Cross-Modal Associations Between Real Tastes and Colors

Cross-Modal Associations Between Real Tastes and Colors Abstract People make reliable and consistent matches between taste and color. However, in contrast to other cross-modal correspondences, all of the research to date has used only taste words (and often color words too), potentially limiting our understanding of how taste–color matches arise. Here, participants sampled the 5 basic tastes, at 3 concentration steps, and selected their best matching color from a color wheel. This test was repeated, and in addition, participants evaluated the valence of the taste and their color choice, as well as the qualities/intensities of the taste stimuli. Participants were then presented with taste names and asked to generate the best matching color name, as well as reporting how they made their earlier choices. Color selections were reliable and consistent and closely followed those based on taste word matches obtained in this and prior studies. Most participants reported basing their color choices on their associated taste-object (often foods). There was marked similarity in valence between taste and color choices, and the saturation of color choices was related to tastant concentration. We discuss what drives color–taste pairings, with learning suggested as one possible mechanism. gustation, color, cross-modal Introduction Cross-modal correspondences have been documented in all sensory domains, but are less explored in the chemical senses. One exception is taste and color. Taste is composed of at least 5 basic qualities, sweet, sour, bitter, salty, and umami (which alone tastes “meaty”; Duffy et al. 2009). Investigators, in 7 studies, have found evidence of consistent mappings between particular tastes and colors (O’Mahony 1983; Heller 1999; Koch and Koch 2003; Tomasik-Krotki and Strojny 2008; Wan et al. 2014; Spence et al. 2015; Woods and Spence 2016). In each of these studies, participants were asked to determine which color (including white, gray, and black), using patches or words, went best with which taste “name” (e.g., sweet) or vice versa. With varying sample sizes and adult participants drawn from many countries, the results have been surprisingly consistent. The most consistent matches (the fastest, Woods et al. 2016, and congruent based on the pattern of event related potentials, Skrandies and Reuther 2008) are as follows: red and pink with sweet, white and blue with salty, green and yellow with sour, and black and green with bitter. Umami (meaty), which has only been included in 2 studies (Tomasik-Krotki and Strojny 2008; Wan et al. 2014), is less consistently linked to any 1 color. What is most surprising about these findings is that they are based on taste “words,” and often color words too, especially as the literature already suggests that different cross-modal effects may ensue when taste words are used rather than tastants (Velasco et al. 2014). This reliance on taste words raises several issues. First, and as already apparent, umami (meaty) is not clearly associated with any particular color, which might have much to do with people being unsure of its specific taste. Another problem emerges for bitter and sour tastes. Although they are readily discriminable, their labels are often confused (e.g., Robinson 1970), which might explain their shared color association—green. Second, taste experiences vary in magnitude. Milk is not very sweet and potatoes are not very salty, but chocolate milk is very sweet and potato chips are very salty—so when participants are making judgments about sweet and salty, and the other tastes too, what particular intensity are they basing their judgments on? Finally, and most importantly, how in particular are participants making these verbally based judgments? Are they imagining a taste, are they visualizing a prototypically tasting food (e.g., sour–“lime”–green), or are they basing their evaluation on common shared features (e.g., I dislike sour and green)? It could be that real tastes would provide a different set of associations, especially when divorced from food and presented as a clear solution in a cup. A further and related issue concerns mechanism. Three explanations, shared across all cross-modal correspondences, have been advanced. First, statistical accounts suggest that people learn environmental contingencies between the properties of objects, including their taste and color (Spence 2011). So, when making a taste–color judgment, this could be based upon the color of the most frequently encountered and consistent taste-object (i.e., sour–lime). Second, semantic accounts rely upon shared meaning between 2 percepts. One form of shared meaning relates to affective or emotional valence (Palmer et al. 2013). Osgood’s early work using the semantic differential identified valence (i.e., pleasantness) as one form of meaning, shared by all sensory continua (see Osgood et al. 1957). As both tastes and colors reliably differ in reported valence (e.g., Palmer and Schloss 2010), it seems plausible that color choices could be based on shared valence between taste and color. Third, there are properties of stimuli that are common across the senses, notably magnitude-related features such as size, duration, and intensity. When the brain encounters information from multiple sensory channels of similar physical magnitude (e.g., a saturated color and a concentrated tastant), they may be perceived as similar due to common neural coding of magnitude (Marks 1989; Shermer and Levitan 2014). This explanation—referred to as the structural account of cross-modal correspondences (see Spence 2011) is less useful in terms of color selection for tastes, but could be very important in determining the saturation of the selected color, as intensity based correspondences may exist across all sensory domains due to common neural structuring of magnitude. In the experiment reported here, we provided participants with the 5 basic tastants at 3 concentrations. First, they sampled each tastant and matched it to a particular color (or white/black/gray) using a color wheel. Ratings of taste characteristics and hedonics were also obtained. Second, in a further round of testing, participants were asked to make color matches again, as well as judging how much they disliked/liked each color choice. Finally, participants were asked to match taste names to color names and to report how they had made their earlier selections. Together, this allowed us to examine the test–retest reliability and consistency of color choices, the reported method used, whether tastant concentration related to saturation, and whether valence was concordant between taste and color choices. Method Participants Fifty students participated for course credit (42 female, 8 male), aged between 17 and 42 years (M = 19.4, standard deviation [SD] = 4.0), with the study advertisement requesting normal color vision as an entry requirement. No participant reported a history of chemosensory impairment. The protocol for this study was approved by the Macquarie University Human Research Ethics Committee and written informed consent was provided by each participant. This study complies with the Declaration of Helsinki for medical research involving human subjects. Materials Tastant solutions were prepared at 3 concentration steps, as determined from prior studies (Yan and Dando 2015; Wang et al. 2016). Tastants and concentrations were: sucrose (CSR; 25 g/L, 50 g/L, 100 g/L), sodium chloride (Woolworths; 2 g/L, 4.6 g/L, 9.6 g/L), citric acid (Sigma Aldrich; 0.56 g/L, 0.80 g/L, 1.5 g/L), quinine (Sigma Aldrich; 0.02 g/L, 0.05 g/L, 0.16 g/L [all initially dissolved in propylene glycol]) and monosodium glutamate (meaty; Ajinomoto; 1.5 g/L, 7 g/L, 45 g/L). All solutions were made up to volume and refrigerated. Tastants, which were visually identical, were presented in 10-mL aliquots in disposable 50-mL cups and allowed to warm to room-temperature prior to testing. Log magnitude rating scales were used to assess intensity, irritancy, and the 5 taste-qualities (ranging from “Barely detectable” to “Strongest imaginable”; see Green et al. 1996). Valence was assessed using the bipolar labeled affective magnitude scale (ranging from “Greatest imaginable dislike” to “Greatest imaginable like”; see Cardello and Schutz 2004). Participants recorded their color choice for each taste using a Microsoft PowerPoint slide, on a Mac laptop (28-cm screen size). Each trial slide contained a white box (4 cm × 4 cm) in the center of the screen. Participants selected their matching color (or gray/black/white) using the computer-HSV color wheel. A brightness slider was located on the right side of the wheel. Together, the wheel and slider provided a measure of color choice (hue), intensity (saturation), and luminance (brightness). Procedure Participants were instructed to refrain from eating, drinking, and smoking 30 min prior to testing and all reported complying. After obtaining basic biographical data, a chemosensory disorder questionnaire and the Ishihara color vision test were completed. The experimental phase then commenced lasting 1 h, with a 5-min interval between part 1 and 2. Part 1 started with training on the HSV-color wheel. Participants then practiced the tasting procedure with a water blank, pouring the whole sample into their mouth, rolling it around, and then expectorating into a spittoon. They then selected their matching color, after which they were shown how to use the log magnitude scales. Eight log magnitude scales were presented on a single page in the same order for each stimulus. Participants were asked to evaluate how intense, sweet, salty, sour, bitter, meaty (i.e., the 5-taste qualities), and irritating and pleasant (i.e., valence), the tastant was. After completing all of the scales, participants rinsed their mouth with water. This same process—sample, expectorate, select color, evaluate taste, and rinse—was then repeated for all 15 tastants and for 3 water blanks, included to reduce adaptation. Stimuli were presented in a different randomized order for each participant. Progression was self-paced. Following a 5-min break, part 2 commenced. The same 18 solutions were presented again, in the same randomized order as part 1, so as to preserve the judgmental context. There was 1 procedural difference. Taste ratings were removed and instead participants