Spectral Imprints of Working Memory for Everyday Associations in the Frontoparietal Network

Abstract

How does the human brain rapidly process incoming information in working memory? In growing divergence from a single-region focus on the prefrontal cortex (PFC), recent work argues for emphasis on how distributed neural networks are rapidly coordinated in support of this central neurocognitive function. Previously, we showed that working memory for everyday "what," "where," and "when" associations depends on multiplexed oscillatory systems, in which signals of different frequencies simultaneously link the PFC to parieto-occipital and medial temporal regions, pointing to a complex web of sub-second, bidirectional interactions. Here, we used direct brain recordings to delineate the frontoparietal oscillatory correlates of working memory with high spatiotemporal precision. Seven intracranial patients with electrodes simultaneously localized to prefrontal and parietal cortices performed a visuospatial working memory task that operationalizes the types of identity and spatiotemporal information we encounter every day. First, task-induced oscillations in the same delta-theta (2-7 Hz) and alpha-beta (9-24 Hz) frequency ranges previously identified using scalp electroencephalography (EEG) carried information about the contents of working memory. Second, maintenance was linked to directional connectivity from the parietal cortex to the PFC. However, presentation of the test prompt to cue identity, spatial, or temporal information changed delta-theta coordination from a unidirectional, parietal-led system to a bidirectional, frontoparietal system. Third, the processing of spatiotemporal information was more bidirectional in the delta-theta range than was the processing of identity information, where alpha-beta connectivity did not exhibit sensitivity to the contents of working memory. These findings implicate a bidirectional delta-theta mechanism for frontoparietal control over the contents of working memory.

Keywords: ECoG; directional connectivity; iEEG; oscillations; parietal cortex; prefrontal cortex; working memory.

Figure 1

Working memory task design and behavior. (A) Single-trial working memory task design. Following a 1-s pretrial fixation interval (−250 to −50 ms pretrial epoch used as baseline), subjects were directed to focus on either IDENTITY or RELATION information. Then, two common shapes were presented for 200 ms each in a specific spatiotemporal configuration (i.e., top/bottom spatial and first/second temporal positions). After a 900- or 1,150-ms jittered pre-cue fixation delay (“maintenance”), the test cue appeared (i.e., one word presented on screen for 800 ms), followed by a post-cue fixation delay of the same length (“processing”). Working memory was tested in a two-alternative forced choice test (0.5 chance rate). In the identity test (top), subjects indicated whether the pair was the SAME pair they just studied (correct response in this example: no). In the spatiotemporal relation test (bottom), subjects indicated which shape fit the TOP/BOTTOM spatial or FIRST/SECOND temporal relation cue (correct response for cue TOP or SECOND: circle). (B) Mean and per-subject task accuracy (top) and response time (RTs; bottom). *Significant result; error bars, SEM; S, subject; orange, spatial trials; blue, identity trials; pink, temporal trials.

Figure 2

Spatial processing power effects. (A) Reconstruction of frontoparietal electrode coverage for all subjects. Electrodes are color-coded according to the results of the per-subject contrast between identity and spatial trials: orange, spatial > identity; blue, identity > spatial; black, no effect. All effects are significant at the cluster-corrected threshold of 0.05. (B) Mean and per-subject time-frequency representations of significant t-statistics (cluster-corrected mask applied) for the electrodes indicated in (A). The black lines to the right of the means indicate the spectral density of the corresponding time-frequency representations (i.e., mean significant t-values per frequency). Frequencies range from 2 Hz to 40 Hz in linearly-spaced steps of 1 Hz. T-statistics are normalized according to the maximum value of each plot (scale: −1 to +1) and color-coded by the direction of effects: cooler, identity > spatial; warmer, spatial > identity. S, subject.

Figure 3

Temporal processing power effects. (A) Reconstruction of frontoparietal electrode coverage for all subjects. Electrodes are color-coded according to the results of the per-subject contrast between identity and temporal trials: pink, temporal > identity; blue, identity > temporal; black, no effect. All effects are significant at the cluster-corrected threshold of 0.05. (B) Mean and per-subject time-frequency representations of significant t-statistics (cluster-corrected mask applied) for the electrodes indicated in (A). The black lines to the right of the means indicate the spectral density of the corresponding time-frequency representations (i.e., mean significant t-values per frequency). Frequencies range from 2 Hz to 40 Hz in linearly-spaced steps of 1 Hz. T-statistics are normalized according to the maximum value of each plot (scale: −1 to +1) and color-coded by the direction of effects: cooler, identity > temporal; warmer, temporal > identity. S, subject.

Figure 4

Bidirectional frontoparietal oscillations for information processing. (A) Delta-theta (2–7 Hz) phase slope index (PSI) shifted from a unidirectional, parietal-led network during maintenance to a bidirectional, frontoparietal network during the processing interval (p < 8 × 10⁻³⁰⁶). Data are represented as mean ± SEM per subject across all trials; positive values indicate that the prefrontal cortex (PFC) leads the parietal cortex and negative values indicate that the parietal cortex leads the PFC. S, subject. (B) Topographical representations of the PSI data depicted in (A) in two subjects. PSI is masked per electrode pair, with significant PFC leads in the top row (z > 1.96, p < 0.05) and parietal leads (z < −1.96, p < 0.05) in the bottom row. S3 was implanted subdurally in the right hemisphere and S5 in the left hemisphere. (C) Equivalent to (A): alpha-beta (9–24 Hz) PSI shifted from a unidirectional, parietal-led network during maintenance to a bidirectional, but still net parietal-led network during the processing interval (p = 0). (D) Equivalent to (B): topographical representations of the PSI data depicted in (C) in two subjects.

Figure 5

Spatiotemporal processing connectivity effects. (A) Grand mean time-resolved delta-theta PSI over the processing interval for the contrast between identity and spatial (left)/temporal (right) conditions. Data are represented as mean ± SEM across subjects; positive values indicate that the PFC leads the parietal cortex and negative values indicate that the parietal cortex leads the PFC. Black marks indicate the timepoints of significant condition effects at the false discovery rate (FDR)-corrected threshold of 0.05. Shaded epochs represent effects that were sustained for >100 ms, color-coded by the direction of effects: cooler, identity > spatial/temporal; warmer, spatial/temporal > identity. Orange, spatial trials; blue, identity trials; pink, temporal trials. (B) Equivalent to (A): alpha-beta PSI did not show sustained effects for one condition or the other in either contrast.