CogSci Colloquium talk: Brain mechanisms that mediate internal and external visual selection

Mode: Online | 26th April 2023 | 7:30 pm

Link: Click to join

Abstract

Our brain is faced with an abundance of information, both from the external world as well as from its own internal milieu. To guide behavior, the brain must “select”, at each moment in time, the most relevant information for prioritized processing and decision-making. How this selection occurs is the central question driving our research. In this talk, I will present two different studies from our lab that explore related facets of this question. First, I will describe a study that explores selection for eye movements in external visual space. Humans, and other foveate mammals, make rapid eye movements (saccades) to sample different parts of our world. Yet, even before the eyes move visual sensitivity improves at the selected target of eye movements, a well-known phenomenon called “presaccadic attention”. We discovered a surprising lack of presaccadic attention benefits in a common, everyday setting: change detection. We show that presaccadic attention, in fact, renders change detection challenging by biasing percepts toward the most recent stimulus presented at the saccade target location. With a Bayesian model we explain how such biases induced by presaccadic attention affect change detection. Second, I will describe a study that explores internal selection in working memory (WM) and a ubiquitous brain signature of attention, called alpha-band (8-12 Hz) oscillations. Often, we have to remember information that has vanished from our immediate environment, like a phone number. Such temporarily stored information in WM drifts toward stable representations, called “attractor” states. I will describe a novel link between alpha oscillations and attractor states in visual WM. Recent experiments in our lab, backed up by secondary analysis of data from other labs, reveals a strong link between alpha oscillations and fidelity of information, governed by attractor states in WM. A dynamical systems model synthesizes these observations and leads to several testable hypotheses. Understanding how the brain selects and maintains relevant information may have critical implications for predicting behavior in real-world settings that require efficiently navigating dynamic environments. This research could also pave the way toward behavioral and neural interventions to improve efficient selection and robust maintenance of relevant information in the brain.

Bionote

Sridharan Devarajan received his Bachelor’s and Master’s (Dual) degrees from the Indian Institute of Technology (IIT), Madras. He completed his Ph.D. and subsequent postdoctoral training at the Stanford University School of Medicine as a Stanford Graduate Fellow and a Dean’s Postdoctoral Fellow. For his Ph.D., he studied the neural basis of attention and executive control with functional neuroimaging (fMRI), electrophysiology, and computational modeling. He is now an Associate Professor at the Centre for Neuroscience and an Associate faculty of the Department of Computer Science and Automation at IISc, Bangalore. His lab studies the neural basis of human attention with a combination of experimental and computational approaches (https://cns.iisc.ac.in/sridhar/). Awards include a Wellcome Trust-DBT India Alliance Fellowship, SERB Early Career award, a Pratiksha Trust Young Investigator award, and a DST Swarna Jayanti Fellowship. He is also a research consultant at Google, where he develops AI models for healthcare applications.