Human observers judge whether there is a faint, foveal, horizontal target bar in a briefly presented stimulus. In some trials, there are task-irrelevant horizontal flanker bars extending horizontally, bilaterally, and collinearly from the target location. Zhaoping and Jingling (2008, PLoS Computational Biology 4 e14) found that low and high contrast flankers bias the perception, such that observers were more likely to see the target as present and absent, respectively. This finding was understood from the framework of Bayesian inference, since the flanker contrast influences our brain's internal likelihood function for the luminance contrast at the target location. In particular, a target is presumed more likely to produce (via noisy image formation) low or even zero image contrast when the flankers have low rather than high contrast, and therefore low contrast flankers tend to make perception fill in the target even when it is absent. We extend the previous study using dichoptic presentation. Each stimulus bar is presented to one eye only; the eye of origins of the left set of the flankers, the right set of the flankers, and the target are randomly and independently chosen for each trial. Preliminary results show that the contextual influence on the perceptual bias is largely independent of the dichoptic condition, as long as the binocular alignment (assessed in a separate experiment) of the dichoptic stimuli is sufficiently adequate. Since there are few monocular neurons beyond the primary visual cortex, our finding implies that extra-striate cortices are largely responsible for the visual inference in our task.
