Representations based on neuronal interactions in motor cortex.

Publisher Summary This chapter describes that large groups of cortical neurons are simultaneously active when a stimulus is perceived or a motor act is planned and executed. All units that are recorded, using a immovable, chronically implanted array in motor cortex, modulated their activity during execution of a simple reaching movement attests to the idea that a very large number of neurons participate in the simplest of behaviors. Population decoding strategies, such as the population vector algorithm assume that neurons are noisy but independent encoders. These types of decoding algorithms demonstrates that pooling the noisy signals from many neurons can reduce the detrimental effects of noise and can predict the direction of movement quite reliably. By recording from multiple neurons simultaneously, neurons are not independent encoders of movement direction. They exhibit statistical dependencies on both time scale. By incorporating these correlations into statistical coding and decoding schemes, it shows that they are not necessarily detrimental to decoding movement direction but can improve predictive power.