Color preference in red–green dichromats

Leticia Álvaro; Humberto Moreira; Julio Lillo; Anna Franklin

doi:10.1073/pnas.1502104112

New Research In

Edited by Dale Purves, Duke University, Durham, NC, and approved June 12, 2015 (received for review February 2, 2015)

Significance

Around 2% of males have red–green dichromacy, which is a genetic disorder of color vision that affects how well certain colors can be seen and discriminated. Humans with normal color vision are known to have systematic and reliable preferences for some colors over others (e.g., blue is liked and yellow-green is disliked). We show that red–green dichromats have a different reliable pattern of color preference in which, for example, yellow is the most, not the least, preferred color. We test current theories of color preference and provide novel evidence that how easily a color can be named is related to how much it is liked. The findings further understanding of dichromacy, color preference, and aesthetics in general.

Abstract

Around 2% of males have red–green dichromacy, which is a genetic disorder of color vision where one type of cone photoreceptor is missing. Here we investigate the color preferences of dichromats. We aim (i) to establish whether the systematic and reliable color preferences of normal trichromatic observers (e.g., preference maximum at blue, minimum at yellow-green) are affected by dichromacy and (ii) to test theories of color preference with a dichromatic sample. Dichromat and normal trichromat observers named and rated how much they liked saturated, light, dark, and focal colors twice. Trichromats had the expected pattern of preference. Dichromats had a reliable pattern of preference that was different to trichromats, with a preference maximum rather than minimum at yellow and a much weaker preference for blue than trichromats. Color preference was more affected in observers who lacked the cone type sensitive to long wavelengths (protanopes) than in those who lacked the cone type sensitive to medium wavelengths (deuteranopes). Trichromats’ preferences were summarized effectively in terms of cone-contrast between color and background, and yellow-blue cone-contrast could account for dichromats’ pattern of preference, with some evidence for residual red–green activity in deuteranopes’ preference. Dichromats’ color naming also could account for their color preferences, with colors named more accurately and quickly being more preferred. This relationship between color naming and preference also was present for trichromat males but not females. Overall, the findings provide novel evidence on how dichromats experience color, advance the understanding of why humans like some colors more than others, and have implications for general theories of aesthetics.

Individuals vary in their perceptual experience of the world, and sometimes this variation is caused by genetic differences (1 ⇓⇓–4). Dichromacy is a form of color-vision deficiency affecting about 2% of human males in which only two of the three types of retinal cone photoreceptors are functional because of genetic factors (1, 2). Protanopes, deuteranopes, and tritanopes lack cone photoreceptors sensitive to long (L), medium (M), and short (S) wavelengths, respectively. Accordingly, dichromats’ color discrimination is poorer, and their spectral sensitivity is slightly shifted to longer wavelengths (deuteranopes) or is moderately shifted to shorter wavelengths (protanopes) compared with that of normal trichromats (common observers; see table 3.6 in ref. 5).

In normal trichromats, cone responses are the input signals for two chromatic cone opponent mechanisms, red–green and yellow–blue, based on L−M and S−(L+M) cone responses, respectively, and one achromatic mechanism, mainly based on L+M responses (1). Traditionally it has been considered that protanopes and deuteranopes lack functionality in the red–green mechanism, because this opponent mechanism is based on the comparison of L and M cone responses, and one of those cone types is affected. (Thus such observers are called “red–green dichromats.”) However, research also has shown that a large proportion of red–green dichromats have residual activity in this mechanism with increasing stimulus size (over 3°; see refs. 6, 7), resulting in surprisingly good color naming (8 ⇓⇓–11). The origin of such red–green residual activity remains unknown and is open to several explanations (7, 10, 12, 13). Other research has shown that protanopes and deuteranopes also exhibit minor alterations in the performance of the achromatic and yellow–blue mechanisms (see figures 4 and 5 in ref. 14). These functional alterations in red–green dichromacy also affect color naming: Moreira et al. (15) have developed a model that explains 94% of the color-naming variance in protanopes and 96% of that in deuteranopes. The color naming of dichromats suggests that, at least in some circumstances, dichromat color perception is supported by red–green residual activity (R–G_res), in addition to yellow–blue and achromatic mechanisms (modeled as s′ and L*_T in ref. 15). These mechanisms also might be relevant for other aspects of dichromats’ color perception, such as color preference.

Although there has been much research on dichromats, their affective response to color has not been systematically investigated previously. Some visual simulations of how different stimuli appear to dichromats (16) have been demonstrated to work reasonably well (17). For example, these simulations suggest that some hues that appear reddish to common observers would appear desaturated and brownish for dichromats—a hue that normal trichromats typically dislike (18). Do dichromats also dislike a brownish appearance? Alternatively, if dichromats perceive a brownish appearance more commonly than trichromats, does this increased perception alter their preference for that appearance relative to other hues? Could dichromats’ and trichromats’ color preferences, like naming, be surprisingly similar if a residual red–green mechanism feeds into dichromats’ color preference?

Investigating color preference in dichromats may provide insight into how they perceive color and also may shed light on the origins of color preference and inform ways to test further theories and models of color preference. Decades of color research have indicated a systematic pattern of preference in trichromats: Blue hues are commonly preferred, and yellow–green hues are commonly disliked (18 ⇓–20). This pattern has been so systematic across studies that some have claimed it is universal, although others have pointed out that cross-cultural differences in color preference can be found also (21). Patterns of color preference have been explained in terms of the emotional response to color (22) and in terms of the valence of objects associated with colors (18).

Another theory is that color preferences can be summarized in terms of the cone contrast between a color and its background (19). Supporting this theory, around 70% of the variance in preference across a set of hues could be explained by L−M and S−(L+M) cone contrasts for British and Chinese trichromats (19). However, other studies have failed to account for this much of the variance with cone contrasts alone (18, 21). Palmer and Schloss (18) extended the model, adding achromatic contrast and saturation as predictors, but accounted for only 37% of the variance. If the cone-contrast model works, we should find altered patterns of color preference in dichromats that reflect their altered color vision. If so, we should be able to model dichromat color preference, as Moreira et al. (15) did for color naming, using a model tailored for dichromats’ altered cone-opponent mechanisms.

The current study investigates color preference in male dichromat (protanopes and deuteranopes) and trichromat observers (males and females). Color preference was measured for a set of 24 colors from the Berkeley Color Project (BCP) (18), comprising saturated, light, and dark versions of eight hues used in prior studies of trichromat color preference (18, 20, 21) and also the focal colors of the basic color categories (Fig. S1). The type of dichromacy was confirmed through the use of a set of color-vision tests including the Nagel anomaloscope. Observers rated their preference for the colors twice and also named the colors. Color preferences were compared across groups. To test the cone-contrast model, trichromat color preferences were modeled with trichromat cone contrasts as in Hurlbert and Ling (19). Dichromat color preferences then were modeled with cone contrasts that take into account dichromats’ altered cone responses. In addition, we explore the relationship between color naming and color preference in dichromats and trichromats, investigating whether color categorization, accuracy, speed, and consistency of naming predict how much a color is preferred.

Fig. S1.

Thirty-five stimuli from the saturated (S), light (L), dark (D), and focal (F) sets presented against the background used in the experiment. Colors are only an approximation of those in the experiment because of reproduction inaccuracies.

Results

Reliability.

To check that participants have reliable color preferences, the individuals’ preference ratings were averaged within each group and were correlated for the two times that colors were rated. Preferences showed high and significant correlations for the four groups (male, r = 0.98; female, r = 0.97; protanope, r = 0.94; deuteranope, r = 0.95, all P < 0.01). Preference also was reliable at the individual level (SI Text).

Normal Trichromat and Dichromat Color-Preference Curves.

Fig. 1 shows color-preference curves (preference plotted as a function of hue). The pattern of preferences for native Spanish normal trichromats in the current study is highly similar to those of normal trichromats from the United States (18) and the United Kingdom (20): Trichromats have the lowest preference for some yellowish-greenish hues, and preference rises steadily with increasing greenness, with a maximum around blue (Fig. 1 male and female). In contrast, dichromats have a distinctly different color-preference pattern (Fig. 1 protanope and deuteranope). For protanopes, preference is fairly flat across dark and light sets, but there is clear variation for saturated colors, for which preference peaks at yellow, with a minimum at cyan and red. Deuteranopes appear more similar than protanopes to male trichromats in their preference, particularly for light and dark sets. However, deuteranopes’ preference curve for saturated colors, like that for protanopes, has a peak at saturated yellow rather than at saturated blue, as seen for trichromatic males (see SI Text for analyses of the similarity in color-preference curves).

