Working memory

Working memory (WM) enables the flexible representation of information over short intervals. It is established that WM performance can be enhanced by a retrospective cue presented during storage, yet the neural mechanisms responsible for this benefit are unclear. Here, we tested several explanations for retrospective cue benefits by quantifying changes in spatial WM representations reconstructed from alpha-band (8 - 12 Hz) EEG activity recorded from human participants before and after the presentation of a retrospective cue. This allowed us to track cue-related changes in WM representations with high temporal resolution. Our findings suggest that retrospective cues engage several different mechanisms such as recovery of information previously decreased to baseline after being cued as relevant and protecting the cued item from temporal decay to mitigate information loss during WM storage. Our EEG findings suggest that participants can supplement active memory traces with information from other memory stores. We next sought to better understand these additional store(s) by asking whether they are subject to the same temporal degradation seen in active memory representations during storage. We observed a significant increase in the quality of location representations following a retrocue, but the magnitude of this benefit was linearly and inversely related to the timing of the retrocue such that later cues yielded smaller increases.

Working memory is a mental workspace which utilizes short and long-term memory to maintain and manipulate information. It is crucial in enabling cognitive control and is largely controlled by interactions within and between frontal and parietal cortices. Recent work has identified visual nonspatial, spatial, and visuospatial working memory spectral characteristics of the local field potential through simultaneous recordings from various areas across the monkey frontoparietal network. However, the reports are minimal in number, and there is no clear narrative tying together the heterogenous functionality of the characteristics. Here, a new spectral model of monkey visual working memory is proposed to address these shortcomings. It highlights functional roles for low, mid, and high frequency bands. Next, the organization of structural connectivity which gives rise to these spectral characteristics is investigated. A new binary association matrix representing connections in the frontoparietal network is proposed. A graph theoretic analysis on the matrix found that a 3-node dynamical relaying M9 motif was a fundamental building block of the network. It is optimally structured for the synchrony found in the spectral model. The network was also found to have a small-world architecture, which confers the integration and specialization of function required by visual working memory. Afterwards, three hypotheses generated by the spectral model are tested on non-spatial data. The low and mid band hypotheses were supported by evidence, while the high band hypothesized activity was not observed. This adds credibility to the roles identified in the model for the low and mid band and identifies a need for further investigation of the high band role. Finally, opportunities to expand the spectral model, analyze the M9 motif, and further test the model are explored. In the future, the spectral model could evolve to apply its predictions to humans in the pursuit of treatments for neurological disorders.

The dissertation discusses various node identification techniques as well as their downstream effects on network characteristics using task-activated fMRI data from two working memory paradigms: a verbal n-back task and a visual n-back task. The three node identification techniques examined within this work include: a group-aggregated approach, a subject-specific approach, and a voxel wise approach. The first chapters highlight crucial differences between group-aggregated and subject-specific methods of isolating nodes prior to undirected functional connectivity analysis. Results show that the two techniques yield significantly different network interactions and local network characteristics, despite having their network nodes restricted to the same anatomical regions. Prior to the introduction of the third technique, a chapter is dedicated to explaining the differences between a priori approaches (like the previously introduced group-aggregated and subject-specific techniques) and no a priori approaches (like the voxel wise approach). The chapter also discusses two ways to aggregate signal for node representation within a network: using the signal from a single voxel or aggregating signal across a group of neighboring voxels. Subsequently, a chapter is dedicated to introducing a novel processing pipeline which uses a data driven voxel wise approach to identify network nodes. The novel pipeline defines nodes using spatial temporal features generated by a deep learning algorithm and is validated by an analysis showing that the isolated nodes are condition and subject specific. The dissertation concludes by summarizing the main takeaways from each of the three analyses as well as highlighting the advantages and disadvantages of each of the three node identification techniques.