Purchase this article with an account.
Oliver Ratcliffe, Bernhard Staresina, Kim Shapiro; The Temporal Dynamics of Working Memory Maintenance in a Category-Based N-Back Task. Journal of Vision 2020;20(11):744. doi: https://doi.org/10.1167/jov.20.11.744.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
A popular theoretical framework suggests theta-gamma coupling facilitates the short-term retention of multiple items in working memory (WM). According to this model, items are represented by gamma bursts occurring sequentially during the upstate of ongoing theta waves. Whereas increasing evidence points to the involvement of these two frequency bands, the evidence does not directly reveal specific content being maintained within one theta cycle. To address this question, we employed multivariate pattern analyses (MVPA) to track reactivation of category-specific working memory representations over the course of an N-Back task.
Twenty-eight subjects (aged 18-35) participated in two EEG tasks. The first task was a delayed-match-to-sample (DMS) task using three stimulus categories: faces, objects, and scenes. The second, an N-Back task, employed two of the previous categories - objects and scenes - so that faces could serve as a neutral category for multivariate analyses.
Behavioural accuracy was high across tasks. Frontal theta power during the delay scaled with WM load. However, whereas power increased from the DMS to the 1-Back, adding an item from the 1-Back to the 2-Back instead elicited a significant shift in peak power to lower theta (4-5Hz). Decoding was successful across the trial period, although maximal during stimulus presentation. On the single trial level, classifier decision values showed significant rhythmicity in theta and beta bands. Homogenous 2-back trials, those where the participant maintained two same-category items, were successfully decoded during the delay period by training classifiers on stimulus presentation in the DMS. Applying this to heterogeneous 2-back trials revealed better decoding of the most recent category over the more behaviourally relevant one. Finally, using independent assessment of object and scene evidence, we show that both category representations are reactivated during the delay period and appear to coincide, thus supporting one formulation of the framework over the proposed alternative.
This PDF is available to Subscribers Only