Electrophysiological correlates of incidentally learned expectations in human vision.

The human visual system is remarkably sensitive to environmental regularities, which can facilitate behavioral performance when sensory events conform to past experience. The point at which prior knowledge is integrated during visual perception is unclear, particularly for incidentally learned associations. One possibility is that expectation shapes neural activity prospectively, in an anticipatory fashion, allowing prior knowledge to affect the earliest stages of sensory processing. Alternatively, cognitive processes underlying object recognition and conflict detection may be necessary precursors, constraining effects to later stages of processing. Here we used electroencephalography (EEG) to uncover neural activity that distinguishes between visual stimuli that match prior exposure and those that deviate from it. Participants identified visual targets that were associated with possible target locations; each location was associated with a high-probability target and a low-probability target. Alongside a behavioral cost for stimuli that had occurred infrequently at a cued location compared with those that had occurred frequently, we observed a focal modulation of the evoked EEG response at 250 ms after target onset. Relative to likely targets, unlikely targets evoked an enhanced negativity at midline frontal electrodes, and individual differences in the magnitude of this effect were correlated with the response time difference between likely and unlikely targets. In contrast, the evoked response at the latency of the P1, a correlate of early sensory processing, was indistinguishable for likely and unlikely targets. Together, these results point to postperceptual processes as a key stage at which experience modulates visual processing. NEW & NOTEWORTHY We combined electroencephalography with an incidental learning paradigm to investigate whether prior knowledge of environmental regularities modulates visual processing at early or late stages of sensory analysis. Our results reveal that modulations of neural activity arising at midlevel processing stages predict behavioral costs for unexpected stimuli rather than effects at early stages of sensory encoding.