Abstract
We perceive the visual world not as a meaningless jumble of colors and shapes, but rather as organized recognizable structure. The obviousness of the distinction between these two alternatives makes it all the more surprising that we do not yet have any reliable EEG-based markers that can distinguish brain activity across the two conditions. The goal of this project is to identify the hitherto elusive neural correlates of perceptual organization. To this end, we recorded high-density (128-channel) event-related potentials (ERPs) while adult observers viewed sequences of images created using Random Image Structural Evolution (RISE - Sadr & Sinha, 2001). In these presentations, an image progressively “evolves” out of noise. For each sequence there is a point at which the observer experiences perceptual onset - i.e., the image becomes coherent and the object is recognized. Our strategy for uncovering neural markers of perceptual organization involved comparing the EEG signals that occur in response to images before versus after perceptual onset. Traditional ERP waveform analyses revealed that early visual components (occurring within the first 300 ms) over posterior regions of the scalp differentiate between images that occur early in the sequence (before perceptual onset) versus late in the sequence (after perceptual onset). Ongoing efforts are focused on validating this marker, disentangling the contribution of attention from perceptual organization, and identifying any additional markers via signal classification techniques drawn from the domain of machine learning. As we demonstrated at last year's meeting (Moulson et al., VSS 2008), the combination of component-based analysis and single-trial classification is a powerful approach that has the potential to provide significant insights into the time course of perceptual organization in the brain.