A long-standing dichotomy is that between short- and long-range motion processing. Here I present evidence for common early analysis, followed by divergent processing sensitive to spatial form for long-range processing. Subjects observed motion sequences consisting of a vertical, sinusoidal luminance grating (40% contrast, 1.0c/deg) divided into horizontal strips of equal height. Alternate strips drifted leftward and rightward in 90 deg steps. The initial relative phase of adjacent strips was either aligned (0 or 180 deg) or non-aligned (90 or 270 deg); in the former case, extended vertical luminance boundaries were present in every other stimulus frame. The single-interval task was to identify the direction of motion of the central strip. An interstimulus interval (ISI), range 0-400ms, was inserted in between consecutive image frames (each presented for 33.3ms) in the motion sequence, and the threshold strip height (TSH) was determined for veridical performance. In the non-aligned condition, the TSH was 7 minarc for all ISIs. In the aligned condition, the TSH was 11 minarc for all ISIs except 0ms (7 minarc) and 50ms (19 minarc). The constant TSH value in the non-aligned condition implies a common physical limit to both processes, perhaps initial processing by the same front-end filters; the figure of 7 minarc agrees well with previous estimates of short-range motion receptive field heights (Anderson & Burr, 1991, JOSA, 8, 1330), indicating similar underlying sensors for the initial stages of the long-range process. The TSH values in the aligned condition reveal (i) the insensitivity of the short-range process to spatial alignment, (ii) the division between short- and long-range processing around 50ms, and (iii) the effect of spatial form on the long-range process. There appear to be common early spatial constraints on both short- and long-range motion processing, at least for narrowband stimuli, and the long-range process is particularly sensitive to spatial form.