”? Are such associations between color
and emotion fundamental to our shared cognitive
architecture? Or are they cultural creations learned
through our languages and traditions? To answer
these questions, we tested the emotional meaning of
colors in 4,598 participants from 30 nations speaking
22 languages. Overall, participants associated similar
emotion concepts with 12 color terms. Moreover,
similarity was higher between nations that share
borders or languages. Color-emotion associations
have universal features, further shaped by a shared
language or geography. These results pose further
theoretical and empirical questions about the
affective properties of color and may inform practice
in applied domains, such as well-being and design.
Color–emotion associations are ubiquitous (Adams &
Osgood, 1973; Hupka, Zaleski, Otto, Reidl, & Tarabrina,
1997; Madden, Hewett, & Roth, 2000; Major, 1895; Palmer,
Schloss, Xu, & Prado-Leon, 2013; Valdez & Mehrabian,
1994; Wexner, 1954; Wilms & Oberfeld, 2018). Common
wisdom would suggest that we “feel blue” when sad,
“see red” when angry, and are “green with envy.” Yet
envy can be yellow or red if we come from Germany
or Poland, respectively (see Hupka etal., 1997). And
although Westerners are likely to wear white to wed-
dings and black to funerals, people from China prefer
red for weddings and white for funerals.
Wherever one comes from, such color-emotion asso-
ciations are intriguing because colors and emotions
seem—at face value—to be fundamentally different
“things.” Colors are visual experiences driven by the
wavelength of light. Emotions are subjective feelings,
cognitions, and physiological responses that signal value.
Are these cross-modal associations cultural creations,
laid down in our languages and traditions? Or are they
fundamental features of our cognitive architecture? Exist-
ing studies have identified both similarities (Adams &
Osgood, 1973; D’Andrade & Egan, 1974; Gao etal., 2007;
Ou etal., 2018) and differences (Hupka etal., 1997;
Madden etal., 2000; Soriano & Valenzuela, 2009) across
cultures. However, they have done so among only a
small number of individual countries, making it nearly
impossible to capture global patterns. In a series of anal-
yses, we examined to what extent color-emotion associa-
tions are universal, testing 4,598 participants from 30
nations on six continents in 22 languages.
There are two theoretical explanations for color-
emotion associations, which make different predictions
about the degree to which the emotional meanings of
color should be shared. According to the first view,
color-emotion associations arise through environmental
experiences. That is, colors may become associated
with emotions because they appear in particular emo-
tional situations of evolutionary significance (e.g., red
face in anger; Benitez-Quiroz, Srinivasan, & Martinez,
2018). If so, color-emotion associations should be
largely universal (in support, see Adams & Osgood,
1973; D’Andrade & Egan, 1974; Gao etal., 2007; Ou
etal., 2018). According to the second theoretical expla-
nation, colors and emotions may become arbitrarily
associated in the language, history, religion, or folklore
of one’s culture. If so, color-emotion associations
should vary between cultures with different languages,
symbolism, and traditions (Evarts, 1919; Soriano &
Valenzuela, 2009). Such cross-cultural variations have
also been reported (Hupka etal., 1997; Madden etal.,

Universal Color-Emotion Associations in 30 Nations 1247
2000; Soriano & Valenzuela, 2009). Although these
views are often cast in opposition to each other, they
are not mutually exclusive. According to the cross-
modal-correspondence framework (Spence, 2011), two
unrelated entities (here, colors and emotions) can
become cross-modally associated when they regularly
appear together in one’s perceptual or linguistic envi-
ronment, whether on a global (shared by all) or local
(shared by some) scale.
It is possible, therefore, that universal tendencies to
associate certain colors with certain emotions are fur-
ther modulated by cultural and individual factors. Con-
sider red, an ambivalent color that has been associated
with both negative and positive emotions depending
on whether one comes from Western countries or China
(Jonauskaite, Wicker, etal., 2019). The existence of
both associations could be explained in evolutionary
terms (e.g., red-blood pairings lead to associations with
both danger and sexuality). In countries such as China,
however, cultural beliefs that red is a symbol of good
fortune might strengthen the link between red and
positive emotions and weaken the link between red
and negative emotions (see Wang, Shu, & Mo, 2014).
In other countries, such as the United States, the strong
link between red and danger or red and failure
(Pravossoudovitch, Cury, Young, & Elliot, 2014) could
strengthen negative associations while weakening posi-
tive associations. Such additional variations might be
maintained through language and geographic locations
(see also Jackson et al., 2019; Jonauskaite, Abdel-
Khalek, etal., 2019).
Existing studies provide examples of both similarities
and differences across countries. But these studies have
focused on just a few countries, languages, or cultures,
and so global patterns are still unknown. To test for the
degree of universality, we performed a large-scale,
cross-cultural survey on color-emotion associations (for
our theoretical motivation, see Mohr, Jonauskaite, Dan-
Glauser, Uusküla, & Dael, 2018). Participants completed
the survey online in their native language. We exceeded
previous investigations in terms of the number of tested
nations, representativeness of participants, and the
number of tested colors and emotions. We collected
data from 4,598 participants from 30 nations located on
all continents but Antarctica (see Fig. 1). Participants
were between the ages of 15 and 87 years and reported
having normal color vision. We used 12 color terms
representing the most common color categories (Berlin
& Kay, 1969; Mylonas & MacDonald, 2015) and an
extensive list of 20 emotion concepts varying in valence
and potency (Scherer, 2005). Participants chose as many
emotion concepts as they thought were associated with
a given color term and rated the intensity of the associ-
ated emotion from weak to strong.
In a series of analyses, we examined the degree of
similarity across the 30 nations in (a) probabilities of
color-emotion associations and (b) intensities of associ-
ated emotions. We then applied a machine-learning
algorithm to quantify the degree of nation specificity
in color-emotion associations. Finally, we assessed how
color-emotion associations varied as a function of lin-
guistic and geographic distances.
Method
Participants
We extracted our data from the ongoing International
Color-Emotion Association Survey (Mohr etal., 2018) per-
formed online. This survey tests participants from a large
age range using predefined age categories (15–29 years,
30–49 years, 50 years and older). We started with the larg-
est possible participant pool (4,883 participants) consist-
ing of data sets from countries for which we had at least
20 usable (e.g., without self-reported problems of color
vision) participants per age category (see also Simmons,
Nelson, & Simonsohn, 2011). We detail additional selec-
tion criteria in the Data Preparation section. Our final
sample (N = 4,598; 1,114 male) consisted of participants
from 30 different nations (see Fig. 1) with a mean age of
35.4 years (SD = 14.5). Counts per nation ranged from
69 to 490 participants. Table S1 in the Supplemental
Material available online provides language informa-
tion, and Table S2 in the Supplemental Material pro-
vides demographic information of the participants of
each nation. Participation was voluntary. The study was
conducted in compliance with the ethical standards
described in the Declaration of Helsinki. Parts of the
data have been reported previously in relation to dif-
ferent research questions (Jonauskaite, Abdel-Khalek,
etal., 2019; Jonauskaite, Parraga, Quiblier, & Mohr,
2020; Jonauskaite, Wicker, etal., 2019).