completed an hedonic evaluation of their selected color choice using a bipolar labeled affective magnitude scale. Participants were reminded to make this rating based on their chosen color not the taste. After completing part 2, participants were asked to tell the experimenter which color, by name, they best associated with each of the primary taste words (sweet, sour, salty, meaty, and bitter). Each participant was then asked verbally by the experimenter whether there were any reasons behind their color selections made during the experiment. The experimenter recorded responses verbatim, which were then coded by the authors. Analyses Psychophysical data To confirm that participants could correctly identify each tastant’s primary quality, a target minus nontarget score was calculated for each tastant at each concentration step. First, the average of all of the nontarget taste qualities for each tastant was calculated (e.g., for sucrose; saltiness, sourness, bitterness, and meatiness ratings were averaged). Second, this average nontarget score was then subtracted from the nominal target quality (e.g., for sucrose–sweetness). As these data were non-normal, a nonparametric (Wilcoxon test) 1-sample test was conducted, with mu set at 0. Alpha was set at 0.017 for each tastant type, following Bonferroni correction by concentration level (i.e., 0.05/3). To examine intensity, irritancy, and valence of the tastants, repeated-measures ANOVAs were used. Concentration step and tastant type served as the independent variables. As these ANOVAs were conducted to confirm the basic character profile of the tastants (e.g., intensity increases with concentration, sweet is liked and bitter is disliked, sour tastes are irritating, etc.), no follow-up tests were conducted. Where violations of sphericity occurred the Huynh-Feldt correction was applied (as for other ANOVAs below). Color–taste pairing data To assess the test–retest reliability of color–taste pairings across part 1 and 2 of the experiment, a method to estimate agreement was used. First, color choices were grouped into 6 categories based upon prior studies (Spence et al. 2015; Woods et al. 2016). Colors with hue values from 15° to 285° were classified as red/pink, 45° to 75° as yellow, 76° to 149° as green, 150° to 255° as blue, with all remaining hue values (namely orange and violet), classified as other. When there was no hue and only brightness was adjusted (i.e., when saturation was 0), responses were grouped into a no-color (black/gray/white) category. Second, for every tastant at each concentration level, participants were either scored a “1” indicating the same color category selected in part 1 and 2, or a “0” indicating no match. As these data were non-normal, a nonparametric procedure was used. To test if agreement rate for each tastant at each concentration step was above a chance agreement rate (calculated as the odds of selecting the same color category in part 1 and 2 [i.e., P = 0.17]), a Wilcoxon test was used. Alpha was set at 0.017 for each tastant type, following Bonferroni correction by concentration level. To assess the consistency of color–taste pairings between participants, we used chi-square tests for each tastant, using the 6 color categories. The null hypothesis is that responses are equally distributed across the 6 categories, and to this end the expected value for each cell of the chi-square test was n divided by the number of color categories. As the color choices were repeated for each tastant and as each tastant had 3 concentration steps, each participant generated 6 color choices per tastant. Thus, to ensure the assumption of independence, 3 different chi-square tests were undertaken for each tastant. The first used color choices made for the medium concentration of each tastant in part 1. The second used the corresponding data from part 2. The third was computed on the modal color choice across all 6 color choices per tastant. If there was no modal choice, the participant’s color choice was classified as other color. Post-hoc chi-square tests were used to explore the source of significant effects, by comparing the target color category of interest with another category composed of all of the remaining categories collapsed together, with the expected values adjusted accordingly (i.e., target category expected = 8.3 vs. the sum of all remaining categories expected = 41.7). To determine if color–taste pairings identified verbally were consistent between participants, a further set of chi-square analyses were conducted. As with the contingency tables used for the actual tastants, participants’ verbal color choices for each taste word were categorized into the same 6 color categories used above. Mechanism-related data Participants’ responses as to how they reportedly made their color choices were coded into statistical, semantic, or ‘no-explanation given’ categories. To establish whether using the most frequently identified strategy (relative to the others) affected test–retest reliability, we used an independent t-test on these data. To assess whether the tastant concentration was lawfully related to the degree of saturation of the selected color (scaled from 0 to 100; all achromatic responses scored 0), a repeated-measures ANOVA was used on these data, with Concentration step, Time (part 1 vs. 2), and Tastant type as within-participant factors. To determine consistency between taste and selected color valence, each participant’s hedonic ratings for their 15 tastants were correlated with their self-reported hedonic ratings for their 15 corresponding selected colors. These correlations were converted to r′ values (i.e., standardized correlation coefficients) prior to a 1-sample t-test, to determine if the average correlation differed from 0. Results Taste properties Participants reported taste qualities, intensities, hedonics, and irritancy as would be expected. As can be seen in Table 1, the target taste quality was dominant in each case (i.e., all values positive) with the score representing a mean difference of around one-third of the rating scale. Table 1. Mean (SD) target–nontarget quality rating for each taste by concentration step Concentration step Taste Sweet Salty Sour Meaty Bitter Low 26.4 (20.4) 14.7 (20.4) 16.6 (22.9) 26.7 (24.2) 36.5 (23.9) Medium 42.1 (19.9) 35.1 (25.6) 18.9 (23.4) 20.0 (30.8) 50.4 (27.5) High 52.0 (21.2) 48.1 (25.3) 29.2 (26.4) 22.3 (33.9) 54.1 (28.1) Concentration step Taste Sweet Salty Sour Meaty Bitter Low 26.4 (20.4) 14.7 (20.4) 16.6 (22.9) 26.7 (24.2) 36.5 (23.9) Medium 42.1 (19.9) 35.1 (25.6) 18.9 (23.4) 20.0 (30.8) 50.4 (27.5) High 52.0 (21.2) 48.1 (25.3) 29.2 (26.4) 22.3 (33.9) 54.1 (28.1) View Large Table 1. Mean (SD) target–nontarget quality rating for each taste by concentration step Concentration step Taste Sweet Salty Sour Meaty Bitter Low 26.4 (20.4) 14.7 (20.4) 16.6 (22.9) 26.7 (24.2) 36.5 (23.9) Medium 42.1 (19.9) 35.1 (25.6) 18.9 (23.4) 20.0 (30.8) 50.4 (27.5) High 52.0 (21.2) 48.1 (25.3) 29.2 (26.4) 22.3 (33.9) 54.1 (28.1) Concentration step Taste Sweet Salty Sour Meaty Bitter Low 26.4 (20.4) 14.7 (20.4) 16.6 (22.9) 26.7 (24.2) 36.5 (23.9) Medium 42.1 (19.9) 35.1 (25.6) 18.9 (23.4) 20.0 (30.8) 50.4 (27.5) High 52.0 (21.2) 48.1 (25.3) 29.2 (26.4) 22.3 (33.9) 54.1 (28.1) View Large The taste intensity ANOVA revealed a significant main effect of Concentration F(2, 96) = 84.15, P < 0.0001, η2 = 0.64, with intensity increasing with the concentration step (low M = 29.7 [SD = 21.5]; medium M = 38.8 [SD = 21.9]; high M = 49.6 [SD = 24.9]), with this occurring to a similar degree for each tastant (i.e., no interaction effect). There was also a main effect of Tastant type F(4, 192) = 31.70 P < 0.0001, η2 = 0.40, with quinine judged most intense and sucrose least, perhaps due to some conflation of intensity and hedonics. For the hedonic data, there was a main effect of Tastant F(3.51, 168.47) = 81.04, P < 0.0001, η2 = 0.63, with sucrose judged as most pleasant (M = 15.5 [SD = 31.4]), sour, and salty as unpleasant (Sour M = −25.3 [SD = 29.1]; Salty M = −27.9 [SD = 29.4]), meaty as more unpleasant still (M = −42.2 [SD = 29.8]), and bitter as the most unpleasant (M = −57.9 [SD = 29.8]). There was also a main effect of Concentration F(1.76, 84.56) = 8.52, P = 0.001, η2 = 0.15, all qualified by an interaction of Tastant and Concentration F(7.33, 351.71) = 4.29, P < 0.001, η2 = 0.08. As concentration increased for salty, sour, bitter, and meaty tastes, pleasantness ratings tended to decrease, whereas for sucrose, pleasantness ratings tended to increase. For irritancy, the ANOVA revealed a main effect of Concentration F(2, 96) = 33.15, P < 0.001, η2 = 0.41, with reported irritancy increasing with concentration step (low M = 20.6 [SD = 24.4], medium M = 26.3 [SD = 26.9], high M = 32.1 [SD = 30.6]). There was also a main effect of Tastant, F(4,192) = 48.35, P < 0.0001, η2 = 0.50 and an interaction between Tastant and Concentration F(4.60, 220.89) = 3.11, P = 0.01, η2 = 0.06. The relationship between Concentration and Irritancy was least pronounced for sucrose, in comparison with the other tastants (see Table 2). Table 2. Mean (SD) irritancy rating for each taste by concentration step Concentration step Taste Sweet Salty Sour Meaty Bitter Low 6.6 (12.8) 16.3 (18.4) 20.9 (20.4) 23.9 (25.6) 35.6 (27.9) Medium 5.9 (9.7) 24.7 (25.6) 23.2 (21.1) 29.6 (24.9) 48.0 (30.9) High 7.7 (14.4) 28.4 (26.8) 30.5 (26.6) 38.6 (31.6) 55.9 (30.1) Concentration step Taste Sweet Salty Sour Meaty Bitter Low 6.6 (12.8) 16.3 (18.4) 20.9 (20.4) 23.9 (25.6) 35.6 (27.9) Medium 5.9 (9.7) 24.7 (25.6) 23.2 (21.1) 29.6 (24.9) 48.0 (30.9) High 7.7 (14.4) 28.4 (26.8) 30.5 (26.6) 38.6 (31.6) 55.9 (30.1) View Large Table 2. Mean (SD) irritancy rating for each taste by concentration step Concentration step Taste Sweet Salty Sour Meaty Bitter Low 6.6 (12.8) 16.3 (18.4) 20.9 (20.4) 23.9 (25.6) 35.6 (27.9) Medium 5.9 (9.7) 24.7 (25.6) 23.2 (21.1) 29.6 (24.9) 48.0 (30.9) High 7.7 (14.4) 28.4 (26.8) 30.5 (26.6) 38.6 (31.6) 55.9 (30.1) Concentration step Taste Sweet Salty Sour Meaty Bitter Low 6.6 (12.8) 16.3 (18.4) 20.9 (20.4) 23.9 (25.6) 35.6 (27.9) Medium 5.9 (9.7) 24.7 (25.6) 23.2 (21.1) 29.6 (24.9) 48.0 (30.9) High 7.7 (14.4) 28.4 (26.8) 30.5 (26.6) 38.6 (31.6) 55.9 (30.1) View Large Color–taste pairings For test–retest reliability (see Table 3), there was better than chance agreement for all concentration steps