Search

It is well known that the colour preferences of ganglion and LGN cells do not match the four perceptually simple colours red, green blue and yellow. It is also known that although colour perception is distorted in the peripheral visual field, there are four hues that appear stable with eccentricity. These are defined as peripherally invariant hues. Both of these observations must in some way reflect the physiological substrate of neurons at different stages of the primary visual pathway. The experiments described here are aimed at understanding the link between the physiology and the perception of colour by studying the characteristics of peripheral colour visionThe following questions have been addressed; i) to what extent does colour matching rely on the retinal physiological substrate? ii) what is the reason for the discrepancy between invariant and unique green and how is cone contrast linked to this paradox? iii) how are the `special' hues (invariant and unique) related to human evolution? iv) how does peripheral colour vision vary between males and females?An asymmetric colour matching paradigm and a colour naming task have been employed. In the colour matching task, 24 chromatic axes of variable purity are used. Observers match the chromaticity of a 3 degree peripheral spot with that of a 1 degree parafoveal spot. The results are expressed in terms of hue rotation, saturation match and cone contrast. In the colour naming experiment the observers name 40 chromatic axes as either red, blue, green or yellow and colour naming functions are derived. The central maxima of these functions are defined as the unique hues. The results suggest that colour matching and cone opponency reflect the characteristics of the retinal neural network as they exhibit nasal-temporal asymmetries, similar to known physiological asymmetries. Although three of the peripherally invariant hues match the unique counterparts, invariant and unique green are markedly different for all observers. In an important control experiment unique hues are shown to be stable with eccentricity and purity. This confirms that these attributes are not confounding factors for the observed discrepancy between invariant and unique green. Unlike for the other 'special' hues the RMS cone contrast of invariant green differs markedly between parafoveal and peripheral targets. It is likely that the cone contrast remains unchanged only if the stimuli excite the same number of cones. Two invariant and two unique hues (blue and yellow) fall on the daylight locus suggesting that discrimination in these regions of the colour space is strongly influenced by terrestrial illumination. Moreover, the inter-individual variability is found to be minimised around the daylight locus showing that the blue-yellow system is more stable across colour normal populations than the red-green system. A statistically significant difference is demonstrated between the peripheral colour vision of males and females. This may be attributed to the M-cone polymorphism which in addition to X-chromosome inactivation, results in more than three cone types in the female retina.

Color scientists have long tried to use geometry to represent human color perception. While color space is not generally Euclidean, Ennis and Zaidi (2019) suggested that perceptual color space may be affine under a Hering color-opponent mental representation. They demonstrated this using Varignon's theorem which states that when the midpoints of adjacent sides of a quadrilateral are connected, the figure formed is always a parallelogram if the intrinsic geometry of the space is at least affine. An important corollary of Varignon's theorem is that if opposite midpoints are bisected, the space can be described using affine geometry if the midpoint-bisections coincide (Todd et al., 2001).In Experiment I, twelve subjects viewed two equiluminant pair of colors and adjusted a third color in hue and saturation to bisect the perceived interval between each pair, following the Varignon's theorem paradigm. We tested subjects on a small predetermined quadrilateral and a large individualized quadrilateral with vertices that were subjects' unique hues. Subjects performed the tasks under two sets of instructions: bisections on perceived differences in hue and saturation (termed color-difference instructions) and bisections based on differences in redness-greenness and blueness-yellowness (termed Hering instructions). Statistical analysis using a 2-way ANOVA indicates that there was no significant difference in the resulting midpoint-bisection differences due to instruction type (p = 0.26, F= 1.3057). There was a statistical difference between quadrilateral size (p < 0.001, F=75.1337), but no statistical interaction between the instruction type and quadrilateral size (p = 0.5687, F= 0.3303).Experiment II further tested Ennis and Zaidi's claims, by testing one large quadrilateral enclosing white (as in their study) and two large quadrilaterals shifted in average redness/greenness. We predicted that if subjects were truly bisecting colors in a metrical Hering color-opponent space, the chosen midpoint-bisections should track with the shifted quadrilaterals. Four subjects received extensive practice and a tutorial on the geometric representation of colors in Hering color-opponent space. Subjects' midpoint-bisections were more nearly coincident than in Experiment I, but there was no shift in the red/green locations of the midpoint-bisections across the quadrilaterals (ChiSq=2.92,p=0.232). Our results fail to replicate those of Ennis and Zaidi's (2019) and add to a growing body of literature that challenge the privilege status of Hering's theory.

The Whorfian linguistic relativity hypothesis is a highly contested model linking cognition and perception, in which substantial cultural-linguistic differences between languages are proposed, which significantly constrain in individual’s perception and world view. Due to the wide variation in color naming in cultures around the world, the constraints on naming imposed by color physics and the physiology of the visual system, and the near-ubiquitous nature of color communication, color naming has often been used as a model for testing linguistic relativity effects. However, by themselves, variations in color naming cannot give any useful information about whether color perception is affected by linguistic relativity. For this reason, we used color naming in conjunction with unique hue selection and multidimensional scaling (MDS) to test for the presence of Whorfian effects related to color naming for speakers of English and Somali, a language previously shown by our lab to show great inter-individual variation in both color naming and non-lexical measures of color perception.In Experiment I we tested English subjects using a non-metric MDS paradigm with heteroluminant stimuli and found it to generally replicate fiducial orderings of stimuli in CIE UV space. Experiment II added a unique hue selection task, and English speakers’ unique hue selections in this task were concordant with those obtained from previous studies. We introduced a new method of MDS data collection, the binary sort protocol, in Experiment III, which allowed us to quickly gather MDS data from English and Somali-speaking subjects. Somali color naming showed similar patterns to previous experiments by our lab, but we were unable to gather data from a sufficient variety of Somali informants to robustly test for Whorfian effects. Somali speakers’ MDS maps conformed more poorly to CIE UV space than English speakers’ maps, though analysis of stress indicated that Somali subjects may use two dimensions of unknown character in their judgments of color difference; additionally, numerous Somali speakers’ MDS maps and/or color naming showed evidence of purple-yellow affiliation, which is contrary to all previous color categorization and perception experiments. Experiment IV focused on English color naming, utilizing the gap statistic and the novel cluster stability analysis to find 15 shared chromatic categories of varying consensus and 4 distinct color-naming motifs among English informants, congruent with previous results from our lab. As in Experiment II and Experiment III, no evidence of linguistic relativity effects was found in our results, though we suggest that our color naming, unique hue, and MDS methodologies may have been insufficiently sensitive to detect these effects if present. We conclude that no definite linguistic relativity effects exist between English and Somali color naming, color difference perception, and unique hue selection, at least as measured by our paradigms.