Spatial and Temporal Properties of Motion Linking

Abstract

Because estimates of velocity are local and inaccurate at early stages, motion linking is necessary to recover global object velocities. Such a process could involve the V1 projections of the magnocellular system into area MT. In the present study, we determined the spatiotemporal characteristics of motion linking and its dependence upon local spatial interactions in order to better understand the dynamics and topology of integration across scales and features. For that purpose, we used structurally different textures comprising nonoverlapping patches of drifting sine-wave gratings of variable spatial and temporal frequencies. The gratings (relative orientation: 90°) were arranged to form either L or T junctions and could either drift smoothly or sinusoidally 90° out of phase. In this latter condition, combining motion signals across space results in clockwise or counterclockwise translations that observers were asked to discriminate. Contrast response functions for both a coherence and a discrimination task indicate that integration is easy at low, but not at high, contrast. In addition, coherence is less likely at low spatial frequencies and high speeds and for small as compared to large patches (for a constant stimulus area). Performance is far better for L than for T junctions, although this effect is attenuated for vertical and horizontal as compared to oblique (±45°) gratings. These results are not compatible with the view that the magnocellular system is primarily responsible for motion linking and we further suggest that complex interactions between the magno and parvo systems are involved in this process. Finally, the better performance observed for L as compared to T junctions brings evidence that spatial grouping provides a handle for motion linking.