for sweet, sour, and meaty. For bitter and salty, only the medium and high concentration steps were significantly better than chance. Table 3. Mean (SD) color–taste test–retest reliability data (percent agreement) across part 1 and 2 of the experiment, based upon color category, with all significantly different from the chance rate of agreement (17%) except as indicated Concentration step Taste Sweet Salty Sour Meaty Bitter Low 68% (47) 44% (50)* 58% (50) 42% (50)* 46% (50) Medium 76% (43) 52% (50) 54% (50) 50% (51) 46% (50) High 72% (45) 52% (50) 50% (51) 52% (50) 56% (50) Concentration step Taste Sweet Salty Sour Meaty Bitter Low 68% (47) 44% (50)* 58% (50) 42% (50)* 46% (50) Medium 76% (43) 52% (50) 54% (50) 50% (51) 46% (50) High 72% (45) 52% (50) 50% (51) 52% (50) 56% (50) *Not significantly different from chance View Large Table 3. Mean (SD) color–taste test–retest reliability data (percent agreement) across part 1 and 2 of the experiment, based upon color category, with all significantly different from the chance rate of agreement (17%) except as indicated Concentration step Taste Sweet Salty Sour Meaty Bitter Low 68% (47) 44% (50)* 58% (50) 42% (50)* 46% (50) Medium 76% (43) 52% (50) 54% (50) 50% (51) 46% (50) High 72% (45) 52% (50) 50% (51) 52% (50) 56% (50) Concentration step Taste Sweet Salty Sour Meaty Bitter Low 68% (47) 44% (50)* 58% (50) 42% (50)* 46% (50) Medium 76% (43) 52% (50) 54% (50) 50% (51) 46% (50) High 72% (45) 52% (50) 50% (51) 52% (50) 56% (50) *Not significantly different from chance View Large For consistency, analysis of the medium concentration of each tastant during part 1 and 2 revealed that participants’ choices were not randomly distributed across color categories (see Table 4). With the exception of meaty, the most common color was the same during part 1 and 2–red-pink for sweet, blue for salty, yellow for sour, and green for bitter. For meaty, the pattern of color choices was evenly distributed between 2 categories, red-pink and other color. Illustrative data are presented in Supplementary Figure 1, for the medium concentration on part 1 of the experiment—the first occasion on which color matches were made. Each panel contains 50 elements, with each element representing the color chosen by each participant for that particular taste. Table 4. Expected and observed values for participants categorized color choices, on part 1 (medium-concentration step), on part 2 (medium-concentration step), and overall with the modal color (across all 3 concentration steps on part 1 and 2) Color categories: Red and pink Green Yellow Blue Other colors Black, white, and gray Expected values (Null model) 8.3 8.3 8.3 8.3 8.3 8.3 Observed values—by taste and color choice measure Sweet1 1 38 3 1 3 4 1 2 40 2 2 3 4 0 Modal 43 1 2 1 3 0 Salty2 1 6 4 7 19 10 4 2 8 6 3 22 10 1 Modal 3 4 3 24 15 1 Sour3 1 4 13 25 2 6 0 2 5 11 21 6 7 0 Modal 3 8 19 4 15 1 Bitter4 1 6 16 5 4 8 11 2 6 17 5 7 4 11 Modal 5 12 7 3 6 17 Meaty5 1 17 5 2 7 16 3 2 14 6 5 10 15 0 Modal 13 5 2 7 23 0 Color categories: Red and pink Green Yellow Blue Other colors Black, white, and gray Expected values (Null model) 8.3 8.3 8.3 8.3 8.3 8.3 Observed values—by taste and color choice measure Sweet1 1 38 3 1 3 4 1 2 40 2 2 3 4 0 Modal 43 1 2 1 3 0 Salty2 1 6 4 7 19 10 4 2 8 6 3 22 10 1 Modal 3 4 3 24 15 1 Sour3 1 4 13 25 2 6 0 2 5 11 21 6 7 0 Modal 3 8 19 4 15 1 Bitter4 1 6 16 5 4 8 11 2 6 17 5 7 4 11 Modal 5 12 7 3 6 17 Meaty5 1 17 5 2 7 16 3 2 14 6 5 10 15 0 Modal 13 5 2 7 23 0 1Chi-square values: 127.6, 145.4, 173.7, P < 0.001; 2chi-square values: 19.4, 33.3, 56.3, P < 0.0017; 3chi-square values: 52.0, 30.6, 31.1, P < 0.001; 4chi-square values: 12.2, 14.3, 16.2, P < 0.02; 5chi-square values: 25.8, 19.8, 43.1, P < 0.002. View Large Table 4. Expected and observed values for participants categorized color choices, on part 1 (medium-concentration step), on part 2 (medium-concentration step), and overall with the modal color (across all 3 concentration steps on part 1 and 2) Color categories: Red and pink Green Yellow Blue Other colors Black, white, and gray Expected values (Null model) 8.3 8.3 8.3 8.3 8.3 8.3 Observed values—by taste and color choice measure Sweet1 1 38 3 1 3 4 1 2 40 2 2 3 4 0 Modal 43 1 2 1 3 0 Salty2 1 6 4 7 19 10 4 2 8 6 3 22 10 1 Modal 3 4 3 24 15 1 Sour3 1 4 13 25 2 6 0 2 5 11 21 6 7 0 Modal 3 8 19 4 15 1 Bitter4 1 6 16 5 4 8 11 2 6 17 5 7 4 11 Modal 5 12 7 3 6 17 Meaty5 1 17 5 2 7 16 3 2 14 6 5 10 15 0 Modal 13 5 2 7 23 0 Color categories: Red and pink Green Yellow Blue Other colors Black, white, and gray Expected values (Null model) 8.3 8.3 8.3 8.3 8.3 8.3 Observed values—by taste and color choice measure Sweet1 1 38 3 1 3 4 1 2 40 2 2 3 4 0 Modal 43 1 2 1 3 0 Salty2 1 6 4 7 19 10 4 2 8 6 3 22 10 1 Modal 3 4 3 24 15 1 Sour3 1 4 13 25 2 6 0 2 5 11 21 6 7 0 Modal 3 8 19 4 15 1 Bitter4 1 6 16 5 4 8 11 2 6 17 5 7 4 11 Modal 5 12 7 3 6 17 Meaty5 1 17 5 2 7 16 3 2 14 6 5 10 15 0 Modal 13 5 2 7 23 0 1Chi-square values: 127.6, 145.4, 173.7, P < 0.001; 2chi-square values: 19.4, 33.3, 56.3, P < 0.0017; 3chi-square values: 52.0, 30.6, 31.1, P < 0.001; 4chi-square values: 12.2, 14.3, 16.2, P < 0.02; 5chi-square values: 25.8, 19.8, 43.1, P < 0.002. View Large Further analyses were conducted on the modal color choice data (see Table 4). For sweet, sour, and salty, modal responses were the same as per the analyses above. For bitter, the black-gray-white category was the modal choice, with gray being most frequent. For meaty, the modal color choice was the other color category. To determine which “other color,” a further set of chi-square analyses were conducted using 8 color categories. This involved splitting responses in the red/pink category into red (values by approach now; 14, 12, 13) and pink (values; 3, 2, 0) and the other color category into violet (values; 1, 0, 9) and orange (values; 15, 15, 14). Orange, closely followed by red, were the colors most frequently chosen for meaty (chi-square values by approach, 32.9, 35.4, 33.8, all P < 0.001). To confirm that the color choice for each taste that we identified above was principally responsible for the nonrandom distribution of color–taste mappings, post-hoc tests were run. The distribution of observed values significantly deviated from the chance expected values for sweet (red/pink vs. remainder; chi square = 173.1, P < 0.001), salty (blue vs. remainder; chi square = 35.3, P < 0.001), sour (yellow vs. remainder; chi square = 16.4, P < 0.001), and bitter (black, gray, and white vs. remainder; chi square = 8.5, P < 0.005). For meaty, this analysis was also conducted but with different expected values based upon 8 color categories, and here too observed values significantly deviated from chance (chi square = 11.0, P < 0.005). Thus, these particular color choices significantly contributed to the nonrandom distribution of color choices made for each taste. To test for consistent pairings across taste-words and color-names, a further set of chi squares were run. As can be seen in Table 5, this revealed a very similar outcome to the data obtained with real tastants and colors. Table 5. Expected and observed values for participants’ categorized color choices when using color names and taste words in the final part of the experiment Color categories Red and pink Green Yellow Blue Other colors Black, white, and gray Expected values (Null model) 8.3 8.3 8.3 8.3 8.3 8.3 Observed values—by taste Sweet1 45 2 2 0 1 0 Salty2 3 2 5 28 5 7 Sour3 2 14 27 0 7 0 Bitter4 3 16 4 2 14 11 Meaty5 28 0 0 0 22 0 Color categories Red and pink Green Yellow Blue Other colors Black, white, and gray Expected values (Null model) 8.3 8.3 8.3 8.3 8.3 8.3 Observed values—by taste Sweet1 45 2 2 0 1 0 Salty2 3 2 5 28 5 7 Sour3 2 14 27 0 7 0 Bitter4 3 16 4 2 14 11 Meaty5 28 0 0 0 22 0 1Chi square = 201.1, P < 0.001; 2chi square = 57.5, P < 0.001; 3chi square = 67.4, P < 0.001; 4chi square = 22.2, P < 0.001; 5chi square = 102.2, P < 0.001, with 22/28 responses for red and pink being red and the other category solely composed of orange and brown. View Large Table 5. Expected and observed values for participants’ categorized color choices when using color names and taste words in the final part of the experiment Color categories Red and pink Green Yellow Blue Other colors Black, white, and gray Expected values (Null model) 8.3 8.3 8.3 8.3 8.3 8.3 Observed values—by taste Sweet1 45 2 2 0 1 0 Salty2 3 2 5 28 5 7 Sour3 2 14 27 0 7 0 Bitter4 3 16 4 2 14 11 Meaty5 28 0 0 0 22 0 Color categories Red and pink Green Yellow Blue Other colors Black, white, and gray Expected values (Null model) 8.3 8.3 8.3 8.3 8.3 8.3 Observed values—by taste Sweet1 45 2 2 0 1 0 Salty2 3 2 5 28 5 7 Sour3 2 14 27 0 7 0 Bitter4 3 16 4 2 14 11 Meaty5 28 0 0 0 22 0 1Chi square = 201.1, P < 0.001; 2chi square = 57.5, P < 0.001; 3chi square = 67.4, P < 0.001; 4chi square = 22.2, P < 0.001; 5chi square = 102.2, P < 0.001, with 22/28 responses for red and pink being red and the other category solely composed of orange and brown. View Large Possible mechanisms When asked, 36/50 (72%) participants reported basing one or more color choices on real-world taste–color associations (i.e., statistical). Two participants reported making choices based upon valence, with the remaining 12 offering no rationale. There was no difference in test–retest reliability between those reporting a statistical explanation versus no explanation. Tastant concentration was strongly related to the saturation of the selected color, F(1.82, 89.07) = 32.10, P < 0.0001, η2 = 0.40, with a highly significant linear trend (P < 0.001, η2 = 0.47) with saturation increasing with tastant concentration (low M = 54.9 [SD = 30.9]; medium M = 63.4 [SD = 28.2]; high M = 66.9 [SD = 29.43]). This relationship was consistent across all tastants (i.e., no interactions). There was also a main effect of Tastant type, F(2.65, 129.85) = 14.43, P < 0.0001, η2 = 0.23, with the most saturated colors chosen for sour (M = 73.7 [SD = 22.3]) followed by meaty (M = 65.3 [SD = 27.7]), salty (M = 63.5 [SD = 28.0]), sweet (M = 53.5 [SD = 23.4]), and then bitter (M = 50.9 [SD = 39.2]). We examined the data derived from each participant’s correlation of their taste and color hedonic ratings. The mean correlation (r′ = 0.65 [SD = 0.54]) was significantly different from 0, t(49) = 8.42, P < 0.001, indicating that selected colors match the valence of their associated