Detecting motion along spatio-temporally coherent vs. incoherent trajectories

Abstract

Purpose: Watamaniuk, McKee & Grzywacz (1995) have reported that observers are surprisingly efficient at detecting a single coherently moving dot within an array of short-lifetime noise dots. A possible explanation is that visual system pools over outputs of local motion detectors along spatio-temporally coherent trajectories (i.e. trajectories consistent with a velocity of the local motion). This strategy would improve signal strength for motion along an extended trajectory. We tested whether observers exhibit sub-threshold summation of motion energy distributed over spatio-temporally coherent trajectories. Method: Subject's task was to detect motion embedded in noise. Dynamic white noise was convolved with a local spatio-temporally oriented filter, resulting in a drifting surface with limited temporal correlations (pixel correlations fell to zero within 70ms, or 0.63° of movement). The drifting surface was blended with background noise using a small gaussian spatial window (s=0.5°), which moved at the same speed as the drifting surface (9.4°/s) along a trajectory extending 6.3°. The gaussian window moved either in the same direction as the surface motion (coherent condition), in the opposite direction (opposite condition), or in an orthogonal direction (orthogonal condition). In the coherent condition, the local motion energy propagates along a global 'trajectory' that is consistent with the motion signal, while in the other conditions it does not. On each trial, observers indicated whether the trajectory was located to the left, to the right, below, or above the fixation point (4AFC). The signal-to-noise ratio was varied to determine thresholds. Results: Detection thresholds were the same across the three conditions. Motion propagating along a consistent trajectory did not improve the detection of the motion. Conclusions: We found no evidence that spatio-temporal coherence facilitates detection of sub-threshold motion energy.