Purchase this article with an account.
Charles H. Anderson, Gregory C. DeAngelis, J. Anthony Movshon; Highly redundant population coding explains the representation of spatial frequency information in primary visual cortex. Journal of Vision 2006;6(13):41. doi: 10.1167/6.13.41.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Information is typically represented in sensory systems by populations of neurons, each of which is selective to a particular range of stimulus space. A fundamental question is what determines how many neurons are allocated to encode the behaviorally relevant information. To address this question, it is shown in this paper that the number of neurons that respond with the same preferred spatial frequency (SF) in V1 is proportional to the signal to noise ratio (SNR) at that SF as conveyed from the retina by the ganglion cells. This quantitatively demonstrates that both ganglion and V1 cortical neurons encode visual information with highly redundant population codes, which is consistent with the idea neurobiological systems achieve high SNR functionality and robustness by pooling over many low SNR neurons. These redundant codes are consistent with experiments demonstrating effects on perception through stimulating the cortex and underlie all neuroprothesis research, but they are not consistent with some of the prevalent theories of optimality in neuronal coding.
This PDF is available to Subscribers Only