Remembering What We've Seen: The Hippocampus and Relational Memory

Abstract

Relational memory is the ability to remember arbitrary associations between objects or events. These memories include things related by location, order, and context. Lesions of the medial temporal lobe (MTL) cause severe relational memory impairments suggesting that the MTL plays an important role in the formation of relational memories. Lesions to the hippocampus, the final output of the MTL, also causes relational memory impairments. Single unit recordings in rodents have identified place and time cells in the hippocampus that are hypothesized to support the spatial and temporal aspects of relational memories. Several studies have also identified neurons with spatial representations in the hippocampus of bats, monkeys, and humans, but it is unclear the extent to which spatial representations in non-rodent species may be sensitive to other parameters such as time and context, which are important components of relational memories. The goal of this thesis is to understand how the primate hippocampus could support the formation of relational memories during natural behavior. I first build the tools necessary to describe natural behavior, specifically focusing on behavior during the free viewing of complex images. I describe free viewing behavior using a foraging model and show that several factors including working memory, oculomotor constraints, and bottom-up salience influence viewing behavior. Additionally, scan paths are composed of short sequences of eye movements, and these sequences may aid in encoding relational memories. Furthermore, relational memory modulates these sequences: at first viewing behavior is well characterized by explorative-type behaviors, but then relational memory shifts viewing behavior towards exploitative-type behaviors. After understanding the behavioral correlates of relational memory, I recorded from neurons in the hippocampus of monkeys freely viewing images. Free viewing trials were interleaved with a directed eye movement task to understand how behavioral context modulates neurons in the hippocampus. Single unit recordings revealed representations of space, time, context, and stimulus novelty. In agreement with previous work, spatial representations were modulated by gaze position. Interestingly, each spatially modulated neuron fired at a consistent latency relative to fixation onset, and the response of the population of spatially modulated neurons fully tiled the duration of the fixation forming a spatiotemporal code that could link information across successive eye movements. Overall, my results highlight the importance of active sensing in organizing spatial representation in the hippocampus. Specifically, my results suggest that the hippocampus encodes information in spatiotemporal sequences reflecting the sequences of eye movements observed in the behavior. Certain patterns of eye movements may even help the hippocampus build relational memories. Finally, a review of the neuroanatomy suggests that there are many pathways in the prefrontal cortex that could explain the bi-directional relationship between relational memory and eye movements observed in this thesis.