Abstract
Contour integration refers to the mid-level visual process in which spatially-separate edge information is grouped together to form object boundaries. Whether contour integration occurs automatically at an early stage of processing in the absence of attention or awareness is currently debated. Previous event-related potential (ERP) experiments have investigated a purported index of contour integration: a negative amplitude deflection over the posterior scalp from ~150-300ms elicited by contour-present versus contour-absent stimuli. Here, we report a series of recent experiments in which this same contrast was made while spatial attention, task relevance, and conscious perception were systematically manipulated. In all experiments, contour-present and contour-absent stimuli were formed by altering the orientation of a subset of lines within larger arrays of randomly oriented line segments. ERPs elicited by the orientation changes were first compared according to whether a contour was present or absent and the resulting ERP differences were then assessed according to attention, awareness, and task-relevance. In one experiment, spatial attention and task relevance were manipulated in a 2x2 crossed design. In another experiment, an inattentional blindness paradigm was employed to render the contours perceptually invisible to roughly half of the participants. Overall, results across a number of experiments converged to indicate that only the early phase of the contour-specific ERP difference reflects contour integration per se, while subsequent phases reflect perceptual, attentional, and task-related processes. The latency of the contour integration component is strongly modulated by the subjects' task, while its amplitude varies according to physical characteristics of the stimuli. Importantly, these experiments demonstrate that while contour integration requires spatial attention it can be carried-out nonconsciously during inattentional blindness and does not require task-based attentional selection.
Meeting abstract presented at VSS 2014