Theories of hippocampal function have consistently focused on this remarkable regions’ important role in spatial navigation and memory. Recent work in cognitive and systems neuroscience have suggested that the hippocampus might support planning, imagination, and navigation by forming “cognitive maps” that capture the abstract structure of physical spaces, tasks, and situations. However, a critical aspect of planning and navigation is that both involve simulating, and acting upon, a sequence of actions to reach a goal. Despite several decades of work investigating hippocampal function, the precise mechanism and neural basis of planning during navigation is still unclear. This dissertation investigated how goals impact activity patterns in the hippocampus during planning and navigation. In Part 1, we examined hippocampal activity patterns in humans, using a goal-directed navigation task, to examine how goal information is incorporated in the construction and execution of navigational plans. Interestingly, we found that hippocampal activity patterns were more similar when participants planned routes that led to the same goal. During navigation, we found that rather than simply representing the current location in space, or the immediate future (as predicted by many theories of hippocampal function), the hippocampus reactivated a key decision point along a route. Building on these findings, in Part 2, a biologically inspired neural network of the hippocampus was used to investigate two plausible theories of the mechanisms underlying planning in humans; chaining and chunking. To simulate chaining during planning, the network was trained to retrieve pairs of overlapping associations in sequence. Using this framework, the model performed quite poorly at moments of high overlap between other sequences. This was in contrast to a model that utilized chunking, which was provided with only the most relevant information for retrieving a navigational memory. Taken together, this dissertation extends prior work in planning and navigation by highlighting the importance of external inputs into the hippocampal circuit to provide both structure and goal-relevant information, which are essential components of human memory.

Optimal behavior in a competitive world requires the flexibility to adapt decision strategies based on recent outcomes. In the present study, we tested the hypothesis that this flexibility emerges through a reinforcement learning process, in which reward prediction errors are used dynamically to adjust representations of decision options. We recorded event-related brain potentials (ERPs) while subjects played a strategic economic game against a computer opponent to evaluate how neural responses to outcomes related to subsequent decision-making. Analyses of ERP data focused on the feedback-related negativity (FRN), an outcome-locked potential thought to reflect a neural prediction error signal. Consistent with predictions of a computational reinforcement learning model, we found that the magnitude of ERPs after losing to the computer opponent predicted whether subjects would change decision behavior on the subsequent trial. Furthermore, FRNs to decision outcomes were disproportionately larger over the motor cortex contralateral to the response hand that was used to make the decision. These findings provide novel evidence that humans engage a reinforcement learning process to adjust representations of competing decision options.

Memory retrieval is well known to modify retention of not only retrieved material, but also related, non-retrieved material. This dissertation consists of two manuscripts that investigated retrieval phenomena in the learning of foreign language vocabulary and science education material. The testing effect–the retention benefit of practicing retrieval compared to studying–and the pretesting effect–incorrectly guessing a target before learning compared to studying–demonstrate the advantages of retrieval-based learning, but no extant theories have accounted for both of these effects. In Chapter 1, we investigated an error-driven learning account whereby retrieval-based learning serves to “stress test” the memory system, allowing it to learn to better predict a target from a cue, whereas in studying, there is no opportunity for the system to form a prediction. We predicted that inserting a small temporal “gap” between a foreign language word and its English translation should enhance retention when compared to simultaneous, “no gap,” presentation. In four experiments (N = 287) we consistently observed that “gap” conditions benefitted retention compared to “no gap” conditions, which supports the error-driven learning account. We observed that a gap as short as 600 ms benefitted retention one day later, one minute later, and in a pure list design. In Chapter 2, we investigated the sequalae of retrieval practice for non-practiced educational science text material. Some evidence suggests that retrieval practice of main ideas would lead to greater retention of non-practiced information, whereas other evidence suggests that retrieval practice of peripheral information would lead to impaired retention of non-practiced information, when compared to a control group that did not practice retrieval. In two experiments (N = 360) we observed robust testing effects, but we did not observe robust differences in non-practiced material, suggesting that the kind of focal retrieval practice used here has focal effects on science material retention.