Material and procedure
Emotion assessment with the Geneva Emotion Wheel.
The Geneva Emotion Wheel (Version 3.0; Scherer, 2005;
Scherer, Shuman, Fontaine, & Soriano, 2013; see Fig. 2) is
a self-report measure designed to assess the feeling com-
ponent of emotional experiences elicited by particular
events. It is based on theoretical categorizations of emo-
tions and validated through research. The Geneva Emo-
tion Wheel represents 20 discrete emotions (e.g., anger,
fear, joy) as spokes on a wheel. Emotion concepts that are
similar in valence (positive or negative) and power (high
or low) are placed close to each other. Each spoke of the
wheel contains five circles that extend from a central square,
representing increasing intensities of each emotion.

1248
Nation-to-Global
Similarity
.65−.70
.70−.75
.80−.85
.85−.90
.90−.95
NA
Fig. 1. World map showing the 30 nations included in the study. The colors indicate nations’ similarity with the global color-emotion association pattern. Redder
nations show color-emotion association patterns more similar to the global mean.

Universal Color-Emotion Associations in 30 Nations 1249
For each color term, participants used a mouse click
to indicate the associated emotions and their intensities
(i.e., they could indicate that a single color term is
associated with more than one emotion concept; see
Fig. 2). At the beginning of the trial, the central square
was selected, indicating no emotion. Participants were
also given the option to select “Different Emotion,”
which produced a pop-up window in which they could
type the name of a different emotion. These responses
were rare, and we did not analyze them.
Participants completed the Geneva Emotion Wheel
in their native language. The translation of the Geneva
Emotion Wheel was available for some languages on
the Swiss Centre for Affective Sciences website. The
remaining translations were created using the back-
translation technique (for further details, see the Trans-
lation of the Geneva Emotion Wheel section in the
Supplemental Material; for emotion terms in each lan-
guage, see Table S3 in the Supplemental Material).
International Color-Emotion Association Survey. We
collected the current data online by sharing the survey link
(http://www2.unil.ch/onlinepsylab/colour/main.php)
with potential participants via university communications,
e-mails, social media, and personal contact, mainly through
our collaborators (coauthors) in each country. The survey
was originally constructed in English and was translated
(without back-translation) by coauthors and collaborators
(see the Acknowledgments section). We used links that
automatically opened in the official language of the country
to encourage participants to complete the survey in their
native language. However, participants could switch to any
other language provided. We analyzed data gathered from
only native speakers. Online data collection naturally
resulted in literate participants with access to the Inter-
net. Some elderly participants were helped with survey
completion.
The first page described the aims of the study and
ethical considerations; participants consented by click-
ing on the “Let’s go” button. The following two instruc-
tion pages explained the task and the use of the Geneva
Emotion Wheel. We then used a manipulation check to
verify that participants understood the task. Participants
were presented with a situation and had to identify the
correct responses. The situation read,
Peter thinks that beige strongly represents intense
compassion, and believes that beige is also
associated with mild relief. Accidentally, he has
selected sadness and wants to correct his choice.
Look at his response in the emotion wheel below
and try to correct it.
Participants saw the largest circle for sadness marked
(Emotion Intensity 5). They could move to the next
page and start the survey only if they successfully cor-
rected Peter’s responses. They had to click on the
square for sadness (no association, rating 0), the largest
Fig. 2. The Geneva Emotion Wheel with the color term red as an example. The wheel was used to assess associations between 20 emotion
concepts and 12 color terms. Participants expressed emotion associations by selecting one of the five circles for each associated emotion. At
the same time, they chose the intensity of the associated emotion, ranging from weak (smallest circle) to strong (largest circle). Participants
could select as many or as few emotions as they thought appropriate. The left panel shows the wheel as it initially appeared. The right
panel shows an example of a response from a participant who associated the color term red with strong love and relatively strong anger.

1250 Jonauskaite et al.
circle for compassion (Emotion Intensity 5), and the
middle circle for relief (Emotion Intensity 3). If partici-
pants made a mistake and tried to move forward, a
pop-up window guided them to the correct responses.
This manipulation check ensured that participants
understood the task.
In the actual task, participants were presented with
12 color terms (not color patches): red, orange, yellow,
green, blue, turquoise, purple, pink, brown, black, gray,
and white (for the color terms in all languages, see Table
S4 in the Supplemental Material). Color terms appeared
one at a time above the Geneva Emotion Wheel in ran-
dom order. For each color term, participants could either
select any number of the emotion concepts they thought
were associated with the given color term or indicate
“none.” They rated the intensity of each chosen emotion
(see Fig. 2). On average, participants associated 3.05
emotion concepts with a color term (95% confidence
interval, or CI = [3.03, 3.08], range = 2.25–3.84; see Table
S5 in the Supplemental Material).
After evaluating the 12 color terms, participants com-
pleted a demographic questionnaire in which they
reported the importance of color in their life, along with
their age, sex, problems with color vision, country of
origin, country of residence, native language, and flu-
ency in the language in which they completed the color-
emotion survey. Participants could select the “do not
want to answer” option for any of the demographic
questions. On the final page, participants were thanked
and received results from a previous, related study in
graphic form. We provided an e-mail address for future
contact. On average, the current sample took 31 min to
complete the survey.
Data preparation
We applied the following inclusion and exclusion criteria
to clean the data. We included participants (a) who fin-
ished the survey, (b) who completed the survey in their
native language, and (c) for whom this language was the
official language of their country of origin. Taking Nor-
way as an example, we included native Norwegian
speakers who completed the survey in Norwegian
(Bokmål) and whose country of origin was Norway. An
exception was made for participants from Nigeria, who
completed the survey in English (the national language).
Nigerian participants had high English proficiency levels
(M = 7.02, SD = 0.29, out of 8; for other languages and
countries, see Table S1 in the Supplemental Material).
As stated above, participants who might have had prob-
lems with color vision were excluded (i.e., responded
“yes,” “do not know,” or “do not want to answer” to the
question “Do you have trouble seeing colors?”). There
were 285 (5.8%) participants who potentially had color
vision problems across all the nations.
Statistical analyses
With 20 emotion concepts and 12 color terms, we
obtained 240 ratings of color-emotion associations per
participant. From these associations, we extracted two
dependent variables. The first dependent variable was
the probability of color-emotion associations. The sec-
ond dependent variable was emotion intensity (see
below). The alpha level was set to .050 for all statistical
analyses. Statistical analyses were performed and
graphs were created with SPSS Version 25 and R Studio
Version 1.1.4 (R Version 3.4.0; R Core Team, 2017).
