We found three dynamic subclusters corresponding to each phase of the trial in the superficial layer, which might indicate separate WM control processes that occur on maintain WM content.

We additionally examined whether there are multivariate layer-specific differences across other phases of the working memory trial: e.g. encoding and retrieval. We saw that superficial layers had preferential decoding (of high vs. low WM load trials) during the retrieval phase.

Since superficial layers of the dlPFC have previously been found to sustain elevated activity for the manipulation of WM, we thought that higher load might also induce higher activity during the delay. And that’s what we found.

So, in this study participants completed a delayed-match-to-sample task while being scanned at 7T. We manipulated the number of items participants had to remember (working memory load) and the motor response and examined the response from superficial and deep layers of the dlPFC.