1. Introduction
Across languages and cultures, colors are often used to convey emotional states (Adams & Osgood, Reference Adams and Osgood1973; Jonauskaite et al., Reference Jonauskaite, Abu-Akel, Dael, Oberfel, Abdel-Khalek, Al-Rasheed and Mohr2020a; Soriano & Valenzuela, Reference Soriano and Valenzuela2009). Culture shapes these color–emotion associations, with differences impacting in turn on the perceptive, cognitive and behavioral dimensions of color experience (Elliot et al., Reference Elliot, Maier, Moller, Friedman and Meinhardt2007; Jiang et al., Reference Jiang, Lu, Yao, Yue and Au2013; Mehta & Zhu, Reference Mehta and Zhu2009). Although color–emotion associations have been described for selected languages and cultures, more comparative research is needed to identify universal and culture-specific factors mediating color–emotion associations in the world’s societies (Hanada, Reference Hanada2018).
More specifically, ongoing research has tried to identify specific dimensions of color that might relate to specific dimensions of emotion. Color dimensions that are typically considered include hue, brightness and saturation, whereas emotion dimensions that are usually analyzed include valence (i.e., the pleasantness of the emotion, ranging from strongly negative to strongly positive), arousal (i.e., the level of activation resulting from the emotion, ranging from strongly calmy to strongly excited) and power (i.e., the sense of dominance brought about by the emotion, ranging from strongly powerless to strongly powerful) (for details, see Fontaine et al., Reference Fontaine, Scherer, Roesch and Ellsworth2007; Fontaine, Reference Fontaine, Fontaine, Scherer and Soriano2013; Forceville & Renckens, Reference Forceville and Renckens2013; Littlemore et al., Reference Littlemore, Bolognesi, Julich-Warpakowski, Leung and Sobrino2023; Safarnejad et al., Reference Safarnejad, Ho-Abdullah and Awal2014; Scherer, Reference Scherer2005; Soriano & Valenzuela, Reference Soriano and Valenzuela2009; Takei & Imaizumi, Reference Takei and Imaizumi2022; Yu, Reference Yu2009, among many others). As a general rule, more saturated and brighter colors tend to result in higher arousal and valence ratings, with hue having a lesser impact (Wilms & Oberfeld, Reference Wilms and Oberfeld2018). Also, according to Soriano and Valenzuela (Reference Soriano and Valenzuela2009), arousal and power ratings tend to be less variable than valence ratings, which appear to be more context-sensitive, seemingly because the former are more dependent on our physiology, that is, on the biological mechanisms underlying emotional experience (particularly, if colors can themselves provoke emotional reactions, as observed by, for example, Elliot et al., Reference Elliot, Maier, Moller, Friedman and Meinhardt2007). Therefore, it frequently happens that different cultures associate the same color with different emotions. Hence, white is typically associated with positive emotions in English, French, Spanish, Turkish or Japanese (Demir, Reference Demir2020; Jonauskaite et al., Reference Jonauskaite, Parraga, Quiblier and Mohr2020b; Kaya & Epps, Reference Kaya and Epps2004; Soriano, Reference Soriano2005), whereas it conveys diverse negative meanings in Mandarin, as in baishi ‘white event’ for funeral; baiyan ‘white eyes’ for scorn; or baiyanlang ‘white eyed wolf’ for ungratefulness (He, Reference He2011; Xing, Reference Xing and Xing2009; Zhou, Reference Zhou2010). Still, cross-cultural differences in color–emotion associations do not appear randomly. A recent study by Jonauskaite et al. (Reference Jonauskaite, Abu-Akel, Dael, Oberfel, Abdel-Khalek, Al-Rasheed and Mohr2020a) has found that these associations are more similar in cultures that are geographically closer (although see Jiang et al., Reference Jiang, Lu, Yao, Yue and Au2013 for some striking differences between Mainland China and Hong Kong).
A related concern is the effect of language on color–emotion associations. Whereas the perception of color can be regarded as similar in all human beings, languages differ in the number of basic color terms, ranging from a couple to more than a dozen (Berlin & Kay, Reference Berlin and Kay1969; Gao & Sutrop, Reference Gao and Sutrop2014; Hsieh et al., Reference Hsieh, Kuriki, Chen, Muto, Tokunaga and Shioiri2020; Lillo et al., Reference Lillo, Moreira, Vitini and Martín2007; Paramei, Reference Paramei2005; Uusküla & Sutrop, Reference Uusküla and Sutrop2007; Wu, Reference Wu2011; Xu et al., Reference Xu, Zhu and Benítez-Burraco2023, among others). Moreover, the color lexicon impacts, even if subtly, on color perception and categorization (González-Perilli et al., Reference González-Perilli, Rebollo, Maiche and Arévalo2017; Winawer et al., Reference Winawer, Witthoft, Frank, Wu, Wade and Boroditsky2007). Language is also known to play a crucial role in shaping emotional associations (Lindquist et al., Reference Lindquist, MacCormack and Shablack2015). Color terms and emotions often intertwine, metaphorically or metonymically (Kövecses, Reference Kövecses and Raymond2000; Soriano & Valenzuela, Reference Soriano and Valenzuela2009). As a reflection, to a great extent, of these cross-cultural differences in color–emotion association, different languages can use different color terms for referring to one specific emotion or associate one particular color term with different emotions. For instance, envy is associated with green in Spanish (e.g., estar verde de envidia ‘to be green with envy’), but with red in Mandarin (e.g., yanhong ‘red eyes’); by contrast, red is associated with anger in Spanish (estar rojo de ira ‘to be incandescent with anger’), whereas in Mandarin, anger is black (as in hei zhe lian ‘black (darker) face’) (Li, Reference Li2020).
