While our three retinal cones allow us to see up to one million different hues, most other mammals make do with two-cone dichromatic colour vision. In horses, the short-wavelength sensitive cone detects blue light, whereas the other cone picks up longer wavelengths - such as long wavelength yellow and medium wavelength green. Comparing the signal from the two cones gives the chromatic space, which is the spectrum of colours a dichromat can see. As Lina Roth from Sweden's Lund University explains, `when both cones are stimulated equally, there is a “neutral” point in the chromatic space'. We would perceive a neutral point as a white or grey colour, but a dichromat could see this point as blue-green. A horse's grey neutral point sits in the chromatic space between green and blue. One question researchers want to answer is whether dichromats see a continuous spectrum of colours or whether they distinguish two colour categories of long and short light wavelengths divided by the neutral point.
To investigate how horses perceive colours, Roth and her colleagues Anna Balkenius and Almut Kelber trained horses to recognise two `positive' colours and a `negative' colour, which were printed on pieces of paper attached to small doors in a screen (p. 2795). By nudging a door with their noses, the horses gained a tasty carrot reward from positive colours but not negative ones. Once the horses reached a 75% or higher success rate in picking out their positive colours, the team started tests.
In the first test they wanted to find out how horses responded to colours on the same side of the neutral point. They trained the horses with long wavelength yellow and a grey as their positive colours, with short wavelength blue as the negative colour. The horses made choices differently to how humans would make them - rather than choose the positive stimulus if it was presented they chose the colour with the longest wavelength, effectively `generalising'their response to `positive equals longer wavelength'. So the horses picked a novel green over shorter wavelength grey and blue, and yellow over grey.
Next the team trained horses with blue and green as the positive colours,and yellow as the negative colour, finding that the horses could also generalise colours either side of the neutral point. The horses chose neutral point grey over yellow, suggesting that they had generalised the grey - in between green and blue - as positive. They preferred blue over both grey and green, showing that the horses were choosing the colour with the shortest wavelength, most different from negative yellow. The results also show that grey is considered a colour in its own right and not as a gap in the chromatic space.
In a final test the team trained the horses with positive green and negative neutral point grey. Given a choice between grey and blue, the horses eventually settled on grey, even though it was the negative stimulus. They chose it because it is closest to green, which was the longer-wavelength positive stimulus.
So horses mostly learn colours in a relative, rather than an absolute manner, but can recognise specific colours too. Because horses have a continuous dichromatic space, the same is likely to be true for other dichromats, and researchers can't assume that grey is a non-colour. Next,though, Roth is going to turn the lights out, `to find out the limits of the horses' colour vision'.