|
Because the coding range of sensory systems is restricted, they have to adapt to changes of the statistics of the relevant stimulus space. Sensory adaptation can ensure precise and reliable encoding of the environment over several orders of magnitude of mean intensity. By removing the mean intensity level, adaptation enables object recognition invariant of the local context. Thus, adaptation removes information from the sensory representation, for example the mean intensity. However, sensory systems have to represent different features of a stimulus in parallel, and the information that is important for detection of these may differ. Consequently, adaptation should act differently in different parts of the pathway. We asked how adaptation mechanisms should act in a pathway that processes two different features among others: in the auditory system of vertebrates and invertebrates, amplitude modulations are used for object recognition, while intensity differences between the two ears are informative on object localization. By analytically calculating the optimal response curves for both tasks we ask how adaptation should ideally affect these response curves and whether adaptation should act primarily peripherally or centrally. The results show that for processing of amplitude modulations, a peripheral removal of mean intensity by adaptation is desirable. However, peripheral adaptation removes information that is important for localization. In invertebrates and vertebrates alike there is evidence for very strong negative feedback in the central units processing intensity differences. We performed numerical simulations of a simplified binaural auditory pathway demonstrating that negative feedback may help to focus computation of locality on the most informative part of a stimulus: the onset. Our model shows that such a focussing can enhance localization performance and greatly increase coding efficiency. Our modeling also makes predictions for the localization performance in response to certain auditory stimuli in humans and grasshoppers alike. |
|