Individual motion measurements are inherently ambiguous since the component of motion parallel to a homogeneous translating edge cannot be measured. Numerous models have proposed that the visual system solves this ambiguity through the integration of motion measurements across disparate contours. To examine this proposal, subjects observed a translating diamond through four stationary apertures. Since the diamond's motion could not be determined from any single contour, motion integration across contours was required to determine the diamond's direction of motion. We demonstrate that observers have difficulty accurately integrating motion information across space. Performance improved when the diamond stimulus was presented at 7 degrees eccentricity, through jagged apertures, or at low contrast. Taken together, these results imply that integration across space is more likely when the motion of contour terminators is less salient or reliable.