Fig. 1.

Mean preference ratings (± SEM) for 24 colors from the saturated, light, and dark sets, averaged for trichromatic males and females and dichromatic protanopes and deuteranopes. The x axis gives the hue: red (R), orange (O), yellow (Y), chartreuse (H), green (G), cyan (C), blue (B), and purple (P). Marker colors are only an approximation of those in the experiment.

The pattern of color preference for focal colors (Fig. 2) is similar to that of the saturated set (Fig. 1) for all four groups. As seen for the BCP stimulus set, preference for focal yellow was elevated for dichromats relative to trichromats. For the desaturated focal color categories (Fig. 2, Center), only females had a high preference for focal pink, and all four groups had a low preference for focal brown. For the achromatic categories, male trichromats appear to like white more than the other groups.

Fig. 2.

Mean preference ratings (± SEM) for 11 colors from the focal set, averaged for trichromatic males (solid line) and females (dotted line), dichromatic protanopes (dashed line), and deuteranopes (dotted and dashed line). The x axis gives the basic color term for the focal saturated color categories (Left), the desaturated color categories of pink and brown (Center), and the achromatic categories (Right): red (R), orange (O), yellow (Y), green (G), blue (B), purple (P), pink (Pi), brown (Br), black (Bk), gray (Gy), and white (W). Marker colors are only an approximation of those in the experiment.

To specify group differences in preference for specific hues, two sets of ANOVAs were conducted, one on the BCP stimulus set (Fig. 1) and another on the focal stimulus set (Fig. 2). For the BCP stimuli, a mixed-model ANOVA with set (saturated, light, and dark), color [red (R), orange (O), yellow (Y), chartreuse (H), green (G), cyan (C), blue (B), and purple (P)], and group (trichromat males, trichromat females, protanopes, and deuteranopes) was conducted on preference ratings, with the Greenhouse–Geisser correction applied when appropriate. This analysis revealed significant main effects of set [F(1.82,60) = 6.49, P < 0.01] and color [F(4.30,60) = 16.62, P < 0.001] and interactions of set and color [F(9.32,60) = 13.39, P < 0.001] and color and group [F(12.88,60) = 5.19, P < 0.001]. To unpack the interaction relevant to our hypotheses, post hoc one-way ANOVAs on the mean preference for each color (mean for saturated, light, and dark sets) were conducted with group as a factor. These analyses revealed group differences for yellow [F(3,60) = 9.07, P < 0.001], cyan [F(3,60) = 6.23, P < 0.01], and purple [F(3,60) = 3.11, P < 0.05]. Bonferroni-corrected pairwise comparisons revealed that protanopes preferred yellow more than trichromat males (P < 0.01), and protanopes preferred cyan less than trichromat males (P < 0.05). Comparisons with females are provided in SI Text.

Another mixed-model ANOVA with color [R, O, Y, G, B, P, pink (Pi), brown (Br), black (Bk), gray (Gy), and white, (W)] and group on focal preference ratings revealed significant main effects of color [F(7.04,60) = 20.64, P < 0.001] and an interaction of color and group [F(21.11,60) = 3.22, P < 0.001]. To investigate this interaction, one-way ANOVAs with group as a factor were conducted on the preference ratings for 11 focal stimuli. These analyses revealed significant group differences for green [F(3,60) = 3.41, P < 0.05], yellow [F(3,60) = 5.23, P < 0.01], pink [F(3,60) = 6.90, P < 0.001], and black [F(3,60) = 3.12, P < 0.05]. Bonferroni-corrected pairwise comparisons did not show significant differences between dichromat and trichromat males for any color. For several colors, however, dichromat and male color preferences differed from those of females (SI Text).

Predictors of Color Preference: Cone Contrast, Lightness, and Saturation.

As in Hurlbert and Ling (19), we analyzed color-preference data (excluding foci) in terms of the two dimensions of opponent cone-contrast space (23). [The focal set has been excluded from these analyses to provide analyses that are comparable to previous research (19, 21), which did not include focal colors. In addition L−M and S−(L+M) will fail to account for the preference of the three achromatic foci.] L−M and S−(L+M) cone contrasts between stimulus and background were calculated (SI Text) and entered as predictors in linear regressions with average color-preference ratings as the outcome variable. We conducted separate regressions for each group of observers. The cone-contrast model significantly accounted for just under half of the variance in the average pattern of color preference for the BCP color set for normal trichromats (Table 1), but for dichromats this model accounted for less variance and was not significant (and therefore is not included in Table 1). S−(L+M) was the only significant predictor for trichromat males [t = 2.65, P < 0.05] and females [t = 3.90, P < 0.001; over a third of the variance was explained by S−(L+M) in a univariate model (Table 1)], but L−M was not a significant predictor for any group (all P > 0.05). Equivalent regressions with predictors drawn from a multistage model that rotates the standard axes into RG, YB (Fig. S2) (24) produced a highly similar pattern of results (SI Text). Additional regressions indicated that lightness (L*) and saturation (s_uv) were not significant predictors of trichromats’ preferences (SI Text).

View this table:

Table 1.

Results for multivariate [L−M and S−(L+M) as predictors (Model)] and univariate [S−(L+M), |s′|, L−M, or |S−(L+M)| as predictors] regressions conducted on trichromat or dichromat preference ratings for various stimulus sets

Fig. S2.

Stimulus–background cone contrast and RG, YB values for saturated (S), light (L), and dark (D) colors computed using the two-stage model L−M (A) and S−(L+M) (B) or the indiscriminate version of the multistage model RG (C) and YB (D).

Hurlbert and Ling (19) applied the cone-contrast model to preference ratings for colors that varied only in hue and accounted for much more variance than seen in subsequent research, which, as here, used BCP stimuli that vary in lightness, saturation, and hue (18, 21). Both Taylor et al. (21) and Palmer and Schloss (18) suggested that this discrepancy could be caused by the cone-contrast model capturing preference better when the stimuli entered into the regression mainly vary in hue. To explore this possibility, we performed separate regressions for each set (saturated, light, and dark) so that the variation in the lightness and the saturation of the stimuli was minimized. Fig. 3 provides the cone contrasts for the three sets (Fig. 3 A–C) and the regression weights when these cone contrasts are entered as predictors of preference for those colors. If cone contrast strongly predicted preference, then the color-preference curves in Fig. 1 should resemble the cone-contrast curves in Fig. 3, and the regression weights should be strong and significant. The cone-contrast model accounted for 75% to almost 100% the variance, depending on color set, for trichromats. S−(L+M) was a significant predictor in these regressions (all P < 0.05) (Fig. 3D) except for trichromat males for the saturated set (P = 0.06; see Fig. 3D). L−M was a significant predictor only for the trichromat males’ preferences in the light set (t = −1.45, P < 0.001) (Fig. 3E).

Fig. 3.

Values for regression predictors (A–C) and resulting regression weights (D–G). (A–C) S−(L+M) (A), L−M (B), and |S−(L+M)| (C) stimulus–background cone-contrast values for saturated (S), light (L), and dark (D) colors. (D–G) Mean (± SEM) individual regression weights when S−(L+M) and L−M are entered as predictors of trichromat preference of saturated, light, and dark colors (D and E) or |S−(L+M)| or L−M is entered as the predictor of dichromat preference (F and G).

The above regressions based on a trichromatic cone-contrast model were repeated for dichromats and, surprisingly, accounted for a significant amount of the variance (78%) for deuteranopes’ preference of the light set, with L−M being a significant predictor (t = −0.97, P < 0.05) (Fig. 3G). Regressions were conducted to investigate whether corrected variables that take into account dichromats’ cone responses (15) for lightness (L*_T) and saturation (s′) (SI Text) could predict dichromat color preferences. Transformed lightness (L*_T) could not account for color preference in dichromats (SI Text). Because dichromats’ preference curves peaked both at yellow and blue (Fig. 1 protanope and deuteranope) and s′ takes positive values for yellowish colors and negative values for bluish colors (SI Text), we thought that absolute s′ (mechanism activation) was more appropriate for saturation estimations for protanope and deuteranope observers. We found |s′| was a significant predictor for protanopes and explained more than 25% of the variance (Table 1). Because s′ is related to yellow–blue mechanism activity, we repeated the cone-contrast regressions but with absolute S−(L+M). As expected, similar to |s′|, the variance explained by |S−(L+M)| was significant and explained almost two thirds of the variance in protanopes’ preference for the full color set and more than three quarters of the preference for saturated and light colors (Fig. 3F). For deuteranopes, |S−(L+M)| was a significant predictor of preference only for saturated colors (Fig. 3F).

