Visual motion is sensed by low-level (energy-based) and high-level (feature-based) mechanisms. Our interest is in the motion detectors underlying the initial ocular following responses (OFR) that are elicited at ultrashort latencies by sudden motions of large images. OFR were elicited in humans by applying horizontal motion to vertical square-wave gratings lacking the fundamental. In the frequency domain, a pure square wave is composed of the odd harmonics—first, third, fifth, seventh, etc.—such that the third, fifth, seventh, etc., have amplitudes that are one-third, one-fifth, one-seventh, etc., that of the first, and the missing fundamental stimulus lacks the first harmonic. Motion consisted of successive quarter-wavelength steps, so the features and 4n+1 harmonics (where n = integer) shifted forward, whereas the 4n-1 harmonics—including the strongest Fourier component (the third harmonic)—shifted backward (spatial aliasing). Thus, the net Fourier energy and the non-Fourier features moved in opposite directions. Initial OFR, recorded with the search coil technique, had minimum latencies of 60 to 70 ms and were always in the direction of the third harmonic, for example, leftward steps resulted in right-ward OFR. Thus, the earliest OFR were strongly dependent on the motion of the major Fourier component, consistent with mediation by oriented spatiotemporal visual filters as in the well-known energy model of motion detection. Introducing interstimulus intervals of 10 to 100 ms (during which the screen was uniform gray) reversed the initial direction of tracking, consistent with extensive neurophysiological and psychophysical data suggesting that the visual input to the motion detectors has a biphasic temporal impulse response.