Past the cones, signals are re-coded into red-green, blue-yellow, and black-white opponent channels

What it is

Before the responses of the three cone types ever leave the retina, they get repackaged into a different code: no longer "how strong is L, how strong is M, how strong is S" individually, but three pairs of mutually opposed difference signals — a red-green channel (L minus M), a blue-yellow channel (S minus L+M), and a black-white channel (total luminance). This is opponent-process theory.

This conversion step is easy to overlook, because everyday talk about color stays unconsciously at the "three primaries" level. But the signal actually carried forward and actually processed by the brain, starting at the retinal ganglion cell stage, is already three difference pairs, not three independent intensity readings. Understanding the phenomena tied to opponent channels covered elsewhere — why certain color combinations simply don't exist, why afterimages reverse hue — starts from this re-coding step.

Why it happens

Getting from the raw responses of the three cone types to the three opponent channels requires a rewiring stage at the retinal ganglion cell level: some ganglion cells set L-cone input as excitatory and M-cone input as inhibitory (or the reverse), combining them into a difference signal along the red-green axis; others set S-cone input against a combined L+M signal with opposite polarity, combining them into a difference signal along the blue-yellow axis; a further class simply pools the total response across all cones without any opponent operation, giving a luminance signal on the black-white axis. This "convert to three colors first, then convert to differences" two-stage architecture isn't an arbitrary engineering choice — it's a more economical way to code information: color signals in natural scenes carry strong inherent correlation (the three cone types' responses tend to move together), and carrying a difference rather than three independent readings compresses the data that needs to be transmitted substantially while preserving what's needed to discriminate colors — a meaningful bandwidth saving for a neural pathway.

Understanding this re-coding is the prerequisite for the phenomena tied to opponent channels covered elsewhere: since the red-green and blue-yellow axes are each a single "up here, down there" difference rather than two independent dimensions, an instruction like "increase the red component and the green component at the same time" simply has no way to be expressed in this code; likewise, the adaptive rebound that follows one channel being pushed toward one pole for a long time naturally shows up as a brief advantage in the opposite direction.

Studying it

Classic evidence for opponent coding comes from color afterimage experiments: participants fixate an adapting stimulus of one color for an extended period, then look at a white or gray field and report the complementary afterimage color they see — the afterimage color systematically appears at the opposite pole of the opponent axis (a red adaptor yields a green afterimage, and vice versa). That regularity is itself behavioral evidence for opponent coding. More direct physiological evidence comes from single-unit electrophysiological recording: recording directly from retinal ganglion cells and lateral geniculate nucleus (LGN) neurons in animals reveals cells that respond with opposite polarity to red and green light (one wavelength excites, the other inhibits), with the same pattern for the blue-yellow axis.

The independent variable is usually the wavelength or hue angle of the adapting or test stimulus; the dependent variable is the afterimage color, or the polarity of a cell's excitatory versus inhibitory response. In interface research, this theory is more often used as the explanatory basis for which color combinations in a palette will pull against or reinforce each other, rather than as something measured directly.

Methodologically: opponent-process theory itself is mature and widely accepted, but applying it to a specific interface color decision still requires validation under the actual viewing conditions (background color, ambient light) — the theoretical framework cannot substitute for measurement in the specific scenario.

Where it stops holding

• Opponent coding happens early, in the retina-to-thalamus pathway. It does not mean higher-level color perception — color naming, cultural differences in category boundaries — is fully determined by the values on these three channels; later cortical processing and cognition build further on top of it.
• The "null point" on each axis (the neutral point that is neither reddish nor greenish, say) varies between individuals, tied to factors like the observer's specific cone pigment genotype and lens pigment density; not everyone perceives exactly the same physical wavelength as "pure red" or "pure green."
• This describes the channel structure of individuals with normal trichromatic vision. People with color vision deficiencies have missing or altered cone input to begin with, so the opponent-channel computation results change accordingly and this normal configuration doesn't transfer directly.

Applying it

• Understanding the opponent-channel structure helps identify which color pairs in a palette are inherently "mutually reinforcing" (colors far apart on an opponent axis, like red vs. green or blue vs. yellow) and which are inherently "hard to tell apart" (colors close together in direction on the same opponent axis) — the former suit status contrast colors that need fast discrimination, the latter tend to be perceived as the same category when placed together.
• When designing an element that requires a sustained gaze followed by a look elsewhere (staring at a saturated color block for a while, then glancing away), expect the user to see that block's complementary afterimage on the opponent axis; avoid having the very next screen happen to use that complementary color as the background for key information, which would let the afterimage interfere with reading it.
• Verification: pair up candidate accent/status colors and check whether they sit on different opponent axes (one red, one green; one blue, one yellow) rather than close together in direction on the same axis, then confirm discriminability against the real background.

Related

• Same group: A1.05.1 The summed response of three cone types produces color perception · A1.05.6 Opponent coding explains why combinations like "reddish green" don't exist · A1.05.7 Color afterimages are caused by adaptive reversal in opponent channels · A1.05.8 Interface color discriminability should be designed along opponent channels, not isolated hue
• Nearby: A1.06 Color vision deficiency · F5.01 Color models and color spaces
• Search terms: opponent-process theory · opponent color channel · red-green channel · blue-yellow channel