Purchase this article with an account.
Lisa E. Williams, Edward M. Hubbard, Vilayanur S. Ramachandran; Postdiction in visual motion perception. Journal of Vision 2004;4(8):210. doi: 10.1167/4.8.210.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
We report a new apparent motion illusion that illustrates postdiction (instead of prediction) in human motion perception, which implies that events in the future can influence the perception of events in the present. Two black dots were flashed simultaneously on diagonally opposite corners of an imaginary square (top-left and bottom-right) in Frame 1 of a movie followed by two dots on the remaining two corners in Frame 2. This creates an ambiguous apparent motion display; one sees either vertical or horizontal motion. We have previously demonstrated visual inertia, a tendency to preserve motion along a straight line (Ramachandran & Anstis, 1986). When the ambiguous display is preceded by a biasing frame composed of two other dots (one appearing to the left of the top-left dot and the other appearing to right of the bottom-right dot) the unambiguous horizontal motion between these priming dots and the dots in Frame 1 strongly biases observers towards seeing straight-line horizontal motion. We now find that such a bias occurs even if the priming dots are presented after Frame 2 so that they continue the motion trajectory. These postdiction dots appear to the right of the top-right dot and to the left of the bottom left dot. We tested 16 observers in two experiments. To control for potential directional bias in the original ambiguous square display, stimuli for the current study were rotated 45 degrees. Postdiction now produced a bias towards the perception of straight-line diagonal motion; the absence of postdiction resulted in the perception of L-shaped motion. We found that postdiction was strongest at faster speeds (66 and 100 ms/frame). We then determined the strength of the postdiction effect using nulling displays that narrowed the distance between the parallel trajectories of the dots, thereby biasing the displays towards the perception of L-shaped motion. We found that postdiction persisted until a bias of 77% of original distance.
This PDF is available to Subscribers Only