Giving one color to your left eye and another to your right does not guarantee that you will see their average. Under suitable viewing conditions, one color can dominate, then the other. Sometimes a mixture appears instead. This is binocular color rivalry: competing input from the two eyes can produce a changing experience even while the physical images remain unchanged.
You may meet the problem in anaglyph pictures viewed through colored glasses, although ordinary 3D images are designed to encourage a coherent scene. Our explanation of red-and-cyan glasses separating two image channels concerns delivering each eye a different view. Rivalry concerns what perception does when those views resist being combined. The glasses route light; they do not dictate a single perceptual outcome.
Mixing light and combining eye views are different problems
When colored lights overlap on one surface, each eye can receive the same mixed light. That physical arrangement is different from presenting separate colors to corresponding locations in the two eyes. In the second arrangement, the visual system must deal with disagreeing inputs. Calling both situations “mixing red and green” hides the difference that matters most.
Hong and Shevell’s published color-rivalry research states the boundary explicitly: different colors presented separately can alternate in perception or, in some cases, yield a mixture. Their work also examines stimuli differing in both color and form. The color experience need not simply follow whichever whole image seems to win.
That qualification protects the title from becoming an instruction with a guaranteed result. The claim is that the two eyes can take turns contributing the dominant color, not that every possible pair of colors will do so, at a fixed rate, for every observer. A simple illustration with two colored patches cannot describe all of binocular vision.
What is present can differ from what is dominant
In ordinary viewing, differences between the eyes often contribute to depth. A near object occupies somewhat different positions in the two views, and the visual system combines the information into a unified scene. The Frontiers study by Andrews and Holmes investigates the more complicated situation in which rivalry and stereoscopic depth coexist.
Their experiments found conditions where depth information remained perceptible during rivalry. The important distinction is between an image being suppressed from awareness and its physical input being removed. The paper discusses evidence that suppressed information can continue to influence the appearance of a dominant stimulus. “I do not currently see it” is therefore not equivalent to “nothing from it is being processed.”
For a color example, imagine that a pattern is most visibly associated with one eye’s view, yet its color appearance is affected by input from the other eye. The details depend on the experiment, but the conceptual lesson is useful: visual features are not always delivered to awareness in a single sealed package labeled left eye or right eye.
Color and pattern can resolve separately
The Hong and Shevell experiments provide a particularly specific case. With rivalrous stimuli differing in form and color, observers could experience alternation between forms while colors from both eyes appeared in distinct parts of the currently perceived form. This does not mean observers freely invent arbitrary combinations. It means the experimental result is more complex than whole images taking turns intact.
It is tempting to explain that complexity with a miniature person inside the brain choosing one eye, then the other. The metaphor fails because it treats the feature relationships as a conscious decision. Rivalry is studied precisely because perception can change while the input stays controlled. The observer’s report describes the experience; it is not a command selecting every feature.
When comparing descriptions of a demonstration, ask whether the person reported a whole-field color, separate patches, a pattern orientation or a depth impression. Those are different observations. Two people saying “it switched” may be reporting different aspects of the experience. Useful descriptions specify what switched rather than assuming there was one all-or-nothing event.
Attention influences rivalry without making it voluntary paint
Attention is part of the research, but it is not an unlimited color selector. Ooi and He found that attention could influence dominance, including making an attended dominant image less likely to be suppressed in their tasks. The findings concern defined experimental conditions rather than a general ability to hold any desired color forever.
Hancock and Andrews also studied attention with rivalrous gratings. Their voluntary-attention result reduced the mean dominance period of the unattended grating rather than increasing the mean dominance period of the attended grating. That detail matters because “attention changes the balance” is a narrower and more accurate account than “attention locks your chosen image on.”
Neither study turns rivalry into a test of concentration or a ranking of eye quality. If an informal demonstration is weak or confusing, its display arrangement and viewing conditions may be unsuitable. Do not infer a diagnosis from whether you can produce a particular effect. The laboratory questions require controlled presentation and carefully defined reports.
The Color Memory task for comparing a remembered target explores a different source of uncertainty: what happens after a color is no longer visible. Rivalry can change which input dominates while competing inputs remain present. Keeping those tasks separate makes it easier to describe the boundary you are noticing.
No forced eye crossing, flashing display or long fixation exercise is needed to understand this result. If you encounter an optional demonstration, keep it brief and stop if it causes discomfort. The useful discovery is conceptual: two working eyes do not merely average two pictures. Color, form and depth can have different relationships to the same competing input.
Editorial backfill for October 7, 2026. Prepared on October 8, 2026.