Relationship of Color Preference and Color Naming.

The naming data were used to group the stimuli according to the 11 terms used (red, green, yellow, blue, orange, pink, purple, brown, black, gray, and white; see Methods for corresponding terms in Spanish). For each observer, each category was composed of stimuli consistently named with the appropriate term. One-way ANOVA analyses of preference ratings when stimuli were classified according to each participant’s naming showed group differences for yellow [F(3,216) = 13.38, P < 0.001], blue [F(3,403) = 8.93, P < 0.001], brown [F(3,109) = 3.31, P < 0.05], black [F(3,60) = 3.32, P < 0.05], and pink [F(3,219) = 8.97, P < 0.001] categories. Bonferroni-corrected comparisons revealed that both groups of dichromats preferred colors named as yellow more than trichromat males (P < 0.05), and protanopes preferred blues less than deuteranopes and trichromat males (P < 0.05). Female comparisons are given in SI Text.

Table S1 gives the frequencies of color names provided by each group. The naming-error rate (a name was considered an error if it was not the modal response of normal trichromats) (11) was significantly larger (t = 4.49 P < 0.001) in protanopes (39.14%) than in deuteranopes (27.56%). Additional analyses are given in SI Text.

View this table:

Table S1.

Naming percentage (>5%) for 35 colors from the focal, saturated, light, and dark sets for normal trichromatic males, females, and dichromatic protanopes and deuteranopes

In addition, for each stimulus and for each group, naming consistency (the percentage of observers using the same color name in both stimulus presentations) and naming consensus (how much observers within a group agree on the name, the percentage frequency of the most frequent term for that group) were computed. We obtained significant correlations between preference and naming response time (negative), naming-error rate (negative), consistency (positive), and consensus (positive) for all groups except females (Table 2 and Fig. S3).

View this table:

Table 2.

Correlation coefficients (r) between preference and naming response time, naming-error rate, naming consistency, and consensus

Fig. S3.

Color-preference ratings (y axis) plotted against the naming response time (RT) (x axis) for 35 colors from the focal (inverted triangles), saturated (circles), light (triangles), and dark (squares) sets, averaged for trichromatic males and females, dichromatic protanopes, and deuteranopes. R² and the best-fitting lines are given also. Marker colors are only an approximation of those in the experiment. ***P < 0.001.

An analysis was conducted to investigate whether dichromats preferred colors more if they gave them the same name as trichromats than if they made naming errors. Unpaired t tests revealed that dichromats preferred colors that they named the same as trichromats (preference for hits, P_h) more than those to which they gave different names on at least one of the two naming trials (preference for errors, P_e), and this effect was significant for both protanopes (mean P_h = 5.41, mean P_e = 4.98, t = 2.33 P < 0.05) and deuteranopes (mean P_h = 5.59, mean P_e = 4.64, t = 5.37 P < 0.001). Additional analyses of error-naming rate are given in SI Text.

A final series of multiple regression analyses was performed to evaluate whether adding naming response times to the cone-contrast regressions accounted for more variance in preference. Naming response time was a significant predictor that increased the explained variance for the three male groups (t = −5.39, explained variance increased from 39.9 to 75.2%, P < 0.001, in trichromat males; t = −2.19, explained variance increased from 60.9 to 68.2%, P < 0.05, in protanopes; t = −3.89, explained variance increased from 14.6 to 50.3%, P < 0.001, in deuteranopes) but not for females (P > 0.05; the increase in explained variance from 46.2 to 51.8% was not significant).

Discussion

Research on dichromacy has covered various topics such as genetics (2), psychophysics (5), color naming (11, 15), and even color-appearance models (7, 16, 25). Here we provide a novel report of the color preference of red–green dichromats and compare their color preference with the color preference and its explicative mechanisms of normal trichromats, including suitable transformations of such mechanisms for dichromats (L*_T and s′). In addition, we investigated the relationship between color naming and color preference. We find that dichromats reliably prefer some colors over others, but these preferences do not fit the systematic pattern found in normal trichromats, who commonly prefer blue hues and commonly dislike yellow-green hues (18 ⇓–20). Instead, saturated yellow was the most preferred color for dichromats, and protanopes’ preference for cyan was muted compared with normal trichromats. The patterns of dichromat and trichromat color preference were more similar across the focal colors than for a set of colors that included less typical examples.

Color-preference patterns were summarized in terms of the cone contrasts between stimulus and background, as in Hurlbert and Ling (19), explaining almost half of the variance in trichromat male and female color preference for all colors, with S−(L+M) being the most significant predictor. As in Palmer and Schloss (18) and Taylor and Franklin (20), less variance was explained than reported in Hurlbert and Ling (19). However, when we modeled preference for colors that varied mainly in hue (as in ref. 19), the explained variance increased substantially, by more than two thirds, depending on the color set, with S−(L+M) remaining the main predictor. L−M also was a significant predictor for trichromat males, but only for lighter colors. Hurlbert and Ling (19) also found that L−M contributed to the male preference for colors that were similar to ours in lightness (our mean L* was ∼80; Hurlbert and Ling’s mean L* was ∼77).

As expected, the cone-contrast model could not explain dichromat color preference: Dichromats supposedly lack L−M cone opponency, and S−(L+M) cone contrast cannot explain high preferences both for blue and yellow hues simultaneously. However, |S−(L+M)|, which ignores the contrast sign but captures its intensity, accounted for almost two thirds of the variance in protanope preference. When colors varied mainly in hue, |S−(L+M)| predicted more than three quarters of the variance in protanope preference. For deuteranopes, neither |S−(L+M)| or |s′| was a significant predictor of preference when colors varied in hue, saturation, and lightness, but |S−(L+M)| was a significant predictor for deuteranope preference of saturated colors that varied mainly in hue. Surprisingly, given the lack of M cones for deuteranopes, L−M significantly predicted more than half of the variance in deuteranope preference for light colors, as it did for the trichromat males in this study and in Hurlbert and Ling (19).

As in prior studies (18 ⇓–20), we modeled color preference using a cone-contrast model [S−(L+M); L−M] developed from psychophysical experiments (i.e., threshold detection; see ref. 23). Some have argued that this kind of model does not fully account for hue perception (SI Text). Alternative models that combine the responses from the three cone types in a different way (e.g., a multistage model described in ref. 24, in which there is an important contribution of S-cone signals for both red–green and yellow–blue axes) have been shown to be more consistent with perceptual experience (26). When we repeated analyses using the multistage model, we found that the two kinds of model equally accounted for the color preference of normal trichromats (SI Text).

Dichromats named colors differently than trichromats about one third of the time. As previously found (11), there were significantly more naming errors by protanopes than by deuteranopes. The difference probably is related to (i) the greater alteration of the achromatic mechanism in protanopes than in deuteranopes (see, for example, the luminous efficiency curves in figure 4 in ref. 14) and (ii) the greater importance of the achromatic mechanism in dichromat than in trichromat naming (see figure 3 in ref. 11). When color preferences were analyzed according to naming, dichromats preferred colors named as yellow more than trichromats, and protanopes preferred blues less than trichromats. Colors that were named more quickly and with less error, consistently, and according to group consensus were most preferred for all groups except females. In addition, dichromats preferred colors that they named the same as trichromats more than colors that they named differently from trichromats.

Combined, the findings have implications for our understanding of (i) how red–green dichromacy affects color perception; (ii) the mechanisms of color preference; and (iii) general principles of aesthetics and perception. We discuss each in turn.

First, the findings suggest that dichromacy affects color perception more in protanopes than in deuteranopes. For both naming and preference, the similarity between dichromats and trichromats was less for protanopes than for deuteranopes. This result supports prior research, which has found that protanopes make more naming errors than deuteranopes (9, 11), and shows for the first time, to our knowledge, that this effect extends to color preference. For deuteranopes, there also was evidence of a residual red–green mechanism affecting color preference, because L−M significantly predicted deuteranope preference for light colors, as it did for male trichromats. Prior research has argued for a residual red–green mechanism influencing dichromat color naming for stimuli over 3° (8 ⇓⇓–11). Here, we show that such a mechanism is likely at play when deuteranopes decide how much they like light colors and provides evidence of residual red–green discrimination in an aesthetic aspect of color perception. Various explanations have been proposed for this residual red–green mechanism, such as rod intrusion (10), gains and nonlinearities applied to the receptor signals (7), or the influence of chromatic aberrations (12). Further research is needed to understand this mechanism and why it appears to influence deuteranope but not protanope preference.

