Temporal dynamics of basal ganglia response and connectivity during verbal working memory

Catherine Chang, Sonia Crottaz-Herbette, Vinod MenonNeuroImage 2007 · DOI

TL;DR

This study investigates the role of the basal ganglia (BG) in verbal working memory (WM) by examining its neural activity and connectivity patterns across different stages of a WM task. The findings reveal that the BG, particularly the anterior caudate, are dynamically involved in encoding, maintaining, and responding to information, and that their connectivity with cortical regions shifts depending on the task phase and cognitive load. This work highlights the BG's crucial, yet often overlooked, role in complex cognitive functions beyond motor control.

What Problem Does This Paper Address?

Working memory (WM) is crucial for many cognitive functions, and while research has extensively explored the involvement of cortical regions like the prefrontal and parietal cortices, the contribution of subcortical structures, specifically the basal ganglia (BG), remains less understood. Existing evidence suggests BG involvement in WM, but the precise temporal dynamics and the specific component processes they support are unclear. Furthermore, previous studies have often focused on visuospatial WM, leaving verbal WM less explored in the context of BG function. This paper aims to bridge this gap by examining the temporal dynamics of BG responses and their effective connectivity with cortical networks during a verbal Sternberg WM task, differentiating between encoding, maintenance, and response phases, and comparing high-load versus low-load conditions.

Methods

The researchers employed event-related functional magnetic resonance imaging (fMRI) to capture the temporal dynamics of brain activity. Participants performed a modified Sternberg task designed to minimize confounds from eye movements and motor demands. The task involved encoding a set of numbers, maintaining them in memory, and then responding to a probe. To analyze the data, they used the General Linear Model (GLM) to model the hemodynamic response for each phase (encoding, maintenance, response) and condition (high-load vs. low-load).

A key methodological innovation was the use of a modified Sternberg task where stimuli were presented one at a time, reducing the need for saccadic eye movements that could confound BG activity. For connectivity analysis, they utilized psychophysiological interaction (PPI) analysis, a form of effective connectivity, to investigate how BG activity influences or is influenced by other brain regions in a load-dependent manner. Specifically, they used the left anterior caudate as a seed region due to its significant activation during the maintenance phase. Simulations were also conducted to validate the GLM's ability to detect phase-specific activations and to assess the impact of temporal spacing between events.

Datasets

The study utilized fMRI data collected from 14 healthy, right-handed adult participants (7 males, 7 females, ages 18-32). No new datasets or benchmarks were introduced; the study relied on data acquired from these participants during the described experimental paradigm.

Key Insights and Ideas

  • Dynamic BG involvement across WM phases: The BG are not just passively involved but actively participate in distinct ways during different stages of verbal WM.
  • Load-dependent BG activation: The intensity of BG activation scales with the cognitive load of the WM task, particularly during encoding and maintenance.
  • Anterior BG for cognition, posterior for motor: The observed BG activations were primarily in the rostral/associative parts of the caudate and putamen, aligning with their proposed roles in cognitive processing rather than purely motor control.
  • Shifting cortical connectivity: The BG's interaction with cortical regions is not static. The anterior caudate dynamically connects with different cortical networks (e.g., prefrontal, parietal, temporal) depending on whether information is being encoded, maintained, or responded to, and whether the task is high or low load.
  • Verbal WM specificity: The study's design, using verbal stimuli and controlling for motor confounds, strongly suggests that the observed BG involvement is specific to verbal, non-spatial WM.

Key Results

  • WM-load-dependent activation: Significant load-dependent activation was observed in the left anterior caudate, anterior putamen, and globus pallidus during both the encoding and maintenance phases. Right anterior caudate activation was seen during maintenance.
  • Response phase activation: During the response phase, BG were active but showed no significant difference in activation between high- and low-load conditions, suggesting a role in motor planning for the response itself rather than WM load.
  • Cortical connectivity: Effective connectivity analyses revealed increased load-dependent interaction of the left anterior caudate with:
  • Left posterior parietal cortex (PPC) during all three phases.
  • Visual association cortex (fusiform gyrus, inferior temporal gyrus) during encoding.
  • Ventrolateral prefrontal cortex (VLPFC) during encoding and maintenance.
  • Pre-supplementary motor area (pre-SMA) during maintenance and response.
  • Dorsolateral prefrontal cortex (DLPFC) and anterior cingulate cortex (ACC) during the response phase.
  • Behavioral performance: Participants performed well in both WM conditions, with accuracy not significantly differing. Reaction times were longer for the high-load condition, as expected.

Limitations and Future Work

The authors acknowledge that the GLM analysis of closely spaced events can be susceptible to multicollinearity. While they addressed this by modeling events with a 4s separation, they note that perfect temporal separation is not always achievable. They also mention that simulations were based on an ideal scenario without HRF mismatches, which could affect real-world detection power.

Future work could explore the precise mechanisms by which the BG influence thalamic disinhibition to facilitate reverberant activation in fronto-parietal loops. The study's findings also suggest that the BG act as a dynamic gating mechanism for controlling and updating WM, which could be further investigated. The authors also highlight the need for further research into the precise roles of VLPFC and DLPFC in WM and how their interactions with the BG evolve. The study's focus on verbal WM opens avenues for comparing these findings with spatial WM tasks to delineate domain-specific BG functions.

Summary updated

BibTeX
@article{chang2007temporal,
  author = {Catherine Chang and Sonia Crottaz-Herbette and Vinod Menon},
  title = {{Temporal dynamics of basal ganglia response and connectivity during verbal working memory}},
  journal = {NeuroImage},
  year = {2007},
  doi = {10.1016/j.neuroimage.2006.08.056},
}
Figure