Changes in attractor dynamics predict altered perceptual decision making with dorsolateral prefrontal tDCS

The left dorsolateral prefrontal cortex (dlPFC) has been linked to the accumulation and comparison of perceptual evidence for decision making independent of sensory and response modalities. We investigated the possible neural dynamics underlying the role of dlPFC in perceptual decision making, through a combination of noninvasive neurostimulation in humans and computational modeling. First, we used an established and biophysically realistic model of a decision making network that employs competition between neural populations. Simulation of depolarizing noninvasive brain stimulation in this model decreased decision time, while hyperpolarizing stimulation increased it. This behavioral effect was caused by an increase in the rate of neural activity integration via recurrent connections, as well as changes in the susceptibility of the network to noisy background inputs which modulated population firing rate differences prior to the onset of the stimulus. These pre-stimulus differences biased the response to one or the other option, thus speeding or slowing decisions.\n\nWe then tested these model predictions in healthy participants performing a perceptual decision making task while receiving transcranial direct current stimulation (tDCS) over the left dlPFC, analogous to our simulated network stimulation. We found a striking match between model predictions and experimental results: depolarizing (inward) currents reduced and hyperpolarizing (outward) currents increased response times, but accuracy remained unaffected. Our results provide interventional evidence for the role of left dlPFC in perceptual decision making, and suggest that this region integrates and compares sensory evidence through competitive interactions between pyramidal cell populations which are selective for each response option. Mechanistically, our model suggests that stimulation of this region changes the rate at which evidence can be accumulated through recurrent activity and its susceptibility to background noise. More generally, our approach demonstrates that a linkage between computational modeling and noninvasive brain stimulation allows mechanistic accounts of brain function to be causally tested.