Global probabilities. To evaluate the probability of
color-emotion associations, we assessed which emotions
are associated with each color term without considering
emotion intensity. To this end, all selected emotion asso-
ciations were coded as 1 (regardless of intensity), and all
nonselected emotion associations were coded as 0. We
used a Bayesian method to estimate probabilities of each
emotion being associated with each color term (see the
Bayesian Probabilities section in the Supplemental Mate-
rial). We used the mean estimated probabilities of all par-
ticipants for each color-emotion pair to construct a global
matrix of color-emotion association probabilities (12 ×
20; see Fig. 3). The same procedure was repeated for
each of the 30 nations separately to obtain mean proba-
bilities of associating every emotion with every color
term in each of the 30 nations (see 30 nation-specific
matrices of color-emotion associations in Table S6 in the
Supplemental Material). We used nation-specific matrices
for further cross-cultural comparisons.
Cultural probabilities and their comparisons. We
first determined the degrees of similarity between nation-
specific patterns of color-emotion associations and the
global pattern of color-emotion associations—nation-to-
global pattern similarity. The underlying values were
Bayesian probabilities. The degrees of similarity were cal-
culated by computing Pearson’s correlations between the
12 × 20 color-emotion association probabilities of each
nation (nation-specific matrix) and the corresponding
global 12 × 20 color-emotion association probabilities
(global matrix without that nation). The global probabili-
ties were always based on data from 29 nations, that is, all
nations but the nation of comparison. These 30 global
matrices including the data from 29 nations correlated
from .9983 to .9993 with the global matrix including the
data from all 30 nations. Hence, no single nation unduly
influenced the global pattern. See the full list of nation-
specific and global matrices in Table S6 in the Supplemen-
tal Material. Next, we estimated nation-to-nation pattern
similarity by correlating all nation-specific matrices with
each other (900 matrix correlations; Table S7 in the Sup-
plemental Material). We also looked at the effects of sex

Universal Color-Emotion Associations in 30 Nations 1251
and age, which are discussed in the Sociodemographic
Factors subsection in Results and reported in full in Tables
S8 and S9, respectively, in the Supplemental Material.
Finally, we repeated the pattern-similarity analyses per
color term. That is, we correlated nation-specific patterns
of color-emotion association probabilities with global pat-
terns, excluding that nation for each color term (e.g.,
nation-specific pattern of red vs. global pattern of red,
excluding that nation; Table S10 in the Supplemental Mate-
rial). In all of these comparisons, a score of 1.0 indicates
perfect similarity in color-emotion association patterns,
whereas a score of 0.0 indicates complete dissimilarity in
color-emotion association patterns.
In addition to calculating similarity in color-emotion
association patterns, we calculated the average prob-
abilities of associating any color with any emotion—
average probability of color-emotion association. The
nation-specific average probability of color-emotion
association was calculated by averaging all 240 Bayes-
ian probabilities of color-emotion associations of each
nation. The unweighted global average probability of
color-emotion association was calculated by averaging
all nation-specific average probabilities of color-emotion
association (global average-probability score = .161,
95% CI = [.150, .174]). We compared the global average
probability of color-emotion association with nation-
specific average probabilities of color-emotion associa-
tion using one-sample t tests. To account for multiple
comparisons, we false discovery rate (FDR) corrected
the p values, using q = .05 as the threshold. As in the
pattern-similarity analyses, we repeated the compari-
sons per color term as well as for sex and age (see
Sociodemographic Factors). An average-probability
score of 1.0 for color-emotion association indicates that
all color terms were associated with all emotion con-
cepts, whereas a score of 0.0 indicates that no color
term was associated with any emotion concept.
The emotion-intensity variable provides information
about the average intensity of all emotions associated
with each color term. To calculate emotion-intensity
similarities, we took all emotion concepts associated
with a given color term (previously coded as 1) by a
given participant and averaged the intensities assigned
to these emotions. Emotion intensities are reported per
color term and not per color-emotion association. They
varied from 1 (weak) to 5 (intense) unless no emotion
was chosen for a given color term (coded as a missing
value). We had 12 emotion-intensity scores per participant
Anger
Hate
Contempt
Disgust
Fear
Disappointment
Shame
Regret
Guilt
Sadness
Compassion
Relief
Love
Admiration
Contentment
Pleasure
Joy
Pride
Amusement
Interest
Red
Orange
Yellow
Green
Turquoise
Blue
Purple
Pink
Brown
White
Gray
Black
0.00
0.35
0.70
Bayesian
Probability
Global Color-Emotion Association
Pattern
Color
Emotion
Fig. 3. Heat map showing the unweighted average probabilities of color-emotion
associations across the 30 nations. More saturated orange or red indicates a higher
probability of a specific color-emotion association. The cells are not exclusive, mean-
ing that the same participant could have contributed to none, one, or several emotion
associations for a given color term (many-to-many associations).

1252 Jonauskaite et al.
(one score per color term) and compared these scores
across nations. We computed Pearson’s correlations
between the 12 emotion-intensity scores of each nation
and the corresponding global emotion-intensity scores,
each time leaving out that nation, when calculating
nation-to-global emotion-intensity similarities (see Table
S11 in the Supplemental Material). The resulting 29
global emotion-intensity matrices including the data
from 29 nations correlated from .9967 to .9999 with the
global emotion-intensity matrix including the data from
all 30 nations. Hence, no single nation unduly influ-
enced the global pattern. A score of 1.0 for emotion-
intensity similarity indicates perfect emotion-intensity
pattern similarity, whereas a score of 0.0 indicates com-
plete pattern dissimilarity.
Multivariate pattern classification. We used a super-
vised machine-learning approach to predict the nation of
each participant from his or her set of 240 ratings of
color-emotion association (see also Jonauskaite, Wicker,
et al., 2019). The accuracy of the classifier provides a
quantitative measure of nation specificity in color-emotion
associations. If the accuracy is not higher than the chance
level, this indicates an absence of nation specificity in the
color-emotion associations (i.e., perfect universality). In
contrast, high accuracy indicates a high degree of nation
specificity. For details of the classifier algorithm, fitting,
and evaluation, see the Multivariate Pattern Classification
section in the Supplemental Material.
A quantitative measure of the similarity between a
pair of nations’ color-emotion associations can be read-
ily computed from the classifiers’ confusion matrix on
the basis of the assumption that nations that are more
similar will be more frequently confused by the classi-
fier than nations that are less similar. We used Luce’s
biased-choice model (Luce, 1963, Equation 5) to esti-
mate similarity values for each pair of nations from the
confusion matrix. By convention, a similarity value
between a nation and itself is set to 1.0 (representing
maximum similarity), whereas a similarity value of 0.0
means that the two nations are completely dissimilar.
The estimated similarity values are displayed in Figure
S1 in the Supplemental Material.
