ABSTRACT
In deriving compass information from the pattern of polarized light in the sky (celestial e-vector pattern), hymenopteran insects like bees and ants accomplish a truly formidable task. Theoretically, one could solve the task by going back to first principles and using spherical geometry to compute the exact position of the sun from single patches of polarized skylight. The insect, however, does not resort to such computationally demanding solutions. Instead, during its evolutionary history, it has incorporated the fundamental spatial properties of the celestial pattern of polarization in the very periphery of its nervous system, the photoreceptor layer. There, in a specialized part of the retina (POL area), the analyser (microvillar) directions of the photoreceptors are arranged in a way that mimics the e-vector pattern in the sky (matched filtering). When scanning the sky, i.e. sweeping its matched array of analysers across the celestial e-vector pattern, the insect experiences peak responses of summed receptor outputs whenever it is aligned with the symmetry plane of the sky, which includes the solar meridian, the perpendicular from the sun to the horizon. Hence, the insect uses polarized skylight merely as a means of determining the symmetry plane of the polarization pattern, and must resort to other visual subsystems to deal with the remaining aspects of the compass problem (parallel coding).
The more general message to be derived from these results is that in small brains sensory coding consists of adapting the peripheral rather than the central networks of the brain to the functional properties of the particular task to be solved. The matched peripheral networks translate the sensory information needed for performing a particular mode of behaviour into a neuronal code that can easily be understood by well-established, unspecialized central circuits. This principle of sensory coding implies that the peripheral parts of the nervous system exhibit higher evolutionary plasticity than the more central ones. Furthermore, it is reminiscent of what one observes at the cellular level of information processing, where the membrane-bound receptor molecules are specialized for particular molecular signals, but the subsequent molecular events are not.
