14 Head Direction and Spatial View Cells in Primates, and Brain Mechanisms for Path Integration and Episodic Memory

The aims of this chapter are to show that there are head direction cells in primates as well as in rats; to describe their properties; to show that a new class of cells found in the primate hippocampus—spatial view cells—are different from head direction cells and from rat place cells; to show the utility of spatial view cells in forming episodic memories; to show how a single network can associate both discrete representations about objects and continuous spatial representations to form episodic memories; and to show how path integration may be performed in continuous attractor networks to update their spatial representations by idiothetic (self-motion) cues in the dark. Head direction cells are described in the presubiculum of the monkey, Macaca mulatta, used as a model of what is likely to be present in humans. The firing rate of these cells is a function of the head direction of the monkey, with a response that is typically 10 to 100 times larger to the optimal as compared to the opposite head direction. The mean half-amplitude width of the tuning of the cells was 76°. The response of head direction cells in the presubiculum was not influenced by where the monkey was located, there being the same tuning-to-head direction at different places in a room, and even outside the room. The response of these cells was also independent of the spatial view observed by the monkey, and also the position of the eyes in its head. The average information about head direction was 0.64 bits, about place was 0.10 bits, about spatial view was 0.27 bits, and about eye position was 0.04 bits. The cells maintained their tuning for periods of at least several minutes when the view details were obscured or the room darkened. This representation of head direction could be useful, together with the hippocampal spatial view cells and whole body motion cells found in primates in such spatial and memory functions as path integration and episodic memory. It is shown that discrete and continuous attractor networks can be combined so that they contain both object and spatial information, and thus provide a model of episodic memory. Self-organizing continuous