Linguistic and geographic distances. In addition to
assessing cultural similarities, we tested whether two
factors—linguistic distance and geographic distance—
explained part of the similarity between the color-emotion
associations of different nations. We extracted linguistic
distances for each nation-nation pair from the study by
Jäger (2018; for language codes, see the Linguistic Dis-
tances section in the Supplemental Material). These dis-
tances are suggested to capture phylogenetic distances
that quantify the degree of similarity between the lan-
guages of our nation pairs.
The linguistic distances in Jäger’s (2018) study range
from 0 to 1, with lower linguistic-distance scores indi-
cating higher linguistic similarities. In this data set, the
linguistic distances are not evenly spread across this
range because there are more unrelated than related
language pairs in the world. This was true in our sample
of languages as well. In fact, the first 25% of distances
fell between 0 and .75, whereas the remaining 75% of
distances were concentrated between .75 and .90. To
make the spread more homogeneous, we used a power
transform of the original distances. At the fourth power,
the transformed linguistic distances resulted in a more
homogeneous spread (quantiles at 0.00, 0.32, 0.41, 0.53,
and 0.65). Jäger proposed that language pairs with dis-
tances below .7 should be considered as related. The
criterion for related languages became .24 (i.e., .74),
using the transformed linguistic distances (hereinafter
referred to simply as linguistic distances; see Table S12
in the Supplemental Material).
We also calculated geographic distances for all
nation pairs. We used population-weighted geographic
centers to reflect the location within each country
where participants were most likely to originate. If we
could not find population-weighted centers, we used
the geographic coordinates of the most populated city
of that nation (see Table S13 in the Supplemental
Material). Using these centers, we calculated distances
(in kilometers) on a sphere between all pairs of
nations (see Table S14 in the Supplemental Material).
In two linear regression models, we used linguistic
and geographic distances to predict (a) nation-to-
nation pattern-similarity scores (see the Cultural Prob-
abilities and Their Comparisons section) and (b) Luce’s
similarity scores (see the Multivariate Pattern Classifi-
cation section). We argue that comparable results
using both approaches provide stronger evidence for
the role of linguistic or geographic distance, not least
because scores are extracted using very different sta-
tistical methods—correlations and multivariate pattern
classification.
Results
Global probabilities
We determined the global matrix of the color-emotion
association probabilities on the basis of the unweighted
means of the estimated Bayesian probabilities for each
color-emotion pair across our 30 nations. Prominent
color-emotion associations (probabilities ≥ .4 on the
basis of our data) were black and sadness, black and
fear, black and hate, red and love, red and anger, pink
and love, pink and joy, pink and pleasure, gray and
sadness, gray and disappointment, yellow and joy,
orange and joy, orange and amusement, and white and
relief (see Fig. 3 & Table S6).

Universal Color-Emotion Associations in 30 Nations 1253
Cultural probabilities
Similarities in color-emotion association patterns.
The nation-to-global similarities in color-emotion associ-
ation patterns were high and significant for all 30 nations.
The average nation-to-global pattern similarity (r) was
.880 (95% CI = [.857, .903], p < .001). All nation-to-global
pattern similarities ranged from .684 (Egypt vs. global) to
.941 (Spain vs. global; all ps < .001, FDR corrected; see
Figs. 1 and 4a). The high pattern similarity indicates that
all individual nations associated color terms with emotion
concepts similarly to the global pattern. Nation-to-nation
pattern similarities were also high and significant (ps <
.001). They had a mean of .781 (95% CI = [.773, .789]) and
ranged from .501 (The Netherlands vs. Azerbaijan) to .951
(Switzerland vs. France; all ps < .001, FDR corrected; see
Fig. S2 and Table S7 in the Supplemental Material). Half
of all nation-to-nation correlations fell between .738 and
.839, with a median correlation of .799. Figure 4b shows
distributions of nation-to-global and nation-to-nation pat-
tern similarities.
Nation-to-global pattern similarities per color term
were also high. Average similarities ranged from .659
(95% CI = [.548, .769]; purple) to .925 (95% CI = [.910,
.940]; pink; see Fig. S3 and Table S10 in the Supple-
mental Material). Across all nations, purple and yellow
had the highest variance in similarities, and pink, green,
turquoise, and black had the lowest variance in similari-
ties, suggesting that associations with the former colors
were the least similar and associations with the latter
colors were the most similar across the 30 nations. We
also observed certain nation-specific color-emotion
associations (see Table S6 and Fig. S3). For instance,
Nigerians associated red with fear in addition to love
and anger; Chinese associated white with sadness in
addition to relief. Unlike other nations, Egyptians did
not associate joy and other positive emotions with yel-
low. Greeks associated purple with sadness, whereas
other nations, on average, associated purple with posi-
tive emotions.
Average probabilities of color-emotion associa-
tions. One-sample t tests showed that the average prob-
abilities of color-emotion association were not significantly
different from the global average probability of color-
emotion association in 25 out of 30 nations (see Fig. 4c;
ps > .604). Only five nations differed significantly from
the global average probability of color-emotion associa-
tion. Relative to the global average probability, partici-
pants from Finland, Lithuania, and New Zealand were
significantly more likely to associate color terms with
emotion concepts, and participants from Azerbaijan and
Egypt were significantly less likely (ps < .005, FDR cor-
rected; see Fig. 4c, nations in green). When visually
inspecting average probabilities of color-emotion associ-
ation per color term (see Fig. S4 in the Supplemental
Material), we found that, in every nation, red and black
had the highest average probability of being associated
with any emotion concept, and brown had the lowest.
Emotion-intensity pattern similarities. Emotion-
intensity pattern similarities were high and significant for
all 30 nations. The average nation-to-global emotion-
intensity similarity was .709 (95% CI = [.666, .752], p <
.001) and ranged from .693 (Azerbaijan vs. global) to .970
(Serbia vs. global; ps ≤ .012, FDR corrected; see Fig. 4d).
Multivariate pattern classification
The machine-learning classifier correctly predicted the
nation for 34.4% of the participants (area under the
receiver-operating-characteristic curve, or AUC = .85).
This proportion of correctly classified instances is well
above the random guessing rate of 9.7% that can be
obtained by always choosing the nation contained most
frequently in our data set (Azerbaijan). The AUC of .85
was also considerably higher than the AUC for the
randomly permuted data sets (.51). Thus, the classifier
performance demonstrates a systematic amount of
nation specificity in color-emotion associations. The
confusion matrix (see Fig. 5) shows that participants
from Nigeria were the easiest to predict (true positive
rate, or TPR = .811), whereas participants from Spain
were the most difficult to predict (TPR = .071).
Linguistic and geographic distances
We fitted a linear regression model with measures of
linguistic and geographic distance as predictors of
nation-to-nation pattern-similarity scores for color-
emotion associations, once with and once without the
interaction between the two distance measures. The
inclusion of the interaction did not improve the model
(p = .389). Therefore, we report the model without the
interaction term. The model was significant overall, F(2,
432) = 39.9, p < .001, and explained 15.2% of the vari-
ance (adjusted R2). A shorter linguistic distance (β =
−0.37, p < .001) and a shorter geographic distance
(β = −0.13, p = .003) both predicted higher nation-to-
nation pattern-similarity scores for color-emotion asso-
ciations (see Figs. 6a and 6b).
