Identifying causal neural oscillations underlying working memory.

Abstract

Electroencephalography is instrumental in understanding neurophysiological mechanisms underlying working memory. While numerous studies have associated electroencephalography features to working memory, understanding causal relationships leads to better characterization of the neurophysiological mechanisms that are directly linked to working memory. Personalized causal modeling is a tool to discover these direct links between brain features and working memory performance. Therefore, we applied this approach to electroencephalography data from 66 adult healthy participants collected while performing a 3-back working memory task. Using graphical causal modeling, we discovered causal neural oscillations of working memory performance and compared the causal features between two groups: high and low performers. Total number of causal features in high performers was higher than low performers. Among the causal features, right temporal gamma oscillation was ~5 times (z-score = 3.87, P = 0.0001) more frequently a causal feature among high performers than low performers. However, the power of causal temporal gamma oscillation was not different between the two groups. Our findings suggest that one potential approach to improve working memory performance is to induce more causal gamma oscillations. This can be achieved by generating more local gamma entrainment over the right temporal cortex, rather than simply increasing gamma power.