Second, the findings have implications for color-preference models. We show that cone contrast can be amazingly effective in capturing color preference when colors vary mainly in hue (explaining almost 100% of the variance in the male preference in light colors) but is less effective for colors that vary in hue and vary strongly in saturation and lightness. Therefore, the model can provide a useful way of comparing the mechanisms of hue preference across groups even if it does not capture preference in a 3D color space. The fact that dichromat color preferences could be modeled using modified variables that consider their altered cone response provides further corroboration that hue preferences can be modeled by the underlying subsystems of color vision. The current findings also suggest that the yellow–blue mechanism is much more important for human color preference than the red–green one, perhaps explaining why the red and green preferences of dichromats with a residual red–green mechanism were so similar to those of trichromats.

Color preference has been explained in terms of the emotional response to color (22) and the valence of objects associated with colors (18). Here, we also show that the degree to which males (but not females) like a color is related to how easy it is to name the color. Response time explained up to half the variance in color preference for trichromat males and for dichromats. Although this variance is less than that explained by the relationship between color preference and the valence of color-associated objects (80%) (18) or emotional response (67%) (22), the relationship between naming speed and preference is still substantial. Furthermore, when we used both the most relevant cone-opponent mechanism and naming response time as predictors, the explained variance increased significantly (by more than 50% for all male groups). Cone contrast and color naming do not account for all the variance in preference. Given that past research has revealed other predictors which also partially explain color preference, future research should aim at understanding the relationships between the various predictors, including those identified in the current study, and should investigate the extent to which each predictor is weighted by different groups. Some combination of predictors may provide a more complete model, although accounting for 100% of the variance in a human judgment such as preference is, of course, a challenge. Future research also should consider whether nonlinear models account for more of the variance, because only linear models have been considered so far.

Third, the finding has broader implications for theories of aesthetics. Since Fechner founded the field of aesthetics research in the late 19th century, experimental research has sought to identify general principles of aesthetics (27). One contemporary theory proposes that the more fluently a stimulus can be processed, the more it is liked (28). It has been proposed that stimuli with greater contrast, clarity, symmetry, familiarity, and prototypicality are processed more fluently by an observer (29), and relationships between fluency and preference have been documented for stimuli such as music (30), faces (31), voices (32), and art (33). Prior work has found greater preference for focal than nonfocal colors (34), but an explicit link between color preference and fluency has not been demonstrated previously. Here we show that, for males, the general principle of fluency (measured by ease of naming, i.e., naming response time) at least partially accounts for color preference. Importantly, we also show that when genetic mutation (e.g., dichromacy) affects stimulus fluency, the relationship between fluency and preference holds for males. Finally, we show that there can be group differences in the extent to which fluency predicts preference, because the relationship between fluency and color preference was present only for tri- and dichromat males, but not for females. Further investigation of such group differences may clarify the conditions under which fluency contributes to preference, as opposed to alternative mechanisms such as complexity or novelty (35). More broadly, the current finding that group differences in color preference (e.g., differences between trichromats and dichromats) are accounted for by biological components of color vision resonates with the debate on the relative contributions of culture, environment, and genes to individual differences in perception (1 ⇓–3).

Combined, the findings on cone contrast and color naming suggest that the more psychological specificity colors have, the more they are preferred. By “psychological specificity” we mean the extent to which colors produce the greatest response in the visual system and the extent to which they are salient, distinctive, and less prone to be confused with other colors. For example, dichromat preference peaks at colors with the strongest response in the yellow–blue mechanism and that are easiest for dichromats to name. As Palmer and Schloss (18) argue, human color preferences could be essentially adaptive, and people could be more likely to interact, survive, and reproduce successfully if they prefer some colors over others (19). Because color perception has adaptive uses related to the speed and ease of processing (e.g., when locating something useful or dangerous), it seems logical that males prefer colors that are easy to name and therefore process. The lack of such a relationship for females deserves further investigation and could reveal important differences in the underlying mechanisms of male and female color preference. The current investigation reveals the aesthetic response of dichromats to their colored world and provides greater understanding of dichromacy, why we like some colors more than others, and general theories of aesthetics.

Methods

Participants.

Participants were 15 males and 17 females with normal color vision and 17 deuteranope males and 15 protanope males with dichromatic color vision. There were no significant group differences in age (mean age: trichromat males = 29.40 y, SD = 6.49; trichromat females = 28.29 y, SD = 9.73; protanopes = 34.00 y, SD = 8.60; deuteranopes = 27.82 y, SD = 8.59), F(3,60) = 1.72, P = 0.17. All participants were native Spanish speakers; 18 were students from Universidad Complutense de Madrid, but most were research volunteers from the local area. Their color vision was tested with a set of psychophysical tests (36 ⇓–38). None had tritan defects. To assign them into groups, participants with color-vision deficiencies also performed a Rayleigh match in a Nagel anomaloscope (Tomey, AF-1). The research was conducted according to the principles of the Declaration of Helsinki, and all participants gave informed consent. This research was approved by the Universidad Complutense de Madrid – Hospital Clínico San Carlos review board.

Stimuli and Experimental Set-Up.

The 24 stimuli (Fig. S1) were close approximations of the saturated (S), light (L), and dark (D) sets from the BCP (18). The saturated set comprised good saturated examples of red (R), orange (O), yellow (Y), chartreuse (H), green (G), cyan (C), blue (B), and purple (P). The light and dark sets were approximately halfway between those of the corresponding hue in the saturated set and Munsell Chroma 1. Munsell value (lightness) was approximately halfway between that of the corresponding hue in the saturated set and Munsell value 9 (light) or 1 (dark). We also used a set of focal stimuli that were the best exemplars of the 11 basic color categories named in Spanish (11) and included the best exemplars of red, green, yellow, blue, orange, pink, purple, brown, black, gray, and white (Fig. S1). The chromaticity coordinates for the 35 stimuli were verified with a Minolta CL-200 luxo-colorimeter (Table S2). All stimuli were presented as circular patches (8°) on a calibrated 24-inch in-place switching LCD monitor with a hood in a dark room (LaCie 324i; resolution of 1,920 × 1,200 pixels; reference white: Y = 18,195 cd/m², x = 0.36, y = 0.35). The center of the monitor was at participants’ eye level, at a distance of 65 cm.

View this table:

Table S2.

CIE (1931) Y (cd/m²), x, y, and L* values of the 35 stimuli

Design and Procedure.

Participants completed a color-preference task and a color-naming task, with task order counterbalanced across participants. For the color-preference task, stimuli were presented individually in the center of a gray background (Y = 32.35 cd/m², x = 0.35, y = 0.35). A line representing a rating scale (22° horizontal, 1° vertical) was presented below the colored circle, and participants were required to move the cursor from its neutral position in the middle of the scale to indicate their degree of preference for the stimuli. Text (12-point Arial font) below each endpoint specified what that end of the scale represented (0, nada, not at all; 10, mucho, very much). A trial began with a 500-ms fixation cross (1° diameter) and a simultaneous whistle, followed by 500 ms of gray background. The stimulus remained on the screen until the participant indicated how much they liked the color by sliding the cursor along the rating scale and clicking the confirmation button. After performing 24 practice trials, participants rated each stimulus twice in a pseudorandom order (without consecutive stimulus repetitions).

The naming task was identical, except that participants were required to name rather than rate the colors, and each stimulus remained on the screen for 2.5 s. Participants were instructed to name aloud the color using one of the 11 Spanish basic color terms: rojo (red), verde (green), amarillo (yellow), azul (blue), naranja (orange), rosa (pink), morado, (purple), marrón, (brown), negro (black), gris (gray), or blanco (white). Response time was registered as the start of the verbal response using a unidirectional condenser microphone (Fonestar FMC-616) and a digital recorder (Olympus VN-711PC) for backup purposes. The experimenter manually registered color names and verbal artifacts (i.e., dubious interjections, changes in response). Participants took an average of 12 min to complete the full experiment. For both tasks, participants were instructed to answer as quickly but precisely as possible.

