Abstract
The development of direction selectivity (DS) in the carnivore primary visual cortex (V1) requires visual experience, although the mechanisms by which visually-driven activity sculpts the cortical circuits are not fully understood. A previous modelling study by Van Hooser et al. (2014) proposed a mechanism for the development of direction selectivity based on changes in the pattern of feedforward inputs from the lateral geniculate nucleus (LGN) onto V1 cells. Recent experimental findings have called into question some of the assumptions that went into the previous model. In that model, prior to visual experience, inputs to V1 cells came from LGN units with a broad range of spatial positions and response latencies. However, recent intracellular recordings from our lab suggest that visually naïve V1 cells receive more narrowly localized input. Additionally, the previous model allowed for all inputs across all positions and latencies to grow, implying that the final input structure could be determined by the properties of the visual training stimulus used. In contrast to this hypothesis, Ritter et al. (2017) found that speed tuning of actual V1 cells following visual training did not depend on the velocity of the training stimulus used. This implies that visual experience plays a permissive, and not instructive, role in the development of direction selectivity. Therefore, in a new model, we implemented a restriction allowing only specific inputs to grow. We also added an inhibitory interneuron that acquired similar tuning properties to the DS cell it connected to. Recent findings by Wilson et al. (2017) show that inhibitory interneurons in ferret V1 are direction-selective and have similar direction tuning to nearby cells, consistent with our new inhibitory paradigm. Thus, our model proposes new mechanisms for the development and sharpening of direction selectivity that agrees with the most recent data from live animal recordings.