taste. Discussion Using real tastants and colors, participants taste–color matches were generally consistent across part 1 and 2 of the experiment and were nonrandomly distributed across color categories. Color choices were very similar to those made using taste and color words both in prior studies (O’Mahony 1983; Heller 1999; Koch and Koch 2003; Tomasik-Krotki and Strojny 2008; Wan et al. 2014; Spence et al. 2015; Woods et al. 2016) and also in our own test of this issue. Many participants reported drawing upon associations between taste and the color of their associated object (generally a food). In addition, there was a strong relationship between the valence of the taste and the selected color. Finally, as with many other cross-modal matches (Marks 1989; Shermer and Levitan 2014), the stimulus magnitude dimensions of color saturation and tastant concentration were strongly related. Most participants in our experiment reported using the statistical approach, for one or more taste–color mapping. That this approach may in fact be relevant to taste–color pairings is suggested by another related literature—namely color-based valence judgments (e.g., the general liking for red and disliking for brown), which are also nonrandomly distributed in Western cultures (Taylor et al. 2013). Correlational evidence suggests that color-based valence judgments may arise from the valence of prototypical objects associated with that color (e.g., brown-feces; Palmer and Schloss 2010). Interestingly, in attempts to determine whether color-based valence judgments are culturally universal—a view long-held in the literature—recent testing in a remote setting suggested otherwise (Taylor et al. 2013) supporting the idea that color-based valence judgments do derive from the valence of objects prototypical for a particular color. As internet-based computer tasks were used to establish cross-cultural color–taste pairings, limited access to the internet in remote/rural settings is likely to generate greater sample homogeneity than the descriptor “tested in different countries” might suggest. Indeed, remote testing of color–taste associations may prove instructive. The statistical view is not mutually exclusive with the strong relationship we observed between taste and selected color valence. This is because shared valence between a taste and a color may be mediated by the prototypical food (or other) object that the taste brings to mind. Conversely, this can also be seen by the aversion to food and drinks, which are colored in an unusual manner (i.e., discordant with their prototypical presentation). Take for example, the declining sales associated with making previously colored beverages clear (e.g., crystal pepsi, Tab Clear) and when tomato sauce (red) is made green (e.g., Heinz’ green ketchup; Velasco et al. 2016). Thus, the valence of the food, its taste, and other properties (including its color) may very well be concordant. Consequently, one would expect to observe a correlation between the valence of tastes and their associated colors. If this is correct, then the relationship we observed emerges as a by-product of the statistical account. In addition to this possibility, a more purely valence-based account could also operate. Wang et al. (2016) suggested for certain taste–pitch matches that participants may match the hedonic valence of the taste with that of the pitch. Similarly, here, participants could be matching the valence of the taste with that arising from viewing the color. This explanation is exclusive, as it presumes that commonality of valence drives color choice, not mediation via a recalled prototypical object. Consistent with prior studies in other modalities (Marks 1989), we found that, as tastant concentration increased, there was also an increase in the saturation of colors selected—and that this relationship was relative and evident to the same degree for all tastants. In a related vein, intensity matching between taste and color can be seen with everyday foods. For instance, in a recent study Shermer and Levitan (2014) showed that the saturation of salsas (i.e., the intensity of their red color) biased participants ratings of their piquancy in taste (i.e., their spice intensity). These lines of evidence are consistent with 2 theories. First, the idea that underlying neural patterns that represent stimulus magnitude are similar across modalities—i.e., the structural account (Spence 2011). Second, the relationship between saturation and taste intensity is also compatible with the statistical account. For example, more intensely tasting foods tend (anecdotally) to have more vividly colored (i.e., saturated) packaging, and vice versa. Pertaining to these findings, 2 additional points need to be considered. First, saturation values differed across tastant type. This effect probably arose because of the greater number of 0 values associated with bitter taste due to achromatic selections. Second, it is possible that irritancy contributed to the concentration–saturation relationship. However, as sucrose does not become more irritating with increasing concentration, in contrast to salty and sour, the absence of a tastant by concentration interaction suggests that any effect of irritancy was probably minor. In conclusion, we find that, when participants are asked to match real tastants to real colors, their choices are generally reliable and consistent, and they are very similar to prior findings based on words alone. We suggest that color–taste relationships and their shared valence may arise from participants basing their color selections on prototypical food objects associated with each taste. Acknowledgements We would like to thank the reviewers for their positive and helpful feedback. The authors would also like to thank Alex Russell for providing the PowerPoint slides used in this study. References Cardello A , Schutz H . 2004 . Research note numerical scale-point locations for constructing the LAM (labeled affective magnitude) scale . J Sens Stud . 19 : 341 – 346 . Google Scholar CrossRef Search ADS Duffy VB , Hayes JE , Bartoshuk LM , Snyder DJ . 2009 . Taste: vertebrate psychophysics . In: Squire LR, editor. Encyclopedia of neuroscience . Oxford : Academic Press . p. 881 – 886 . Green BG , Dalton P , Cowart B , Shaffer G , Rankin K , Higgins J . 1996 . Evaluating the ‘Labeled Magnitude Scale’ for measuring sensations of taste and smell . Chem Senses . 21 : 323 – 334 . Google Scholar CrossRef Search ADS PubMed Heller E . 1999 . Wie farben wirken [How color works] . Reinbek bei Hamburg . Koch C , Koch EC . 2003 . Preconceptions of taste based on color . J Psychol . 137 : 233 – 242 . Google Scholar CrossRef Search ADS PubMed Marks LE . 1989 . On cross-modal similarity: the perceptual structure of pitch, loudness, and brightness . J Exp Psychol Hum Percept Perform . 15 : 586 – 602 . Google Scholar CrossRef Search ADS PubMed O’Mahony M . 1983 . Gustatory responses to nongustatory stimuli . Perception . 12 : 627 – 633 . Google Scholar CrossRef Search ADS PubMed Osgood CE, Suci GJ, Tannenbaum PH. 1957. The measurement of meaning. Urbana: University of Illinois Press. Palmer SE , Schloss KB . 2010 . An ecological valence theory of human color preference . Proc Natl Acad Sci USA . 107 : 8877 – 8882 . Google Scholar CrossRef Search ADS PubMed Palmer SE , Schloss KB , Xu Z , Prado-León LR . 2013 . Music-color associations are mediated by emotion . Proc Natl Acad Sci USA . 110 : 8836 – 8841 . Google Scholar CrossRef Search ADS PubMed Robinson JO . 1970 . The misuse of taste names by untrained observers . Br J Psychol . 61 : 375 – 378 . Google Scholar CrossRef Search ADS PubMed Shermer DZ , Levitan CA . 2014 . Red hot: the crossmodal effect of color intensity on perceived piquancy . Multisens Res . 27 : 207 – 223 . Google Scholar CrossRef Search ADS PubMed Skrandies W , Reuther N . 2008 . Match and mismatch of taste, odor, and color is reflected by electrical activity in the human brain . J Psychophys . 22 : 175 – 184 . Google Scholar CrossRef Search ADS Spence C . 2011 . Crossmodal correspondences: a tutorial review . Atten Percept Psychophys . 73 : 971 – 995 . Google Scholar CrossRef Search ADS PubMed Spence C , Wan X , Woods A , Velasco C , Deng J , Youssef J , Deroy O . 2015 . On tasty colours and colourful tastes? Assessing, explaining, and utilizing crossmodal correspondences between colours and basic tastes . Flav . 4 : 23 . Google Scholar CrossRef Search ADS Taylor C , Clifford A , Franklin A . 2013 . Color preferences are not universal . J Exp Psychol Gen . 142 : 1015 – 1027 . Google Scholar CrossRef Search ADS PubMed Tomasik-Krotki J , Strojny J . 2008 . Scaling of sensory impressions . J Sen Stud . 23 : 251 – 266 . Google Scholar CrossRef Search ADS Velasco C , Michel C , Youssef J , Gamez X , Check A , Spence C . 2016 . Using single colours and colour pairs to communicate basic tastes . Int J Food Design . 1 : 83 – 102 . Google Scholar CrossRef Search ADS Velasco C , Woods A , Deroy O , Spence C . 2014 . Hedonic mediation of the crossmodal correspondence between taste and shape . Food Qual Pref . 41 : 151 – 158 . Google Scholar CrossRef Search ADS Wan X , Woods AT , van den Bosch JJ , McKenzie KJ , Velasco C , Spence C . 2014 . Cross-cultural differences in crossmodal correspondences between basic tastes and visual features . Front Psychol . 5 : 1365 . Google Scholar CrossRef Search ADS PubMed Wang QJ , Wang S , Spence C . 2016 . “Turn up the taste”: assessing the role of taste intensity and emotion in mediating crossmodal correspondences between basic tastes and pitch . Chem Senses . 41 : 345 – 356 . Google Scholar CrossRef Search ADS PubMed Woods AT , Marmolejo-Ramos F , Velasco C , Spence C . 2016 . Using single colors and color pairs to communicate basic tastes II: foreground-background color combinations . Iperception . 7 : 2041669516663750 . Google Scholar PubMed Woods AT , Spence C . 2016 . Using single colors and color pairs to communicate basic tastes . Iperception . 7 : 2041669516658817 . Google Scholar PubMed Yan KS , Dando R . 2015 . A crossmodal role for audition in taste perception . J Exp Psychol Hum Percept Perform . 41 : 590 – 596 . Google Scholar CrossRef Search ADS PubMed © The Author(s) 2018. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Chemical Senses Oxford University Press

Cross-Modal Associations Between Real Tastes and Colors

Loading next page...
