To characterize affective responses to the ends of the visible wavelength spectrum, and their consistency across individuals, we began by simply examining where red and blue are situated in an individual’s two-dimensional affect space. Participants rated valence from −3 to +3 (extremely negative to positive) and arousal from 1 to 7 (not at all to extremely activated), to represent pleasure and activation, respectively. In valence-arousal space, individuals’ ratings of blue and red occupied were largely linearly separable in this affect space, with 87.5% correct color classification based on arousal and valence. In general, blue was more pleasant and less arousing than red (Fig. 1A). Confirming that blue has a calming and more pleasant value, arousal responses to red were significantly higher than blue, t(19) = 5.51, p < 0.001, Cohen’s d = 1.23; by contrast, blue was more pleasant than red, t(19) = 2.59, p = 0.018, Cohen’s d = 0.58. Despite large individual differences, these results are consistent with blue and red occupying distinct locations in affective space, with blue evoking, on average, more pleasure than red.
Colors have idiosyncratic influences on the individual and we found substantial individual differences in affective responses to color. If the underlying basis for these judgments is affective and related to the rewarding nature of color experience, then these individual differences may be explained, in part, by approach motivation and sensitivity toward reward29. We repeated the affect assessment in a separate larger set of participants (N = 31), which allowed us to assess individual differences related to two motivational systems examining approach and avoidance behavior and affect; we employed the sensitivity to punishment and sensitivity to reward questionnaire (SPSRQ)29. In this group, we replicated that red was significantly more arousing than blue, t(30) = 8.32, p < 0.001, Cohen’s d = 1.50, while blue was significantly more pleasant than red, t(30) = 2.55, p = 0.016, Cohen’s d = 0.60. Further, we found that reward sensitivity was related to differential valence responses between blue and red (robust fit, R2 = 0.147, p = 0.033). Those with the highest reward sensitivity had up to 3 times greater positive affective responses to blue over red (Fig. 1B). We confirmed the independence of reward and punishment sensitivity, robust fit, R2 = 0.026, p = 0.384. No relationship was found between punishment sensitivity and differential color valence (R2 = 0.072, p = 0.143) or red valence specifically (robust fit, R2 = 0.019, p = 0.457), suggesting a unique role of reward sensitivity in affective responses to color.
As blue and red occupy largely distinct portions of two-dimensional affect space, they appear to evoke qualitatively distinct forms of affective experience. Collapsing across data sets from Experiments 1 and 2, we found color dependent relationships between valence and arousal, robust fit, R2 = 0.168, p = 0.003 (Fig. 1C). Increasing pleasure to blue was associated with decreased arousal; conversely, increasing pleasure to red was associated with increased arousal (also see Fig. 1A). Even when blue and red similarly evoke pleasurable valence, these responses appear qualitatively distinct, representing increasing calm versus stimulation.
If colors are genuinely more or less rewarding, then they should be differentially susceptible to positive reinforcement, demonstrating differential incentive salience52. Nineteen participants took part in an experiment where money could be earned by either red or blue circles. Participants indicated whether each circle was blue or red as quickly and accurately as possible, meeting a response time (RT) threshold of 650 ms to win $0.30 on a correct trial (Fig. 2A).
For RT analysis, error and outlier trials were excluded (<3%). First, we show that red and blue RGB values we selected reliably generated different affective responses above, were also highly reliably categorized as red and blue (>97%), even under time pressure response conditions. Across both colors, reward resulted in reinforcement of color discriminations, i.e., a facilitated behavioral response as evidenced by faster response times (RT), F(1,18) = 338.83, p < 0.001, η2 = 0.95, and greater accuracy, F(1,18) = 14.76, p = 0.001, η2 = 0.45, when colors were rewarded (Reward) compared to unrewarded (No Reward). Critically, there was also a robust interaction between reward and color for RT, F(1,17) = 342.99, p < 0.001, η2 = 0.95, and accuracy, F(1,17) = 15.04, p = 0.001, η2 = 0.47 (Fig. 2B). As shown in Fig. 2B (left panel), the RT difference (indexed by No Reward – Reward) for red was smaller than the corresponding RT difference for blue. While we did not find red or blue resulted in differential fast response times, when paired with monetary reward, blue resulted in faster performance compared to red, consistent with red and blue having differential capacity for reward. Aligning with subjective self-report, colors had objective consequences for behavior, with blue acquiring a stronger association with reward than red, enhancing speed of color discrimination.
