Cortical direction selectivity increases from the input to the output layers of visual cortex

How is motion direction selectivity processed and relayed across the laminar structure of the early visual cortex? While it is well-established that direction selectivity exists in the visual system, debate has persisted over whether this selectivity is simply inherited from earlier stages or if it is refined and re-established through local cortical circuitry. This study investigates the transmission of direction-selective signals from input layers (L4 and L6) to output layers (L2/3 and L5) within the cat visual cortex.

Using multielectrode linear probes in anesthetized cats, the researchers recorded laminar neural responses to drifting grating stimuli. By ensuring the probes were perpendicular to the cortical surface, they could isolate activity within a single cortical column. They calculated the direction selectivity index (DSI) and response latency for each layer. To understand the underlying mechanisms, they developed computational models comparing untuned versus direction-tuned gain mechanisms and used Granger causality analysis to identify the contributions of feedforward and recurrent connections to signal transmission.

The researchers found that while direction selectivity in the output layers is strongly correlated with that in the input layers, it is significantly enhanced in the output layers. This enhancement is not a passive inheritance; rather, the output layers exhibit a direction-tuned gain mechanism. Specifically, neuronal responses to the preferred direction are selectively preserved, while responses to non-preferred directions and blank stimuli are reduced. Computational modeling revealed that this enhancement is driven by a combination of feedforward inputs from input layers and recurrent connections within the output layers, which together act as a direction-tuned gain filter that sharpens the neural representation of motion.