Time-course of cortical networks involved in working memory

Abstract

Working memory (WM) is one of the most studied cognitive constructs. Although many neuroimaging studies have identified brain networks involved in WM, the time course of these networks remains unclear. In this paper we use dense-array electroencephalography (dEEG) to capture neural signals during performance of a standard WM task, the n-back task, and a blend of principal components analysis and independent components analysis (PCA/ICA) to statistically identify networks of WM and their time courses. Results reveal a visual cortex centric network, that also includes the posterior cingulate cortex, that is active prior to stimulus onset and that appears to reflect anticipatory, attention-related processes. After stimulus onset, the ventromedial prefrontal cortex, lateral prefrontal prefrontal cortex, and temporal poles become associated with the prestimulus network. This second network appears to reflect executive control processes. Following activation of the second network, the cortices of the temporo-parietal junction with the temporal lobe structures seen in the first and second networks re-engage. This third network appears to reflect activity of the ventral attention network involved in control of attentional reorientation. The results point to important temporal features of network dynamics that integrate multiple subsystems of the ventral attention network with the default mode network in the performance of working memory tasks.

Keywords: attention; dense-array EEG; frontal lobe; temporal parietal junction; working memory.

FIGURE 1

Task schematic (see Stimuli and Experimental Design).

FIGURE 2

Behavioral data from all participants included in the EEG analyses. Accuracy (A) and response time (B) data. Note that response time data are from correctly detected match trials only.

FIGURE 3

Original grand-average ERP waveforms at FCz and Pz. Note that in this figure, a baseline correction (between -200 ms and stimulus onset) was applied, in contrast to all PCA/ICA components displayed in Figures 4–6.

FIGURE 4

Component (C1). (A) Component waveform. Green boxes represent the time windows used for statistical analysis. (B) Topographic map showing voltage distribution of C1. Orientation is top looking down with nose at the top. Positive voltage values are red and negative values are blue, with the zero crossing in white. Dots on the map are channel locations and white dot represents channel used for the waveform in (A). (C) Source results projected onto a schematic flat map (unfolded cortex) to show activity at all cortical sites. The area within the white border represents the lateral surface (outside of this is the medial wall); the left side is the left hemisphere. Activity was thresholded to show the top 10% of source activity. (D) This component waveform was re-computed with temporospatial analysis using a long prestimulus interval [see Component 1 (C1)]. (E) Topographic map showing voltage distribution, with the same topography as when computed with a shorter baseline. This component also showed a similar pattern of cortical source activity as the C1 in Figure 4C.

FIGURE 5

Component 1 (C1) waveforms with error trials.

FIGURE 6

Component 7 (C7). (A–C) are as in Figure 4.

FIGURE 7

Component 9 (C9). For each component, (A–C) are as in Figure 4.

FIGURE 8

Time course of each temporospatial component in relation to stimulus onset and mean reaction times for each experimental condition. Note that border of source maps are color coded to match waveforms.