A final concern is whether the mode of presentation of color impacts color–emotion associations. A study by Wang et al. (Reference Wang, Shu and Mo2014) with Mandarin speakers found differences in valence ratings for red and blue depending on whether the stimuli encompassed color patches or color terms. These authors further found that whereas some color patches activate both biologically based associations (i.e., metonymical, for example, red to blood, fire or danger) and culturally based associations (i.e., metaphorical, for example, red to envy), others can activate only one type. By contrast, a recent study by Jonauskaite et al. (Reference Jonauskaite, Parraga, Quiblier and Mohr2020b) with French speakers found no differences between color patches and color terms in a color–emotion association task.
In this paper, we aim to delve into the potential differences in color–emotion association between native speakers of Spanish and Mandarin. Contrary to previous research that focused either on a few colors (e.g., red and blue in Wang et al., Reference Wang, Shu and Mo2014) or a few emotions (e.g., happiness and sadness in Takei & Imaizumi, Reference Takei and Imaizumi2022), here we examine an ample set of colors, as well as a comprehensive set of emotions. In addition, we try to determine the factor(s) that promote(s) the use of one (or more) specific color(s) for referring to a particular emotion. In order to do so, we consider the three fundamental dimensions of emotion as characterized above: valence, arousal and power. Finally, we try to clarify whether color–emotion associations differ when colors are presented visually or verbally. Overall, we expect to find some (remarkable) differences in the color–emotion association between Spanish and Mandarin, considering the notable typological differences between the two languages, the noteworthy cultural differences between speakers and the noticeable distance between the areas where the two languages are spoken.
2. Materials and methods
For examining the research questions posited above, we mostly replicated the study by Jonauskaite et al. (Reference Jonauskaite, Parraga, Quiblier and Mohr2020b) with our own sample of monolingual speakers of Spanish and Mandarin. In brief, we asked participants to associate color terms in their language (Task 1) or color patches (Task 2) with a set of emotion concepts (and intensities of emotion) as found in the Geneva Emotion Wheel (GEW), which is a tested instrument aimed to measure emotional reactions to objects, events and situations (Scherer, Reference Scherer2005; Scherer et al., Reference Scherer, Shuman, Fontaine, Soriano, Fontaine, Scherer and Soriano2013).
2.1. Stimuli
For Task 1, we used the 11 basic color terms in Spanish and their Mandarin equivalents (Spanish – negro ‘black’, blanco ‘white’, rojo ‘red’, amarillo ‘yellow’, verde ‘green’, azul ‘blue’, marrón ‘brown’, naranja ‘orange’, rosa ‘pink’, morado ‘purple’, gris ‘gray’; Mandarin – hei ‘black’, bai ‘white’, hong ‘red’, huang ‘yellow’, lü ‘green’, lan ‘blue’, zong/he ‘brown’, cheng/jü ‘orange’, fen ‘pink’, zi ‘purple’, hui ‘gray’). We also included Spanish celeste ‘light blue’ (Mandarin qianlan) and Spanish lila ‘light purple’ (Mandarin qianzi), since in our previous research we found evidence that celeste could be a 12th basic color in Spanish, whereas lila is quite consistently used by Spanish speakers to refer to light purple (with morado being used for darker tones) (see Xu et al., Reference Xu, Zhu and Benítez-Burraco2023 for details). This differential usage of lila and morado has been confirmed by independent research by Jonauskaite and colleagues (e.g., Epicoco et al., Reference Epicoco, Mohr, Uusküla, Quiblier, Meziane, Laurent and Jonauskaite2024; Uusküla et al., Reference Uusküla, Mohr, Epicoco and Jonauskaite2023). For Task 2, we presented participants with 13 color patches corresponding to the best examples of the color terms used in Task 1. These are the color patches that subjects use more consistently for referring to a specific color, as found in our previous research too (see Xu et al., Reference Xu, Zhu and Benítez-Burraco2023 for details). Color patches were retrieved from The World Color Survey (Kay et al., Reference Kay, Berlin, Maffi, Merrifield and Cook2010; https://www1.icsi.berkeley.edu/wcs/data.html). The Munsell values for each of the color patches used in this task can be found in Supplementary Table S1. With regard to the set of emotions considered in both tasks and their distinctive dimensions (i.e., valence: positive–negative, arousal: high–low and power: strong–weak), we made use, as noted, of the GEW, version 3.0 (Scherer, Reference Scherer2005; Scherer et al., Reference Scherer, Shuman, Fontaine, Soriano, Fontaine, Scherer and Soriano2013) (see Supplementary Table S2 for details), via the interface that is usually considered as the easiest to use and understand by participants (see Caicedo & Van Beuzekom, Reference Caicedo and Van Beuzekom2006; Warpechowski et al., Reference Warpechowski, Orzeszek and Nielek2019 for details). As for the Spanish and Mandarin versions of the emotion terms, we did not translate them by ourselves, but used instead the versions provided by Jonauskaite et al. (Reference Jonauskaite, Parraga, Quiblier and Mohr2020b) in their online survey (http://www2.unil.ch/onlinepsylab/colour/main.php).