SI Text

SI Stimuli

The stimuli used are shown in Fig. S1.

Reliability.

When individual data were correlated (rather than taking group averages), all individuals in the four groups were reliable in their preference, with significant correlations (all P < 0.01) between first and second rating for normal trichromats (mean r_xx: r = 0.84 for males and r = 0.88 for females) and dichromats (mean r_xx: r = 0.77 for protanopes and r = 0.76 for deuteranopes).

Similarity in Group Color-Preference Curves.

To quantify the similarity of dichromats’ and trichromats’ color preferences across all stimuli, preference ratings of protanopes and deuteranopes were correlated with ratings of male trichromats (because the dichromats also were male). Male trichromat color preference was significantly related to both dichromat groups but was more highly related to deuteranopes (r = 0.81, P < 0.001) than to protanopes (r = 0.46, P < 0.01). Significant differences between both correlations were confirmed after Fisher’s Z transformation (Z = 2.52, P < 0.01), suggesting that color preference is affected more by protanopia than by deuteranopia. Interestingly, male and female trichromat color preference (r = 0.64, P < 0.001) is less similar than male trichromat and deuteranope color preference, but the difference between both correlations was not statistically significant (Z = 1.48; P = 0.07).

Female Trichromat and Dichromat Color Preference for BCP Stimuli: Pairwise Comparisons.

Bonferroni-corrected pairwise comparisons revealed that trichromat females preferred yellow less than protanopes (P < 0.001) and deuteranopes (P < 0.05); protanopes preferred cyan less than trichromat females (P < 0.01); deuteranopes preferred purple less than females (P < 0.05).

Female Trichromat Color Preference for Focal Colors Versus Dichromat or Male Color Preference for Focal Colors: Pairwise Comparisons.

Focal yellow preference was significantly higher in protanopes and deuteranopes than in trichromat females; focal pink preference was significantly lower in protanopes and deuteranopes than in trichromat females; and focal black preference was significantly higher in deuteranopes than in trichromat females. Normal male and female trichromats differed only in their preference for green: focal green preference was significantly higher in males than females (all P < 0.05).

Colorimetric Computations: Cone Contrast.

The L−M and S−(L+M) cone-opponent color contrasts between stimulus and background were calculated using the procedure described by Hurlbert and Ling (19). Beginning with x, y, Y [International Commission on Illumination (CIE) 1931] values and the Smith and Pokorny (39) cone fundamentals, the L, M, and S cone excitations were calculated. Cone-contrast values (ΔL, ΔM, and ΔS) were calculated as ΔL = (L_s − L_b)/L_b; ΔM = (M_s − M_b)/M_b; and ΔS = (S_s − S_b)/S_b, where the subscript “s” represents the stimulus, and the subscript “b” represents the background color. Following Eskew et al. (23), the L−M cone contrast is equal to LM_c = 0.7*ΔL − 0.72ΔM + 0.02*ΔS, and the S−(L+M) cone contrast is equal to S_c = 0.8*ΔS − 0.55ΔL − 0.25*ΔM.

Colorimetric Computations: Multistage Model.

To account for some perceptual phenomena unexplained by the standard two-stage model [L−M and S−(L+M)], De Valois and De Valois (24) proposed an alternative multistage color-vision model that incorporates the rotation of the standard chromatic axes. To investigate whether this multistage model can account for color preference to a greater extent than the standard model, the RG and YB systems were modeled using the following procedure. First, beginning with x, y, Y (CIE 1931) values and the Smith and Pokorny (39) cone fundamentals, the L, M, and S cone excitations (relative to the background color) were calculated. Second, RG and YB perceptual opponency was calculated using the computations given in table 1 in ref. 24. They suggested two possibilities that differ in the nature of the inputs to the surrounds of the midget and bipolar ganglion cells: indiscriminate surrounds (inputs coming from L, M, and S cones) or cone-specific (discrete) surrounds (L-cone center cells receive an antagonistic input only from M cones, and vice versa). In the indiscriminate version, the red–green system is equal to RG = 90*L − 115*M + 25*S and the yellow–blue system is equal to YB = −130*L + 95*M + 35*S; In the discrete version, RG = 90*L − 115*M + 30*S and YB = −130*L + 95*M + 30*S. Fig. S2 compares the L−M and S−(L+M) values (23) of the saturated, light, and dark sets of colors and the corresponding RG and YB values (24).

Regression on Color Preference: Multistage Model.

The indiscriminate multistage color model significantly accounted for less than half of the variance in the average pattern of color preference for trichromat males [39.3%, F(2,21) = 6.79, P < 0.01] and females [46.2%, F(2,21) = 9.02, P < 0.01] but accounted for less variance and was not statistically significant for protanopes [5.3%, F(2,21) = 0.59, P = 0.57] and deuteranopes [11.6%, F(2,21) = 1.37, P = 0.28]. The YB system was the only significant predictor for trichromat males [t = 3.68, P < 0.01; the univariate linear regression model accounted for 39.1% of the variance, F(1,22) = 14.11, P < 0.01] and females [t = 4.06, P < 0.01; 38.3%, F(1,22) = 13.63, P < 0.01], but RG was not a significant predictor for any group (all P > 0.05). The results of the discrete multistage color model were similar to those of the indiscriminate model for all four groups: trichromat males [41.0%, F(2,21) = 7.29, P < 0.01], females [45.9%, F(2,21) = 8.91, P < 0.01], protanopes [5.3%, F(2,21) = 0.58, P = 0.57], and deuteranopes [12.3%, F(2,21) = 1.48, P = 0.25]. The YB system was the only significant predictor for trichromat males [t = 3.82, P < 0.01; the univariate linear regression model accounted for 40.9% of the variance, F(1,22) = 15.20, P < 0.01] and females [t = 3.85, P < 0.01; 36.5%, F(1,22) = 12.62, P < 0.01], but RG was not a significant predictor for any group (all P > 0.05).

Colorimetric Computations: L*_T.

Two L*_T values were calculated for each stimulus, bearing in mind protanopes’ lack of functionality in L cones and deuteranopes’ lack of functionality in M cones. The procedure was similar to that of Moreira et al. (15). For each stimulus L and M cone responses were estimated from x, y, Y (CIE 1931) values and the Smith and Pokorny (39) cone fundamentals. The relative contribution of L and M cones to luminosity (2:1) was used to compute a specific transformed luminance for protanopes (Y_p) and deuteranopes (Y_d), assuming that the luminance of the reference white is exactly the same both for normal trichromats and dichromats, so that it is possible to compute one scaling factor for protanopes and another for deuteranopes. Finally, we computed the corresponding (L*_T) values for both types of dichromats.

Colorimetric Computations: s′.

Two s′ values were computed for each stimulus following a procedure analogous to that of Moreira et al. (15). Transformed saturation identifies hue and quantifies saturation using the confusion lines in the chromaticity diagram. A confusion line includes all the pseudoisochromatic stimuli for a dichromat type. All confusion lines radiate from a dichromatic confusion point (protanope: x = 0.75, y = 0.25; deuteranope: x = 1.4, y = −0.40). Every confusion line has a specific s′ value. This value is zero for the line that includes the achromatic point (x = 0.36, y = 0.35), because that line goes through all the stimuli that dichromats must perceive as achromatic. To compute s′ for the rest of the confusion lines, each line defines an intersection point with the yellow line (λ_d = 575 nm) or with the blue line (λ_d = 475 nm). The variable s′ is the distance between the achromatic point and the intersection point. This distance coincides (except for a scalar) with the value of s_uv for the intersection point. To identify hue, s′ is positive for intersection points located on the yellow line and negative for those located on the blue line.

Regression on Color Preference: Lightness and Saturation as Predictors.

For lightness, regressions were conducted for the four groups with standard lightness [L*: male 14.7%, F(1,22) = 3.80; female 1.6%, F(1,22) = 0.36; protanopia 6.9%, F(1,22) = 1.63; deuteranopia 0.1%, F(1,22) = 0.03] and for both dichromat groups with transformed lightness [L*_T: protanopia 5.7%, F(1,22) = 1.34; deuteranopia 0.4%, F(1,22) = 0.76] as predictors, but none was significant (all P > 0.05). For saturation, regressions were conducted on the average preference ratings for the four groups with standard saturation (s_uv) as predictor [male 14.7%, F(1,22) = 3.81; female 15.9%, F(1,22) = 4.17; protanopia 0.2%, F(1,22) = 0.45; deuteranopia 2.6%, F(1,22) = 0.58], but none were significant (all P > 0.05).