The analogous linear regression model with linguis-
tic and geographic distances as predictors of Luce’s
similarity scores in multivariate pattern classification
was also significant, F(2, 432) = 37.4, p < .001. The
model explained 14.4% of the variance (adjusted R2).
Again, shorter linguistic (β = −0.36, p < .001) and

1254 Jonauskaite et al.
Azerbaijan
Egypt
Nigeria
United States
Italy
China
Iran
New Zealand
Israel
Germany
Georgia
Saudi Arabia
Netherlands
United Kingdom
France
Lithuania
Ukraine
Belgium
Estonia
Finland
Switzerland
Mexico
Norway
Russia
Colombia
Greece
Sweden
Poland
Spain
Serbia
0.00 0.25 0.50 0.75 1.00
Emotion-Intensity Pattern Similarity (r)
Egypt
Azerbaijan
Saudi Arabia
Russia
Iran
Ukraine
United Kingdom
United States
Georgia
Netherlands
Spain
Mexico
Belgium
Italy
Estonia
Poland
Norway
Serbia
Israel
France
China
Greece
Colombia
Sweden
Germany
Switzerland
Nigeria
Finland
New Zealand
Lithuania
0.0 0.1 0.2 0.3
Average Probability of All Associations
Egypt
Nigeria
Azerbaijan
China
Iran
Saudi Arabia
Georgia
Netherlands
Ukraine
Italy
Greece
Estonia
New Zealand
Russia
United States
Israel
Serbia
Belgium
Colombia
Finland
Mexico
United Kingdom
Lithuania
France
Poland
Switzerland
Germany
Sweden
Norway
Spain
0.5 0.6 0.7 0.8 0.9 1.0
Similarity in Color-Emotion Association Pattern (r)
0
3
6
9
0.00 0.25 0.50 0.75 1.00
Density
Nation-to-Global Similarity Nation-to-Nation Similarity
ab
cd
Similarity in Color-Emotion Association Pattern (r)
Fig. 4. (continued on next page)

Universal Color-Emotion Associations in 30 Nations 1255
Fig. 4. Nation comparisons. Nation-to-global similarity in color-emotion association patterns (a) is shown for each of the 30 nations. The
dotted line marks perfect pattern similarity (r = 1). Density plots (b) show nation-to-global and nation-to-nation similarity in color-emotion
association patterns. Average probabilities of all color-emotion associations (c) are shown for each nation. The average probability of
each nation was compared with the global average probability, which is the unweighted average of all average probabilities (dark green
line; gray area = 95% confidence interval, or CI). Nations marked in green are significantly different from the global average probability
after false-discovery-rate correction. A higher score indicates a higher probability of associating any color term with any emotion concept.
Nation-to-global emotion-intensity pattern similarity (d) is shown for each nation. The dotted line marks perfect pattern similarity (r = 1).
Error bars in (a), (c), and (d) represent 95% CIs.
geographic (β = −0.13, p = .003) distances predicted
higher Luce’s similarity scores (see Figs. 6c and 6d).
Sociodemographic factors
We examined the influence of two key sociodemo-
graphic factors—sex and age—on similarities in color-
emotion association patterns and on average probabilities
of color-emotion associations. Color-emotion association
patterns of men and women were almost identical (r =
.987, p < .001; see Table S8) and were very similar
across age groups (r = .901–.991; ps < .001; see Table
S9). Men and women also did not differ in their average
probability of color-emotion associations, t(478) = 0.49,
p = .624 (see Fig. S5a in the Supplemental Material).
Notably, however, age was nonlinearly related with
average probabilities of color-emotion associations. A
curve-estimation analysis revealed that the association
of age with average probabilities followed a U-shaped
pattern; the average probability gradually decreased
from early adulthood, that is, from 15 to 20 years of
age to 50 to 60 years of age, and then started increasing
from 50 to 60 years of age onward, F(2, 1677) = 55.22,
p < .001, adjusted R2 = .061 (see Fig. S5b in the Supple-
mental Material). In other words, 50- to 60-year-old par-
ticipants were the least likely to associate any color term
with any emotion concept.
Discussion
The cross-modal association of color with emotion is a
universal phenomenon. Moreover, there is global simi-
larity in how specific emotion concepts are associated
with specific color terms, although these universal asso-
ciations are modulated by geographic and linguistic
factors. Across 30 nations and 22 languages on six con-
tinents, the pattern of color-emotion associations in
each country highly coincided with the global pattern
(mean r = .88). In other words, participants from dif-
ferent nations shared the relative tendencies to favor
certain color-emotion associations (e.g., love and anger
with red) over others (e.g., shame with red). Further-
more, participants from different nations agreed on
which colors were the most (i.e., black and red) and
the least (i.e., brown) emotional. Finally, they rated
emotion intensities in a similar manner. Hence, we dem-
onstrated robust agreement across 30 nations in color-
emotion associations, providing strong evidence that
such associations might represent a human psychologi-
cal universal (in agreement with Adams & Osgood,
1973; D’Andrade & Egan, 1974; Gao etal., 2007; Ou
etal., 2018). Potential mechanisms for these universal
associations may be found in a lasting shared human
history, regularities in human languages and environ-
ments, and shared cognitive biases.
But beyond these global similarities, certain color-
emotion associations additionally varied locally (see
also Hupka etal., 1997; Madden etal., 2000; Soriano &
Valenzuela, 2009). In particular, nations that were lin-
guistically or geographically closer had more similar
color-emotion association patterns. Such nations were
predicted with lower accuracy by the machine-learning
algorithm, even though the algorithm could still predict
any participant’s nation from the ratings of color-emotion
associations above chance level (see also Jonauskaite,
Wicker, etal., 2019). These variations might originate
from cultural or linguistic differences in how emotion
terms or color terms are understood across nations
(Jackson etal., 2019). But these variations might also
stem from differences in the physical environments
themselves. For instance, we recently reported that
exposure to sunshine modulated the degree to which
yellow was perceived as a color of joy (Jonauskaite,
Abdel-Khalek, etal., 2019).
Although the majority of nations did not vary in the
extent to which color-emotion associations were
endorsed, specific variations were nevertheless
observed. Finns, Lithuanians, and New Zealanders
endorsed color-emotion associations to a greater extent,
whereas Azerbaijanis and Egyptians did so to a lesser
extent than the global average. The source of these
differences requires further study. Moreover, some
nations exhibited idiosyncratic color-emotion associa-
tions. For instance, although sadness was universally
associated with black, Greeks also associated it with
purple, and Chinese also associated it with white.
Likely, these divergent color-emotion associations
reflect different cultural traditions. White is commonly
worn at funerals in China, whereas Greeks occasionally
wear darker shades of purple during mourning periods.