2.2. Participants
The Spanish sample consisted of 50 monolingual speakers of Castilian (i.e., European) Spanish (22 males) with a mean age of 28.56 years (95% confidence interval [CI] = [26.04, 31.08]). All of them lived in Barcelona (a big city in Northeastern Spain), the place where they were tested, although they had different origins (Madrid, Barcelona or Seville). 25 individuals (10 males) participated in Task 1 (M age = 28.53, 95% CIage = [24.36, 32.70], range: 18-58) and the remaining 25 individuals (12 males) participated in Task 2 (M age = 28.6, 95% CIage = [23.92, 33.28], range: 18-62). An independent samples t test showed that the age was comparable between both groups (t(48) = 0.03, p = .98, d = .009) as well as the gender (χ2 (1) = .08, p = .78, V = .04). The Mandarin sample consisted of 50 monolingual speakers of Mandarin (25 males) with a mean age of 29.34 years (95% confidence interval [CI] = [26.69, 31.99]). All of them lived in Chengdu (a big city in the province of Sichuan, Southwestern China), where they were tested, although they also had different origins (Chengdu, Wuhan, Qingdao and Beijing). 25 subjects (10 males) took part in Task 1 (M age = 29.04, 95% CIage = [24.95, 33.13], range: 18-59), and the remaining 25 subjects (15 males) participated in Task 2 (M age = 29.64, 95% CIage = [25.96, 33.32], range: 20-55). An independent samples t test showed that the age was comparable between groups (t(48) = .23, p = .82, d = .07), as also was the gender distribution (χ2 (1) = 1.28, p = .26, V = .16). Additionally, an independent samples t test showed that the age of the participants was comparable between the two groups (Spanish vs. Mandarin) participating in Task 1 (t(48) = 0.18, p = .85, d = .053) and Task 2 (t(48) = 0.36, p = .72, d = .10). The same was true for gender distribution (Task1: (χ2 (1) = 0, p = 1, V =0); Task 2: (χ2 (1) = .32, p = .57, V = .08)). Using different subsets of participants for these two tasks helps avoiding biases and confounding effects. Participants were not introduced to the subject of the research until the end of the experiment, in order to avoid potential priming effects.
2.3. Procedure
Before starting with the two tasks, the experimenter first introduced the procedure and participants signed an informed consent document. Participants then received a booklet (15 pages in total) containing a one-page introduction to the GEW, 13 identical wheels and a sheet for gathering some personal data: age, sex, education level, the language(s) they spoke and a confirmation that they did not suffer from color vision problems.
In Task 1, color terms were randomly presented to the participants in writing, using white pieces of sheet and black ink. In Task 2, color patches (15 x 15 cm) were also presented randomly on the screen of a laptop (resolution: 2560x1600 (227 ppi); brightness: 300 nits; contrast ratio: 900:1; standard color gamut (sRGB)) in sufficient natural daylight (thus avoiding direct sunlight or deep shade) and using a gray base as the background (N5, Munsell color notation). Participants observed color patches at a distance of around 60 cm with an average visual angle of 3.5 degrees.
During both tasks, Correlated Color Temperature (CCT) oscillated between 4000 and 4500 K and illumination ranged from 1750 to 2000 lux. CCT is a measure of the quality of the white light, with low values corresponding to ‘warm’ lights (which are richer in low-frequency wavelengths like reds and yellows) and with high values being typical of ‘cool’ lights (that are richer in high-frequency blues and violets). Illumination is a measure of the intensity of the light. We secured the same experimental conditions in both testing places (Spain and China).
As noted, depending on the task, participants had to associate color terms or color patches with one (or more than one) of the 20 types of emotion included in the GEW (for details, see Figure 1 and Supplementary Table S2). Participants were allowed to associate colors with emotions not included in the GEW. They could also refuse to associate colors with any emotion. Additionally, they were requested to rate the intensity of the emotion(s). For this, they had to choose one among the different circle sizes also included in the GEW, which range from the smallest one (for the weakest emotion, coded as 1) to the biggest (for the strongest emotion, coded as 5). For recording their choices, they just needed to make a checkmark on the circles in the GEW sheets. Since color terms or color chips were presented in a randomized order that changed from participant to participant, the experimenter labeled the order of each playback as well as the participant’s results.
