Organization of spatiotemporal information and relational memory in the hippocampus

neuroscience;hippocampus;relational memory;memory;cognitive science;artificial intelligence;psychology;navigation;reconstruction;time travel

Horecka, Kevin Michael

关键词: neuroscience;    hippocampus;    relational memory;    memory;    cognitive science;    artificial intelligence;    psychology;    navigation;    reconstruction;    time travel;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/100899/HORECKA-DISSERTATION-2018.pdf?sequence=1&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

This work examines the role of the hippocampus and relational memory in organizing episodic memory during navigation and reconstruction. Navigation is a critical component in most organisms’ survival. Reconstruction, on the other hand, provides an incredibly rich method of evaluating the precise information remembered by an individual after attempting to learn and remember that information. Through validating the computational framework in this work on amnesic patients with hippocampal damage, an understanding of some of the specific types of relations which rely on the hippocampus can be established. Then, this framework can be applied to a much more complex, spatiotemporal navigation and reconstruction task in healthy individuals to gain a wider perspective on the organization of episodic memory, which is known to critically rely on the hippocampus.The first experiment and associated analysis framework presented in this document (Chapter 2) uses spatial reconstruction to establish that not all types of spatial relations are impaired in hippocampal damaged patients. In particular, the arbitrary, identity-location relations (i.e. those relationships where the element being bound could have just as easily been anything) are critically impaired in hippocampal damaged patients while location information, disregarding identity, is not. The use of reconstruction in this context allows for the establishment of a set of critical computational metrics which relate to hippocampal function in reconstruction which can then be applied to other reconstruction tasks in healthy individuals to learn more about the wider structure and organization of memory.In the second experiment (Chapters 3 and 4), the methodologies which were applied to hippocampal damaged patients in the first experiment are applied to a novel Spatiotemporal Navigation Task in healthy young adults. In this task, participants are not just asked to study and reconstruct items in space, but instead, participants are asked to, in Virtual Reality, navigate space and time (via normal movement and simulated Time Travel) and study, then reconstruct the locations of events in spacetime. The computational framework established in the previous chapter is then applied to show that relational memory errors in time are far more common in this task than in space, suggesting differences in representations between these two domains even when the navigation and exploration of the domains are put on a more equal footing. Additionally, in time, these relational memory errors are far more likely to occur within a shared contextual region than should occur by chance. In fact, this error (temporal relational memory error within a context) gets worse across the first 3 trials, suggesting a systematic bias due to context. Finally, a more traditional bias, the context boundary effect (i.e. a “squishing” of within context temporal locations and “stretching” of across context temporal locations) is observed even though participants are allowed to reexplore the contexts arbitrarily, multiple times. This suggests that the context boundaries are having a profound impact on both the distance judgements and relational memory structure associated with events in spacetime.Finally, in the fourth chapter, the navigation component of the previous Spatiotemporal Navigation Task is examined to determine if changes in study time navigation and exploration relate to changes in the various test metrics discussed in the previous chapter. More rapid improvements in spatial and temporal navigation are shown to relate to more rapid improvements in memory in those domains, separably, suggesting that spatial and temporal representations may in some way be separable in this task in both the relational representations and the navigation strategies supporting those representations. Relational memory improvements are shown to be uniquely tied to changes in navigation complexity and systematicity, pointing to an interplay between in-the-moment, memory-guided decision making and subsequent relational memory efficacy. Context boundaries are suggested to act as more of a discriminatory feature (at least in this task) than one used to strengthen within-context relational memory organization accuracy as there is a significant relationship between changes in context boundary crossing and both the context boundary effect and across-context temporal relational memory errors. Finally, a preference towards exploring an otherwise temporally-flexible environment in the implied, forward order with increasing contiguity is suggested to be a critical element in improving temporal, relational, and contextual memory organization.Taken together, this work shows the richness of spatiotemporal navigation and reconstruction in observing the complex interplay between navigation in space, navigation in time and how these ultimately may relate to navigation in memory. Through embracing principled approaches to analysis of behavioral data, and the inclusion of complex behavioral mechanics (such as simulated time travel), this work extends our understanding of the role of hippocampal relational memory and overall memory organization.