Beyond behavioral reinforcement, the incentive salience of reward is associated with enhanced attentional salience53. If individuals perceive colors as differing in their rewarding properties, then colors should have an unequal priority for attention resulting in differing salience for perception. To investigate the differential attentional salience of blue and red, we examined their competition for attention, employing a temporal order judgment (TOJ) task. Red and blue circles appeared first or second across a range of stimulus onset asynchronies (SOA), ranging from 8 to 98 ms, to the left or right of fixation (Fig. 2C). Participants indicated, as accurately as possible, which side the color circle appeared first. The task had self-paced responses to emphasize perceptual salience and there was no feedback. The mean percent of correct responses were analyzed for each SOA and by which color was presented first.
Accuracy increased with increasing temporal delay between events, i.e., SOA, F(3,47) = 75.16, p < 0.0001, η2 = 0.83. There was a significant effect of color, F (1,16) = 4.67, p = 0.046, η2 = 0.23, particularly at the shortest SOA, where temporal uncertainty was greatest, t(16) = 2.95, p = 0.009, Cohen’s d = 0.72 (Fig. 2D). Compared to red, when blue was presented first it was much more likely to win the competition, consistent with a blue attentional priority. To further quantify the increased attentional salience, we estimated the point of subjective simultaneity (PSS): the temporal delay estimated between red and blue where they would be judged to be simultaneous. On average blue was perceived as arriving 6.1 ms prior to red stimuli, t(16) = 2.26, p = 0.038, Cohen’s d = 0.55 (Fig. 2E). Although this is small in magnitude of time, the effect size suggests a moderate effect size of color on attentional salience54. This blue-red temporal perceptual asymmetry, however, is not necessarily diagnostic of differential reward. Given that we found differences in self-reported affective responses to color were related to trait reward sensitivity, we further examined how red-blue temporal asymmetry similarly followed individual differences in reward sensitivity. Indeed, we found the magnitude of blue-shift in temporal onset increased with an individual’s reward sensitivity, robust fit, R2 = 0.334, p = 0.031 (Fig. 2F).
If blue and red represent differential for reward, then by manipulating reward we should be able to further increase their incentive salience and the temporal asymmetry it supports. In a separate study, we rewarded color discrimination through monetary reinforcement, whereby speeded color circle discriminations earned money ($0.30) on each trial. These monetary reinforcement trials were intermixed with TOJ trials, which were unrewarded. This source of extrinsic reward further magnified the color temporal asymmetry, with blue having an even greater accuracy advantage over red, F(2, 38) = 4.46, p = 0.001, η2 = 0.23, and further extended to longer SOAs between colors, SOA < 40 ms, t(18) = 2.65, p = 0.016, Cohen’s d = 0.85, (Fig. 2D). Estimating the point of subjective simultaneity (PSS), revealed that reward doubled the existing temporal priority, with blue judged on average as appearing 12.9 ms prior to red stimuli, t(18) = 2.35, p = 0.031, Cohen’s d = 0.54, (Fig. 2E), an effect size consistent with findings in how reward influences temporal judgments55.
If the visual experience of colors differ biologically in the capacity for reward and engagement of the behavioral activation system, then colors should be associated with altered activity within visual and reward circuits, as well as their interaction. We conducted an fMRI experiment to examine the neural bases of differential color reward as well as individual differences in their expression. We examined a priori (1) color sensitive visual regions and (2) affective-motivational circuitry demonstrating reward sensitivity and (3) the functional coupling between the two, in the context of reward learning. Within visual regions, we examined differential color activity in the primary visual area V1 and the higher-order cortical color region V447. Within reward circuitry, we examined differential color activity in the ventral striatum (VS), amygdala, and orbitofrontal cortex (OFC). While the VS supports reward learning and its motivational states56, the amygdala supports value-based stimulus salience32, the OFC supports visual reward associations, the representation of valence and the subjective experience of pleasure57. Regions of interest were predefined, independently of the data under consideration, for purposes of unbiased hypothesis testing.
