On-line (Sign-up required): 45th CiNet Monthly Seminar: Gouki Okazawa “Context dependent geometry of the representation of decision variable in the parietal cortex”

CiNet Monthly Seminar

September 30, 2020
Apply for participation from here.
You will be notified of participation details by e-mail on Sept. 29.

“Context dependent geometry of the representation of decision variable in the parietal cortex”

Gouki Okazawa
Center for Neural Science
New York University

Host : Nobuhiro Hagura (Ikegaya G)

During perceptual decision making, neurons in multiple brain regions represent decision formation. For example, lateral intraparietal (LIP) neurons increase their average firing rate in response to sensory evidence supporting their preferred saccade target during a motion direction discrimination task. These findings have been explained using circuit models, in which neural ensembles encoding actions compete to form decisions.
Two key assumptions underlie these models: (1) decision variables (DVs) are represented as partially potentiated action plans, as ensembles increase their average responses for stronger evidence supporting their preferred actions. (2) DV representation and readout are implemented similarly for decisions with identical competing actions, irrespective of task context differences.
Here, we show experimental results challenging these core assumptions. In a novel face-discrimination task, we found that LIP firing rates decrease with stronger evidence supporting the preferred saccade target, contrary to conventional motion discrimination tasks. These opposite response patterns arise from similar mechanisms in which DV representation forms along curved population-response manifolds in neural state space. These manifolds are misaligned with action representations and rotate in state space depending on task context, necessitating distinct readouts. We show similar manifolds in lateral and medial prefrontal cortices, suggesting a ubiquitous representational geometry across decision-making circuits. Our findings invite a major update in existing computational frameworks of perceptual decision making.