Color awareness has long been a puzzle for researchers in neuroscience and psychology, who debate over how much color observers really perceive. A study from Dartmouth in collaboration with Amherst College finds that people are aware of surprisingly limited color in their peripheral vision; much of our sense of a colorful visual world is likely constructed by our brain.
The findings are published in the Proceedings of the National Academy of Sciences .
To test people’s visual awareness of color during naturalistic viewing, the researchers used head-mounted virtual reality displays installed with eye-trackers to immerse participants in a 360-degree real-world environment.
The virtual environments included tours of historic sites, a street dance performance, a symphony rehearsal and more, where observers could explore their surroundings simply by turning their heads.
With the eye-tracking tool, researchers knew exactly where an observer was looking at all times in the scene and could make systematic changes to the visual environment so that only the areas where the person was looking were in color.
The rest of the scene in the periphery was desaturated so that it had no color and was just in black and white.
After a series of trials, observers were asked a series of questions to gauge if they noticed the lack of color in their periphery. A supplemental video from the study illustrates how the peripheral color was removed from various scenes.
In your visual field, your periphery extends approximately 210 degrees, which is similar to if your arms are stretched out on your left and right. The study’s results showed that most people’s color awareness is limited to a small area around the dead center of their visual field.
When the researchers removed most color in the periphery, most people did not notice. In the most extreme case, almost a third of observers did not notice when less than five percent of the entire visual field was presented in color (radius of 10 degrees visual angle).
Participants were astonished to find out later that they hadn’t noticed the desaturated periphery, after they were shown the changes that were made to a virtual scene that they had just explored.
A second study tasked the participants to identify when color was desaturated in the periphery. The results were similar in that most people failed to notice when the peripheral color had been removed.
A large number of people participated in the two studies, which featured nearly 180 participants in total.
“We were amazed by how oblivious participants were when color was removed from up to 95 percent of their visual world,” said senior author, Caroline Robertson, an assistant professor of psychological and brain sciences at Dartmouth.
“Our results show that our intuitive sense of a rich, colorful visual world is largely incorrect. Our brain is likely filling-in much of our perceptual experience.”
Previous studies evaluating the limitations of visual awareness often relied on participants staring at video content on computer screens directly in front of them.
By leveraging the virtual reality experience, this research approach is novel, as the 360-degree environment is more similar to the way people experience the real-world.
There is a widespread misconception even among vision scientists, and hence the population in general, that the high cone density in the fovea implies that color vision is restricted to the central vision, and conversely that the high density of rods in the periphery implies a lack of color vision in the periphery. For example, the Wikipedia article on peripheral vision says
Peripheral vision is weak in humans, especially at distinguishing colour and shape … rod cells are unable to distinguish colour and are predominant at the periphery, while cone cells are concentrated mostly in the centre of the retina, the fovea. (https://en.wikipedia.org/wiki/Peripheral_vision)
In fact, however, both historical (Østerberg, 1935) and more recent (Curcio, Sloan, Packer, Hendrickson, Kalina, 1987; Song, Chui, Zhong, Elsner, & Burns, 2011) measurements of photoreceptor densities indicate otherwise.
Despite the high concentration of cones in the fovea, even the central 5° of the retina contains only about 50,000 cones (1% of the total), while the remainder of the total population of about 5 million cones is distributed throughout the peripheral retina with an average density of about 5,000 cones/mm2 (beyond about 10° eccentricity).
Since the cone inner segments act as their light-catching apertures, and since their diameter is about 10 mm in peripheral retina (Jonas, Schneider & Naumann, 1992), this density implies that the light-catching area of the cones is about 0.3 mm2 per mm2 of peripheral retina, while the rod light-catching area accounts for most of the rest. Thus, about one third of the peripheral retina should be considered to support color vision (Williams, 1991).
The mapping from retina to cortex can be approximated as a linear scaling from the fovea to the periphery, particularly for the cortical mappings of V2 and V3 (Schira, Tyler, Breakspear, & Spehar, 2009).
To project from the retina to equal regions of early visual cortex, therefore, the stimuli should be scaled in proportion to eccentricity, and studies of peripheral color processing should use such scaling in order to assess the cortical capabilities of color processing.
Indeed, with suitable areal scaling, color discrimination can be equated at all eccentricities This is not the place for an extensive review, but it should be noted that many studies of peripheral processing have used constant stimulus size and report progressive declines in hue discrimination (Nagy & Wolf, 1993; McKeefry, Murray & Parry, 2007; Mullen, 1991), chromatic saturation (Stabell & Stabell, 1982; Abramov, Gordon & Chan, 1992, McKeefry, Murray & Parry, 2007; Volbrecht & Nerger, 2012), and conspicuity (Gunther, 2014).
Those studies using appropriate size scaling of the stimuli generally find approximate invariance of the processing properties as a function of eccentricity (Noorlander, Koenderink, den Ouden, & Edens, 1983; Rovamo & Iivanainen, 1991; Abramov, Gordon & Chan, 1991, 1992; Sakurai & Mullen, 2006).
Where significant effects are reported, one may question the precise choice of scaling factor. Tyler (1987a), for example, proposed that the appropriate scaling factor should be based on stimulating the same number of cones at each eccentricity, whereas most scaling studies attempt to equate the number of ganglion cells stimulated or the cortical magnification factor per se (although some such studies do not apply this logic to the central foveal stimuli).
These properties are indicated by the following demonstration images, which should be viewed at a distance so as to make the width of the central disk measure about one twelfth of the viewing distance (e.g., 1 inch diameter at a 12 inch viewing distance).
This ratio corresponds to about the most liberal definition of the foveal region (5° diameter). Figure 1 shows an array of multicolored “balloons” within this foveal region scaled to stimulate about 1 cm2 of visual cortex at each eccentricity.
In Figure 2, the same form of array is scaled up to project to the periphery beyond the foveal limit (2.5° eccentricity) into the periphery. If periphery color vision had weaker color vision, the colors when viewing Figure 2 should appear desaturated relative to those in Figure 1, but inspection verifies that they do not.
(To reach the farther periphery, the viewing distance should be reduced to 5 inches with the 1 inch central disk region, when the outer rim will reach out to 45° eccentricity with central fixation.)
Figure 2 also includes an overlay of a horizontal line of unscaled disks typical of unscaled studies of peripheral color vision, to allow comparison of its degradation with eccentricity under these conditions.
Given that the natural units of cortical processing characterized by the concept of the “hypercolumn” are of the order of 2 mm wide in human visual cortex, the disks in Figures 1 and and22 should each stimulate about 25 such units.
To check whether color processing is similarly uniform at a grain of about 1 hypercolumn, the sizes of each disk are reduced by a factor of five for the peripheral version in Figure 3.
It can be seen that color is again visible out to the edge of the image without noticeable desaturation under these reduced stimulation conditions, so integration across multiple hypercolumnar units is not required to support peripheral color processing.
If anything, inspection of the figures shows the color perception is more vivid in the periphery, as might be expected from the fact that the cone density decreases at a slower rate than linear reciprocity with eccentricity (Curcio et al., 1987).
In fact, the cone density scales with approximately the −2/3 power of eccentricity out to 20° (Tyler, 1987b). Thus, the linear scaling of the disk sizes should result in the stimulation of about five times [(10−2/3/10−1)2] more cones by 20° than 2° in areal terms, giving scope for cortical processing to account for the extra vividness that is perceptually observed.
More information: Michael A. Cohen el al., “The limits of color awareness during active, real-world vision,” PNAS (2020). www.pnas.org/cgi/doi/10.1073/pnas.1922294117