Poster A38, Saturday, March 24, 1:30–3:30 pm, Exhibit Hall C
Predicting task performance with multivariate pattern decoding using EEG oscillatory activity
Elaine Astrand1; 1Mälardalen University, Västerås, Sweden
Working Memory (WM) is central for goal-directed behavior. It is the ability to remember and use relevant information during a short period of time. Tightly coupled with attention, the two processes allow us to tackle many of the tasks that we face every day by filtering the information flow to select only relevant information to process and maintaining and manipulating that information in memory in order to produce an appropriate behavior. As task demands increase, inducing higher cognitive load, more mental resources are required for successful performance. This study seeks to extract a continuous measure from recorded brain activity that correlates to task performance during a dynamic computer game in which relevant and irrelevant distracters appear. ElectroEncephaloGram (EEG) oscillatory activity was recorded from healthy participants while they were engaged in different versions of a visual n-back task. We show that a decoder constructed from two discrete levels of WM load can generalize to WM load on a continuous scale that correlates to trial-by-trial task performance before action. Moreover, this measure allows to assess the impact that an upcoming distracter will have on attention and working memory processing during the task. These results constitute an important contribution towards a wide range of applications in the field of cognitive brain-machine interfaces, particularly towards developing rich neurofeedback techniques to train attention and working memory.
Topic Area: EXECUTIVE PROCESSES: Working memory