Interactions Between Memories During Learning: A Behavioural Measure of Representational Plasticity
Loading...
Authors
Ali, Youssef
Date
Type
thesis
Language
eng
Keyword
memory , learning , hippocampus , representational change , plasticity
Alternative Title
Abstract
The capacity of human memory is enormous, but not unlimited. Because of the finite neural tissue in the hippocampus, memories are often encoded with overlapping patterns of activation, which can facilitate the cohesion of related memories, but also evoke competition between them. One theory of neuroplasticity, the non-monotonic plasticity hypothesis (NMPH), predicts that the initial degree of overlap or co-activation amongst memories may drive learning-related change; memories integrate at high co-activation, but due to competition, differentiate at moderate co-activation. As an alternate theory, recent evidence suggests a competitive differentiation (CD) mechanism whereby greater memory co-activation results in differentiation, not integration. We bridge these hypotheses by curating a stimulus set that samples a comprehensive range of visual similarity as a proxy for memory co-activation, and by using participants’ selections of incorrect memory lures as a novel behavioural index of integration and differentiation. Because they are theorized to differentially recruit trisynaptic and monosynaptic hippocampal pathways, respectively, we measured patterns of representational change in two major learning paradigms: episodic and statistical learning. We used a convolutional neural network to generate image pairs, which were embedded into a sequence where participants learned paired associates either episodically or statistically. Representational change was measured using a four-alternative forced choice task. Importantly, none of the options were the correct pairmate; rather, they were subtly shifted to be either more or less visually similar to the cue, signaling integration or differentiation respectively. Statistical learning followed an NMPH-consistent pattern, while episodic learning followed a more idiosyncratic pattern, where integration was attenuated at higher co-activation. When the experiment was repeated with a narrower similarity range, statistical learning followed a linear pattern that fits into the broader NMPH pattern, whereas episodic learning again showed integration attenuation at the end of the range, inconsistent with the broader pattern. This may suggest an additive influence of NMPH and CD at high co-activation. We also found that differentiation occurs at a lower level of co-activation in episodic compared to statistical learning. This work established a novel behavioural task for measuring subtle shifts in memory, and builds toward a general framework for learning-related representational change.
Description
Citation
Publisher
License
Queen's University's Thesis/Dissertation Non-Exclusive License for Deposit to QSpace and Library and Archives Canada
ProQuest PhD and Master's Theses International Dissemination Agreement
Intellectual Property Guidelines at Queen's University
Copying and Preserving Your Thesis
This publication is made available by the authority of the copyright owner solely for the purpose of private study and research and may not be copied or reproduced except as permitted by the copyright laws without written authority from the copyright owner.
Attribution 3.0 United States
ProQuest PhD and Master's Theses International Dissemination Agreement
Intellectual Property Guidelines at Queen's University
Copying and Preserving Your Thesis
This publication is made available by the authority of the copyright owner solely for the purpose of private study and research and may not be copied or reproduced except as permitted by the copyright laws without written authority from the copyright owner.
Attribution 3.0 United States