Relationship of Color Preference and Color Naming: Pairwise Comparisons Between Female Trichromats and Dichromats or Male Trichromats.

Dunnett's C-corrected comparisons revealed that protanopes prefer colors named as brown or black significantly more than females do (P < 0.05); deuteranopes prefer colors named as black significantly more than females do (P < 0.05); females prefer colors named as pink significantly more than males and deuteranopes do (P < 0.05) but not more than protanopes (P > 0.05).

Error-Naming Rate.

As expected, the error-naming rate (Table S1) was significantly lower (P < 0.001) for focal colors than for the BCP stimuli both for protanopes (15.76 and 49.86%, t = −9.70) and deuteranopes (12.83 and 34.31%, t = −7.08).

SI Relationship of Color Preference and Color-Naming Response Time.

The relationship between color preference and color-naming response time is shown in Fig. S3.

Relationship of Color Preference and Color-Naming Categories.

An analysis also was conducted to establish whether, when dichromats made consistent errors in naming (e.g., a saturated purple stimulus identified twice as blue), dichromat preference was more similar to trichromat preference for the same stimulus (|Δ_s|, i.e., mean preference of all trichromat males for saturated purple) or for stimuli that share the incorrect color name (|Δ_n|, mean preference of all trichromat males for stimuli which were most frequently named blue). Paired t tests revealed that, when consistent errors are considered, the similarity between deuteranopes and trichromats was greater if the actual stimulus was the same than if the given color name was the same (mean |Δ_s| = 1.43, mean |Δ_n| = 1.71, t = −2.13 P < 0.05). The same tendency was present for protanopes (mean |Δ_s| = 1.83, mean |Δ_n| = 1.96), but the difference was not significant (t = −1.08, P > 0.05).

When deuteranopes made errors in color naming (e.g., a deuteranope called dark yellow “green”), their preference was significantly more similar to that of male trichromats’ preference for the same actual stimulus (dark yellow) than for other colors that male trichromats named the same as the dichromats (greens). This finding suggests that the categories of common observers do not predict how much dichromats like specific colors.

SI Chromatic Coordinates and Luminance for the Full Set of Colors.

Chromatic coordinates and luminance for the full set of colors are shown in Table S2.

Acknowledgments

We thank Jenny Bosten for helpful comments on an earlier version of this manuscript. This work was funded by Universidad Complutense de Madrid Research Grant BE48/09 (to L.Á.), European Research Council Starting Grant (project “CATEGORIES”) 283605 (to A.F.) and Ministerio de Economía y Competitividad Grant PSI2012-37778 (to J.L., H.M., and L.Á.).

Footnotes

↵¹To whom correspondence may be addressed. Email: lalvaro{at}ucm.es or anna.franklin{at}sussex.ac.uk.

Author contributions: L.Á., H.M., J.L., and A.F. designed research; L.Á. and J.L. performed research; L.Á. and H.M. analyzed data; and L.Á., H.M., J.L., and A.F. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1502104112/-/DCSupplemental.

View Abstract

References

↵
1. Gegenfurtner KR,
2. Sharpe LT
1. Sharpe LT,
2. Stockman A,
3. Jagle H,
4. Nathans J
(1999) Opsin genes, cone photopigments, color vision, and color blindness. Color Vision From Genes to Perception, eds Gegenfurtner KR, Sharpe LT (Cambridge Univ Press, Cambridge, UK), pp 3–51
.
↵
1. Neitz J,
2. Neitz M
(2011) The genetics of normal and defective color vision. Vision Res 51(7):633–651
.
OpenUrl CrossRef PubMed
↵
1. Wilmer JB
(2008) How to use individual differences to isolate functional organization, biology, and utility of visual functions; with illustrative proposals for stereopsis. Spat Vis 21(6):561–579
.
OpenUrl CrossRef PubMed
↵
1. Gingras B,
2. Honing H,
3. Peretz I,
4. Trainor LJ,
5. Fisher SE
(2015) Defining the biological bases of individual differences in musicality. Philos Trans R Soc Lond B Biol Sci 370(1664):20140092
.
OpenUrl Abstract/FREE Full Text
↵
1. Shevell SK
1. Smith VC,
2. Pokorny J
(2003) Color matching and color discrimination. The Science of Color, ed Shevell SK (Optical Society of America, Amsterdam), 2nd Ed, pp 103–148
.
↵
1. Smith VC,
2. Pokorny J
(1977) Large-field trichromacy in protanopes and deuteranopes. J Opt Soc Am 67(2):213–220
.
OpenUrl CrossRef PubMed
↵
1. Wachtler T,
2. Dohrmann U,
3. Hertel R
(2004) Modeling color percepts of dichromats. Vision Res 44(24):2843–2855
.
OpenUrl CrossRef PubMed
↵
1. Scheibner HM,
2. Boynton RM
(1968) Residual red-green discrimination in dichromats. J Opt Soc Am 58(8):1151–1158
.
OpenUrl CrossRef PubMed
↵
1. Nagy AL,
2. Boynton RM
(1979) Large-field color naming of dichromats with rods bleached. J Opt Soc Am 69(9):1259–1265
.
OpenUrl CrossRef PubMed
↵
1. Montag ED,
2. Boynton RM
(1987) Rod influence in dichromatic surface color perception. Vision Res 27(12):2153–2162
.
OpenUrl CrossRef PubMed
↵
1. Lillo J,
2. Moreira H,
3. Álvaro L,
4. Davies I
(2014) Use of basic color terms by red–green dichromats: 1. General description. Color Res Appl 39(4):360–371
.
OpenUrl CrossRef
↵
1. Forte JD,
2. Blessing EM,
3. Buzás P,
4. Martin PR
(2006) Contribution of chromatic aberrations to color signals in the primate visual system. J Vis 6(2):97–105
.
OpenUrl Abstract
↵
1. Cohen JD,
2. Matthen M
1. Broackes J
(2010) in What Do the Colour-Blind See? Color Ontology and Color Science, eds Cohen JD, Matthen M (MIT Press, London), pp 291–405
.
↵
1. Pridmore RW
(2014) Orthogonal relations and color constancy in dichromatic colorblindness. PLoS ONE 9(9):e107035
.
OpenUrl CrossRef PubMed
↵
1. Moreira H,
2. Lillo J,
3. Álvaro L,
4. Davies I
(2014) Use of basic color terms by red–green dichromats. II. Models. Color Res Appl 39(4):372–386
.
OpenUrl CrossRef
↵
1. Brettel H,
2. Viénot F,
3. Mollon JD
(1997) Computerized simulation of color appearance for dichromats. J Opt Soc Am A Opt Image Sci Vis 14(10):2647–2655
.
OpenUrl CrossRef PubMed
↵
1. Lillo J,
2. Álvaro L,
3. Moreira H
(2014) An experimental method for the assessment of color simulation tools. J Vis 14(8):15
.
OpenUrl Abstract/FREE Full Text
↵
1. Palmer SE,
2. Schloss KB
(2010) An ecological valence theory of human color preference. Proc Natl Acad Sci USA 107(19):8877–8882
.
OpenUrl Abstract/FREE Full Text
↵
1. Hurlbert AC,
2. Ling Y
(2007) Biological components of sex differences in color preference. Curr Biol 17(16):R623–R625
.
OpenUrl CrossRef PubMed
↵
1. Taylor C,
2. Franklin A
(2012) The relationship between color-object associations and color preference: Further investigation of ecological valence theory. Psychon Bull Rev 19(2):190–197
.
OpenUrl CrossRef PubMed
↵
1. Taylor C,
2. Clifford A,
3. Franklin A
(2013) Color preferences are not universal. J Exp Psychol Gen 142(4):1015–1027
.
OpenUrl CrossRef
↵
1. Ou L-C,
2. Luo MR,
3. Woodcock A,
4. Wright A
(2004) A study of colour emotion and colour preference. Part III: Colour preference modeling. Color Res Appl 29(5):381–389
.
OpenUrl CrossRef
↵
1. Gegenfurtner KR,
2. Sharpe LT
1. Eskew RT,
2. McLellan JS,
3. Giulianini F
(1999) Chromatic detection and discrimination. Color Vision: From Genes to Perception, eds Gegenfurtner KR, Sharpe LT (Cambridge Univ Press, Cambridge, UK), pp 345–368
.
↵
1. De Valois RL,
2. De Valois KK
(1993) A multi-stage color model. Vision Res 33(8):1053–1065
.
OpenUrl CrossRef PubMed
↵
1. Viénot F,
2. Brettel H,
3. Ott L,
4. Ben M’Barek A,
5. Mollon JD
(1995) What do colour-blind people see? Nature 376(6536):127–128
.
OpenUrl CrossRef PubMed
↵
1. Schmidt BP,
2. Neitz M,
3. Neitz J
(2014) Neurobiological hypothesis of color appearance and hue perception. J Opt Soc Am A Opt Image Sci Vis 31(4):A195–A207
.
OpenUrl CrossRef PubMed
↵
1. Palmer SE,
2. Schloss KB,
3. Sammartino J
(2013) Visual aesthetics and human preference. Annu Rev Psychol 64:77–107
.
OpenUrl CrossRef PubMed
↵
1. Reber R,
2. Schwarz N,
3. Winkielman P
(2004) Processing fluency and aesthetic pleasure: Is beauty in the perceiver’s processing experience? Pers Soc Psychol Rev 8(4):364–382
.
OpenUrl Abstract/FREE Full Text
↵
1. Shimamura AP,
2. Palmer SE
1. Reber R
(2012) Processing fluency, aesthetic pleasure, and culturally shared taste. Aesthetic Science: Connecting Mind, Brain, and Experience, eds Shimamura AP, Palmer SE (Oxford Univ Press, New York), pp 223–249
.
↵
1. Smith JD,
2. Melara RJ
(1990) Aesthetic preference and syntactic prototypicality in music: ’Tis the gift to be simple. Cognition 34(3):279–298
.
OpenUrl CrossRef PubMed
↵
1. Langlois JH,
2. Roggman LA
(1990) Attractive faces are only average. Psychol Sci 1(2):115–121
.
OpenUrl Abstract/FREE Full Text
↵
1. Babel M,
2. McGuire G
(2015) Perceptual fluency and judgments of vocal aesthetics and stereotypicality. Cogn Sci 39(4):766–787
.
OpenUrl CrossRef PubMed
↵
1. Halberstadt J,
2. Hooton K
(2008) The affect disruption hypothesis: The effect of analytic thought on the fluency and appeal of art. Cogn Emotion 22(5):964–976
.
OpenUrl CrossRef
↵
1. Martindale C,
2. Moore K
(1988) Priming, prototypicality, and preference. J Exp Psychol Hum Percept Perform 14(4):661–670
.
OpenUrl CrossRef
↵
1. Hekkert P,
2. Snelders D,
3. van Wieringen PC
(2003) ‘Most advanced, yet acceptable’: Typicality and novelty as joint predictors of aesthetic preference in industrial design. Br J Psychol 94(Pt 1):111–124
.
OpenUrl CrossRef PubMed
↵
1. Ishihara MD
(1996) Ishihara’s Tests for Colour Deficiency (Kanehara Trading, Tokyo)
.
↵
1. Fletcher R
(1980) The City University Colour Vision Test (Keeler, London), 2nd Ed
.
↵
1. Lanthony P
(1985) Album Tritan (Luneau Ophtalmologie, Paris)
.
↵
1. Smith VC,
2. Pokorny J
(1975) Spectral sensitivity of the foveal cone photopigments between 400 and 500 nm. Vision Res 15(2):161–171
.
OpenUrl CrossRef PubMed

