The Contribution of Alpha Oscillations to Working Memory Processing

Abstract

Working memory, which enables the temporary storage of information in an active “online” state, is an exceptionally capacity limited system. Given this capacity limit, irrelevant information in our environment must be filtered out, while relevant representation is maintained. Research has shown that neural oscillations in the alpha frequency range (8-12Hz) are greatly influenced by the number of items in memory. Most work has argued that alpha oscillations primarily support working memory processing by suppressing information that could interfere with items already in memory, as indexed by an increase in alpha power. However, other work has shown that decreases in alpha power, with little evidence of concurrent increases, support the maintenance of working memory representations.

In this thesis we show that, in the context of visual working memory, the primary role of alpha oscillations is to maintain distinct working memory representations, rather than to suppress irrelevant information. This is shown in a series of three experiments all indicating that as the number of relevant items increases, the power of alpha oscillations systematically decreases. In the first experiment, we use a whole report and change detection task to examine how the number of items in memory influences alpha oscillations. In the second experiment, we use a cuing (Experiment 2A) and filtering (Experiment 2B) paradigm to demonstrate that alpha power tracks the number of remembered items instead of the number of total items on the screen. Lastly, by presenting items sequentially (Experiment 3A) or in overlapping locations (Experiment 3B), we see evidence that decreases in alpha power are related to the maintenance of relevant spatial locations, instead of the number of items in memory. The results of the experiments suggest that alpha power reflects the maintenance of relevant working memory representations, rather than the suppression of irrelevant external distractors or the inhibition of task-irrelevant neural areas. Furthermore, our last experiment indicates that the alpha frequency band is especially sensitive to the maintenance of spatial information.