In natural vision, recognizing objects based on the retinal image is challenging and is often an ill-posed problem because a single image is compatible with multiple interpretations. Nevertheless, the primate brain has a remarkable ability to understand ambiguous scenes and solve difficult object recognition problems. Converging evidence suggests that this process, especially in challenging contexts—e.g., occlusion or low-visibility environments—is based on the integration of sensory information with prior knowledge built from experience. Our goal is to develop circuit diagrams at a cellular level that specify how inter-areal interactions support the integration of sensory signals related to the visual image with internal models that represent prior knowledge, thereby revealing the computations that underlie scene understanding, object recognition, and perceptual decision making in the primate brain. To achieve this goal, we have assembled a synergistic team of experts to bring together, (i) viral-based circuit tracing and optogenetic methods to identify connected neurons; (ii) multiphoton imaging and high-density electrode recordings to functionally characterize neurons and signaling motifs in the awake macaque monkey; (iii) behavioral manipulations and (iv) cutting-edge computational modeling to reveal how systems of connected neurons across brain regions interact and support complex perceptual processes. Our proposal includes four projects. In Project 1, PI Briggs will lead an effort to establish circuit tracing protocols to support dense, reliable, and long-term tracking of connected neurons in the macaque monkey. We will histologically compare lentivirus and AAVretro constructs in terms of their efficacy, toxicity, directional reliability, layering, and spread in labeling connected neurons, and we will test opto-tagging using high density neurophysiology. In Projects 2 & 3, PI Bair will lead the effort to implement multiphoton imaging in the awake monkey to identify projecting neurons in vivo during the simultaneous physiological characterization of 100s of neurons down to a depth of ~1 mm in cortex. In Project 4, PI Pasupathy will lead the effort to apply the viral methods and physiological characterization with high-density neuropixels probes and multiphoton imaging to study neurons in visual cortex (area V4), prefrontal cortex and the visual pulvinar as macaque monkeys perform shape detection in impoverished images. PI Wu will lead the effort to interpret the population dynamics in the context of communication subspace models and reveal how connected neurons in three brain regions underlie the multiplexing of sensory signals and prior knowledge to facilitate object detection and scene understanding.

We will develop a set of cutting-edge methods to: (i) reliably identify connected neurons in superficial and deep brain structures, (ii) visualize and physiologically characterize these neurons within a larger population using advanced scanning laser imaging and high density multielectrode recordings, and (iii) analyze the population dynamics in the context of an ethologically relevant behavioral task. Our findings will reveal how brain regions communicate and integrate sensory signals and prior knowledge to support complex perceptual behavior. Ultimately, our studies will bring us closer to devising strategies to alleviate and treat brain disorders of impaired object recognition resulting from dysfunctions in the occipito-temporal pathway (e.g. agnosia) and from diminished communication between visual and prefrontal cortex (e.g. autism).