ERP time course of perceptual and post-perceptual mechanisms of spatial selection

Abstract

Event-related potentials (ERPs) were recorded from volunteers performing a task requiring simple judgements about the spatial location of a single target that could appear with equal probability to the left or right of fixation. A robust finding in the ERP literature is a dichotomy between attentional selection for spatial and non-spatial features. Visual spatial selection is manifest as a modulation of early components (P1, N1) that reveal exogenous processes, while non-spatial selection is revealed by the presence of longer latency endogenous components (N2). We present an analysis of several conditions that require different degrees of visual analysis to confirm the location of the single target, and show that spatial selection can be manifest at early (N1) or later (N2) stages. Observers identified the location of targets that were more salient (2D line drawings with abrupt onset) or less salient (2D line drawings without abrupt onset or 3D objects embedded in random-dot stereograms). We examined differences in amplitude, latency, and topography of early ERP components (P1, N1, P2, N2), and compared responses measured over the left and right hemispheres in response to left and right targets. The results support the hypothesis that the processes involved in spatial selection can be manifest at early or late stages, dependent on the quality of the incoming data. Moreover, the iterative process by which the percept is established benefits from a change in the visual input that is specific to the target.

Introduction

Many studies have examined event-related potentials (ERP) to reveal the time course of visual attention. The tasks used have primarily been spatial cueing tasks in which attended and unattended stimuli are compared, or visual search tasks in which the target is presented in a visual array with one or more distractors. It is fairly well accepted that the mechanisms of attention for selection of spatial location are distinct from the mechanisms for selection of other cues such as colour, form, orientation, spatial frequency, and feature conjunctions [15], [26]. A robust finding is that there are timing differences between spatial and non-spatial selection. Spatial selection appears to have precedence in processing over attentional selection based on non-spatial cues [11], [17]. One way to characterize these differences is that spatial cues benefit from early attentional selection mechanisms and are more influenced by bottom-up exogenous input, whereas non-spatial cues rely on later mechanisms and are linked with top-down endogenous processes.

Although there is some evidence to suggest that it is possible to reverse the order of selection (e.g., colour prior to location) [17], the mechanisms which determine early or late selection of spatial location per se are not yet clear. This paper examines the nature of early perceptual and attentional processing of object information under conditions that make the spatial location of the target easy or difficult to extract from the visual display.

In Experiment 1, observers made simple decisions about the spatial location of object parts embedded in random-dot autostereograms. Extracting information about an object embedded in a random-dot stereogram requires processing that is additional to that needed to recognize a 2D line drawing of the same object. Form perception in random-dot stereograms is not available in the 2D representation of the image, but is based on the correlation between the information presented to each eye. Thus, the cyclopean object contours are defined by retinal disparity, and corresponding monocular dot patterns must be fused before the depth object can be identified [20].

Recent single cell work from monkeys suggests that, although V1 neurons respond to stereoscopic surfaces related to receptive field location [4], [40], stereoscopic edges and edge orientation are explicitly represented in area V2 [40]. This is interesting in light of the evidence that V2 cells are also important in the response to illusory contours [30], for example, the contours of a Kanizsa triangle (but see [9]). Essentially, the contours of the cyclopean object embedded in a random-dot stereogram are illusory in the same sense that the contours of a Kanizsa triangle are illusory. In such cases, the second-order features depend on the configuration of first-order features. The first-order features of random-dot stereograms are the dots, and the second-order features are the contours of the object perceived once the correct correspondence has been attained between the dot patterns represented on the left and right retinae.

The results of Experiment 1 suggested that simple selection of spatial location is delayed in time when the stimulus is defined by second-order features. The results may provide insight into the interpretation of ERP components for which the generator sites are not easily localized. Note that cells in V2 are selective for complex shapes [13], and could be important to timing differences for object analysis.