Supplementary Figure 6: A model for bidirectional control of the speed of response generation and inhibition by BF neurons

(a) Simple decision-making processes, such as those measured by RT, are commonly modeled by an activity accumulation process of a hypothesized decision unit that is likely located in the fronto-basal-ganglia circuit. In response to the go signal, a behavioral response is initiated once the activity of the decision unit reaches a threshold. Given the same go signal, the activity accumulation slope is variable from trial to trial (middle panel), resulting in variable RTs (top panel). This activity accumulation process likely depends on BF bursting activity (bottom panel) because successful detection in a near-threshold auditory detection task is coupled with the presence of BF bursting, while failed detection is coupled with the absence of BF bursting²³. (b) Blue and green traces depict neuronal and behavioral responses to two go signals that differ only in the associated amount of reward. A stronger BF bursting responses to the go signal is coupled with a faster and more precise RT distribution, suggesting that the strength of BF bursting modulates the slope of activity accumulation in the decision unit²⁵. (c) In the SST, the stop signal elicits near complete BF neuronal inhibition (red trace, bottom panel), which is coupled with and slightly precedes rapid behavioral stopping and SSRT (top panel). We hypothesize that BF neuronal inhibition leads to rapid behavioral stopping by preventing further activity accumulation in trials that have not reached threshold (middle panel). The short delay (~10ms) between the onset of BF neuronal inhibition and SSRT requires that BF can modulate cortical activity within the same delay. Consistent with such a requirement, BF bursting neurons can rapidly modulate cortical activity within 5-10ms to generate an event-related potential²⁸. (d) When the stop signal is presented earlier (compared to panel (c)), BF neuronal inhibition also occurs earlier and prevents more trials from reaching the threshold, resulting in more successful stop trials. Our results also suggest that the BF bursting response and BF neuronal inhibition can be independently controlled by the go signal and the stop signal, respectively, such that SSRT is not affected by the delay between go and stop signals. Together, these results support the unified hypothesis that BF bursting neurons serve as a bidirectional gain modulation signal for the decision-making process.