A topic in the Open Knowledge Graph — a free, open map of 15,290 topics and the order to learn them in.
The ventral visual stream (occipital → inferior temporal cortex) transforms low-level visual features into high-level object representations. Neurons become progressively more selective: V4 encodes colors and textures, IT cortex encodes object categories (faces, bodies, scenes) that are invariant to size, position, and viewing angle. Damage to inferior temporal cortex produces visual agnosia—inability to recognize objects despite intact sensation.
From your prerequisite on the visual processing pathway, you know that visual information from the retina travels through the LGN to primary visual cortex (V1), and that it then splits into two major processing streams. The dorsal stream heads toward parietal cortex and handles spatial location and action guidance — the "where/how" pathway. The ventral stream heads from V1 down through V2, V4, and into the inferior temporal (IT) cortex — and this is the "what" pathway, responsible for recognizing objects, faces, and scenes.
The ventral stream's architecture embodies a fundamental principle of feature hierarchies. V1 neurons respond to simple oriented edges and gratings — they are sensitive to local contrasts at specific positions in the visual field. Neurons in V2 and V3 respond to slightly more complex features like curves and corners. V4 neurons respond to color, texture, and moderate curvature — the building blocks of surfaces and shapes. By the time you reach inferotemporal (IT) cortex, neurons respond to complete objects: a specific face, a hand, a car, a tool. Critically, these IT neurons have achieved invariance: the same neuron fires to a face regardless of whether it's large or small, centered or off to one side, upright or slightly tilted. This invariance is computationally powerful — it means the brain can recognize your friend's face across vastly different viewing conditions using the same neural response.
The organization within IT cortex has striking spatial structure. Face-selective regions (fusiform face area, occipital face area) respond preferentially to faces. The parahippocampal place area responds to scenes and spatial layouts. The extrastriate body area responds to bodies and body parts. These are not perfectly modular — they overlap and interact — but they reveal that IT cortex is organized by visual category, with different object domains clustered in different anatomical zones. This organization likely emerges from the statistics of natural visual experience during development.
The clinical importance of the ventral stream is illustrated by visual agnosia: the inability to recognize objects despite intact basic vision. A patient with IT cortex damage can describe what they see — "there is a round, shiny, metal object" — but cannot identify it as a coin. They can copy a drawing accurately but not recognize what they've drawn. This dissociation — intact sensation, impaired recognition — is the signature of ventral stream damage. The most famous subtype is prosopagnosia, the selective inability to recognize faces, often produced by bilateral fusiform damage. These patients may recognize their own spouse only by voice or gait, unable to extract identity from the face that appears visually intact to them. The ventral stream, then, is not just about seeing — it is about knowing what you see.