Event:

Bridging the gap between neurophysiological and behavioral descriptions of a cognitive agent requires an understanding of how higher cognitive functions relate to neuronal response properties. Spatial memory constitutes a rare case of cognition for which we can draw on a rich catalog of behavioral and neuropsychological findings on one side and a vast literature of electrophysiological research and modeling studies on the other. Based on previous work (Byrne et al. Psych Rev 2007) we propose a model that gives a functional account of (human) spatial memory and imagery (scene construction, object-place memory, and mental navigation) in terms of single neuron responses, including place cells, head direction cells, and boundary vector cells. Using these cell types, the model shows how an egocentric representation of the local sensory environment (i.e. a specific point of view) can be transformed into a viewpoint-independent (allocentric) representation for long-term memory. The egocentric-allocentric transformation is assigned to a gain-field circuit in retrosplenial cortex. We include egocentric and allocentric representations which code for discrete objects within an environment and embed them in the spatial context of extended boundaries. Object representations are encoded into spatial memory during exploration of a familiar environment with the help of a rudimentary form of attention, and can populate the geometric context in imagery. We show how mental navigation (imagined movement in familiar environments) might occur with the help of grid cell circuitry. Driving the point of view in imagery with the help of grid cells also suggests a mechanistic interpretation of the phenomenon known as preplay.