The GEW used for testing the association of color terms or patches with emotions by participants. Emotions are displayed in the outer part of the wheel (in English (A), Spanish (B) and Mandarin (C)). The intensity of each emotion correlates with the size of the circles in the inner part of the wheel (reproduced from Jonauskaite et al., Reference Jonauskaite, Abu-Akel, Dael, Oberfel, Abdel-Khalek, Al-Rasheed and Mohr2020a).
2.4. Data analysis
For both tasks, we first determined the number of participants associating a specific color term (Task 1) or color patch (Task 2) with a specific emotion. We also calculated the number of participants associating a specific color with a specific emotion regardless of the mode of presentation. If a particular color term or color patch was not associated with any emotion, we coded this as 0. If a color was associated with an emotion not included in the GEW, we coded this as a missing value. We then used autocluster analyses to uncover the most prominent color–emotion associations for each task and for each language. Autocluster analyses are aimed to find and categorize inherent patterns within datasets in the absence of manual input. We employed Fisher’s test to look for significant differences between conditions (i.e., between types of emotion, between languages and between modes of presentation). We further used the Pearson Correlations Coefficient (PCC) to determine the similarity in color–emotion associations between languages and between tasks. In this latter case, the higher the PCC value, the more similarity (for p < 0.05). Second, for each task, we calculated the average intensity assigned to each color–emotion association. If no emotion was associated with the color, or if the subject associated the color with an emotion not shown in the GEW, we coded this as a missing value, since intensity ranges from 1 to 5. We then conducted a series of independent samples t tests to find differences in color–emotion associations and intensity ratings between speakers of Spanish and Mandarin, and also between Task 1 and Task 2. Finally, we conducted a mixed-design multivariate analysis of variance (MANOVA) to identify the dimension(s) that impact(s) mostly on color–emotion association. Accordingly, we set the broadness of valence (i.e., the number of positive emotions and negative emotions), arousal (i.e., the number of high emotions and low emotions) and power (i.e., the number of strong emotions and weak emotions) as dependent variables, with color presentation mode, color type and language as independent variables (see Supplementary Table S2 for details). Additionally, we calculated the biases in valence, arousal and power when it comes to rate emotions. Our aim was to determine whether specific colors tend to significantly evoke specific emotional responses in any of these two languages. Valence bias is the difference between the positive mean and the negative mean for valence ratings by participants and was calculated as follows: Valence bias = (sum of the intensity of all negative emotions / number of negative emotions) – (sum of the intensity of all positive emotions / number of positive emotions). We calculated the arousal bias and the power bias similarly: Arousal bias = (sum of the intensity of all low arousal emotions / number of low emotions) - (sum of the intensity of all high arousal emotions / number of high emotions) and Power bias = (sum of the intensity of all weak power emotions / number of weak emotions) – (sum of the intensity of all strong power emotions / number of strong emotions). The intensity of negative emotions, low arousal or weak power all received a negative sign, whereas the intensity of positive emotions, high arousal or strong power, all received a positive sign. Individual mixed-design ANOVAs were then conducted for each of these three dimensions of emotion to determine the effect of diverse factors (languages, color and mode of presentation), with the aim of understanding how these factors impact, solely or coordinately, on each dimension of emotion.
3. Results
3.1. Type of emotion
Autocluster analyses revealed that Spanish speakers tend to agree more than Mandarin speakers when it comes to associating specific colors with specific emotions in general, with this difference being significant ((F = .43, 95% CI = [0.34, 0.56], p < .001). The strongest differences between languages were observed for white (F = .26, 95% CI = [0.10, 0.65], p = .005), red (F = .27, 95% CI = [0.11, 0.69], p = .007), yellow (F = .25, 95% CI = [0.10, 0.64], p = .004), blue (F = .29, 95% CI = [0.12, 0.73], p = .009) and light blue (F = .33, 95% CI = [0.13, 0.85], p = .02). For example, Spanish speakers tend to associate the color white with relief (62% of cases) and contentment (44%), whereas Mandarin speakers are more likely to associate it with sadness (34%) and disappointment (22%). Likewise, Spanish speakers tend to associate blue with sadness (54% of cases) and relief (44%), whereas Mandarin speakers preferentially associate it with pleasure (24%) and compassion (22%) (see Supplementary Table S3 for details). At the same time, both Spanish speakers and Mandarin speakers converge more within groups on color–emotion associations when colors are presented verbally instead of visually (F = .61, 95% CI = [0.48, 0.78], p < .001 for Spanish; F = .55, 95% CI = [0.43, 0.71], p < .001 for Mandarin). This difference is mostly attributable to the remarkable differences observed for some specific colors: red and light purple for Spanish speakers, and red, white, yellow and orange for Mandarin speakers. Accordingly, when red is presented verbally, 88% of Spanish speakers associate this color with anger, but this number goes down to only 28% when red is presented visually. Similarly, Mandarin speakers strongly associate white with sadness, but only if the color is presented verbally (52% vs 16% if it is presented as a color patch) (see Supplementary Table S3 for details).
Figure 2 shows the heatmaps for general color–emotion associations (above), color term–emotion associations (middle) and color patch–emotion associations (below) for both languages.
