Model-based analysis of excitatory lateral connections in the visual cortex

Excitatory lateral connections within the primary visual cortex are thought to link neurons with similar receptive field properties. Here we studied whether this rule can predict the distribution of excitatory connections in relation to cortical location and orientation preference in the cat visual cortex. To this end, we obtained orientation maps of areas 17 or 18 using optical imaging and injected anatomical tracers into these regions. The distribution of labeled axonal boutons originating from large populations of excitatory neurons was then analyzed and compared with that of individual pyramidal or spiny stellate cells. We demonstrate that the connection patterns of populations of nearby neurons can be reasonably predicted by Gaussian and von Mises distributions as a function of cortical location and orientation, respectively. The connections were best described by superposition of two components: a spatially extended, orientation-specific and a local, orientation-invariant component. We then fitted the same model to the connections of single cells. The composite pattern of nine excitatory neurons (obtained from seven different animals) was consistent with the assumptions of the model. However, model fits to single cell axonal connections were often poorer and their estimated spatial and orientation tuning functions were highly variable. We conclude that the intrinsic excitatory network is biased to similar cortical locations and orientations but it is composed of neurons showing significant deviations from the population connectivity rule.