Motion detection requires the comparison of spatially and temporally displaced samples of the image. Here, we discuss the problems associated with measuring the delay between spatially displaced signals within biological motion detectors. Data are presented from direction-selective neurons in the nucleus of the optic tract of the wallaby, Macropus eugenii. Their motion responses depend on stimulus contrast and the adapted state of the cells. At low contrasts or in an adapted state, it appears that the input to the motion detectors is a temporally low-passed version of the image. At high contrasts or in the unadapted state, the input signals appear to be temporally band-pass-filtered. Contrary to previous claims, we find that neither the response to stimulation with apparent motion nor measurements of temporal frequency response functions provide a direct estimate of the delay filter time constants. Instead, we find that both measures are also dependent on the temporal characteristics of prefiltering stages. A model is proposed to account for the responses of the neurons and their contrast dependence.
