Scene perception in early vision: Figure-ground organization in the lateral geniculate nucleus

The visual image that we perceive is initially analyzed in the retina, the lateral geniculate nucleus of the thalamus (LGN), and then in the primary visual cortex (V1). Neurons in these regions have small receptive fields (RFs). Initially, they respond to image elements within a small region of the visual field. This local analysis contrasts drastically with our perception, which depends only partially on local visual information from a small patch of the visual scene. Perception is dramatically affected by the global interpretation of the scene (1, 2): for example, on how we segment the scene into recognizable objects (“figures”) and background. Perceptually, the regions that belong to figures undergo enhanced processing, and they may even appear to be higher in contrast (3). Image regions that are assigned to figures also provide input for the cortical stages that contribute to shape recognition, unlike the regions that are assigned to the background and remain unrecognized (4). Thus, the final perception of an image patch depends on an intricate interplay between the retinal input and the contextual information present in the scene. Previously, neuroscientists focused on the visual cortex of awake, behaving monkeys to study figure-ground segregation. These studies revealed that the neuronal activity elicited by a figure in V1 is stronger than that elicited by the background, even if precisely the same image elements fall into the neurons’ RFs (5⇓–7). This enhancement of responses evoked by figures is known as figure-ground modulation (FGM) and has a characteristic temporal profile. The initial visually driven activity depends only on the image elements inside the RF. FGM takes more time to develop, presumably because it requires a loop through higher visual areas that then provide feedback to earlier cortical regions to boost the figure response and to suppress activity at …