Abstract
We have previously shown that randomly refreshing textures presented at a constant frame rate give rise to illusory percepts of coherent apparent motion, a phenomenon we term motion pareidolia (Davidenko et al., VSS 2015, CogSci 2015). Here we report the effects of frame rate (Study 1) and number of frames (Study 2) on the strength of this phenomenon. Participants observed brief movie clips (between 0.625 and 10 seconds long) and classified each clip as vertical motion, horizontal motion, or random. Vertical motion clips consisted of a fixed random texture of 140x140 pixels alternately shifting up and down by 4 pixels, with 40% of pixels re-randomized at each frame (40% noise). Similarly, horizontal motion clips alternately shifted right and left with 40% noise, and random clips refreshed randomly at each frame (100% noise). Across studies, participants were very accurate at classifying vertical and horizontal motion trials (average proportion correct ~0.94), but very inaccurate at classifying random trials as random (average proportion correct ~0.57), replicating our previous motion pareidolia findings. In Study 1 (N=65), we found that whereas performance on motion trials increased monotonically as frame rate increased from 0.5Hz to 8Hz, performance on random trials showed a U-shape pattern, significantly dipping between 1Hz and 2Hz, suggesting that motion pareidolia peaks around this frequency. In Study 2 (N=85) we kept the frame rate constant at 2Hz and varied the movie clip durations from 3 to 9 frames. We predicted that as the number of frames increased, participants would accumulate more evidence for or against coherent motion, and performance would increase substantially, especially for random trials. Instead, performance on random trials stayed relatively constant, fluctuating between 0.60 and 0.65, across movie clip durations. Our results suggest that once illusory motion is perceived, it is likely to persist across many subsequent frame transitions.
Meeting abstract presented at VSS 2016