Heatmaps of color–emotion associations for Spanish and Mandarin. (A) Color–emotion associations. (B) Color term–emotion associations. (C) Color patch–emotion associations. In the heatmaps, more intense reds correspond to stronger associations between colors and emotions.
We performed Pearson Correlations analyses and found a similarity between both languages in overall color–emotion associations (r(258) = .69, p < .001), color term–emotion associations (r(258) = .55, p < .001) and color patch–emotion associations (r(258) = .62, p < .001). However, significant differences were observed for specific colors (see Table 1). Across these three conditions, differences were significant for light purple only.
Pearson Correlation Coefficients (PCCs) for the similarity in color–emotion associations between speakers of Spanish vs Mandarin
Note: Three conditions are compared: overall color–emotion associations, color term–emotion associations (Task 1) and color patch–emotion associations (Task 2). Significance levels are expressed with asterisks (* for p ≤ 0.05, ** for p ≤ 0.01 and *** for p ≤ 0.001). Non-significant correlations are highlighted in bold.
With regard to the mode of presentation, Pearson Correlations analyses indicate that both Spanish speakers (r(258) = .66, p < .001) and Mandarin speakers (r(258) = .73, p < .001) tend to associate the same emotions with the same colors, with the exception of purple and light purple, which are associated with different emotions in each language, whether presented verbally or visually (see Supplementary Table S4 for details).
3.2. Intensity of emotion
Concerning the intensity of emotion, an independent samples t test showed that Spanish speakers attribute a higher mean intensity to colors than Chinese speakers do (M = 3.72, SD = .86 for Spanish; M = 3.27, SD=1.0 for Mandarin), with this difference being statistically significant between groups (t (1164) = −8.17, p < .001, d = −.48). At the same time, both groups tend to attribute a higher mean intensity to color terms than to color patches (M = 3.82, SD = .85 vs M = 3.62, SD = .87 for Spanish; M = 3.43, SD = .94 vs. M = 3.12, SD = 1.04 for Mandarin), with this difference being statistically significant between groups too (t (615) = −2.78, p = .006, d = −.22 for Spanish; t (547) = −3.78, p = .0001, d = −.32 for Mandarin). Still, this disparity is mostly driven by differences in selected colors that differ between the two languages: black, gray and yellow for Spanish; green, orange and purple for Mandarin (see Supplementary Table S5 for details).
3.3. Dimensions of emotion
Finally, a mixed-design MANOVA analysis of our data indicated that Spanish speakers tend to associate more different emotion concepts with each individual color (M= 3.46, SD = 2.24 for Spanish; M = 1.84, SD = 1.93 for Mandarin), F(1, 1229) = 36.41, p < .001 (see Supplementary Table S6 for details). The Pillai’s trace values of the MANOVA (which are used to measure the effect of sets of independent variables on sets of dependent variables in small samples like ours) further suggested that this difference is also significant for the three basic dimensions considered in our study: valence, arousal and power (Pillai’s trace value = .15, F(6, 1224) = 34.61, p < .001, ηp2 = .15). Likewise, our analysis suggested that for both languages, there is a significant effect of the mode of presentation (Pillai’s trace value = .21, F(18, 3672) = 15.54, p < .001, ηp2 = .07), so that participants tend to give higher numbers to color terms than to color patches. Also, for both languages, the interaction between color and mode of presentation was significant too (Pillai’s trace value = .11, F(72, 7086) = 1.8, p < .001, ηp2 = .02). This means that the effect of the mode of presentation depends on the color. For example, Spanish speakers significantly associate more emotions with red when the color is presented verbally (N = 107, M = 4.28, SD = 2.03) compared to visually (N = 81, M = 3.24, SD = 1.71). This is also observed among Mandarin speakers (Color term: N = 76, M = 3.17, SD = 3.29, Color patch: N = 42, M = 1.68, SD = 1.07). Lastly, for both languages, the three-way interaction between language, mode of presentation and color was also significant (Pillai’s trace value = .09, F(66, 7086) = 1.66, p = .001, ηp2 = .02). This means that the number of emotion types depends on the combined effect of language, mode of presentation and color.
At the same time, an individual mixed-design ANOVA on valence bias revealed no significant difference between groups (F (1, 1138) = .49, p = .48). This means that colors evoke, on average, the same positive or negative emotions in both languages. Still, we found a significant interaction between language and color (F (12, 1138) = 3.68, p < .001). This suggests that some specific colors impact valence bias differently depending on the language. To identify the most impacting colors in each language, we carried out post hoc comparisons. We found that red evokes more negative emotions in Spanish speakers compared to Mandarin speakers (β = −1.76, SE = .71, p = .01), whereas white evokes more positive emotions in speakers of Spanish (β = 2.12, SE = .75, p = .004) (see Figure 3A and Supplementary Table S7). We also found a significant interaction between color and mode of presentation, but only for the Spanish group (F (12, 591) = 2.48, p = .004), not for the Mandarin group (F (12, 523) = 1.13, p = .34). For instance, Spanish speakers significantly associate blue with more negative emotions when the color is presented verbally (M = −0.37, SD = 2.87) compared to visually (M = 2.78, SD = 2.5) (see Figure 3B and Supplementary Table S7).
