Selecting visual information from complex, multi-object scenes for further analysis and action is necessary because of the imitations of the human visual system. Recent accounts of visual attention acknowledge that preferential processing may be afforded to information selected on the basis of its physical location (space- based) and/or on the basis of its shape or object properties (object- based). Our hypothesis is that both spatial and object attentional effects emerge in a unitary model from competitive and cooperative interactions between representations, and that these representations are significantly influenced by interactions with task demands. In this mechanistic account, then, attention which is defined as the selection of a small set of active units, arises as part of the process of scene interpretation in conjunction with fulfilling the demands of a particular task. We adopt two complementary experimental paradigms: an object cost paradigm that reveals the difficulty attending to two objects simultaneously, and a distractor paradigm that reflects the inability to ignore elements grouped together with relevant information. Simulations and experiments will address three main issues: (1) We will explore the nature of the representations derived from perceptual experience, how these representations transfer and generalize to novel stimuli and how they serve to complete patterns that are impoverished or degraded (including occlusion and amodal completion). We will also explore how varying sources of perceptual information combine and become grouped, and how grouping might bread down in neuropsychological patients. (2) The emergence of space- and object based effects. In simulation work, we will evaluate the differences between architectures with modular or distributed space and object representation, and assess whether they can account for a range of empirical data that reflect the simultaneous contribution of space- and object- processes. We will also examine how object representations modulate location based attention in patients with a spatial deficit and simulate the effect of a spatial bias in the ~lesioned~ network to provide comparisons between the network and the empirical data. (3) Modulation of visual selection by task demands. We will explore how task requirements interact with space and location based representations, how task knowledge is learned and how the visual representations are reshaped or reconfigured under the pressure of differing task demands. Moreover, we map out the time course of processing and evaluate whether the standard time course can be reversed by manipulating task complexity. Finally, the three-way interaction between space-based, object based and task based influences are explored in patients with brain damage to reveal the full interactivity of the selection process.
Showing the most recent 10 out of 49 publications
