Humans can compare the orientations and locations of two motion-defined test bars several degrees apart so as to rapidly encode and place in memory their mean orientation, orientation difference, separation and mean location, while ignoring stimuli located between the two test bars. Performance is not impaired by randomly varying the location of the bars. We conclude that the two test bars are not compared by shifting gaze location or attention from one test bar to the other, nor by attending to two spatial locations. In addition, observers can discriminate the orientation difference and mean orientation of two test bars that, each of which is rendered visible by a different sub-modality (motion, disparity or luminance). Taking into account the findings reported here and previously reported findings on the early processing of luminance-defined form (Vis. Res. 40 (2000) 2291, Vis. Res. 42 (2002) 49) and cyclopean form (Proc. Roy. Soc. Lond. B 268 (2001) 213) we propose that the human visual system contains a fast long-distance comparator that compares the orientation and locations of two test bars while being insensitive to stimuli in the space between the test bars, and that this process is independent of whether the test bars are rendered visible by only one of three kinds of contrast (luminance, disparity, motion) or by combinations of the three. One role of this comparator mechanism may be to rapidly bind the spatial aspects of the retinal image across sub-modalities immediately after each saccade.