Wang TINS 2001 Wang Neuron 2002 An integrated microcircuit model of working memory and decision making.

Wang TINS 2001 Wang Neuron 2002

An integrated microcircuit model of working memory and decision making

Lorente de Nó’s reverberatory circuit

Roitman and Shadlen 2002

Roitman and Shadlen 2002

Reaction Time Task

• 2-population excitatory and inhibitory neurons (integrate-and-fire or conductance-based Hodgkin-Huxley neurons)

• Biologically realistic synaptic kinetics (AMPA, NMDA and GABAA)

• Structured network connectivity

MT output

Reaction Time Simulations

Model

Data

Data by J Roitman, J Ditterich and M Shadlen

Reaction time decreases with increasing coherence Weber's Law

Integrate-and-Decide (diffusion) Model

R Ratcliff (Psychol Rev 1978)J Schall (Nature Rev Neurosci 2001)Mazurek et al (Cereb Cortex 2003)

Kong-Fatt Wong

c’=6.4%

c’=51.2% c’=100%

D Munoz and R Wurtz

Fixation

Target

A bursty neuron in superior colliculus

But how is threshold-crossing readout by downstream neurons?

Bistability: hard threshold detection

Chung-Chuan Lo

Feedback Inhibition: Bursting

A Large-scale Network Model of Decision-Making

Cortex

Caudate

SNr

SC

LIP Network

SC Network

Threshold can be effectively modulated by the cortico-striatal synaptic pathway

Adjusting the threshold to optimize rewards: Speed-accuracy tradeoff

Optimal threshold should be adjustable according tothe distribution of coherence levels in the environment

End