Layer and rhythm specificity for predictive routing

Abstract

In predictive coding, experience generates predictions that attenuate the feeding forward of predicted stimuli while passing forward unpredicted "errors". Different models have different neural implementations of predictive coding. We recorded spikes and local field potentials from laminar electrodes in five cortical areas (V4, LIP, area 7A, FEF, and PFC) while monkeys performed a task that modulated visual stimulus predictability. Pre-stimulus predictions were associated with increased alpha/beta (8-30 Hz) power/coherence that fed back the cortical hierarchy primarily via deep-layer cortex. Unpredictable stimuli were associated with increases in spiking and in gamma-band (40-90 Hz) power/coherence that fed forward up the cortical hierarchy via superficial-layer cortex. Area 7A uniquely showed increases in high-beta (~22-28 Hz) power/coherence to unpredicted stimuli. These results suggest that predictive coding may be implemented via lower-frequency alpha/beta rhythms that "prepare" pathways processing predicted inputs by inhibiting feedforward gamma rhythms and associated spiking.