Article Alerts

Email Article

Citation Tools

Request Permissions

Proceedings of the National Academy of Sciences: 112 (30)

Table of Contents

Submit

Article Classifications

Social Sciences
Psychological and Cognitive Sciences

[1] ↵
Gegenfurtner KR,
Sharpe LT
Sharpe LT,
Stockman A,
Jagle H,
Nathans J
(1999) Opsin genes, cone photopigments, color vision, and color blindness. Color Vision From Genes to Perception, eds Gegenfurtner KR, Sharpe LT (Cambridge Univ Press, Cambridge, UK), pp 3–51
.

[2] Gegenfurtner KR,

[3] Sharpe LT

[4] Sharpe LT,

[5] Stockman A,

[6] Jagle H,

[7] Nathans J

[8] ↵
Neitz J,
Neitz M
(2011) The genetics of normal and defective color vision. Vision Res 51(7):633–651
.
OpenUrl CrossRef PubMed

[9] Neitz J,

[10] Neitz M

[11] ↵
Wilmer JB
(2008) How to use individual differences to isolate functional organization, biology, and utility of visual functions; with illustrative proposals for stereopsis. Spat Vis 21(6):561–579
.
OpenUrl CrossRef PubMed

[12] Wilmer JB

[13] ↵
Gingras B,
Honing H,
Peretz I,
Trainor LJ,
Fisher SE
(2015) Defining the biological bases of individual differences in musicality. Philos Trans R Soc Lond B Biol Sci 370(1664):20140092
.
OpenUrl Abstract/FREE Full Text

[14] Gingras B,

[15] Honing H,

[16] Peretz I,

[17] Trainor LJ,

[18] Fisher SE

[19] ↵
Shevell SK
Smith VC,
Pokorny J
(2003) Color matching and color discrimination. The Science of Color, ed Shevell SK (Optical Society of America, Amsterdam), 2nd Ed, pp 103–148
.

[20] Shevell SK

[21] Smith VC,

[22] Pokorny J

[23] ↵
Smith VC,
Pokorny J
(1977) Large-field trichromacy in protanopes and deuteranopes. J Opt Soc Am 67(2):213–220
.
OpenUrl CrossRef PubMed

[24] Smith VC,

[25] Pokorny J

[26] ↵
Wachtler T,
Dohrmann U,
Hertel R
(2004) Modeling color percepts of dichromats. Vision Res 44(24):2843–2855
.
OpenUrl CrossRef PubMed

[27] Wachtler T,

[28] Dohrmann U,

[29] Hertel R

[30] ↵
Scheibner HM,
Boynton RM
(1968) Residual red-green discrimination in dichromats. J Opt Soc Am 58(8):1151–1158
.
OpenUrl CrossRef PubMed

[31] Scheibner HM,

[32] Boynton RM

[33] ↵
Nagy AL,
Boynton RM
(1979) Large-field color naming of dichromats with rods bleached. J Opt Soc Am 69(9):1259–1265
.
OpenUrl CrossRef PubMed

[34] Nagy AL,

[35] Boynton RM

[36] ↵
Montag ED,
Boynton RM
(1987) Rod influence in dichromatic surface color perception. Vision Res 27(12):2153–2162
.
OpenUrl CrossRef PubMed

[37] Montag ED,

[38] Boynton RM

[39] ↵
Lillo J,
Moreira H,
Álvaro L,
Davies I
(2014) Use of basic color terms by red–green dichromats: 1. General description. Color Res Appl 39(4):360–371
.
OpenUrl CrossRef

[40] Lillo J,

[41] Moreira H,

[42] Álvaro L,

[43] Davies I

[44] ↵
Forte JD,
Blessing EM,
Buzás P,
Martin PR
(2006) Contribution of chromatic aberrations to color signals in the primate visual system. J Vis 6(2):97–105
.
OpenUrl Abstract

[45] Forte JD,

[46] Blessing EM,

[47] Buzás P,

[48] Martin PR

[49] ↵
Cohen JD,
Matthen M
Broackes J
(2010) in What Do the Colour-Blind See? Color Ontology and Color Science, eds Cohen JD, Matthen M (MIT Press, London), pp 291–405
.

