Equine colour vision was measured under conditions that minimised the possibility of animals using brightness cues to make chromatic discriminations. In a two-stage study, we first obtained luminance discrimination functions for achromatic targets then tested for chromatic discrimination over a range of target luminances. Horses were trained on a two-choice discrimination task. The positive stimulus was varied in luminance and/or colour using neutral density and broad band colour filters. The negative stimulus appeared as a uniform grey. In the brightness discrimination task, the horses performed well at large luminance differences but their percentage of correct responses declined to near chance levels at differences of less than 0.2 log units. In addition, a decrement in performance was noted at luminance differences of less than 0.2 log units for green and yellow chromatic discrimination functions, suggesting that horses cannot easily discriminate yellow and green from grey. However, the chromatic discrimination functions for red and blue showed that animals performed very well across the full range of target luminances. These results suggest that horses are at least dichromats.