Next time someone asks you what your favorite color is, you might want to think twice before answering—your brain might be playing a trick on you. What we see when we look at an object is not its “true” physical color, says David Brainard, RRL Professor of Psychology, but our brain’s subjective reading of spectrum. “There are creatures out there that are able to perceive even more colors than humans,” says Brainard. “Some birds, insects, and mantis shrimps, for instance, would say we were somewhat colorblind in comparison.”
The process begins with illumination. When light arrives at an object, some fraction of that power is absorbed at varying wavelengths. The light is then reflected back to the eye where different cone cells, sensitive to either short, medium, or long wavelengths (referred to colloquially as blue, green, and red cones) receive it. When the light source is varied—a cloudy sky versus fluorescent lights, for instance—the eye and brain are able to correct the color. Brainard says we can see this same kind of process, called color constancy in the case of color, at work in the way we perceive size.
“If an object is moved from across the room to right in front of you, it just got twice as big on the retina. To the viewer, however, it doesn’t appear to have changed in size,” says Brainard, who is the director of both the Vision Research Center and the Institute for Research in Cognitive Science at Penn. “You’re not aware of the fact that your brain is constantly factoring the distance to generate perceived size, much like you’re not aware of it correcting colors.” In some instances the brain is even more advanced than its machine counterparts. “If you watch tennis or football on TV, sometimes you’ll see color artifacts in the white lines,” he says. “Those arise because the way that color cameras process images is not as smart as the way brain does it.”
In the blocks-copying task, a model of colored blocks is provided. These serve as the targets. The subject selects a block from the source by clicking on it and then deposits it in the workspace by clicking on one of the black blocks there. The subjects’ task is to reconstruct the model, “as closely as possible”, in the workspace. At the time point depicted by the schematic, the subject has placed one block in the workspace and has placed the cursor (black circle) over a second block in the source. The model is rendered under a reference illuminant, while the illumination used to render the workspace and the source is the test illuminant. The text shown above and below the stimuli is for expository purposes and is not present during the experiment. (view larger image)
The brain can be tricked, however. Brainard cites the recent Internet debate surrounding a picture of a two-tone dress that evoked different perceptions of color. “If you juxtapose the image of that dress over two different background colors, your brain thinks the illumination is blue in one instance and yellow in the other, altering perception,” says Brainard, who commented extensively on the phenomenon in the media and is currently editing a special issue of a scientific journal that will feature papers on this topic. “So the same process that stabilizes the colors of real objects against changes in the illumination may be leading us astray in the case of the picture of the dress.”
Over the years, Brainard has invited student volunteers to participate in similar tests at his lab. “If I give you a series of lights that range from greenish yellow to reddish yellow and ask you to pick the ‘best’ yellow, it turns out almost everyone will agree,” Brainard says. “If I were to do the same thing but ask you to pick a green that’s neither blue-ish nor yellow-ish, there is enormous individual variability in that.”
Illustrative stimuli for Aim 2. The montage shows stimuli that vary in material (specular parameters vary across top row, with diffuse spectral reflectance held fixed) and in color (diffuse spectral reflectance varies across bottom row, with specular parameters held fixed). The blobs are members of a nine-parameter family of shapes obtained by sinusoidally perturbing the positions of the vertices of a sphere around three axes. When the target is T (same in both rows), the set of color lures is shown along the bottom row (including the target T, which corresponds to ∆M = ∆C = 0). Each color lure is paired with a blob with each value of ∆M shown along the top row (although we do not run the target against itself). (view larger image)
Brainard is bringing his expertise to a variety of different projects within the Penn community.
He recently collaborated with Assistant Professor of Psychology Coren Apicella to better understand how different cultures assign color labels. Apicella, who works in the field with the Hadza, a population of nomadic hunter-gatherers in Tanzania, presented a palette to them and asked them to identify colors by regional name. “Every Hadza agreed on the name for red, black, and white, but beyond that, the most frequent response was: ‘That color doesn’t have a name’ or “I don’t know the name of that color,” says Brainard. “Nonetheless, their color vision is perfect. It’s yet another layer of subjectivity.”
Brainard is also working with Assistant Professor of Ophthalmology Jessica Morgan to investigate ways in which measuring perception might eventually be used to help guide treatment of blinding diseases. The project is supported by Research to Prevent Blindness, which earlier this year presented Brainard with the Stein Innovation Award. “We’re going to make measurements of how well people can see if we stimulate individual cells in the retina,” says Brainard. “We hope to link those measurements to structural changes in the retina, as a way to characterize disease progression and remission.”
