Monkey Hippocampal Neurons Track Ongoing Experiences

Abstract

The hippocampus and surrounding structures of the medial temporal lobe are necessary for the formation of new episodic memories, that is, memories of experiences. It remains unclear, however, how neural activity within the hippocampus supports this mnemonic function. Decades of recordings from the rodent hippocampus have documented neural activity that tracks the animal’s location in space, though it has yet to be determined if or how such navigational correlates might contribute to memory. An increasing body of work suggests that the precise navigational readout seen in rodent work may not be conserved across species. Navigational correlates in the monkey, for example, can be seen under the right conditions but often either offer poor spatial resolution or are better described by other features of a behavioral task. Together with results showing coding outside the spatial domain in rodents and other species, these data suggest hippocampal neurons reliably respond as an animal progresses through an orderly task, suggesting that salient task events are sufficient to structure hippocampal activity. This raises the compelling possibility that the ongoing activity in the hippocampus functions to link salient events into a continuous, unitary episode.To characterize task-relevant activity in the monkey hippocampus, I trained monkeys to perform a variety of tasks in virtual reality while using chronic recording techniques to record from their hippocampus. In my primary experiment, monkeys performed a spatial delayed alternation task in virtual reality. I used a chronically implanted hyperdrive targeting the medial temporal lobe with 124 independently movable single wire electrodes to record neural activity from thousands of cells over hundreds of behavioral sessions. I find neurons with significantly reliable and informative about the animal’s position in virtual space. However, the majority of these spatially correlated neurons cluster around the key events in the maze, including the start zone, the choice point and the goals. In fact, the population activity in this task appears to be better described in terms of maze segments or task phase than it is in terms of spatial location. Some of these neurons maintain their task-specific responses irrespective of the visual cues defining the virtual environment, while other neurons “remap” when the monkey performs the same task in a visually distinct context. The stability of activity across mazes suggest that the hippocampus may abstract task structure across individual exemplars. In preliminary data from other virtual reality tasks, we find additional examples of task relevant information encoded even in the absence of explicit spatial coding. Taken together, these findings emphasize the important role that task structure plays in shaping hippocampal activity, and are consistent with the theory that activity in the hippocampus continuously tracks task-relevant events, joining them into a cohesive episode.