A model of temporal adaptation in fly motion vision

A computational model is proposed to account for the adaptive properties of the fly motion system. The response properties of motion-sensitive neurons in the fly are modelled using an underdamped adaptive scheme to adjust the time constants of delay filters in an array of Reichardt detectors. It is shown that the increase in both temporal resolution and sensitivity to velocity change observed following adaptation to constant motion can be understood as a consequence of local adaptation of the filter time constants on the basis of the outputs of elementary motion detectors.