Colour Vision in Birds : Comparing behavioural thresholds and model predictions

Sammanfattning: Birds use colour vision for many biologically relevant behaviours such as foraging and mate choice. Bird colour vision is mediated by four types of single cones, giving them an extra dimension of colour information compared to trichromatic humans. The cone photoreceptors of birds have coloured oil droplets that are assumed to increase the discriminability of colours in bright light at the cost of dim light sensitivity. In this thesis I present four studies where we have trained chickens to perform colour discrimination and tested the limits of their behavioural performance. In paper I we tested how small colour differences chickens can discriminate. This allowed us to test the predictions of the most well established model for bird colour vision, the receptor noise limited model. There was a reasonably good fit between model and behaviour. Furthermore, we tested in how dim light chickens could discriminate colours and found that the intensity threshold was affected by the colour difference between the stimuli and their intensity. In Paper II we continued testing colour discrimination in dim light and tested the hypothesis that chickens sum the signals from many photoreceptors to increase contrast sensitivity at the cost of spatial resolution in dim light, so called spatial pooling. We used food containers covered with larger, smaller, more or fewer colour patches. Supporting the hypothesis, the containers covered by more colour could be discriminated in dimmer light. In Paper III we tested colour constancy, the ability to maintain colour perception in different spectral illuminations that would otherwise confuse colour perception. Our aim was to find the largest illumination change that chicken colour constancy could tolerate. We found that chicken colour constancy could tolerate larger illumination changes when discriminating stimuli that were more different from each other. In paper IV we continued the work on colour constancy but allowed the chickens to use relative colour learning, which was specifically excluded in paper III. In Paper IV we found that their colour constancy could tolerate larger illumination changes. In nature relative colour cues are available and may be an important aspect of colour learning and perception. These results suggest that such cues can make colour constancy more robust to larger illumination changes. In both experiments chicken colour constancy was improved if they were adapted for 5 minutes in the tested illumination before performing the discrimination task. We compared the illuminations for which chickens retained colour constancy, to the difference between natural illuminations and we can conclude that chickens are well equipped to maintain accurate colour perception when changing between habitats in the wild. Objects are detected both by their chromatic and achromatic contrasts. The receptor noise limited model can be used to predict discriminability through both chromatic and achromatic vision. To use the model reliably its assumptions and predictions must be compared to behavioural results. This has been done for the chromatic version of the model but not the achromatic. In Paper V we compiled all known chromatic and achromatic contrast detection thresholds, and used them to derive the limiting noise level to be used when predicting visual discrimination in a range of animals. We discuss the limitations of using modelling in the wild such as the need to consider the spatial pattern of the stimuli and the light intensities in which the modelling occurs.