Attention and Memory: Neural Relationships and Differences Using RSA

This research aims to explore a relatively new analysis methodology in the field of neuroscience called Representational Similarity Analysis (RSA). This methodology allows for the identification of relationships between brain representations of different natures, the generalization of results across multiple users, and even the comparison of electrophysiological results with behavioral data or computational model outcomes. Specifically, the goal is to uncover the underlying neural relationships between attentional and memory cognitive processes. A sample of 21 users had their brain electrical activity recorded during the performance of cognitive and memory tasks, which have been validated in the state of the art. These EEG signals were pre-processed to remove artifacts that could affect the study. Afterward, the signals were divided into frequency bands and segmented into seven time windows. For each of the 7 generated time windows, 38 power features and statistics moments were extracted to represent each EEG record. The results showed that the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\theta$</tex> and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\alpha$</tex> frequency bands have a high monotonic correlation between attentional and memory processes. Additionally, considering the general cognitive processes and focusing only on their neural activation power features, no relationships between the two processes were observed. This indicates that their brain activation patterns are independent of each other.