1256 Jonauskaite et al.
AZ BE CH CN CO DE EE EG ES FI FR GB GE GR IL IR IT LT MX NG NL NO NZ PL RS RU SA SE UA US
AZ
BE
CH
CN
CO
DE
EE
EG
ES
FI
FR
GB
GE
GR
IL
IR
IT
LT
MX
NG
NL
NO
NZ
PL
RS
RU
SA
SE
UA
US
244 100 215 229 109 198 171 266 79 179 102 89 144 201 134 198 156 108 116 210 124 95 163 133 100 94 138 98 93 124
427
87
192
149
100
218
135
193
156
139
92
150
121
269
66
112
105
123
117
127
70
208
171
171
95
125
139
192
65
96
Probability
00.20.40.60.8
Predicted Nation
Actual Nation
Fig. 5. Confusion matrix for the prediction of the participants’ nation (machine-learning multivariate-pattern-classification approach).
Rows represent the actual nation that participants were from, and columns represent the predicted nation. Cells represent the probability
that participants originating from the nations specified in rows were classified by the machine-learning algorithm as originating from the
nations specified in columns, on the basis of their 240 individual color-emotion associations. Thus, probabilities on the main diagonal rep-
resent the true positive rate, or recall. The numbers to the right of the matrix represent the absolute frequency of participants originating
from a given nation. The numbers above the matrix represent the absolute frequency of participants predicted to originate from a given
nation. AZ = Azerbaijan, BE = Belgium, CN = China, CO = Colombia, EG = Egypt, EE = Estonia, FI = Finland, FR = France, GE = Georgia,
DE = Germany, GR = Greece, IR = Iran, IL = Israel, IT = Italy, LT = Lithuania, MX = Mexico, NL = The Netherlands, NZ = New Zealand,
NG = Nigeria, NO = Norway, PL = Poland, RU = Russia, SA = Saudi Arabia, RS = Serbia, ES = Spain, SE = Sweden, CH = Switzerland, UA
= Ukraine, GB = United Kingdom, US = United States.

Universal Color-Emotion Associations in 30 Nations 1257
Hence, cultural pairings of white, purple, or black with
funerals may explain why specific colors are associated
with sadness in some nations but not others.
In this study, we asked participants about their asso-
ciations between color terms and emotion terms, allow-
ing us to capture the conceptual relationship between
them (see also Hupka etal., 1997; Ou etal., 2018;
Palmer etal., 2013; Wexner, 1954). However, we do not
know whether that relationship also plays out in emo-
tional experiences associated with color perception.
That is, people may universally associate the concepts
of red and anger but may not universally feel angry
when seeing red objects. Within cultures, colors do
induce specific subjective and physiological emotional
R2 = .139
0.5
0.6
0.7
0.8
0.9
1.0
0.0 0.2 0.4 0.6
Linguistic Distance
Similarity in Color-Emotion Association Pattern
R2 = .131
0.0
0.2
0.4
0.6
0.8
0.0 0.2 0.4 0.6
Linguistic Distance
Similarity in Multivariate Pattern Classification
R2 = .017
0.5
0.6
0.7
0.8
0.9
1.0
0 5 10 15 20
Geographic Distance (1,000 km)
Similarity in Color-Emotion Association Pattern
R2 = .016
0.0
0.2
0.4
0.6
0.8
0 5 10 15 20
Geographic Distance (1,000 km)
Similarity in Multivariate Pattern Classification
b
cd
a
Fig. 6. Scatterplots showing linguistic and geographic distance as predictors of nation-to-nation similarities. The top row shows similarity
between color-emotion association patterns and both (a) linguistic distance and (b) geographic distance (see also Fig. S2 in the Supplemental
Material available online and Fig. 4b). The bottom row shows similarity between multivariate pattern classifications and both (c) linguistic
distance and (d) geographic distance according to Luce’s biased-choice model applied to the classifier confusion matrix (see also Fig. S1 in
the Supplemental Material). In all the plots, solid lines indicate best-fitting regressions, and shaded areas indicate 95% confidence intervals.

1258 Jonauskaite et al.
responses (e.g., Wilms & Oberfeld, 2018), and similar
emotion concepts are associated with color terms and
their best perceptual examples (Jonauskaite et al.,
2020). It remains to be seen whether the direct associa-
tion between color and emotion shows the same pat-
terns of linguistic and geographic modulation that we
have described here.
Our results suggest that there is a universal basis for
color-emotion associations that is shared by everyone.
Numerous other human universals exist (Brown, 1991).
In the domains of color and affect, such universals
include the shared understanding of facial emotion
expressions (Ekman, Sorenson, & Friesen, 1969, but see
Gendron, Roberson, van der Vyver, & Barrett, 2014), of
emotions perceived in music (Cowen, Fang, Sauter, &
Keltner, 2020), and of emotions expressed in human
songs (Mehr etal., 2019) as well as the shared loci of
focal colors (Regier, Kay, & Cook, 2005, but see Uusküla
& Bimler, 2016). This universal foundation of color-
emotion association is further modulated by language,
geography, and culture.
Some researchers might understand the modulation
as evidence against universality because color-emotion
associations were not shared at 100%. Yet no human
psychological universal is shared at 100% (Mehr etal.,
2019; Norenzayan & Heine, 2005; Regier etal., 2005).
Luckily, they are not. The scope for dissimilarities seems
essential for dynamic adaptations to immediate and
lasting changes in one’s environment (Lupyan & Dale,
2016). Other researchers might interpret our overall
conclusions as evidence of a globalized world. This
concern might be justified because we mainly tested
computer-literate participants who completed the sur-
vey online. Potentially, our color-emotion associations
become increasingly similar as we share more and more
information globally via the Internet and other com-
munication channels. To test the generalizability of our
results, we would need further data from small-scale
societies (e.g., Davidoff, Davies, & Roberson, 1999;
Groyecka, Witzel, Butovskaya, & Sorokowski, 2019).
Given our current knowledge, we suggest that color-
emotion associations represent a human psychological
universal that likely contributes to shared communica-
tion and comprehension. Thus, the next time you feel
blue or see red, know that the world is with you.
Transparency
Action Editor: Ayse K. Uskul
Editor: D. Stephen Lindsay
Author Contributions
A. Abu-Akel, N. Dael, and D. Oberfeld contributed equally
to this study and share joint second authorship. The study
was conceptualized by D. Jonauskaite, N. Dael, and
C. Mohr. Data were curated by D. Jonauskaite, A. Chamsed-
dine, and Y. Schrag. Data were analyzed by D. Jonauskaite,
A. Abu-Akel, D. Oberfeld, and J.-P. Antonietti. Data were
collected by D. Jonauskaite, A. Abu-Akel, N. Dael, D.
Oberfeld, A. M. Abdel-Khalek, A. S. Al-Rasheed, V.
