Open access

Hit refresh : decoding the neural signatures of refreshing in human working memory with EEG

Master program titleMaîtrise universitaire en neurosciences
Defense date2023-01-27

Working memory is a limited-capacity system that holds information in short-term memory according to behavioral goals and demands. Current models propose several mechanisms in which humans maintain information; one of which is known as refreshing. Assumed to be used spontaneously, refreshing is also assumed to allow us to retain visual, verbal, and spatial memoranda by directing our attention to this information in memory – what we may understand as mental reactivations of information. Recent findings have sparked a theoretical debate concerning the existence of refreshing: some have found no evidence for refreshing, others propose alternative accounts for behavioral effects that have typically been interpreted as refreshing (i.e., verbal rehearsal). Moreover, it appears we have reached the limits of behavioral methods to investigate refreshing. In this master’s thesis, we have begun to address this issue by proposing an alternative way to study the occurrence of refreshing in visual working memory using machine learning, specifically, by applying multivariate pattern analysis (MVPA) on electroencephalographic (EEG) data. In Step 1, we sought to build a pattern classifier to decode delay-period EEG activity, one that would ultimately allow us to assess whether we can successfully track mental reactivations of distinct categorical representations during maintenance in a refreshing task. In Step 2, we sought to test this classifier to detect guided refreshing of categories prior to testing it on the spontaneous refreshing of categories in Step 3. To accomplish the first step, we designed a paradigm that would allow us to obtain the necessary data to obtain a pattern classifier, which included a two-phase task for training and testing of data, as has been used in the relevant literature. Next, we developed a pipeline that would allow us to classify EEG data, also based on previous decoding studies having used MVPA on neuroimaging data. Several of our planned attempts, however, did not result in successful decoding in Step 1. Consequently, rather than moving on to Steps 2 and 3, the remainder of this thesis focused on Step 1: exploring different analysis pipelines to decode categorical representations from delay-period activity, which has led us to arrive mainly at asking whether a mnemonic or perceptual classifier is more useful for our overall research question. Further, through this master’s project, we have been able to reflect on the challenges of applying decoding methods to experimental paradigms used in cognitive science. Nevertheless, these explorations into brain data may be used to detect guided and spontaneous refreshing for future research seeking to clarify our understanding of refreshing, and more broadly, the interplay between working memory and attention.

Citation (ISO format)
JEANNERET CUERVO, Stéphanie. Hit refresh : decoding the neural signatures of refreshing in human working memory with EEG. 2023.
Main files (1)
Master thesis
  • PID : unige:168875

Technical informations

Creation04/26/2023 7:36:33 AM
First validation05/19/2023 12:12:13 PM
Update time05/19/2023 12:12:13 PM
Status update05/19/2023 12:12:13 PM
Last indexation05/06/2024 3:55:00 PM
All rights reserved by Archive ouverte UNIGE and the University of GenevaunigeBlack