[50] Cohen JD,

[51] Matthen M

[52] Broackes J

[53] ↵
Pridmore RW
(2014) Orthogonal relations and color constancy in dichromatic colorblindness. PLoS ONE 9(9):e107035
.
OpenUrl CrossRef PubMed

[54] Pridmore RW

[55] ↵
Moreira H,
Lillo J,
Álvaro L,
Davies I
(2014) Use of basic color terms by red–green dichromats. II. Models. Color Res Appl 39(4):372–386
.
OpenUrl CrossRef

[56] Moreira H,

[57] Lillo J,

[58] Álvaro L,

[59] Davies I

[60] ↵
Brettel H,
Viénot F,
Mollon JD
(1997) Computerized simulation of color appearance for dichromats. J Opt Soc Am A Opt Image Sci Vis 14(10):2647–2655
.
OpenUrl CrossRef PubMed

[61] Brettel H,

[62] Viénot F,

[63] Mollon JD

[64] ↵
Lillo J,
Álvaro L,
Moreira H
(2014) An experimental method for the assessment of color simulation tools. J Vis 14(8):15
.
OpenUrl Abstract/FREE Full Text

[65] Lillo J,

[66] Álvaro L,

[67] Moreira H

[68] ↵
Palmer SE,
Schloss KB
(2010) An ecological valence theory of human color preference. Proc Natl Acad Sci USA 107(19):8877–8882
.
OpenUrl Abstract/FREE Full Text

[69] Palmer SE,

[70] Schloss KB

[71] ↵
Hurlbert AC,
Ling Y
(2007) Biological components of sex differences in color preference. Curr Biol 17(16):R623–R625
.
OpenUrl CrossRef PubMed

[72] Hurlbert AC,

[73] Ling Y

[74] ↵
Taylor C,
Franklin A
(2012) The relationship between color-object associations and color preference: Further investigation of ecological valence theory. Psychon Bull Rev 19(2):190–197
.
OpenUrl CrossRef PubMed

[75] Taylor C,

[76] Franklin A

[77] ↵
Taylor C,
Clifford A,
Franklin A
(2013) Color preferences are not universal. J Exp Psychol Gen 142(4):1015–1027
.
OpenUrl CrossRef

[78] Taylor C,

[79] Clifford A,

[80] Franklin A

[81] ↵
Ou L-C,
Luo MR,
Woodcock A,
Wright A
(2004) A study of colour emotion and colour preference. Part III: Colour preference modeling. Color Res Appl 29(5):381–389
.
OpenUrl CrossRef

[82] Ou L-C,

[83] Luo MR,

[84] Woodcock A,

[85] Wright A

[86] ↵
Gegenfurtner KR,
Sharpe LT
Eskew RT,
McLellan JS,
Giulianini F
(1999) Chromatic detection and discrimination. Color Vision: From Genes to Perception, eds Gegenfurtner KR, Sharpe LT (Cambridge Univ Press, Cambridge, UK), pp 345–368
.

[87] Gegenfurtner KR,

[88] Sharpe LT

[89] Eskew RT,

[90] McLellan JS,

[91] Giulianini F

[92] ↵
De Valois RL,
De Valois KK
(1993) A multi-stage color model. Vision Res 33(8):1053–1065
.
OpenUrl CrossRef PubMed

[93] De Valois RL,

[94] De Valois KK

[95] ↵
Viénot F,
Brettel H,
Ott L,
Ben M’Barek A,
Mollon JD
(1995) What do colour-blind people see? Nature 376(6536):127–128
.
OpenUrl CrossRef PubMed

[96] Viénot F,

[97] Brettel H,

[98] Ott L,

[99] Ben M’Barek A,

[100] Mollon JD

[101] ↵
Schmidt BP,
Neitz M,
Neitz J
(2014) Neurobiological hypothesis of color appearance and hue perception. J Opt Soc Am A Opt Image Sci Vis 31(4):A195–A207
.
OpenUrl CrossRef PubMed

[102] Schmidt BP,

[103] Neitz M,

[104] Neitz J

[105] ↵
Palmer SE,
Schloss KB,
Sammartino J
(2013) Visual aesthetics and human preference. Annu Rev Psychol 64:77–107
.
OpenUrl CrossRef PubMed

[106] Palmer SE,

[107] Schloss KB,

[108] Sammartino J

[109] ↵
Reber R,
Schwarz N,
Winkielman P
(2004) Processing fluency and aesthetic pleasure: Is beauty in the perceiver’s processing experience? Pers Soc Psychol Rev 8(4):364–382
.
OpenUrl Abstract/FREE Full Text

[110] Reber R,

[111] Schwarz N,

[112] Winkielman P

[113] ↵
Shimamura AP,
Palmer SE
Reber R
(2012) Processing fluency, aesthetic pleasure, and culturally shared taste. Aesthetic Science: Connecting Mind, Brain, and Experience, eds Shimamura AP, Palmer SE (Oxford Univ Press, New York), pp 223–249
.

[114] Shimamura AP,

[115] Palmer SE

[116] Reber R

[117] ↵
Smith JD,
Melara RJ
(1990) Aesthetic preference and syntactic prototypicality in music: ’Tis the gift to be simple. Cognition 34(3):279–298
.
OpenUrl CrossRef PubMed

[118] Smith JD,

[119] Melara RJ

[120] ↵
Langlois JH,
Roggman LA
(1990) Attractive faces are only average. Psychol Sci 1(2):115–121
.
OpenUrl Abstract/FREE Full Text

[121] Langlois JH,

[122] Roggman LA

[123] ↵
Babel M,
McGuire G
(2015) Perceptual fluency and judgments of vocal aesthetics and stereotypicality. Cogn Sci 39(4):766–787
.
OpenUrl CrossRef PubMed

[124] Babel M,

[125] McGuire G

[126] ↵
Halberstadt J,
Hooton K
(2008) The affect disruption hypothesis: The effect of analytic thought on the fluency and appeal of art. Cogn Emotion 22(5):964–976
.
OpenUrl CrossRef

[127] Halberstadt J,

[128] Hooton K

[129] ↵
Martindale C,
Moore K
(1988) Priming, prototypicality, and preference. J Exp Psychol Hum Percept Perform 14(4):661–670
.
OpenUrl CrossRef

[130] Martindale C,

[131] Moore K

[132] ↵
Hekkert P,
Snelders D,
van Wieringen PC
(2003) ‘Most advanced, yet acceptable’: Typicality and novelty as joint predictors of aesthetic preference in industrial design. Br J Psychol 94(Pt 1):111–124
.
OpenUrl CrossRef PubMed

[133] Hekkert P,

[134] Snelders D,

[135] van Wieringen PC

[136] ↵
Ishihara MD
(1996) Ishihara’s Tests for Colour Deficiency (Kanehara Trading, Tokyo)
.

[137] Ishihara MD

[138] ↵
Fletcher R
(1980) The City University Colour Vision Test (Keeler, London), 2nd Ed
.

[139] Fletcher R

[140] ↵
Lanthony P
(1985) Album Tritan (Luneau Ophtalmologie, Paris)
.

[141] Lanthony P

[142] ↵
Smith VC,
Pokorny J
(1975) Spectral sensitivity of the foveal cone photopigments between 400 and 500 nm. Vision Res 15(2):161–171
.
OpenUrl CrossRef PubMed

[143] Smith VC,

[144] Pokorny J

Main menu

User menu

Search

New Research In

Physical Sciences

Featured Portals

Articles by Topic

Social Sciences

Featured Portals

Articles by Topic

Biological Sciences

Featured Portals

Articles by Topic

Color preference in red–green dichromats

Significance

Abstract

Results

Reliability.

Normal Trichromat and Dichromat Color-Preference Curves.

Predictors of Color Preference: Cone Contrast, Lightness, and Saturation.

Relationship of Color Preference and Color Naming.

Discussion

Methods

Participants.

Stimuli and Experimental Set-Up.

Design and Procedure.

SI Text

SI Stimuli

Reliability.

Similarity in Group Color-Preference Curves.

Female Trichromat and Dichromat Color Preference for BCP Stimuli: Pairwise Comparisons.

Female Trichromat Color Preference for Focal Colors Versus Dichromat or Male Color Preference for Focal Colors: Pairwise Comparisons.

Colorimetric Computations: Cone Contrast.

Colorimetric Computations: Multistage Model.

Regression on Color Preference: Multistage Model.

Colorimetric Computations: L*T.

Colorimetric Computations: s′.

Regression on Color Preference: Lightness and Saturation as Predictors.

Relationship of Color Preference and Color Naming: Pairwise Comparisons Between Female Trichromats and Dichromats or Male Trichromats.

Error-Naming Rate.

SI Relationship of Color Preference and Color-Naming Response Time.

Relationship of Color Preference and Color-Naming Categories.

SI Chromatic Coordinates and Luminance for the Full Set of Colors.

Acknowledgments

Footnotes

References

Citation Manager Formats

Sign up for Article Alerts

Article Classifications

Jump to section

You May Also be Interested in

Similar Articles

Articles

PNAS Portals

Information

Colorimetric Computations: L*_T.