Bogushevskaya, E. Chkonia, V. Corona, E. Fonseca-Pedrero,
Y. A. Griber, G. Grimshaw, A. A. Hasan, J. Havelka, M.
Hirnstein, B. S. A. Karlsson, E. Laurent, M. Lindeman, L.
Marquardt, P. Mefoh, M. Papadatou-Pastou, A. Pérez-Albéniz,
N. Pouyan, M. Roinishvili, L. Romanyuk, A. Salgado Montejo,
Y. Schrag, A. Sultanova, M. Uusküla, S. Vainio, G.
Wa˛sowicz, S. Zdravkovic´, M. Zhang, and C. Mohr. D.
Jonauskaite, N. Dael, and C. Mohr served as project admin-
isters. Software was developed by A. Chamseddine. Trans-
lation of the survey was overseen by D. Jonauskaite and
implemented by D. Jonauskaite and Y. Schrag. The study
was supervised by C. Mohr. The figures were created by
D. Jonauskaite, A. Abu-Akel, and D. Oberfeld. The original
draft of the manuscript was written by D. Jonauskaite, A.
Abu-Akel, D. Oberfeld, and C. Mohr. The manuscript was
reviewed and edited by all of the authors, and all the
authors approved the final version for submission.
Declaration of Conflicting Interests
The author(s) declared that there were no conflicts of
interest with respect to the authorship or the publication
of this article.
Funding
This research was supported by the Swiss National Science
Foundation, which provided a Doc.CH fellowship grant to
D. Jonauskaite (P0LAP1_175055) and a project funding
grant to C. Mohr and N. Dael (100014_182138). M. Hirnstein
was supported by a research grant from the Bergen
Research Foundation (BFS2016REK03). Y. A. Griber was
supported by the Russian Foundation for Basic Research
(17-29-09145). The initiation of this research was made
possible through the support of AkzoNobel, Imperial
Chemical Industries, David Elliott and Tom Curwen (Color
R&I Team, Slough, United Kingdom), and Stephanie
Kraneveld (Sassenheim, The Netherlands).
Open Practices
All data and materials have been made publicly available
and can be accessed at https://forsbase.unil.ch/datasets/
dataset-public-detail/15126/2060/ and https://forsbase
.unil.ch/project/study-public-overview/15126/1474/,
respectively. The complete Open Practices Disclosure for
this article can be found at http://journals.sagepub.com/
doi/suppl/10.1177/0956797620948810. The design and
analysis plans for this study were not preregistered. The
complete Open Practices Disclosure for this article can be
found at http://journals.sagepub.com/doi/suppl/10.1177/
0956797620948810. This article has received the badges
for Open Data and Open Materials. More information
about the Open Practices badges can be found at http://
www.psychologicalscience.org/publications/badges.

Universal Color-Emotion Associations in 30 Nations 1259
ORCID iDs
Domicele Jonauskaite https://orcid.org/0000-0002-7513-9766
Ahmad Abu-Akel https://orcid.org/0000-0003-3791-8643
Daniel Oberfeld https://orcid.org/0000-0002-6710-3309
Victoria Bogushevskaya https://orcid.org/0000-0001-8147-3800
Acknowledgments
For their help with the translation of the International Color-
Emotion Survey into their respective languages, we thank
Agnieszka Gawda (Polish), Aurika Jonauskiene˙ (Lithuanian),
Afrodite Kapsaridi (Greek), Bruno Kemm (Spanish and Por-
tuguese), Richard Klein (French), Riina Martinson (Estonian),
Galina Paramei (Russian), Angeliki Theodoridou (Greek),
Evelina Thunell (Swedish), Alessandro Tremea (Italian), and
Yaffa Yeshurun (Hebrew). For their help in distributing and
promoting the survey in their countries, we thank Sanne Bred-
eroo (The Netherlands), Cornelis B. Doorenbos (The Nether-
lands), Tinatin Gamkrelidze (Georgia), Lise Lesaffre (France),
Arzu Memmedova (Azerbaijan), Mariam Okruashvili (Georgia),
C. Alejandro Párraga (Spain), Vilde Johanna Solheim Lie (Nor-
way), Halvor Stavland (Norway), Hedda Andrea Struksnæs
Sørdal (Norway), and Zumrud Sultanova (Azerbaijan).
Supplemental Material
Additional supporting information can be found at http://
journals.sagepub.com/doi/suppl/10.1177/0956797620948810
References
Adams, F. M., & Osgood, C. E. (1973). A cross-cultural study of
the affective meanings of color. Journal of Cross-Cultural
Psychology, 4, 135–157. doi:10.1177/002202217300400201
Benitez-Quiroz, C. F., Srinivasan, R., & Martinez, A. M. (2018).
Facial color is an efficient mechanism to visually transmit
emotion. Proceedings of the National Academy of Sciences,
USA, 115, 3581–3586. doi:10.1073/pnas.1716084115
Berlin, B., & Kay, P. (1969). Basic color terms: Their univer-
sality and evolution. Berkeley: University of California
Press.
Brown, D. (1991). Human universals. New York, NY: McGraw-
Hill.
Cowen, A. S., Fang, X., Sauter, D., & Keltner, D. (2020). What
music makes us feel: At least 13 dimensions organize sub-
jective experiences associated with music across different
cultures. Proceedings of the National Academy of Sciences,
USA, 117, 1924–1934. doi:10.1073/pnas.1910704117
D’Andrade, R., & Egan, M. (1974). The colors of emotion.
American Ethnologist, 1, 49–63. doi:10.1525/ae.1974.1.1
.02a00030
Davidoff, J. B., Davies, I., & Roberson, D. (1999). Colour
categories in a stone-age tribe. Nature, 398, 203–204.
doi:10.1038/18335
Ekman, P., Sorenson, E. R., & Friesen, W. V. (1969). Pan-
cultural elements in facial displays of emotion. Science,
164, 86–88. doi:10.1126/science.164.3875.86
Evarts, A. B. (1919). Color symbolism. The Psychoanalytic
Review, 6, 124–157.
Gao, X., Xin, J. H., Sato, T., Hansuebsai, A., Scalzo, M.,
Kajiwara, K., . . . Billger, M. (2007). Analysis of cross-
cultural color emotion. Color Research and Application,
32, 223–229. doi:10.1002/col.20321
Gendron, M., Roberson, D., van der Vyver, J. M., & Barrett,
L. F. (2014). Perceptions of emotion from facial expres-
sions are not culturally universal: Evidence from a remote
culture. Emotion, 14, 251–262. doi:10.1037/a0036052
Groyecka, A., Witzel, C., Butovskaya, M., & Sorokowski, P.
(2019). Similarities in color preferences between women
and men: The case of Hadza, the hunter-gatherers from
Tanzania. Perception, 48, 428–436. doi:10.1177/03010
06619840937
Hupka, R. B., Zaleski, Z., Otto, J., Reidl, L., & Tarabrina, N. V.
(1997). The colors of anger, envy, fear, and jealousy.