(A) Valence of the emotions associated with different colors in Spanish (above) and Mandarin (below). (B) Comparison of valence according to the two modes of presentation (color terms vs. color patches) in Spanish (above) and Mandarin (below).
Likewise, the difference was not significant for arousal bias (F (1, 1138) = 8.48, p = .52), which means that colors evoke, on average, the same arousal (either high or low) in both languages. However, as with valence, we also observed a significant interaction between language and color (F (12, 1138) = 2.12, p < .001). Post hoc comparisons revealed that several colors evoke a lower arousal in the Mandarin group compared to Spanish speakers: red (β = −1.49, SE = .7, p = .03), yellow (β = −1.69, SE = .72, p = .02) and purple (β = −1.47, SE = .73, p = .04) (see Figure 4 and Supplementary Table S8). However, contrary to valence, we found no significant interaction between color and mode of presentation (F (12, 523) = 1.13, p = .34 for Spanish; F (12, 523) = 1.13, p = .34 for Mandarin).
(A) Arousal of the emotions associated with different colors in Spanish (above) and Mandarin (below). (B) Comparison of arousal according to the two modes of presentation (color terms vs. color patches) in Spanish (above) and Mandarin (below).
Finally, the difference was not significant either for power (F (1, 1138) = 1.71, p = .19). This entails that for both languages, color evokes, on average, the same emotional strength. Nonetheless, we found a significant interaction between language and color (F (12, 1138) = 2.35, p = .006). Post hoc comparisons revealed that light blue evokes weaker emotions in Spanish speakers compared to Mandarin speakers (β = −1.75, SE = .72, p = .02) (see Figure 5 and Supplementary Table S9). Similarly to arousal (but not to valence), we found no significant interaction between color and mode of presentation (F (12, 591) = 1.21, p = .27 for Spanish; F (12, 523) = 1.69, p = .06 for Mandarin).
(A) Power of the emotions associated with different colors in Spanish (above) and Mandarin (below). (B) Comparison of power according to the two modes of presentation (color terms vs. color patches) in Spanish (above) and Mandarin (below).
4. Discussion
Overall, our findings point to some commonalities and some differences in how colors are used to convey emotions by speakers of Spanish and Mandarin.
To begin with, Mandarin speakers tend to be more variable than Spanish speakers when it comes to associating specific colors with specific emotions, to the extent that only two color–emotion pairs (gray–sadness and black–fear) are used by more than 50% of our participants. One reason for this could be that basic color terms are also employed less consistently by Mandarin speakers generally (Xu et al., Reference Xu, Zhu and Benítez-Burraco2023). Another reason could be that Chinese people are not traditionally in the habit of telling others what they are thinking or feeling, mostly because of their Confucian cultural background, which praises zhongyong or ‘middle way’, that is, being moderate and balanced in life (Fu, Reference Fu2012).
We further found that overall color–emotion associations are similar in both groups, although significant differences can be observed for specific colors. This finding is in line with previous findings by Adams and Osgood (Reference Adams and Osgood1973), who identified similar patterns of color–emotion associations across 23 different cultures, but also with more recent research by Jonauskaite et al. (Reference Jonauskaite, Epicoco, Al-Rasheed, Aruta, Bogushevskaya, Brederoo, Corona, Fomins, Gizdic, Griber, Havelka, Hirnstein, John, Jopp, Karlsson, Konstantinou, Laurent, Marquardt, Mefoh, Oberfeld and Mohr2024). In our study, significant differences concern white and blue. White was overwhelmingly associated with positive emotions by Spanish speakers, including relief, contentment, pleasure, interest or compassion, as previously observed by Soriano (Reference Soriano2005), and similarly to English (Kaya & Epps, Reference Kaya and Epps2004) or Turkish (Demir, Reference Demir2020). By contrast, Mandarin speakers tend to associate white with negative emotions, including sadness or disappointment, as also observed by Jonauskaite et al. (Reference Jonauskaite, Wicker, Mohr, Dael, Havelka, Papadatou-Pastou, Zhang and Oberfeld2019, Reference Jonauskaite, Abu-Akel, Dael, Oberfel, Abdel-Khalek, Al-Rasheed and Mohr2020a) or Wang (Reference Wang2013). In Chinese culture, white is also linked to death or despise (He, Reference He2011; Wang, Reference Wang2013; Xing, Reference Xing and Xing2009; Zhou, Reference Zhou2010). For blue, the trend was the opposite. It was associated with positive feelings by Mandarin speakers, including pleasure or compassion, while it was associated with negative emotions by Spanish speakers, including sadness and disappointment. This finding is also in line with previous research by Soriano and Valenzuela (Reference Soriano and Valenzuela2009), but contrasts with findings by Bazán (Reference Bazán2018), who attested the association of blue with concepts like calm, peace or relaxation. Interestingly, red evoked potentially contradictory emotions in our Spanish speakers, including love, anger and hate, this is also in line with previous studies (e.g., Soriano, Reference Soriano2005). By contrast, it evoked more similar (and positive) emotions in the Mandarin group, like joy, love, contentment or admiration, in line with previous findings too (Jiang et al., Reference Jiang, Lu, Yao, Yue and Au2013). In Chinese culture, red is often used metaphorically to represent positive things, as in hongxishi ‘red affair’ (related to wedding) or hongren ‘red man’ (to refer to a popular man) (Lee, Reference Lee2017), whereas in Spanish, as in many other languages, it is frequently used to refer metonymically to anger, as in verlo todo rojo (‘to see in anger’) or estar rojo de ira (‘to be in anger’) (see Soriano, Reference Soriano2005 for discussion). Interestingly too, in our sample, colors like purple, light purple and brown did not get the negative connotations found by previous research, as observed for example in English, in which purple is commonly associated with rage (see Steinvall, Reference Steinvall, MacLaury, Paramei and Dedrick2007). Actually, 32% of our Mandarin speakers did not associate brown with any emotion.
