Romanski Published in Neuron
Monday, January 11, 2021
"A View from the Top: Prefrontal Control of Object Recognition"
Object recognition occurs effortlessly. We easily name objects and recognize them as members of a class. We retrieve our knowledge about the properties of an object class and integrate that knowledge with goals to orchestrate our actions. Although object recognition seems automatic, easy, and rapid, the underlying computational problem that must be solved is daunting, and the neural mechanisms are correspondingly complex. The visual system must map many highly variable patterns of sensory input to one stable pattern of population activity at higher levels of the visual system constituting the neural code for the object’s identity. Despite changes in view, color, size, and orientation, the visual system recognizes the object. In neural terms, this means the system has found a “many-to-one” mapping, in the form of synaptic connections that allow many patterns of population activity at lower levels of the visual processing hierarchy (V1, for example) to propagate through sequential stages (V2 and above) in such a way that they converge onto, and recruit, a pattern of population activity at the top of the visual processing hierarchy, presumed to be inferotemporal cortex (IT), that corresponds to the object’s identity.Read More: Romanski Published in Neuron