Four Arabian horses and one Thoroughbred were presented with a series of two-choice color vs. gray discrimination problems. Testing was done in a stall containing a wall with two translucent panels that were illuminated from behind by light projected through color or gray filters to provide the discriminative stimuli. Horses first learned to push one of the panels in order to receive the food reward behind the positive stimulus in an achromatic light–dark discrimination task, and were then tested on their ability to discriminate between gray and four individual colors: red (617 nm), yellow (581 nm), green (538 nm), and blue (470 nm). The criterion for learning was set at 85% correct responses, and final testing for all color vs. gray discriminations involved grays of varying intensities, making brightness an irrelevant cue. Three subjects were tested with all four colors. Two of those subjects successfully reached the criterion for learning on all four color vs. gray discriminations, while the third reached criterion with red and blue, but performed at chance levels for yellow and green. A fourth horse was only tested with green and yellow, and a fifth only with blue, and both of those horses successfully reached criterion on the discriminations they attempted. With the exception of the one subject's poor performance with yellow and green, there was no significant difference between horses on any of the discrimination tasks, and no significant difference in their performance with different colors. The results suggest that horses have color vision that is at least dichromatic, although partial color-blindness may occur in some individuals.