What moves in this display? The motion cue is rather abstract: In each frame, a vertical column or bar of dots in the pattern reverses in contrast (i.e. all black dots become white, and all white dots become black). As the frame sequence advances, each successive bar reverses in contrast. This kind of motion display is called 'second-order' because the motion is carried by the textural properties of pairs of points, rather than by the luminance properties of individual points (known as 'first-order' motion). In this case the pairs of points which distinguish the bar from its background are defined by flicker (points inside the bar reverse in contrast, those outside do not). In the corresponding first-order display, each point inside the bar would remain fixed at one intensity, while each point outside would remain at another.
Perception of first-order motion can be explained by activity in conventional neural motion detectors (also called 'first-order' or 'energy-based' motion detectors). Perception of second-order motion seems to involve two mechanisms; a more sophisticated form of neural motion detector that can respond to moving texture boundaries ('second-order motion detector'), and another process that tracks corresponding features in the pattern as they move.
See:
Cavanagh, P. & Mather, G. (1989) Motion: the long and short of it. Spatial Vision, 4, 103-129.
Smith, A. T. (1994). Correspondence-based and energy-based detection of second-order motion in human vision. JOSA A, 11, 1940-1948.