Abstract
Neural field theory is used to analyze attention by extending an existing model of the large-scale activity in the corticothalamic system to incorporate local feedbacks that modulate the gains of neural connectivity as part of the response to incoming stimuli. Treatment of both activity changes and connectivity changes as part of a generalized response enables generalized linear transfer functions of the combined response to be derived. These are then analyzed and interpreted via control theory in terms of stimulus-driven changes in system resonances that were recently shown to implement data filtering and prediction of the inputs. Using simple visual stimuli as a test case, it is shown that the gain response can implement attention by evaluating two main features of the stimuli: the magnitude and the rate of change, by increasing the weight placed on the rate of change in response to sudden changes, while reducing the contribution of stimuli value in tandem. These changes of filter parameters are shown to improve the prediction of the upcoming stimuli based on its recent time course. This outcome is analogous to controller-parameter tuning for performance enhancement in engineering control theory.
Highlights
Attention is implemented in brain by dynamically selecting specific neural activities for further evaluation, in which irrelevant sensory information is filtered out in favor of the behaviorally relevant
(ii) Once gain modulation was incorporated as part of the system response, the system transfer functions witch were shown to differ from fixed term transfer functions only in their now-dynamic gain coefficients
As in recent work (Babaie-Janvier and Robinson, 2018), these corresponded to slow-wave, alpha, and beta responses each of which can be interpreted as being a standard second-order PID filter, but the gain parameters of these filters can change dynamically in response to stimuli. (iv) Notably, it was shown that each resonance can be interpreted as a filter that responds to two features of the incoming stimulus in its resonant frequency range: the stimulus value and rate of change
Summary
Attention is implemented in brain by dynamically selecting specific neural activities for further evaluation, in which irrelevant sensory information is filtered out in favor of the behaviorally relevant. Luck et al (2000) studied the same phenomenon for single unit cells in macaque monkeys and concluded that when several simultaneous stimuli are present, a neuron’s sensory response to the attended signal is significantly enhanced. This amplification is analogous to improving the signal-to-noise ratio which enhances the extraction of pertinent features of the external stimulus (Morgan et al, 1996; Hillyard and Anllo-Vento, 1998; Hillyard et al, 1998). The examples of such sensory gain control mechanisms involved in visual attention, were measured via event-related potentials (ERPs) and cerebral blood flow
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call