 
/lp/ou_press/cross-modal-associations-between-real-tastes-and-colors-0DMs3spy6y
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com
ISSN
0379-864X
eISSN
1464-3553
D.O.I.
10.1093/chemse/bjy033
Publisher site
See Article on Publisher Site

Abstract

Abstract People make reliable and consistent matches between taste and color. However, in contrast to other cross-modal correspondences, all of the research to date has used only taste words (and often color words too), potentially limiting our understanding of how taste–color matches arise. Here, participants sampled the 5 basic tastes, at 3 concentration steps, and selected their best matching color from a color wheel. This test was repeated, and in addition, participants evaluated the valence of the taste and their color choice, as well as the qualities/intensities of the taste stimuli. Participants were then presented with taste names and asked to generate the best matching color name, as well as reporting how they made their earlier choices. Color selections were reliable and consistent and closely followed those based on taste word matches obtained in this and prior studies. Most participants reported basing their color choices on their associated taste-object (often foods). There was marked similarity in valence between taste and color choices, and the saturation of color choices was related to tastant concentration. We discuss what drives color–taste pairings, with learning suggested as one possible mechanism. gustation, color, cross-modal Introduction Cross-modal correspondences have been documented in all sensory domains, but are less explored in the chemical senses. One exception is taste and color. Taste is composed of at least 5 basic qualities, sweet, sour, bitter, salty, and umami (which alone tastes “meaty”; Duffy et al. 2009). Investigators, in 7 studies, have found evidence of consistent mappings between particular tastes and colors (O’Mahony 1983; Heller 1999; Koch and Koch 2003; Tomasik-Krotki and Strojny 2008; Wan et al. 2014; Spence et al. 2015; Woods and Spence 2016). In each of these studies, participants were asked to determine which color (including white, gray, and black), using patches or words, went best with which taste “name” (e.g., sweet) or vice versa. With varying sample sizes and adult participants drawn from many countries, the results have been surprisingly consistent. The most consistent matches (the fastest, Woods et al. 2016, and congruent based on the pattern of event related potentials, Skrandies and Reuther 2008) are as follows: red and pink with sweet, white and blue with salty, green and yellow with sour, and black and green with bitter. Umami (meaty), which has only been included in 2 studies (Tomasik-Krotki and Strojny 2008; Wan et al. 2014), is less consistently linked to any 1 color. What is most surprising about these findings is that they are based on taste “words,” and often color words too, especially as the literature already suggests that different cross-modal effects may ensue when taste words are used rather than tastants (Velasco et al. 2014). This reliance on taste words raises several issues. First, and as already apparent, umami (meaty) is not clearly associated with any particular color, which might have much to do with people being unsure of its specific taste. Another problem emerges for bitter and sour tastes. Although they are readily discriminable, their labels are often confused (e.g., Robinson 1970), which might explain their shared color association—green. Second, taste experiences vary in magnitude. Milk is not very sweet and potatoes are not very salty, but chocolate milk is very sweet and potato chips are very salty—so when participants are making judgments about sweet and salty, and the other tastes too, what particular intensity are they basing their judgments on? Finally, and most importantly, how in particular are participants making these verbally based judgments? Are they imagining a taste, are they visualizing a prototypically tasting food (e.g., sour–“lime”–green), or are they basing their evaluation on common shared features (e.g., I dislike sour and green)? It could be that real tastes would provide a different set of associations, especially when divorced from food and presented as a clear solution in a cup. A further and related issue concerns mechanism. Three explanations, shared across all cross-modal correspondences, have been advanced. First, statistical accounts suggest that people learn environmental contingencies between the properties of objects, including their taste and color (Spence 2011). So, when making a taste–color judgment, this could be based upon the color of the most frequently encountered and consistent taste-object (i.e., sour–lime). Second, semantic accounts rely upon shared meaning between 2 percepts. One form of shared meaning relates to affective or emotional valence (Palmer et al. 2013). Osgood’s early work using the semantic differential identified valence (i.e., pleasantness) as one form of meaning, shared by all sensory continua (see Osgood et al. 1957). As both tastes and colors reliably differ in reported valence (e.g., Palmer and Schloss 2010), it seems plausible that color choices could be based on shared valence between taste and color. Third, there are properties of stimuli that are common across the senses, notably magnitude-related features such as size, duration, and intensity. When the brain encounters information from multiple sensory channels of similar physical magnitude (e.g., a saturated color and a concentrated tastant), they may be perceived as similar due to common neural coding of magnitude (Marks 1989; Shermer and Levitan 2014). This explanation—referred to as the structural account of cross-modal correspondences (see Spence 2011) is less useful in terms of color selection for tastes, but could be very important in determining the saturation of the selected color, as intensity based correspondences may exist across all sensory domains due to common neural structuring of magnitude. In the experiment reported here, we provided participants with the 5 basic tastants at 3 concentrations. First, they sampled each tastant and matched it to a particular color (or white/black/gray) using a color wheel. Ratings of taste characteristics and hedonics were also obtained. Second, in a further round of testing, participants were asked to make color matches again, as well as judging how much they disliked/liked each color choice. Finally, participants were asked to match taste names to color names and to report how they had made their earlier selections. Together, this allowed us to examine the test–retest reliability and consistency of color choices, the reported method used, whether tastant concentration related to saturation, and whether valence was concordant between taste and color choices. Method Participants Fifty students participated for course credit (42 female, 8 male), aged between 17 and 42 years (M = 19.4, standard deviation [SD] = 4.0), with the study advertisement requesting normal color vision as an entry requirement. No participant reported a history of chemosensory impairment. The protocol for this study was approved by the Macquarie University Human Research Ethics Committee and written informed consent was provided by each participant. This study complies with the Declaration of Helsinki for medical research involving human subjects. Materials Tastant solutions were prepared at 3 concentration steps, as determined from prior studies (Yan and Dando 2015; Wang et al. 2016). Tastants and concentrations were: sucrose (CSR; 25 g/L, 50 g/L, 100 g/L), sodium chloride (Woolworths; 2 g/L, 4.6 g/L, 9.6 g/L), citric acid (Sigma Aldrich; 0.56 g/L, 0.80 g/L, 1.5 g/L), quinine (Sigma Aldrich; 0.02 g/L, 0.05 g/L, 0.16 g/L [all initially dissolved in propylene glycol]) and monosodium glutamate (meaty; Ajinomoto; 1.5 g/L, 7 g/L, 45 g/L). All solutions were made up to volume and refrigerated. Tastants, which were visually identical, were presented in 10-mL aliquots in disposable 50-mL cups and allowed to warm to room-temperature prior to testing. Log magnitude rating scales were used to assess intensity, irritancy, and the 5 taste-qualities (ranging from “Barely detectable” to “Strongest imaginable”; see Green et al. 1996). Valence was assessed using the bipolar labeled affective magnitude scale (ranging from “Greatest imaginable dislike” to “Greatest imaginable like”; see Cardello and Schutz 2004). Participants recorded their color choice for each taste using a Microsoft PowerPoint slide, on a Mac laptop (28-cm screen size). Each trial slide contained a white box (4 cm × 4 cm) in the center of the screen. Participants selected their matching color (or gray/black/white) using the computer-HSV color wheel. A brightness slider was located on the right side of the wheel. Together, the wheel and slider provided a measure of color choice (hue), intensity (saturation), and luminance (brightness). Procedure Participants were instructed to refrain from eating, drinking, and smoking 30 min prior to testing and all reported complying. After obtaining basic biographical data, a chemosensory disorder questionnaire and the Ishihara color vision test were completed. The experimental phase then commenced lasting 1 h, with a 5-min interval between part 1 and 2. Part 1 started with training on the HSV-color wheel. Participants then practiced the tasting procedure with a water blank, pouring the whole sample into their mouth, rolling it around, and then expectorating into a spittoon. They then selected their matching color, after which they were shown how to use the log magnitude scales. Eight log magnitude scales were presented on a single page in the same order for each stimulus. Participants were asked to evaluate how intense, sweet, salty, sour, bitter, meaty (i.e., the 5-taste qualities), and irritating and pleasant (i.e., valence), the tastant was. After completing all of the scales, participants rinsed their mouth with water. This same process—sample, expectorate, select color, evaluate taste, and rinse—was then repeated for all 15 tastants and for 3 water blanks, included to reduce adaptation. Stimuli were presented in a different randomized order for each participant. Progression was self-paced. Following a 5-min break, part 2 commenced. The same 18 solutions were presented again, in the same randomized order as part 1, so as to preserve the