Opponent processing is widely accepted as providing a general framework for the standard model of human color vision. After the cones' responses are transmitted to second stage neurons, however, there is no consensus on exactly how synaptic connections are organized. The Relative Absorption Model introduced here is an explicit neural network that generates neural correlates of color vision. The model makes detailed predictions of known color and neural phenomena, including familiar aspects of color perception. Until now these phenomena have not had an explicit neural explanation. The model's simplicity shows that color does not require complex processing of spectral information. The network receives excitatory and inhibitory input from three classes of spatially proximate photoreceptors with different spectral sensitivities. Four second stage neurons provide symmetric input to four third stage neurons, whose outputs are correlates of red, green, blue, and yellow. These color cells identify which receptor type has the greatest absorption (of photons and which has the least. Their response intensities correspond to the differences between those absorptions and the middle absorption. A single second stage neuron computes violet and purple information that is then transmitted through the red and blue channels, the only channels in the network capable of conveying the information. Five additional neurons produce correlates of black and white. The white cell's response intensity measures the smallest of the three absorptions, and the black response measures how far the largest absorption is from full saturation. (c) 2005 Wiley Periodicals, Inc.