Donnelly, G. , Bergmann, J. , Bennett, M. , Petro, L. & Muckli, L.
University of Glasgow
When a sensory input is weak or ambiguous, the brain uses surrounding contextual information to form predictions about this input to facilitate perception. We investigated how top-down predictions based on contextual cues influence the processing of degraded feedforward input in visual scenes.
The lower right quadrant (our ?Target? region) of natural scene images was presented at 8 low contrast levels within conditions where the remaining full-contrast quadrants either completed the ?Target? image region, contained a different image, or were absent. Subjects were asked to classify if the Target region image was natural or manmade (identification task), and if it was present or not (detection task).
Results show that identification accuracies are high when the image surrounding the Target region shows the same type of scene (natural or manmade); with the highest level of accuracy when the Target region and surround contain the same image. When the surrounding scene is absent and only the low contrast Target image region is present, subjects perform at chance level until the contrast is high enough for the Target image to be clearly visible. When the surround and Target region are inconsistent at the group level (natural vs manmade), subjects? accuracies drop to below chance, indicating a tendency to identify the Target region as the type of scene that the context implies. Interestingly, subjects? ability to detect whether a scene was present remained at chance level at the lowest contrast levels, despite identification remaining above chance. This suggests that identification might stay largely intact even in the absence of conscious perception. Increase in contrast resulted in increases in both identification and detection accuracy.
Our results show top-down predictions facilitate the processing of weak consistent feedforward information and interfere with that of inconsistent bottom-up signals, reinforcing the notion of predictive processing within the visual system.