judgmental context. There was 1 procedural difference. Taste ratings were removed and instead participants completed an hedonic evaluation of their selected color choice using a bipolar labeled affective magnitude scale. Participants were reminded to make this rating based on their chosen color not the taste. After completing part 2, participants were asked to tell the experimenter which color, by name, they best associated with each of the primary taste words (sweet, sour, salty, meaty, and bitter). Each participant was then asked verbally by the experimenter whether there were any reasons behind their color selections made during the experiment. The experimenter recorded responses verbatim, which were then coded by the authors. Analyses Psychophysical data To confirm that participants could correctly identify each tastant’s primary quality, a target minus nontarget score was calculated for each tastant at each concentration step. First, the average of all of the nontarget taste qualities for each tastant was calculated (e.g., for sucrose; saltiness, sourness, bitterness, and meatiness ratings were averaged). Second, this average nontarget score was then subtracted from the nominal target quality (e.g., for sucrose–sweetness). As these data were non-normal, a nonparametric (Wilcoxon test) 1-sample test was conducted, with mu set at 0. Alpha was set at 0.017 for each tastant type, following Bonferroni correction by concentration level (i.e., 0.05/3). To examine intensity, irritancy, and valence of the tastants, repeated-measures ANOVAs were used. Concentration step and tastant type served as the independent variables. As these ANOVAs were conducted to confirm the basic character profile of the tastants (e.g., intensity increases with concentration, sweet is liked and bitter is disliked, sour tastes are irritating, etc.), no follow-up tests were conducted. Where violations of sphericity occurred the Huynh-Feldt correction was applied (as for other ANOVAs below). Color–taste pairing data To assess the test–retest reliability of color–taste pairings across part 1 and 2 of the experiment, a method to estimate agreement was used. First, color choices were grouped into 6 categories based upon prior studies (Spence et al. 2015; Woods et al. 2016). Colors with hue values from 15° to 285° were classified as red/pink, 45° to 75° as yellow, 76° to 149° as green, 150° to 255° as blue, with all remaining hue values (namely orange and violet), classified as other. When there was no hue and only brightness was adjusted (i.e., when saturation was 0), responses were grouped into a no-color (black/gray/white) category. Second, for every tastant at each concentration level, participants were either scored a “1” indicating the same color category selected in part 1 and 2, or a “0” indicating no match. As these data were non-normal, a nonparametric procedure was used. To test if agreement rate for each tastant at each concentration step was above a chance agreement rate (calculated as the odds of selecting the same color category in part 1 and 2 [i.e., P = 0.17]), a Wilcoxon test was used. Alpha was set at 0.017 for each tastant type, following Bonferroni correction by concentration level. To assess the consistency of color–taste pairings between participants, we used chi-square tests for each tastant, using the 6 color categories. The null hypothesis is that responses are equally distributed across the 6 categories, and to this end the expected value for each cell of the chi-square test was n divided by the number of color categories. As the color choices were repeated for each tastant and as each tastant had 3 concentration steps, each participant generated 6 color choices per tastant. Thus, to ensure the assumption of independence, 3 different chi-square tests were undertaken for each tastant. The first used color choices made for the medium concentration of each tastant in part 1. The second used the corresponding data from part 2. The third was computed on the modal color choice across all 6 color choices per tastant. If there was no modal choice, the participant’s color choice was classified as other color. Post-hoc chi-square tests were used to explore the source of significant effects, by comparing the target color category of interest with another category composed of all of the remaining categories collapsed together, with the expected values adjusted accordingly (i.e., target category expected = 8.3 vs. the sum of all remaining categories expected = 41.7). To determine if color–taste pairings identified verbally were consistent between participants, a further set of chi-square analyses were conducted. As with the contingency tables used for the actual tastants, participants’ verbal color choices for each taste word were categorized into the same 6 color categories used above. Mechanism-related data Participants’ responses as to how they reportedly made their color choices were coded into statistical, semantic, or ‘no-explanation given’ categories. To establish whether using the most frequently identified strategy (relative to the others) affected test–retest reliability, we used an independent t-test on these data. To assess whether the tastant concentration was lawfully related to the degree of saturation of the selected color (scaled from 0 to 100; all achromatic responses scored 0), a repeated-measures ANOVA was used on these data, with Concentration step, Time (part 1 vs. 2), and Tastant type as within-participant factors. To determine consistency between taste and selected color valence, each participant’s hedonic ratings for their 15 tastants were correlated with their self-reported hedonic ratings for their 15 corresponding selected colors. These correlations were converted to r′ values (i.e., standardized correlation coefficients) prior to a 1-sample t-test, to determine if the average correlation differed from 0. Results Taste properties Participants reported taste qualities, intensities, hedonics, and irritancy as would be expected. As can be seen in Table 1, the target taste quality was dominant in each case (i.e., all values positive) with the score representing a mean difference of around one-third of the rating scale. Table 1. Mean (SD) target–nontarget quality rating for each taste by concentration step Concentration step Taste Sweet Salty Sour Meaty Bitter Low 26.4 (20.4) 14.7 (20.4) 16.6 (22.9) 26.7 (24.2) 36.5 (23.9) Medium 42.1 (19.9) 35.1 (25.6) 18.9 (23.4) 20.0 (30.8) 50.4 (27.5) High 52.0 (21.2) 48.1 (25.3) 29.2 (26.4) 22.3 (33.9) 54.1 (28.1) Concentration step Taste Sweet Salty Sour Meaty Bitter Low 26.4 (20.4) 14.7 (20.4) 16.6 (22.9) 26.7 (24.2) 36.5 (23.9) Medium 42.1 (19.9) 35.1 (25.6) 18.9 (23.4) 20.0 (30.8) 50.4 (27.5) High 52.0 (21.2) 48.1 (25.3) 29.2 (26.4) 22.3 (33.9) 54.1 (28.1) View Large Table 1. Mean (SD) target–nontarget quality rating for each taste by concentration step Concentration step Taste Sweet Salty Sour Meaty Bitter Low 26.4 (20.4) 14.7 (20.4) 16.6 (22.9) 26.7 (24.2) 36.5 (23.9) Medium 42.1 (19.9) 35.1 (25.6) 18.9 (23.4) 20.0 (30.8) 50.4 (27.5) High 52.0 (21.2) 48.1 (25.3) 29.2 (26.4) 22.3 (33.9) 54.1 (28.1) Concentration step Taste Sweet Salty Sour Meaty Bitter Low 26.4 (20.4) 14.7 (20.4) 16.6 (22.9) 26.7 (24.2) 36.5 (23.9) Medium 42.1 (19.9) 35.1 (25.6) 18.9 (23.4) 20.0 (30.8) 50.4 (27.5) High 52.0 (21.2) 48.1 (25.3) 29.2 (26.4) 22.3 (33.9) 54.1 (28.1) View Large The taste intensity ANOVA revealed a significant main effect of Concentration F(2, 96) = 84.15, P < 0.0001, η2 = 0.64, with intensity increasing with the concentration step (low M = 29.7 [SD = 21.5]; medium M = 38.8 [SD = 21.9]; high M = 49.6 [SD = 24.9]), with this occurring to a similar degree for each tastant (i.e., no interaction effect). There was also a main effect of Tastant type F(4, 192) = 31.70 P < 0.0001, η2 = 0.40, with quinine judged most intense and sucrose least, perhaps due to some conflation of intensity and hedonics. For the hedonic data, there was a main effect of Tastant F(3.51, 168.47) = 81.04, P < 0.0001, η2 = 0.63, with sucrose judged as most pleasant (M = 15.5 [SD = 31.4]), sour, and salty as unpleasant (Sour M = −25.3 [SD = 29.1]; Salty M = −27.9 [SD = 29.4]), meaty as more unpleasant still (M = −42.2 [SD = 29.8]), and bitter as the most unpleasant (M = −57.9 [SD = 29.8]). There was also a main effect of Concentration F(1.76, 84.56) = 8.52, P = 0.001, η2 = 0.15, all qualified by an interaction of Tastant and Concentration F(7.33, 351.71) = 4.29, P < 0.001, η2 = 0.08. As concentration increased for salty, sour, bitter, and meaty tastes, pleasantness ratings tended to decrease, whereas for sucrose, pleasantness ratings tended to increase. For irritancy, the ANOVA revealed a main effect of Concentration F(2, 96) = 33.15, P < 0.001, η2 = 0.41, with reported irritancy increasing with concentration step (low M = 20.6 [SD = 24.4], medium M = 26.3 [SD = 26.9], high M = 32.1 [SD = 30.6]). There was also a main effect of Tastant, F(4,192) = 48.35, P < 0.0001, η2 = 0.50 and an interaction between Tastant and Concentration F(4.60, 220.89) = 3.11, P = 0.01, η2 = 0.06. The relationship between Concentration and Irritancy was least pronounced for sucrose, in comparison with the other tastants (see Table 2). Table 2. Mean (SD) irritancy rating for each taste by concentration step Concentration step Taste Sweet Salty Sour Meaty Bitter Low 6.6 (12.8) 16.3 (18.4) 20.9 (20.4) 23.9 (25.6) 35.6 (27.9) Medium 5.9 (9.7) 24.7 (25.6) 23.2 (21.1) 29.6 (24.9) 48.0 (30.9) High 7.7 (14.4) 28.4 (26.8) 30.5 (26.6) 38.6 (31.6) 55.9 (30.1) Concentration step Taste Sweet Salty Sour Meaty Bitter Low 6.6 (12.8) 16.3 (18.4) 20.9 (20.4) 23.9 (25.6) 35.6 (27.9) Medium 5.9 (9.7) 24.7 (25.6) 23.2 (21.1) 29.6 (24.9) 48.0 (30.9) High 7.7 (14.4) 28.4 (26.8) 30.5 (26.6) 38.6 (31.6) 55.9 (30.1) View Large Table 2. Mean (SD) irritancy rating for each taste by concentration step Concentration step Taste Sweet Salty Sour Meaty Bitter Low 6.6 (12.8) 16.3 (18.4) 20.9 (20.4) 23.9 (25.6) 35.6 (27.9) Medium 5.9 (9.7) 24.7 (25.6) 23.2 (21.1) 29.6 (24.9) 48.0 (30.9) High 7.7 (14.4) 28.4 (26.8) 30.5 (26.6) 38.6 (31.6) 55.9 (30.1) Concentration step Taste Sweet Salty Sour Meaty Bitter Low 6.6 (12.8) 16.3 (18.4) 20.9 (20.4) 23.9 (25.6) 35.6 (27.9) Medium 5.9 (9.7) 24.7 (25.6) 23.2 (21.1) 29.6 (24.9) 48.0 (30.9) High 7.7 (14.4) 28.4 (26.8) 30.5 (26.6) 38.6 (31.6) 55.9 (30.1) View Large Color–taste pairings For test–retest reliability (see Table 