Colour Constancy in Birds: An Alternative Mechanism?

Abstract

When recognising a surface colour, the visual system discounts the illumination, apparently by using some reference surfaces (like a spectrophotometer). To recover the illuminant colour it uses signals from different, sometimes remote, parts of the scene viewed either in sequence or in parallel. As a result, humans and animals fail to recognise the colour of a patch that is locally illuminated with a narrow light beam different in colour from the ambient illumination, but show good colour constancy when the beam envelops the surrounding scene.

The results obtained in birds distinguish them from humans and all animals hitherto investigated. The behaviour of hole-dwelling birds was studied in the wild by the method of alternative choice of entrance into experimental nesting-boxes having three entrances marked with coloured stimuli made from papers painted in different shades of blue, grey, or orange (see Maximov and Derim-Oglu, 1996 Perception 25 Supplement, 98, and the corresponding WWW site: http://www.digipark.com/science/meta). The spectral content of the direct sunlight illumination was changed by filters, either locally at the stimuli or over the whole front panel of the experimental box. In such artificial lighting conditions the birds proved to be incapable of using a neighbouring white surface as a sign of the illumination to discount its effect on the colour of objects. These unexpected results can be explained by the ability of animals with colour vision of rather high dimensionality to take recourse to local mechanisms of colour constancy that extract the colour information which is invariant to changes of the illumination, using certain a priori constraints on environmental light spectra. Natural daylight spectra have been shown to satisfy these constraints.