Abstract
Visual attention is frequently described as a spotlight that focuses on one location, but the underlying neural mechanisms of such an attentional system are not well established. We have been developing a candidate architecture for an attentional control system that exhibits the tendency for attention to converge at one location, but is also consistent with findings of divided attention in specific cases, as well as the structural and neurophysiological properties of the human visual system. This convergent gradient field (CGF) model simulates paradigms in which attention converges sharply at one location to the detriment of other locations (e.g. attentional capture), but can also exhibit multiple, stable attractors, when the triggering cues appear simultaneously. The model provides a unifying explanation of several attention effects, including attentional capture, attentional cueing, and some aspects of the attentional blink. The model predicts that the first step in deploying attention is to reconstruct the location of a target that has been detected by the ventral visual stream. This process of localizing a target produces a transient burst of neural activity within an attention map that ends when attention has successfully locked-on to the target's location. This transient burst of neural activity may be the neural source of the N2pc component of the EEG. Accordingly, the model predicts that two sequential targets in the same location will produce an N2pc that is identical to that produced by a single target, which our experimental data have confirmed. The model also predicts that the offset of the N2pc marks the initiation of attentional selection within the visual field. This work places new constraints on our understanding of attention and suggests that the inhibitory connections that cause convergence of attention may have an important role in segmenting a continuous input into discrete packets of information.
Meeting abstract presented at VSS 2014