Neural mechanisms of context-driven conscious visual perception

There is an extensive neural puzzle to be solved between the moment that patterns of light first excite the photoreceptors in our retinas and the moment that we become aware of a visual scene. The effortlessness with which the brain usually solves this puzzle indicates that there must be an elaborate functional organization underlying conscious visual perception. Since every single neuron in early visual cortex only receives information from limited portions of space, and the construction of coherent global percepts requires the spatiotemporal integration of information from many neurons. The brain thus uses context to infer the structure of the global scene and create conscious visual experiences. A better understanding of the neural mechanisms by which the brain uses context to create conscious perceptual experience is crucial for the treatment of visual disorders and the development of efficient interfaces for interaction with visually oriented organisms such as humans and other primates. In this thesis, we present experiments that investigate the neural mechanisms of such context-driven conscious visual perception. We used ambiguous visual stimuli that have multiple, mutually exclusive perceptual interpretations when they are presented in isolation. However, if context is added, this ambiguity may be resolved and perception stabilizes into one perceptual interpretation. Combining psychophysical and neurophysiological experiments with computational modeling studies and knowledge of the visual system’s functional anatomy, we demonstrate several mechanisms of temporal, spatial, attentional and crossmodal contextual facilitation of local ambiguity resolution. By temporarily removing an ambiguous stimulus from view for predefined intervals, we demonstrate that human visual perception stabilizes with long interruptions and destabilizes with short interruptions. Observers were able to influence this process with attention in a way that suggests early, low-level interactions between attention and a neural trace of recent perceptual history. In subsequent neurophysiological experiments, we recorded neuronal response patterns from a motion sensitive region of visual cortex in rhesus macaques, while they viewed intermittently presented ambiguous and unambiguous motion stimuli. These responses patterns showed stabilization effects that depended on the temporal characteristics of stimulus presentation in similar ways as the perceptual stabilization shown in humans. Furthermore, this neuronal response stabilization was accompanied by an increased power in the high gamma range of the local field potential, which suggests that local neural networks increase their coherence in order to encode repeatedly presented stimuli with higher signal-to-noise ratios. With respect to spatial context, we provide human behavioral and computational evidence for the idea that the well-known horizontal connections between cortical columns may distribute motion information over space in a way that depends on the depth location of the visual motion stimulus. This possibly provides an efficient mechanism to resolve partial occlusion situations. Other studies included in this thesis demonstrate that the computational operations with which the brain resolves isolated visual ambiguities are surprisingly generic, that the adult brain recalibrates its functional connectivity for binocular vision based an recent sensory experience, and that auditory and visual information are involuntary grouped in order to establish a sense of event duratio