Measuring long-term memories at the feature level reveals mechanisms of interference resolution
Loading...
Date
2023-07-06
Authors
Drascher, Maxwell
Journal Title
Journal ISSN
Volume Title
Publisher
University of Oregon
Abstract
When memories share similar features, this can lead to interference, and ultimately forgetting. At the same time, many highly similar memories are remembered vividly for years to come. Understanding what causes interference and how it is overcome is key to understanding the vast human memory capacity. One unresolved challenge is that interference has primarily been studied with dichotomous measures of memory (“remembered”, “forgotten”). This limits our understanding because memories are not all-or-none, they are comprised of multiple features, each of which can be recalled with different levels of detail or bias. In order to investigate this issue, this dissertation focuses on the use of face stimuli. Faces are a unique class of stimuli for studying memory interference in that they are readily parameterizable and humans are experts at perceiving them. This means that they can be manipulated to be similar enough to cause interference, but subtle differences can also be stored and later probed from long-term memory. This dissertation develops a methodology to create synthetic faces that can be manipulated and probed along a set of perceptually-important feature dimensions. This development process included documenting face landmark positions, sorting faces based on perceived similarity, and collecting subjective ratings on a corpus of 1,148 face images. In a series of three experiments, I then applied this novel methodology to understand how memories change at the feature level when there is interference between highly similar memories. I found two memory changes that specifically occurred when there was interference between highly similar stimuli: (1) during recollection there was a bias to exaggerate the subtle differences and (2) distinguishing features were recalled with greater consistency. Critically, these memory changes were adaptive in that they were associated with less interference-related errors. Finally, in a separate fMRI experiment, I used the same corpus of faces and feature dimensions to reconstruct faces based on patterns of fMRI activity evoked while viewing them. I argue that this approach can be utilized in the future to measure neural representational changes during interference resolution. Together our findings provide important insights into how the memory system resolves interference between highly similar memories.
Description
Keywords
Episodic memory and recall, Faces, Forgetting, Interference, Memory distortion, Neural reconstruction