Abstract
The primary visual cortex is considered to represent visual information efficiently, while aiming to reduced redundancy in the sensory code. These functions suggest attenuation of correlated signals to increase statistical independence of component responses. Here we describe spatial interactions in human V1, which show a non-linear spatial summation of BOLD signals. Nine subjects participated in visual fMRI experiment. First we compared checkerboard stimulation of a single visual field position centred in upper quadrant at 4.8° eccentricity to parallel stimulation of nine positions at and around the centre position. The multiple regions were stimulated with multifocal fMRI paradigm using orthogonal timing. The non-linearity of the compound responses was successfully modelled using a two-phase process, in which linear spatial interaction is followed by a non-linear saturation of the response. One open question is whether the non-linearity is due to non-specific saturation of BOLD or neural response, or due to efficiency principle in the sensory coding. After determining the strongest BOLD responses, we will repeat the experiment with low contrast stimuli. Our model predicts that non-specific saturation would become linear at low contrast, because the response maximum is not reached. If non-linearity persists in the low contrast condition, it can be due to a functional role. Finally, we are going to decorrelate stimuli in the neighbouring visual field positions, which should enhance the centre response, if saturation aims to reduce redundancy in neighbouring visual field positions. This would be in line with Barlow's principle of efficient coding in human V1.
This work has been supported by Finnish Academy grant N:o 105628 and Finnish Cultural Foundation