Regarding the different dimensions of emotions considered in our study, our results are in line with the study by Soriano and Valenzuela (Reference Soriano and Valenzuela2009), who also found that valence is more variable than arousal and power. These authors explained this variability by the greater sensitivity of valence to cultural differences. Overall, we found in both groups of speakers a negative bias toward black, gray and brown, but a positive bias toward chromatic colors. This is also in line with previous findings by Jonauskaite et al. (Reference Jonauskaite, Abu-Akel, Dael, Oberfel, Abdel-Khalek, Al-Rasheed and Mohr2020a, Reference Jonauskaite, Parraga, Quiblier and Mohr2020b, Reference Jonauskaite, Camenzind, Parraga, Diouf, Ducommun, Müller and Mohr2021), and Wilms and Oberfeld (Reference Wilms and Oberfeld2018). These latter authors argue that valence depends on a combination of hue, saturation and brightness, with increased lightness resulting in higher positive ratings. We have found evidence of this effect, since in our sample, for example, pink is associated with more positive emotions than red, pretty much as light blue is associated with more positive emotions than blue. Still, we detected some potential effect of cultural factors on valence, since, for example, red receives quite positive appreciations by both groups, particularly by the Mandarin group. If the valence of a color depended on environmental factors only, we would expect that red received more negative values (and more similarity between samples), since red is universally related to danger (Elliot et al., Reference Elliot, Maier, Moller, Friedman and Meinhardt2007; Elliot & Maier, Reference Elliot and Maier2014; Pravossoudovitch et al., Reference Pravossoudovitch, Cury, Young and Elliot2014), and particularly, makes animals and humans alert, nervous and anxious (Wells et al., Reference Wells, McDonald and Ringland2008). This potential effect of culture on color valence is particularly notable in (and expected for) the Mandarin sample, since in Chinese culture red is highly appreciated (see Chung, Reference Chung2011; Lee, Reference Lee2017 for discussion and examples). Regarding arousal and power, we also found a mixture of universality and cultural dependency. Accordingly, on average, colors evoke the same emotional arousal and power in both groups of speakers, but at the same time there are significant differences for some specific colors, which seemingly have a cultural origin. For instance, both groups give high arousal values to red, as observed in many other languages and cultures (Soriano & Valenzuela, Reference Soriano and Valenzuela2009; Wilms & Oberfeld, Reference Wilms and Oberfeld2018; Zieliński, Reference Zieliński2016). However, Mandarin speakers tend to give lower arousal values to yellow and purple compared to the Spanish-speaking group. In Chinese culture, yellow and purple are the colors related to the Imperial power (e.g., the emperor’s clothes were called huangpao ‘yellow robes’) (Xing, Reference Xing and Xing2009). Yellow is also a symbol of harvest. These circumstances might contribute to explaining why speakers tend to associate yellow with contentment. By contrast, Spanish speakers tend to associate this color with joy, hence probably the higher arousal values it gets by this group.
