Countermanding Perceptual Decision-Making

Abstract

We investigated whether a task requiring concurrent perceptual decision-making and response control can be performed concurrently, whether evidence accumulation and response control are accomplished by the same neurons, and whether perceptual decision-making and countermanding can be unified computationally. Based on neural recordings in a prefrontal area of macaque monkeys, we present behavioral, neural, and computational results demonstrating that perceptual decision-making of varying difficulty can be countermanded efficiently, that single prefrontal neurons instantiate both evidence accumulation and response control, and that an interactive race between stochastic GO evidence accumulators for each alternative and a distinct STOP accumulator fits countermanding choice behavior and replicates neural trajectories. Thus, perceptual decision-making and response control, previously regarded as distinct mechanisms, are actually aspects of a common neuro-computational mechanism supporting flexible behavior.

Keywords: Behavioral Neuroscience; Biological Sciences; Cognitive Neuroscience; Neuroscience; Sensory Neuroscience.

Graphical abstract

Figure 1

Perceptual Decision Countermanding Task (A) No-stop trials (top panel) began by fixating a central spot. After a variable delay, two targets appeared in the periphery. After another variable delay, a 10 × 10 checkerboard choice stimulus (magnified inset) appeared 3° above the fixation spot, while the fixation spot simultaneously disappeared. The checkerboard was composed of variable fractions of cyan and magenta squares. Fluid reward was delivered if monkeys shifted gaze to the target assigned to the majority color. On a minority of trials (stop-signal trials, bottom panel), the fixation spot reappeared after a variable stop-signal delay. Reward was delivered if monkeys canceled the planned saccade. (B) Performance of monkey Br during neural recordings. Upper left: Mean ± SEM choice probability as a function of color coherence for no-stop (black) and non-canceled (red) trials. Gray and red lines plot values predicted by best fit model. Upper right: Inhibition function averaged across all sessions. Solid line plots the inhibition function predicted by the best fit model. Lower left: Mean ± SEM of response time (RT) as a function of color coherence for correct choice no-stop (black) and non-canceled (red) trials. Non-canceled RT was systematically shorter than no-stop RT, justifying the application of the Logan and Cowan race model. Gray and red lines plot values predicted by the best fit model. Lower right: Mean ± SEM stop-signal reaction time (SSRT) derived from race model as a function of color coherence (black). SSRT did not vary with decision-making difficulty. Mean ± maximum and minimum SSRT predicted by the best fit model (gray).

Figure 2

Neural Mechanism of Perceptual Decision Countermanding Average discharge rates of neurons representing the perceptual decision variable and instantiating race model GO process. Discharge rate is plotted as a function of time relative to presentation of the choice stimulus (left), stop-signal (middle), and saccade (right). The thick lines mark times when stop-signal occurred across trials. Dashed lines mark corresponding SSRT across sessions. On trials with no stop-signal (thin), discharge rates accumulated faster with higher (black) relative to lower (gray) color coherence but reached the same trajectory immediately before saccade initiation. On successfully canceled stop-signal trials (thick), accumulating discharge rates were inhibited before SSRT.

Figure 3

Interactive Race Model of Countermanding Perceptual Decision-Making (A) Model architecture. Alternative decisions are committed when the accumulated activation of one of the GO units reaches a threshold. The accumulation of each GO unit was driven by the evidence supporting each alternative. This accumulation could be reduced by feedforward inhibition proportional to the evidence supporting the other alternative and by lateral inhibition from the other GO unit. Decisions are canceled if the activation of a STOP unit inhibits the GO units. This inhibition arises after a delay, which preserves the stochastic independence of the race finish times. (B) Comparison of model SSRT for each monkey (circles) with observed SSRT average (bars) and range of minimum and maximum values across sessions (error bars). (C) Representative trajectories of GO unit and STOP unit on trials with a stop-signal when the response was successfully canceled. (D) Observed (solid) and simulated (dashed) cancel times.