Non-linearity of spatial integration varies across layers of primary visual cortex

The receptive field (RF) of visual cortical neurons is highly dynamic and context-dependent, shaped by both the spatial and temporal properties of stimuli and the complex architecture of cortical circuits. While classical RF mapping through extracellular recordings reveals only the area triggering spiking responses, intracellular recordings reveal a much broader region of subthreshold synaptic input. We investigated how neurons in different cortical layers integrate visual input across space, with a focus on the linearity of spatial summation. Using intracellular recordings, we found that supragranular complex cells integrate input in a highly sublinear manner, in contrast to infragranular complex cells and simple cells, which exhibited near-linear summation. To understand the underlying mechanisms, we employed a large-scale recurrent spiking model of cat primary visual cortex (V1). Modeling results point to the differential patterning of long-range horizontal connections—particularly their targeting of excitatory versus inhibitory neurons—as a potential source of the observed layer-specific integration properties. These findings suggest that RFs emerge from interaction of feedforward, horizontal, and possibly feedback inputs, that are continuous in space, challenging the conventional notions of fixed spatial RF boundaries in early visual processing.