Finally, we found no significant impact of how colors are presented (verbally vs. visually) on the type of emotion they are associated with. Certainly, language plays an important role in how emotions are construed, and ultimately, in how we interpret and understand sensations of the body (Giraud et al., Reference Giraud, Marelli and Nava2023; Jonauskaite et al., Reference Jonauskaite, Wicker, Mohr, Dael, Havelka, Papadatou-Pastou, Zhang and Oberfeld2019, Reference Jonauskaite, Abu-Akel, Dael, Oberfel, Abdel-Khalek, Al-Rasheed and Mohr2020a; Lindquist et al., Reference Lindquist, MacCormack and Shablack2015). However, this effect seems to be subtle, and it is more noticeable with regard to the intensity of emotion than to the type of emotion. Accordingly, we found that both Spanish speakers and Mandarin speakers tend to associate more intense emotions with color terms than with color patches, although this difference depends on the effect of different colors for each group. This somehow contrasts with previous results by Jonauskaite et al. (Reference Jonauskaite, Parraga, Quiblier and Mohr2020b), who found that for their French speakers, a difference between terms and patches is only observed with black. Uusküla and Eessalu (Reference Uusküla and Eessalu2018) have suggested that this difference might result from some methodological caveat, particularly, if the color patch used in the experiment is not representative of the color. We do not think this is the case with our samples, since in our previous research we found that both Spanish and Mandarin speakers are able to name color patches consistently (see Xu et al., Reference Xu, Zhu and Benítez-Burraco2023 for details). In our opinion, this differential effect of color terms vs color patches might derive from the richer connotative meanings associated with the former, since there are usually dozens of linguistic expressions in which color terms appear, each with a different meaning. Also, color terms normally refer to a whole portion of the chromatic space, so that different shades can be associated with different types of emotions while still being referred to with the same color term. By contrast, when confronted with a color patch, subjects see one specific shade only (see Jonauskaite et al., Reference Jonauskaite, Abu-Akel, Dael, Oberfel, Abdel-Khalek, Al-Rasheed and Mohr2020a for discussion). Finally, and related to this last concern, our finding that both groups tend to attribute higher mean intensity values to color terms than to color patches might be also due to the more abstractness of the former. Previous research suggests that abstractness (e.g., why things happen) results in more intense self-conscious emotions (like guilt and shame) and less intense basic emotions (like anger and fear), whereas concreteness (e.g., how things happen) provokes the opposite response (see Bornstein et al. Reference Bornstein, Katzir, Simchon and Eyal2021 for discussion). Still, in our research, we found no evidence of a significant interaction between the mode of presentation and the type of emotion as far as intensity is concerned. Lastly, regarding the potential effect of the mode of presentation on valence bias, our study is in line with Wang et al.’s (Reference Wang, Shu and Mo2014), who found that red is rated equally (positive) under both modes of presentation.
For future studies aimed to delve into this intriguing cross-cultural pattern, and specifically, into the similarities and differences between Spanish and Mandarin in color–emotion association, the color purple should be of particular interest. As noted, the status of light purple and dark purple is still evolving, and there is low agreement between the two languages on the emotions associated with these two colors. Furthermore, this has been confirmed by the more comprehensive and insightful research conducted by Jonauskaite and colleagues (e.g., Epicoco et al., Reference Epicoco, Mohr, Uusküla, Quiblier, Meziane, Laurent and Jonauskaite2024; Jonauskaite et al., Reference Jonauskaite, Abu-Akel, Dael, Oberfel, Abdel-Khalek, Al-Rasheed and Mohr2020a, Reference Jonauskaite, Parraga, Quiblier and Mohr2020b, Reference Jonauskaite, Camenzind, Parraga, Diouf, Ducommun, Müller and Mohr2021, Reference Jonauskaite, Epicoco, Al-Rasheed, Aruta, Bogushevskaya, Brederoo, Corona, Fomins, Gizdic, Griber, Havelka, Hirnstein, John, Jopp, Karlsson, Konstantinou, Laurent, Marquardt, Mefoh, Oberfeld and Mohr2024; Uusküla et al., Reference Uusküla, Mohr, Epicoco and Jonauskaite2023).
5. Conclusion
Summarizing our results and our discussion above, the main findings of our research about color–emotion association in Spanish vs Mandarin are the following:
-
- Spanish speakers tend to agree more than Mandarin speakers when it comes to associating specific colors with specific emotions;
-
- there are no overall significant differences between both groups with regard to the type of emotion associated with colors (when presented either verbally or visually); at the same time, significant differences exist for specific colors;
-
- as a general rule, Spanish speakers attribute higher intensity values to colors; at the same time, both groups attribute higher intensity values to color terms compared to color patches, although this difference is driven by different colors in each group;
-
- as a general rule, Spanish speakers tend to associate a higher number of emotion values (valence, arousal and power) with colors; at the same time, both groups attribute higher numbers to color terms than to color patches;
-
- on average, colors evoke the same emotional polarity in both languages, although significant differences exist for specific colors;
-
- on average, colors evoke the same emotional arousal in both languages, although significant differences exist for specific colors;
-
- on average, colors evoke the same emotional strength in both languages, although significant differences exist for specific colors.
Overall, our study points to a mixed pattern of universality and culture-specificity regarding how colors are used for conveying emotions.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/langcog.2024.52.
Data availability statement
The raw data generated by this research are presented as Supplementary Materials.
Author contribution
M.X. conducted the research, analyzed the data, reviewed the available literature and wrote the paper. J.Z. analyzed the data, reviewed the available literature and wrote the paper. A.B-B. conceived the paper, reviewed the available literature and wrote the paper. All authors approved the final version of the manuscript.
CRediT author statement
Mingshan Xu: Methodology, Investigation, Data curation, Formal analysis, Validation, Writing; Jingtao Zhu: Formal analysis, Writing; Antonio Benítez-Burraco: Conceptualization, Writing, Supervision.
Funding statement
This research was conducted in the absence of external funding.
Competing interest
The authors declare none.
Ethical standard
The participants in the experiments discussed in the paper consented to be part of the research and authorized the authors to publicize the results (anonymously) in scientific presentations and publications.
Open access
