Inhibitory control by an integral feedback signal in prefrontal cortex: a model of discrimination between sequential stimuli.

The prefrontal cortex (PFC) is known to be critical for inhibitory control of behavior, but the underlying mechanisms are unclear. Here, we propose that inhibitory control can be instantiated by an integral signal derived from working memory, another key function of the PFC. Specifically, we assume that an integrator converts excitatory input into a graded mnemonic activity that provides an inhibitory signal (integral feedback control) to upstream afferent neurons. We demonstrate this scenario in a neuronal-network model for a temporal discrimination task. The task requires the working memory of the vibrational frequency (f1) of an initial stimulus (stimulus 1), followed by comparison of the frequency (f2) of a second stimulus (stimulus 2) with the stored f1 and a binary decision (f2 > f1 or f2 < f1). The integral feedback signal generated by stimulus 1 gates the later inputs based on the amplitude difference (f2 - f1). The feedback control signal enables a subset of neurons to reverse their tuning to f1 between stimulus 1 and stimulus 2, when they become tuned to the difference, f2 - f1. These neurons maintain a lower firing rate during the delay compared with their peak rate during stimulus 1. A second subset of neurons, tuned to f1 during the delay, reaches a rate during stimulus 2 that depends on the maximum of f1 and f2. Our work suggests a circuit mechanism for discrimination across time and predicts neuronal behavior that can be tested experimentally.