Understanding the computational function of the primary visual cortex (V1) in low-level perception has historically been relegated to a passive role, limited to the reception of feedforward inputs. The dominant view holds that the integration of complex spatial and temporal information would be mediated by higher cortical areas, through top-down processes. However, the results presented by Gerard-Mercier et al. (2016) challenge this perspective by showing that V1 has complex associative capabilities, already in the early stages of visual processing, even in the absence of conscious attention.
Using intracellular recordings in anesthetized cats, the authors quantitatively mapped intracortical horizontal connections and their synaptic impact. By systematically stimulating the silent (nonspiking surround) of receptive fields (RFs), they identified subthreshold responses spatially organized according to the preferred orientation of each neuron. Remarkably, these synaptic responses mirror the “perceptual association” described psychophysically in humans for the detection of collinear contours, pointing to the existence of a static synaptic association field (S-AF) intrinsic to V1. This emergent anisotropic structure supports that V1 already incorporates, in its local connectivity, a sensitivity to coherent spatial configurations, even when stimuli are presented in regions distant from the fovea and outside the classical firing field.
Another notable finding concerns dynamic processing. When the authors presented apparent motion (AM) stimuli in two-dimensional Gabor stimulus sequences, they observed facilitatory synaptic integration when the motion pattern was collinear and centripetal (i.e., convergent toward the center of the receptive field). This synaptic facilitation was maximal when the horizontal inputs, propagating slowly through unmyelinated connections in V1, arrived in phase with the feedforward drive. The optimal lag time was estimated to be around 16 ms, which is equivalent to the speed of saccadic movements (~300°/s), typical of natural visual exploration. Such nonlinear responses – with a supralinear character – reinforce the hypothesis that V1 operates as a spatiotemporal integrator sensitive to visual flow and contour configuration, something previously attributed to areas such as MT.
This property has been formalized as a dynamic association field (D-AF), whose topology changes transiently as a function of the direction of visual flow. Computationally, this suggests that V1 is capable of generating local predictions about global motion patterns based solely on its intracortical connectivity and synaptic properties, without the need for top-down modulation. It is particularly interesting to note that the dynamic association field was robust even in anesthetized animals without any behavioral involvement, suggesting that such mechanisms are endogenous to the functional architecture of V1.
Another relevant aspect is the quantification of intracortical horizontal propagation. The latencies of synaptic responses increased linearly with retinotopic distance, evidencing propagation velocities of 0.1–0.5 m/s, compatible with conduction by unmyelinated horizontal axons. These velocities, significantly slower than feedforward and feedback pathways, are sufficient to generate temporal windows of functional synaptic interaction, whose coincidence with feedforward inputs is critical for the observed facilitation.
The work of Gerard-Mercier et al. (2016) provides a solid empirical basis for revisiting the role of V1 in primary visual processing. The notion that the primary visual cortex is merely a hierarchical relay station needs to be reconsidered. The data point to an active role of V1 in constructing integrated perceptual representations of form and motion, providing evidence that the encoding of spatial and temporal regularities may occur early in the visual system.
From a computational neuroscience perspective, these findings provide a functional model that approximates the principles of V1 synaptic connectivity to the Gestalt laws of perception, especially those associated with collinearity, proximity, and common fate. It is plausible to propose that the static and dynamic association fields form a basic operational architecture for figure-ground segmentation and moving object detection.
In summary, this study represents a turning point in the understanding of the functional architecture of V1, by demonstrating that the spatiotemporal integration of visual stimuli occurs already in the first cortical synapses, mediated by dynamic properties of horizontal connections. This reinforces the need to include the intracortical mechanisms of V1 in models of visual perception, especially with regard to contour integration and global motion detection.
Reference:
GERARD-MERCIER, Florian et al. Synaptic correlates of low-level perception in V1. The Journal of Neuroscience, v. 36, n. 14, p. 3925–3942, 2016. DOI: 10.1523/JNEUROSCI.4492-15.2016. Available at: https://www.jneurosci.org/content/36/14/3925. Accessed on: June 17, 2025.