Journal of Cross-Cultural Psychology, 28, 156–171. doi:10
.1177/0022022197282002
Jackson, J. C., Watts, J., Henry, T. R., List, J.-M., Forkel, R.,
Mucha, P. J., . . . Lindquist, K. A. (2019). Emotion seman-
tics show both cultural variation and universal structure.
Science, 366, 1517–1522. doi:10.1126/science.aaw8160
Jäger, G. (2018). Global-scale phylogenetic linguistic infer-
ence from lexical resources. Scientific Data, 5(1), Article
180189. doi:10.1038/sdata.2018.189
Jonauskaite, D., Abdel-Khalek, A. M., Abu-Akel, A.,
Al-Rasheed, A. S., Antonietti, J.-P., Ásgeirsson, Á. G., . . .
Mohr, C. (2019). The sun is no fun without rain: Physical
environments affect how we feel about yellow across
55 countries. Journal of Environmental Psychology, 66,
Article 101350. doi:10.1016/j.jenvp.2019.101350
Jonauskaite, D., Parraga, C. A., Quiblier, M., & Mohr, C.
(2020). Feeling blue or seeing red? Similar patterns of
emotion associations with colour patches and colour
terms. i-Perception, 11(1). doi:10.1177/2041669520902484
Jonauskaite, D., Wicker, J., Mohr, C., Dael, N., Havelka, J.,
Papadatou-Pastou, M., . . . Oberfeld, D. (2019). A machine
learning approach to quantify the specificity of colour–
emotion associations and their cultural differences. Royal
Society Open Science, 6(9), Article 190741. doi:10.1098/rsos
.190741
Luce, R. D. (1963). Detection and recognition. In R. D. Luce,
R. R. Bush, & E. Galanter (Eds.), Handbook of mathemati-
cal psychology (Vol. 1, pp. 103–189). Hoboken, NJ: Wiley.
Lupyan, G., & Dale, R. (2016). Why are there different lan-
guages? The role of adaptation in linguistic diversity.
Trends in Cognitive Sciences, 20, 649–660. doi:10.1016/j
.tics.2016.07.005
Madden, T. J., Hewett, K., & Roth, M. S. (2000). Managing
images in different cultures: A cross-national study of
color meanings and preferences. Journal of International
Marketing, 8, 90–107. doi:10.1509/jimk.8.4.90.19795
Major, D. R. (1895). On the affective tone of simple sense-
impressions. The American Journal of Psychology, 7,
57–77.
Mehr, S. A., Singh, M., Knox, D., Ketter, D. M., Pickens-Jones,
D., Atwood, S., . . . Glowacki, L. (2019). Universality
and diversity in human song. Science, 366(6468), Article
eaax0868. doi:10.1126/science.aax0868

1260 Jonauskaite et al.
Mohr, C., Jonauskaite, D., Dan-Glauser, E. S., Uusküla, M., &
Dael, N. (2018). Unifying research on colour and emotion:
Time for a cross-cultural survey on emotion associations
with colour terms. In L. W. MacDonald, C. P. Biggam, &
G. V. Paramei (Eds.), Progress in colour studies: Cognition,
language, and beyond (pp. 209–222). Amsterdam, The
Netherlands: John Benjamins. doi:10.1075/z.217.11moh
Mylonas, D., & MacDonald, L. (2015). Augmenting basic
colour terms in English. Color Research and Application,
41, 32–42. doi:10.1002/col.21944
Norenzayan, A., & Heine, S. J. (2005). Psychological univer-
sals: What are they and how can we know? Psychological
Bulletin, 131, 763–784. doi:10.1037/0033-2909.131.5.763
Ou, L. C., Yuan, Y., Sato, T., Lee, W. Y., Szabó, F., Sueeprasan,
S., & Huertas, R. (2018). Universal models of colour emo-
tion and colour harmony. Color Research and Application,
43, 736–748. doi:10.1002/col.22243
Palmer, S. E., Schloss, K. B., Xu, Z., & Prado-Leon, L. R.
(2013). Music-color associations are mediated by emotion.
Proceedings of the National Academy of Sciences, USA,
110, 8836–8841. doi:10.1073/pnas.1212562110
Pravossoudovitch, K., Cury, F., Young, S. G., & Elliot, A. J.
(2014). Is red the colour of danger? Testing an implicit
red–danger association. Ergonomics, 57, 503–510. doi:10
.1080/00140139.2014.889220
R Core Team. (2017). R: A language and environment for statis-
tical computing. Retrieved from http://www.R-project.org
Regier, T., Kay, P., & Cook, R. S. (2005). Focal colors are
universal after all. Proceedings of the National Academy
of Sciences, USA, 102, 8386–8391. doi:10.1073/pnas.05
03281102
Scherer, K. R. (2005). What are emotions? And how can they
be measured? Social Science Information, 44, 695–729.
doi:10.1177/0539018405058216
Scherer, K. R., Shuman, V., Fontaine, J. R. J., & Soriano, C. (2013).
The GRID meets the Wheel: Assessing emotional feeling via
self-report. In J. R. J. Fontaine, K. R. Scherer, & C. Soriano
(Eds.), Components of emotional meaning: A sourcebook
(pp. 281–298). Oxford, England: Oxford University Press.
Simmons, J. P., Nelson, L. D., & Simonsohn, U. (2011).
False-positive psychology: Undisclosed flexibility in
data collection and analysis allows presenting anything
as significant. Psychological Science, 22, 1359–1366.
doi:10.1177/0956797611417632
Soriano, C., & Valenzuela, J. (2009). Emotion and colour
across languages: Implicit associations in Spanish colour
terms. Social Science Information, 48, 421–445. doi:10
.1177/0539018409106199
Spence, C. (2011). Crossmodal correspondences: A tutorial
review. Attention, Perception, and Psychophysics, 73,
971–995. doi:10.3758/s13414-010-0073-7
Uusküla, M., & Bimler, D. (2016). How universal are focal col-
ors after all? A new methodology for identifying focal color
terms. In G. Paulsen, M. Uusküla, & J. Brindle (Eds.), Color
language and color categorization (pp. 2–39). Newcastle
Upon Tyne, England: Cambridge Scholars Publishing.
Valdez, P., & Mehrabian, A. (1994). Effects of color on emo-
tions. Journal of Experimental Psychology: General, 123,
394–409. doi:10.1037/0096-3445.123.4.394
Wang, T., Shu, S., & Mo, L. (2014). Blue or red? The effects of
colour on the emotions of Chinese people. Asian Journal
of Social Psychology, 17, 152–158. doi:10.1111/ajsp.12050
Wexner, L. B. (1954). The degree to which colors (hues)
are associated with mood-tones. Journal of Applied
Psychology, 38, 432–435. doi:10.1037/h0062181
Wilms, L., & Oberfeld, D. (2018). Color and emotion: Effects
of hue, saturation, and brightness. Psychological Research,
82, 896–914. doi:10.1007/s00426-017-0880-8