We measured evoked neuromagnetic responses to the direction-reversal of translating visual gratings. The direction reversal itself is designed to eliminate flicker cues during the transient reversal and provide a pure motion signal. The gratings are highly similar in appearance (same carrier) but carefully calibrated to stimulate only the first- and only the second-order motion systems. Magnetoencephalographic (MEG) data was recorded from three subjects as they passively viewed these direction reversing, first- and second-order motion displays. Visually evoked magnetic fields (VEFs) yielded high amplitude focal activation, with dipolar sensor distributions, located over lateral sensors in all subjects. Data were additionally analyzed by Second-Order Blind Identification (SOBI), Independent Component Analysis (ICA), and Cortically-constrained Current Density (CCD) source localization. The data reveal a characteristic profile of MEG activity accompanying the change in motion direction. Activated sites include extrastriate dorsal and lateral visual areas (putatively V3/MT+), and portions of the inferior parietal lobe, but canceling the flicker cue seems to have eliminated VEFs in early visual cortical areas (V1/V2). The spatiotemporal VEFs for first- and second-order stimuli are quite similar; the SOBI analysis suggests that second-order responses are delayed relative to first-order responses by about 20 msec (which may be due to uncontrolled differences in stimulus strength between the first- and second-order stimuli). The similarity of VEFs for both kinds of stimuli, the high correlations between first- and second-order derived SOBI and ICA components, and the large degree of overlap in CCD sources indicate that spatial differences —if any—between first- and second-order visual motion sources are not resolvable by MEG.