Abstract
A commonly held view about neurons in early visual cortex is that they serve as localized feature detectors. Here, however, we demonstrate that the responses of neurons in early visual cortex are sensitive to global visual patterns. Using multiple methodologies–psychophysics, fMRI, and EEG–we measured neural responses to an oriented Gabor (“target”) embedded in various orientation patterns. Specifically, we varied whether a central target deviated from its context by changing distant orientations while leaving the immediately neighboring flankers unchanged. The results of psychophysical contrast adaptation and fMRI experiments show that a target that deviates from its context results in more neural activity compared to a target that is grouped into an alternating pattern. For example, the neural response to a vertically oriented target was greater when it deviated from the orientation of flankers (HHVHH) compared to when it was grouped into an alternating pattern (VHVHV). We then found that this pattern-sensitive response manifests in the earliest sensory component of the event-related potential to the target. Finally, in a forced-choice classification task of “noise” stimuli, perceptions are biased to “see” an orientation that deviates from its context. Our results show that neurons in early visual cortex are sensitive to large-scale global patterns in images in a way that is more sophisticated than localized feature detection. Our results showing a reduced neural response to statistical redundancies in images is not only optimal from an information theory perspective but also takes into account known energy constraints in neural processing.