3), there was better than chance agreement for all concentration steps for sweet, sour, and meaty. For bitter and salty, only the medium and high concentration steps were significantly better than chance. Table 3. Mean (SD) color–taste test–retest reliability data (percent agreement) across part 1 and 2 of the experiment, based upon color category, with all significantly different from the chance rate of agreement (17%) except as indicated Concentration step Taste Sweet Salty Sour Meaty Bitter Low 68% (47) 44% (50)* 58% (50) 42% (50)* 46% (50) Medium 76% (43) 52% (50) 54% (50) 50% (51) 46% (50) High 72% (45) 52% (50) 50% (51) 52% (50) 56% (50) Concentration step Taste Sweet Salty Sour Meaty Bitter Low 68% (47) 44% (50)* 58% (50) 42% (50)* 46% (50) Medium 76% (43) 52% (50) 54% (50) 50% (51) 46% (50) High 72% (45) 52% (50) 50% (51) 52% (50) 56% (50) *Not significantly different from chance View Large Table 3. Mean (SD) color–taste test–retest reliability data (percent agreement) across part 1 and 2 of the experiment, based upon color category, with all significantly different from the chance rate of agreement (17%) except as indicated Concentration step Taste Sweet Salty Sour Meaty Bitter Low 68% (47) 44% (50)* 58% (50) 42% (50)* 46% (50) Medium 76% (43) 52% (50) 54% (50) 50% (51) 46% (50) High 72% (45) 52% (50) 50% (51) 52% (50) 56% (50) Concentration step Taste Sweet Salty Sour Meaty Bitter Low 68% (47) 44% (50)* 58% (50) 42% (50)* 46% (50) Medium 76% (43) 52% (50) 54% (50) 50% (51) 46% (50) High 72% (45) 52% (50) 50% (51) 52% (50) 56% (50) *Not significantly different from chance View Large For consistency, analysis of the medium concentration of each tastant during part 1 and 2 revealed that participants’ choices were not randomly distributed across color categories (see Table 4). With the exception of meaty, the most common color was the same during part 1 and 2–red-pink for sweet, blue for salty, yellow for sour, and green for bitter. For meaty, the pattern of color choices was evenly distributed between 2 categories, red-pink and other color. Illustrative data are presented in Supplementary Figure 1, for the medium concentration on part 1 of the experiment—the first occasion on which color matches were made. Each panel contains 50 elements, with each element representing the color chosen by each participant for that particular taste. Table 4. Expected and observed values for participants categorized color choices, on part 1 (medium-concentration step), on part 2 (medium-concentration step), and overall with the modal color (across all 3 concentration steps on part 1 and 2) Color categories: Red and pink Green Yellow Blue Other colors Black, white, and gray Expected values (Null model) 8.3 8.3 8.3 8.3 8.3 8.3 Observed values—by taste and color choice measure Sweet1 1 38 3 1 3 4 1 2 40 2 2 3 4 0 Modal 43 1 2 1 3 0 Salty2 1 6 4 7 19 10 4 2 8 6 3 22 10 1 Modal 3 4 3 24 15 1 Sour3 1 4 13 25 2 6 0 2 5 11 21 6 7 0 Modal 3 8 19 4 15 1 Bitter4 1 6 16 5 4 8 11 2 6 17 5 7 4 11 Modal 5 12 7 3 6 17 Meaty5 1 17 5 2 7 16 3 2 14 6 5 10 15 0 Modal 13 5 2 7 23 0 Color categories: Red and pink Green Yellow Blue Other colors Black, white, and gray Expected values (Null model) 8.3 8.3 8.3 8.3 8.3 8.3 Observed values—by taste and color choice measure Sweet1 1 38 3 1 3 4 1 2 40 2 2 3 4 0 Modal 43 1 2 1 3 0 Salty2 1 6 4 7 19 10 4 2 8 6 3 22 10 1 Modal 3 4 3 24 15 1 Sour3 1 4 13 25 2 6 0 2 5 11 21 6 7 0 Modal 3 8 19 4 15 1 Bitter4 1 6 16 5 4 8 11 2 6 17 5 7 4 11 Modal 5 12 7 3 6 17 Meaty5 1 17 5 2 7 16 3 2 14 6 5 10 15 0 Modal 13 5 2 7 23 0 1Chi-square values: 127.6, 145.4, 173.7, P < 0.001; 2chi-square values: 19.4, 33.3, 56.3, P < 0.0017; 3chi-square values: 52.0, 30.6, 31.1, P < 0.001; 4chi-square values: 12.2, 14.3, 16.2, P < 0.02; 5chi-square values: 25.8, 19.8, 43.1, P < 0.002. View Large Table 4. Expected and observed values for participants categorized color choices, on part 1 (medium-concentration step), on part 2 (medium-concentration step), and overall with the modal color (across all 3 concentration steps on part 1 and 2) Color categories: Red and pink Green Yellow Blue Other colors Black, white, and gray Expected values (Null model) 8.3 8.3 8.3 8.3 8.3 8.3 Observed values—by taste and color choice measure Sweet1 1 38 3 1 3 4 1 2 40 2 2 3 4 0 Modal 43 1 2 1 3 0 Salty2 1 6 4 7 19 10 4 2 8 6 3 22 10 1 Modal 3 4 3 24 15 1 Sour3 1 4 13 25 2 6 0 2 5 11 21 6 7 0 Modal 3 8 19 4 15 1 Bitter4 1 6 16 5 4 8 11 2 6 17 5 7 4 11 Modal 5 12 7 3 6 17 Meaty5 1 17 5 2 7 16 3 2 14 6 5 10 15 0 Modal 13 5 2 7 23 0 Color categories: Red and pink Green Yellow Blue Other colors Black, white, and gray Expected values (Null model) 8.3 8.3 8.3 8.3 8.3 8.3 Observed values—by taste and color choice measure Sweet1 1 38 3 1 3 4 1 2 40 2 2 3 4 0 Modal 43 1 2 1 3 0 Salty2 1 6 4 7 19 10 4 2 8 6 3 22 10 1 Modal 3 4 3 24 15 1 Sour3 1 4 13 25 2 6 0 2 5 11 21 6 7 0 Modal 3 8 19 4 15 1 Bitter4 1 6 16 5 4 8 11 2 6 17 5 7 4 11 Modal 5 12 7 3 6 17 Meaty5 1 17 5 2 7 16 3 2 14 6 5 10 15 0 Modal 13 5 2 7 23 0 1Chi-square values: 127.6, 145.4, 173.7, P < 0.001; 2chi-square values: 19.4, 33.3, 56.3, P < 0.0017; 3chi-square values: 52.0, 30.6, 31.1, P < 0.001; 4chi-square values: 12.2, 14.3, 16.2, P < 0.02; 5chi-square values: 25.8, 19.8, 43.1, P < 0.002. View Large Further analyses were conducted on the modal color choice data (see Table 4). For sweet, sour, and salty, modal responses were the same as per the analyses above. For bitter, the black-gray-white category was the modal choice, with gray being most frequent. For meaty, the modal color choice was the other color category. To determine which “other color,” a further set of chi-square analyses were conducted using 8 color categories. This involved splitting responses in the red/pink category into red (values by approach now; 14, 12, 13) and pink (values; 3, 2, 0) and the other color category into violet (values; 1, 0, 9) and orange (values; 15, 15, 14). Orange, closely followed by red, were the colors most frequently chosen for meaty (chi-square values by approach, 32.9, 35.4, 33.8, all P < 0.001). To confirm that the color choice for each taste that we identified above was principally responsible for the nonrandom distribution of color–taste mappings, post-hoc tests were run. The distribution of observed values significantly deviated from the chance expected values for sweet (red/pink vs. remainder; chi square = 173.1, P < 0.001), salty (blue vs. remainder; chi square = 35.3, P < 0.001), sour (yellow vs. remainder; chi square = 16.4, P < 0.001), and bitter (black, gray, and white vs. remainder; chi square = 8.5, P < 0.005). For meaty, this analysis was also conducted but with different expected values based upon 8 color categories, and here too observed values significantly deviated from chance (chi square = 11.0, P < 0.005). Thus, these particular color choices significantly contributed to the nonrandom distribution of color choices made for each taste. To test for consistent pairings across taste-words and color-names, a further set of chi squares were run. As can be seen in Table 5, this revealed a very similar outcome to the data obtained with real tastants and colors. Table 5. Expected and observed values for participants’ categorized color choices when using color names and taste words in the final part of the experiment Color categories Red and pink Green Yellow Blue Other colors Black, white, and gray Expected values (Null model) 8.3 8.3 8.3 8.3 8.3 8.3 Observed values—by taste Sweet1 45 2 2 0 1 0 Salty2 3 2 5 28 5 7 Sour3 2 14 27 0 7 0 Bitter4 3 16 4 2 14 11 Meaty5 28 0 0 0 22 0 Color categories Red and pink Green Yellow Blue Other colors Black, white, and gray Expected values (Null model) 8.3 8.3 8.3 8.3 8.3 8.3 Observed values—by taste Sweet1 45 2 2 0 1 0 Salty2 3 2 5 28 5 7 Sour3 2 14 27 0 7 0 Bitter4 3 16 4 2 14 11 Meaty5 28 0 0 0 22 0 1Chi square = 201.1, P < 0.001; 2chi square = 57.5, P < 0.001; 3chi square = 67.4, P < 0.001; 4chi square = 22.2, P < 0.001; 5chi square = 102.2, P < 0.001, with 22/28 responses for red and pink being red and the other category solely composed of orange and brown. View Large Table 5. Expected and observed values for participants’ categorized color choices when using color names and taste words in the final part of the experiment Color categories Red and pink Green Yellow Blue Other colors Black, white, and gray Expected values (Null model) 8.3 8.3 8.3 8.3 8.3 8.3 Observed values—by taste Sweet1 45 2 2 0 1 0 Salty2 3 2 5 28 5 7 Sour3 2 14 27 0 7 0 Bitter4 3 16 4 2 14 11 Meaty5 28 0 0 0 22 0 Color categories Red and pink Green Yellow Blue Other colors Black, white, and gray Expected values (Null model) 8.3 8.3 8.3 8.3 8.3 8.3 Observed values—by taste Sweet1 45 2 2 0 1 0 Salty2 3 2 5 28 5 7 Sour3 2 14 27 0 7 0 Bitter4 3 16 4 2 14 11 Meaty5 28 0 0 0 22 0 1Chi square = 201.1, P < 0.001; 2chi square = 57.5, P < 0.001; 3chi square = 67.4, P < 0.001; 4chi square = 22.2, P < 0.001; 5chi square = 102.2, P < 0.001, with 22/28 responses for red and pink being red and the other category solely composed of orange and brown. View Large Possible mechanisms When asked, 36/50 (72%) participants reported basing one or more color choices on real-world taste–color associations (i.e., statistical). Two participants reported making choices based upon valence, with the remaining 12 offering no rationale. There was no difference in test–retest reliability between those reporting a statistical explanation versus no explanation. Tastant concentration was strongly related to the saturation of the selected color, F(1.82, 89.07) = 32.10, P < 0.0001, η2 = 0.40, with a highly significant linear trend (P < 0.001, η2 = 0.47) with saturation increasing with tastant concentration (low M = 54.9 [SD = 30.9]; medium M = 63.4 [SD = 28.2]; high M = 66.9 [SD = 29.43]). This relationship was consistent across all tastants (i.e., no interactions). There was also a main effect of Tastant type, F(2.65, 129.85) = 14.43, P < 0.0001, η2 = 0.23, with the most saturated colors chosen for sour (M = 73.7 [SD = 22.3]) followed by meaty (M = 65.3 [SD = 27.7]), salty (M = 63.5 [SD = 28.0]), sweet (M = 53.5 [SD = 23.4]), and then bitter (M = 50.9 [SD = 39.2]). We examined the data derived from each participant’s correlation of their taste and color hedonic ratings. The mean correlation (r′ = 0.65 [SD = 0.54]) was significantly different from 0, t(49) = 8.42, P < 0.001, indicating that selected colors match the