Limits to Animal Discrimination and Recognition in a Noise-Free External Environment

One and two dimensional sensory mechanisms, corresponding to one patch and a line array, respectively, of sensory cells are analyzed. A fundamental stimulus conversion law to average neural pulse rate is employed to explain several phenomena in psychophysics. A theory for neural noise is developed which is shown to specify the limits to animal discrimination and to explain certain observed neural mechanisms. Finally, concepts of optimum detection are employed to specify the ideal animal recognition sensory schema as limited by constraints associated with unavoidable neural noise. Actual animal mechanisms appear to follow the ideal. Certain aspects of Gestalt psychology are also evidenced. The general theory provides quantitative insight into artificial animal recognition and information handling systems and indicates the extent to which the capability of a mechanical device can equal or exceed that of its living counterpart. Explicit design considerations for implementation of the ideal artificial animal system are provided.