A ventral stream-prefrontal cortex processing cascade enables working memory gating dynamics

Abstract

The representation of incoming information, goals and the flexible processing of these are required for cognitive control. Efficient mechanisms are needed to decide when it is important that novel information enters working memory (WM) and when these WM ‘gates’ have to be closed. Compared to neural foundations of maintaining information in WM, considerably less is known about what neural mechanisms underlie the representational dynamics during WM gating. Using different EEG analysis methods, we trace the path of mental representations along the human cortex during WM gate opening and closing. We show temporally nested representational dynamics during WM gate opening and closing depending on multiple independent neural activity profiles. These activity profiles are attributable to a ventral stream-prefrontal cortex processing cascade. The representational dynamics start in the ventral stream during WM gate opening and WM gate closing before prefrontal cortical regions are modulated. A regional specific activity profile is shown within the prefrontal cortex depending on whether WM gates are opened or closed, matching overarching concepts of prefrontal cortex functions. The study closes an essential conceptual gap detailing the neural dynamics underlying how mental representations drive the WM gate to open or close to enable WM functions such as updating and maintenance.

Neurophysiological effective network connectivity supports a threshold-dependent management of dynamic working memory gating

“…Beyond motor control and sequence processing, 36 the SMA is essential in WM maintenance and updating 37 , 38 , 39 , 40 as well as cognitive flexibility (task switching) 41 and has previously been shown to be involved in WM gating. 10 , 12 The current finding that the SMA is only involved in gate closing is in line with the role of theta band activity in medial and superior frontal cortices to support cognitive control. 42 , 43 , 44 , 45 , 46 , 47 , 48 Cognitive control is central for shielding information stored in WM, and this is what WM gate closing is about.…”

“… 2 , 3 , 4 The closed state is assumed to reflect the default WM state, which allows the maintenance of representation through shielding WM from distraction. 5 , 6 So far, several studies have showed WM gate opening and closing effects and delineated some underlying neural principles, 7 , 8 , 9 , 10 , 11 , 12 which are triggered by useful or distracting perceptual information, respectively. 13 However, it is still unclear how the decision to open or close the WM gate is implemented on a computational neural level.…”

“…However, when considering the neurophysiological dynamics, it is relevant to consider that WM processes reflect a prime example of distributed neural processes. 12 , 28 It has been outlined that there is distributed, content-specific WM activity in brain regions such as the lateral prefrontal cortex and the posterior parietal cortex as well as in ventral visual stream regions, because of a division of labor. 28 Sensory regions likely encode low-level details of WM information, while more abstract information unrelated to a specific modality is processed in parietal and prefrontal regions.…”

Response stopping under conflict: The integrative role of representational dynamics associated with the insular cortex

“…The group‐ICA was employed to analyze EEG data and extract subcomponents, as done in previous work (Gholamipourbarogh et al, 2022; Gholamipourbarogh et al, 2023; Yu et al, 2022). The use of group‐ICA was motivated by the fact that traditional ICA is not directly applicable to group data, as it estimates different sets of components for each individual or run, making it challenging to draw inferences about the entire group.…”

“…We employed the MVPA light toolbox to conduct MVPA on the back‐projected matching component pairs obtained from the CORRMAP results (Yu et al, 2022). At the single‐subject level, two analyses were performed: a binary classification to identify time points with different patterns between IC pairs for congruent and incongruent trials and a temporal generalization analysis to examine the temporal dynamics of the representational content at the component ERP level.…”