Perceived Velocity in Dynamic Random-Dot Patterns is Influenced by Noise Type

Abstract

Local motion signals can be pooled to detect the direction of coherent motion in random-dot kinematograms (RDKs) having a high proportion of random noise. Noise type does not appear to affect direction discrimination in such displays (Scase et al, 1996 Vision Research36 2579 – 2586). We have observed that RDKs with low coherence yet an obvious global direction appear to move slower than similar RDKs with high coherence. Using judgements of relative speed between RDKs containing different proportions of noise in a 2AFC paradigm we have quantified this effect and sought to determine if the type of noise influences perceived velocity. Levels of coherence in all dot patterns were well above the thresholds for directional judgements. Dots were assigned as ‘Noise vs Signal’ randomly on each frame of the RDK. Noise dots were either of type ‘random position’ or of type ‘random walk’. Position noise dots were randomly repositioned within the area of the display on each frame and had an isotropic distribution of directions and variable speeds. Random-walk dots moved at the same speed on successive frames (their displacement matched to that of the signal dots) but in a randomly chosen direction. The two noise types yielded statistically different results. In RDKs containing random-walk noise, decreasing the coherence of the display (30% signal, 70% noise) reduced perceived velocity (on average to 0.75 of the actual velocity), while increasing the coherence of the display increased perceived velocity until at high coherence levels (80% signal, 20% noise) the perceived velocity approximated the veridical velocity (on average 0.96). The proportion of position noise in a display had no effect on perceived velocity. These basic results are discussed in relation to current models of motion detectors and velocity perception.