Cognitive & Neural Science

Overview

The Cognitive & Neural Science division takes a deliberately layered approach to the brain: from the biophysics of single synapses, through the dynamics of local circuits, to the behavior of organisms making decisions under uncertainty. CNS researchers hold that understanding cognition requires theory at every level, and that theory is only credible when it makes quantitative predictions about data.

The Computational Neuroscience Group builds mathematical models of neural circuit function — rate models, spiking network models, normative frameworks — and tests them against large-scale multi-electrode recordings. The Perception & Decision Lab studies how people and animals integrate noisy sensory signals, form beliefs and commit to choices, using psychophysics, neuroimaging and computational modeling to decompose these processes into their constituent mechanisms. The Neural Engineering Group designs and implants hardware for closed-loop brain-machine interfaces: flexible electrode arrays that conform to cortical geometry, real-time decoding algorithms that translate neural activity into device commands, and stimulation protocols that write information back into the brain.

CNS shares the Advanced Microscopy Centre for histological validation and uses the Meridian HPC Cluster for large-scale neural data analysis and model fitting. The division has clinical research agreements with two regional hospital networks; patient data is handled under strict ethics protocols and held in an air-gapped secure enclave on campus.

Research themes

• Population coding and neural geometry — how information is represented in the joint activity of many neurons, and how it transforms as it passes through circuits.
• Normative theories of perception — Bayesian and near-optimal accounts of multisensory integration, prior learning and confidence, tested against behavior and neural data.
• Decision under uncertainty — value-based choice, cognitive effort, metacognition and the neural mechanisms of second-order representations.
• Closed-loop neurostimulation — real-time decoding of motor and cognitive intent, feedback-driven stimulation and neural prosthetics for motor rehabilitation.
• Neural data science — dimensionality reduction, manifold inference, population dynamics and machine-learning tools for large-scale electrophysiology datasets.