Spatio-temporal dynamics of object recognition: A study of "closure" processes

Abstract

Humans recognize visual objects, often despite highly adverse viewing conditions (e.g. occlusion, poor lighting or novel orientations). The term "perceptual closure" has been used to refer to the neural processes responsible for "filling-in" missing information in the visual image under such conditions. In a series of high-density electrophysiological studies from our laboratory (Doniger et al., 2000, 2001, 2002), line drawings of common objects, which were systematically fragmented to varying degrees, were used to study the timing of the neural processes responsible for perceptual closure. Presently we use high-density electrophysiologic recordings using scalp electrodes in conjunction with source analysis and functional magnetic resonance imaging to study the closure phenomenon and its spatiotemporal dynamics. Our studies revealed a robust event-related potential component termed the NCL (for negativity associated with closure) that tracked the neural processes related to perceptual closure. This component was manifest as a relative negativity over bilateral occipito-temporal scalp and occurred in the 230-400 ms timeframe typically peaking at 290-300 ms. The scalp topography of this component suggested that it largely reflected neural activity from a group of object recognition areas within the so-called lateral occipital complex (LOC). Our investigations have shown that the critical generators of NCL reside in the regions of LOC; further the investigations indicate that the process of closure also involves other nodes of object recognition, involving feedback, feedforward processes and characterizes their interactions.;Additionally, to assess the relative contributions of the generators involved in closure, intracranial recordings were made using the same stimulus conditions that were used for the scalp recordings. Principal component analysis of the ERP response to each condition (unscrambled & scrambled) revealed well-defined functional components and the electrode clusters that loaded onto them. In addition, for the first time we were able to characterize the activity arising from the depth structures such as hippocampus in a closure task. The analysis revealed a robust differential activity in the 230-400 ms timeframe. These activities are related to the well-characterized scalp-recorded negativity associated with closure (the NCL component).