Multiple-module network mechanism for two alternative forced choice tasks

In a seminal work of Wang [1] it is described the neurobiological structure involved in DM during oculomotor tasks. He proposed a multiple-module network mechanism for two alternative forced choice tasks (Figure 3). 

Figure 3. Model of Wang (2008). Neuronal population in the cortex rules the decision. Visual stimuli are integrated by neurons in the cortex (Cxe). They projects both to the SC and to the caudate (CD) in the BG. CD inhibits neurons in the SNr, which project inhibitory synapses to the movement neurons in the SC. The  growth of the firing rates in the Cx neurons increases their inputs to the SC and CD. As these inputs cross a certain threshold level, the movement neurons in the SC start bursting spike trains which lead to the saccade output (e.g., toward the A or B targets).

A top-down command flows such that when the time integration of evidence makes a certain level of firing rate of the decision neurons reached, then an entirely response in the downstream neurons is triggered, whereby the command output (motor action) is produced.  For the oculomotor tasks the candidate would be the neurons in the frontal eye field (FEF) and superior colliculus (SC) (but these areas are also involved in the decision and response selection for saccadic movements).  In the model of Wang, a decision threshold, i.e., a boundary of firing rates in the decision neurons, is given by a threshold of the projected inputs for eliciting the burst of spikes in the downstream movement neurons. A question now is raised whether the decision threshold may change adaptively. Interestingly, even though Wang had defined the decision threshold as the minimum cortical firing needed to cause a burst spiking in the downstream SC neurons, he argued that the cortico-striatal pathway, rather than the cortico-collicular pathway, facilitates adaptive tuning of the decision threshold. The explanation for this deduction comes from the switch role of the BG. In fact, neurons in SNr normally fire tonically at a high rate, which results in the sustained inhibition to the SC movement neurons. In turn, the activation of CD neurons by the ramping firing rates of decision neurons in the cortex, determines the inhibition of SNr and consequently the bursting of SC which gets the action output. The role of the DA neurons in the circuit makes the decision threshold more sensitive to the tuning of the synaptic weights in the cortico-striatal pathways than to the changes in the cortico-collicular pathway [2]. Hence, according to Wang, the cortico-striatal synapses, whose plasticity is modulated by the dopamine, represents the neurobiological structure for the adaptive tuning of the decision threshold. This finding is coherent with the reinforcement learning model of BG, where the dopamine signal gates the transmission of cortical-striatal signals between the direct and the indirect pathway (see above models of basal ganglia in saccadic movement).

1. Wang, X.J. (2008). Decision making in recurrent neuronal circuits. Neuron 60: 215-234.
2. Reynolds, J.N.J., Hyland, B.I., Wickens, J.R. (2001). A cellular mechanism of reward-related learning. Nature 413: 67-70.