valence of their associated taste. Discussion Using real tastants and colors, participants taste–color matches were generally consistent across part 1 and 2 of the experiment and were nonrandomly distributed across color categories. Color choices were very similar to those made using taste and color words both in prior studies (O’Mahony 1983; Heller 1999; Koch and Koch 2003; Tomasik-Krotki and Strojny 2008; Wan et al. 2014; Spence et al. 2015; Woods et al. 2016) and also in our own test of this issue. Many participants reported drawing upon associations between taste and the color of their associated object (generally a food). In addition, there was a strong relationship between the valence of the taste and the selected color. Finally, as with many other cross-modal matches (Marks 1989; Shermer and Levitan 2014), the stimulus magnitude dimensions of color saturation and tastant concentration were strongly related. Most participants in our experiment reported using the statistical approach, for one or more taste–color mapping. That this approach may in fact be relevant to taste–color pairings is suggested by another related literature—namely color-based valence judgments (e.g., the general liking for red and disliking for brown), which are also nonrandomly distributed in Western cultures (Taylor et al. 2013). Correlational evidence suggests that color-based valence judgments may arise from the valence of prototypical objects associated with that color (e.g., brown-feces; Palmer and Schloss 2010). Interestingly, in attempts to determine whether color-based valence judgments are culturally universal—a view long-held in the literature—recent testing in a remote setting suggested otherwise (Taylor et al. 2013) supporting the idea that color-based valence judgments do derive from the valence of objects prototypical for a particular color. As internet-based computer tasks were used to establish cross-cultural color–taste pairings, limited access to the internet in remote/rural settings is likely to generate greater sample homogeneity than the descriptor “tested in different countries” might suggest. Indeed, remote testing of color–taste associations may prove instructive. The statistical view is not mutually exclusive with the strong relationship we observed between taste and selected color valence. This is because shared valence between a taste and a color may be mediated by the prototypical food (or other) object that the taste brings to mind. Conversely, this can also be seen by the aversion to food and drinks, which are colored in an unusual manner (i.e., discordant with their prototypical presentation). Take for example, the declining sales associated with making previously colored beverages clear (e.g., crystal pepsi, Tab Clear) and when tomato sauce (red) is made green (e.g., Heinz’ green ketchup; Velasco et al. 2016). Thus, the valence of the food, its taste, and other properties (including its color) may very well be concordant. Consequently, one would expect to observe a correlation between the valence of tastes and their associated colors. If this is correct, then the relationship we observed emerges as a by-product of the statistical account. In addition to this possibility, a more purely valence-based account could also operate. Wang et al. (2016) suggested for certain taste–pitch matches that participants may match the hedonic valence of the taste with that of the pitch. Similarly, here, participants could be matching the valence of the taste with that arising from viewing the color. This explanation is exclusive, as it presumes that commonality of valence drives color choice, not mediation via a recalled prototypical object. Consistent with prior studies in other modalities (Marks 1989), we found that, as tastant concentration increased, there was also an increase in the saturation of colors selected—and that this relationship was relative and evident to the same degree for all tastants. In a related vein, intensity matching between taste and color can be seen with everyday foods. For instance, in a recent study Shermer and Levitan (2014) showed that the saturation of salsas (i.e., the intensity of their red color) biased participants ratings of their piquancy in taste (i.e., their spice intensity). These lines of evidence are consistent with 2 theories. First, the idea that underlying neural patterns that represent stimulus magnitude are similar across modalities—i.e., the structural account (Spence 2011). Second, the relationship between saturation and taste intensity is also compatible with the statistical account. For example, more intensely tasting foods tend (anecdotally) to have more vividly colored (i.e., saturated) packaging, and vice versa. Pertaining to these findings, 2 additional points need to be considered. First, saturation values differed across tastant type. This effect probably arose because of the greater number of 0 values associated with bitter taste due to achromatic selections. Second, it is possible that irritancy contributed to the concentration–saturation relationship. However, as sucrose does not become more irritating with increasing concentration, in contrast to salty and sour, the absence of a tastant by concentration interaction suggests that any effect of irritancy was probably minor. In conclusion, we find that, when participants are asked to match real tastants to real colors, their choices are generally reliable and consistent, and they are very similar to prior findings based on words alone. We suggest that color–taste relationships and their shared valence may arise from participants basing their color selections on prototypical food objects associated with each taste. Acknowledgements We would like to thank the reviewers for their positive and helpful feedback. The authors would also like to thank Alex Russell for providing the PowerPoint slides used in this study. References Cardello A , Schutz H . 2004 . Research note numerical scale-point locations for constructing the LAM (labeled affective magnitude) scale . J Sens Stud . 19 : 341 – 346 . Google Scholar CrossRef Search ADS Duffy VB , Hayes JE , Bartoshuk LM , Snyder DJ . 2009 . Taste: vertebrate psychophysics . In: Squire LR, editor. Encyclopedia of neuroscience . Oxford : Academic Press . p. 881 – 886 . Green BG , Dalton P , Cowart B , Shaffer G , Rankin K , Higgins J . 1996 . Evaluating the ‘Labeled Magnitude Scale’ for measuring sensations of taste and smell . Chem Senses . 21 : 323 – 334 . Google Scholar CrossRef Search ADS PubMed Heller E . 1999 . Wie farben wirken [How color works] . Reinbek bei Hamburg . Koch C , Koch EC . 2003 . Preconceptions of taste based on color . J Psychol . 137 : 233 – 242 . Google Scholar CrossRef Search ADS PubMed Marks LE . 1989 . On cross-modal similarity: the perceptual structure of pitch, loudness, and brightness . J Exp Psychol Hum Percept Perform . 15 : 586 – 602 . Google Scholar CrossRef Search ADS PubMed O’Mahony M . 1983 . Gustatory responses to nongustatory stimuli . Perception . 12 : 627 – 633 . Google Scholar CrossRef Search ADS PubMed Osgood CE, Suci GJ, Tannenbaum PH. 1957. The measurement of meaning. Urbana: University of Illinois Press. Palmer SE , Schloss KB . 2010 . An ecological valence theory of human color preference . Proc Natl Acad Sci USA . 107 : 8877 – 8882 . Google Scholar CrossRef Search ADS PubMed Palmer SE , Schloss KB , Xu Z , Prado-León LR . 2013 . Music-color associations are mediated by emotion . Proc Natl Acad Sci USA . 110 : 8836 – 8841 . Google Scholar CrossRef Search ADS PubMed Robinson JO . 1970 . The misuse of taste names by untrained observers . Br J Psychol . 61 : 375 – 378 . Google Scholar CrossRef Search ADS PubMed Shermer DZ , Levitan CA . 2014 . Red hot: the crossmodal effect of color intensity on perceived piquancy . Multisens Res . 27 : 207 – 223 . Google Scholar CrossRef Search ADS PubMed Skrandies W , Reuther N . 2008 . Match and mismatch of taste, odor, and color is reflected by electrical activity in the human brain . J Psychophys . 22 : 175 – 184 . Google Scholar CrossRef Search ADS Spence C . 2011 . Crossmodal correspondences: a tutorial review . Atten Percept Psychophys . 73 : 971 – 995 . Google Scholar CrossRef Search ADS PubMed Spence C , Wan X , Woods A , Velasco C , Deng J , Youssef J , Deroy O . 2015 . On tasty colours and colourful tastes? Assessing, explaining, and utilizing crossmodal correspondences between colours and basic tastes . Flav . 4 : 23 . Google Scholar CrossRef Search ADS Taylor C , Clifford A , Franklin A . 2013 . Color preferences are not universal . J Exp Psychol Gen . 142 : 1015 – 1027 . Google Scholar CrossRef Search ADS PubMed Tomasik-Krotki J , Strojny J . 2008 . Scaling of sensory impressions . J Sen Stud . 23 : 251 – 266 . Google Scholar CrossRef Search ADS Velasco C , Michel C , Youssef J , Gamez X , Check A , Spence C . 2016 . Using single colours and colour pairs to communicate basic tastes . Int J Food Design . 1 : 83 – 102 . Google Scholar CrossRef Search ADS Velasco C , Woods A , Deroy O , Spence C . 2014 . Hedonic mediation of the crossmodal correspondence between taste and shape . Food Qual Pref . 41 : 151 – 158 . Google Scholar CrossRef Search ADS Wan X , Woods AT , van den Bosch JJ , McKenzie KJ , Velasco C , Spence C . 2014 . Cross-cultural differences in crossmodal correspondences between basic tastes and visual features . Front Psychol . 5 : 1365 . Google Scholar CrossRef Search ADS PubMed Wang QJ , Wang S , Spence C . 2016 . “Turn up the taste”: assessing the role of taste intensity and emotion in mediating crossmodal correspondences between basic tastes and pitch . Chem Senses . 41 : 345 – 356 . Google Scholar CrossRef Search ADS PubMed Woods AT , Marmolejo-Ramos F , Velasco C , Spence C . 2016 . Using single colors and color pairs to communicate basic tastes II: foreground-background color combinations . Iperception . 7 : 2041669516663750 . Google Scholar PubMed Woods AT , Spence C . 2016 . Using single colors and color pairs to communicate basic tastes . Iperception . 7 : 2041669516658817 . Google Scholar PubMed Yan KS , Dando R . 2015 . A crossmodal role for audition in taste perception . J Exp Psychol Hum Percept Perform . 41 : 590 – 596 . Google Scholar CrossRef Search ADS PubMed © The Author(s) 2018. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

Chemical SensesOxford University Press

Published: Jun 2, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off