Color was incidental to the task: with no requirement to attend to or program behavioral responses related to color. Participants indicated whether a gap in colored circles faced left or right (i.e., a C or its reverse) as quickly and as accurately as possible, meeting a response time threshold to earn (gain) or save (avoid losing) $0.30. Gains and averted losses similarly engage the mesolimbic reward system58, and afforded a broader and more generalizable manipulation of reward. Both reward types yielded above 93% correct responses and thus positive outcomes (earning, 93%; saving money, 94%), consistent with a manipulation reward rather than punishment. As such, we collapsed across reward type (earn vs save) and restricted our analysis to the comparison of red and blue. Even though color was task-irrelevant, there was a trend toward faster responses to blue than red trials, t(30) = 1.91, p = 0.066, Cohen’s d = 0.34, with no significant difference in accuracy, t(30) = 0.98, p = 0.333, Cohen’s d = 0.18 or average earnings between red and blue, t(30) = 0.77, p = 0.446, Cohen’s d = 0.14.
We next examined neural responses. There were prominent individual differences in color response to blue vs. red within each region (Fig. 3A). These variable color responses were not noise, however. When we assessed color-dependent functional connectivity between brain regions, this revealed individual differences in color tuning responses were associated with altered coupling within reward circuitry components. The VS (robust R2 = 0.170, p = 0.022), amygdala (robust R2 = 0.199, p = 0.015), and OFC (robust R2 = 0.255, p = 0.005), were coupled with area V4 (Fig. 3B), demonstrating reward coupling with color cortex.
While there appeared to be an equal propensity toward blue and red tuning across individuals, a major source of differential color response was individual differences in reward sensitivity. Those higher in reward sensitivity demonstrated differential blue tuning (i.e., more activated for blue than red) and those lower in reward sensitivity tended to show differential red tuning (i.e., less activated for blue than red). Within visual regions, reward sensitivity was related to color region V4, contributing to a higher blue-shift in tuning associated with higher reward sensitivity, robust R2 = 0.211, p = 0.016 (Fig. 3C). Within reward regions, greater reward sensitivity was associated with greater blue-shift in each the OFC (robust R2 = 0.178, p = 0.027), amygdala (robust R2 = 0.206, p = 0.010), and at trend level, in the VS (R2 = 0.133, p = 0.062) (Fig. 3B,D). There was no evidence of a relationship between color tuning and punishment sensitivity, all R2 < 003, smallest p = 0.770).
To examine the central role of V4 cortex in color reward, we conducted two mediation analyses (for a similar approach, see46,59,60,61). We chose the VS as the dependent variable, given its role in reward learning and motivation33. We found that V4 mediated color-specific coupling between the OFC and the VS (Fig. 4). A bias-corrected 95% bootstrap confidence interval (CI) for the indirect effect (a1*b1 = 0.669) based on 10,000 samples was entirely above zero (0.1753~1.7033), such that the total color effect between OFC and the VS was dependent on V4. To address the psychological and motivational dimension of color based affect, we tested an additional mediation model with reward sensitivity. We found V4 mediated the association between VS response and reward sensitivity. A bias-corrected 95% bootstrap CI for the indirect effect (a2*b2 = 0.209) was entirely above zero (0.0124 ~ 0.5037). These results suggest that color-based affect is related to the neural bases of color perception. In particular, V4 is central to the expression of color based reward, both in terms of orchestrating interactions within reward circuits and individual differences in reward experience. In sum, these findings support that affective, motivational and perceptual representations underlie individual differences in the experience of color affect.