Josh McDermott continues MIT's course 9.35, Perception, with a lecture on how the visual system resolves ambiguity in the images striking the retina. He introduces the generic viewpoint assumption, the brain's implicit bet that it is not viewing a scene from a rare, coincidental angle that would make separate objects look aligned or edges look connected. McDermott works through the visual illusions and geometric examples that expose this assumption at work, showing what happens when a scene actually is captured from an improbable vantage point. The lecture then turns to object recognition, laying out the computational problem of identifying objects despite changes in viewpoint, lighting, and pose, and previewing the perceptual and neural mechanisms that let the brain treat wildly different retinal images as the same object. Delivered as a standard lecture-hall talk with diagrams and example images, aimed at students already grounded in the course's earlier material on vision.