- 95 views
- 137 downloads
Accounting for Order and Mental Imagery within Mathematical Models of Association Memory
-
- Author / Creator
- Thomas, Jeremy J.
-
The general goal of this thesis was to uncover the computational characteristics of verbal association memory by focusing on two specific topics.
We first examined the role of mental imagery in association memory. One of the most effective ways to improve verbal association memory is to ask participants to form mental images of verbal stimuli. However, the functional role of mental imagery in cognition has been a subject of debate (Pearson & Kosslyn, 2015; Pylyshyn, 2002). An idea we test in the following work is if conscious mental imagery is an essential component of interactive imagery instructions. We tested this idea in chapter 2 by examining whether individual differences in both mental imagery vividness, or mental imagery ability predicted the benefit due to imagery instruction. We also examined how imagery instructions benefited a sub-population of individuals who report little to no imagery experience at all (aphantasia). We found that individual differences in visual imagery vividness and skill did not co-vary with the effectiveness of interactive imagery, and self-identified aphantasics benefited equally from imagery instructions. These results suggest that the visual image is not necessary for interactive imagery effects, and opens the possibility for alternative explanations of this effect, such as interactive imagery leading participants to encode more pair-unique representations of items.
Next, we examined memory for the constituent order of associations (AB versus BA). Existing mathematical models of association memory predict that associations are remembered with perfect order, or no order at all. However, empirical data indicates memory for the constituent-order of associations is moderate (Rehani & Caplan, 2011; Kato & Caplan, 2017). To help resolve this challenge to models, we first tested the possibility that imagery instructions could improve memory for constituent-order, perhaps to the levels predicted by perfect-order models. In chapter 2, we found that imagery instructions did not improve the ability to judge constituent-order (AB versus BA), nor the moderate relationship between order memory and association memory. This result increased the need to modify the mathematical models themselves. In chapter 3, we attempted to address this by extending convolution-based models, which normally store associations with no order, to store order in four ways. We evaluated these extended models against several behavioural benchmarks. We found that these extensions could account for moderate performance on order judgments; however, only one out of four could solve the additional benchmark of double function lists. This latter result suggests that to account for the full set of benchmark data, one needs to adopt specific assumptions about how constituent-order is represented in memory.
In the final chapter, we discussed how we might synthesize both of the major topics examined, considering our finding that interactive imagery instructions could not improve order recognition performance. This finding indicates that any account of imagery effects, or order memory, must also explain the in-variance of order memory to imagery instructions, providing an additional constraint for models. If a model can satisfy this constraint, it would simultaneously inform our computational account of both imagery effects and memory for constituent-order. -
- Graduation date
- Fall 2022
-
- Type of Item
- Thesis
-
- Degree
- Master of Science
-
- License
- This thesis is made available by the University of Alberta Library with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.