Purchase this article with an account.
Nikos Gekas, Kyle McDermott, Pascal Mamassian; Perceptual effects of adaptation over multiple timescales. Journal of Vision 2017;17(10):489. doi: 10.1167/17.10.489.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
It is well known that adaptation to a visual stimulus leads to a negative correlation between the current percept and previous percepts. However, there are diverging views on how stimuli further in the past affect the current percept. We have argued that the negative correlation between the current percept and recent ones is accompanied by a positive correlation with events occurring further in the past (Chopin and Mamassian, Current Biology, 2012; McDermott et al., VSS, 2015). Here, we design a novel psychophysical experiment to measure the perceptual effects of adapting to dynamically changing stimulus statistics. Observers are presented with a series of oriented Gabor patches from a range of orientations and are asked to judge whether the orientation of some test patches in the series is clockwise or counter-clockwise from a reference orientation. Unbeknownst to the observers, the test stimuli are always at the point of subjective equality as measured at the start of the experiment, i.e. they are highly ambiguous. The orientations of the stimuli before each response are drawn from a Gaussian distribution whose mean changes slowly in time following a sinusoidal pattern. Moreover, the frequency of the sine wave increases as the experiment progresses, thus affecting the rate of mean orientation change. We measure the bias in observers' responses over the course of thousands of trials and hundreds of responses. Our results suggest that a negative tilt after-effect for short timescales gradually changes into a positive effect for trials further in the past. In addition, there is a weak positive correlation with stimuli seen hundreds of trials before the current percept. We present a computational model that illustrates how a combination of negative and positive correlations can predict the psychophysical data more accurately that a single negative correlation with recent stimuli.
Meeting abstract presented at VSS 2017
This PDF is available to Subscribers Only