Parallel frontoparietal-centered connectivity axes across cortex and cerebellum support cognitive control

Technical Summary: Parallel Frontoparietal-Centered Connectivity Axes Across Cortex and Cerebellum Support Cognitive Control

Problem Statement
Cognitive control operates across a hierarchy ranging from lower-level sensorimotor control to higher-level cognitive flexibility, exhibiting substantial inter-individual variability. While the frontoparietal network (FPN) is established as a central hub with widespread resting-state functional connectivity (RSFC) to both the cerebral cortex and the cerebellum, the relationship between individual variability along FPN-centered connectivity axes and distinct levels of this cognitive control hierarchy remains unclear. Specifically, it is unknown how variability in FPN-centered connectivity across the cortex and cerebellum maps onto different control levels, and whether cerebellar contributions to higher-level control are independent or mediated by lower-level sensorimotor processes.

Methodology
The study utilized resting-state fMRI (rs-fMRI) and behavioral data from the Human Connectome Project Young Adult (HCP-YA) cohort ($N=997$).

• Connectivity Axis Derivation: Researchers defined the FPN using the Cole–Anticevic Brain-wide Network Partition (CAB-NP). For each participant, mean time series were extracted from FPN regions and correlated with time series from every cortical vertex and cerebellar voxel. These individual connectivity maps were concatenated into participant-by-vertex/voxel matrices. Principal Component Analysis (PCA) was applied to extract the first principal component (PC1), defining the dominant low-dimensional axis of inter-individual variation for both the FPN–cortex and FPN–cerebellum systems.
• Behavioral Measures: Five tasks indexed a hierarchy of cognitive control: motor control (9-hole Pegboard), inhibitory control (Flanker test), working memory (List Sorting), cognitive flexibility (Dimensional Change Card Sort), and delay discounting.
• Statistical Analysis: Bivariate Pearson correlations assessed associations between individual axis scores and behavioral performance. Mediation analyses (using the lavaan package in R with 10,000 bootstrap resamples) tested whether motor control mediated the link between connectivity axes and higher-level control. A serial mediation model was employed to test the pathway: FPN–cortex axis $\rightarrow$ FPN–cerebellum axis $\rightarrow$ motor control $\rightarrow$ higher-level cognitive control.

Key Results

1. Spatial Organization: Both the FPN–cortex and FPN–cerebellum axes exhibited a unimodal-to-transmodal organization, spanning from primary sensorimotor regions to transmodal association areas. The FPN itself was located at the transmodal end of these axes, maximally distant from primary sensorimotor cortex. At the network level, corresponding cortical and cerebellar networks showed strong correlation ($r=0.96$) in their axis positions, indicating a shared low-dimensional organization.
2. Behavioral Associations: Individual variability along both axes was positively associated with motor control, cognitive flexibility, and delay discounting. Neither axis showed significant associations with inhibitory control or working memory.
3. Mediation by Motor Control:
   • FPN–Cortex Axis: The association with higher-level control (cognitive flexibility and delay discounting) was partially mediated by motor control.
   • FPN–Cerebellum Axis: The association with cognitive flexibility was fully mediated by motor control, whereas the association with delay discounting was only partially mediated.
4. Serial Mediation: A serial mediation model revealed that the FPN–cerebellum axis fully mediated the relationship between the FPN–cortex axis and motor control. Furthermore, the association between the FPN–cortex axis and cognitive flexibility was serially mediated by the FPN–cerebellum axis and subsequent motor control. In this serial pathway for cognitive flexibility, the direct effect of the FPN–cortex axis was non-significant, indicating full serial mediation. This serial effect was not significant for delay discounting.

Key Contributions

• Unified Framework: The study provides a shared low-dimensional framework for understanding how FPN-centered connectivity varies across both the cortex and cerebellum, demonstrating that these systems are organized along parallel unimodal-to-transmodal gradients.
• Hierarchical Mapping: It systematically maps individual variability along these connectivity axes to a behavioral hierarchy, identifying specific links to motor control, cognitive flexibility, and delay discounting, while noting the absence of links to inhibitory control and working memory.
• Mechanistic Insight: Through mediation analysis, the paper elucidates a specific statistical pathway where lower-level motor control acts as a bridge between distributed cortico-cerebellar connectivity and higher-level cognitive flexibility. It further suggests a sequential dependency where cortical FPN organization relates to flexibility via cerebellar organization and motor performance.

Significance and Claims
The authors claim that their findings offer a systems-level account of how individual variability in low-dimensional FPN-centered connectivity axes relates to hierarchical cognitive control. They propose that the FPN–cortex and FPN–cerebellum axes provide a "shared low-dimensional framework" for understanding these behaviors.

Specifically, the paper argues that:

• The cerebellum's contribution to higher-level cognitive flexibility is largely expressed through its support of lower-level motor control, consistent with views of the cerebellum performing domain-general computations (e.g., prediction, sequencing) that underpin cognitive processes.
• The cortical axis may reflect a combination of motor-control-related variance and additional higher-order processes (e.g., rule updating), given its partial mediation pattern.
• The dissociation between cognitive flexibility (fully mediated by motor control in the serial model) and delay discounting (not significantly serially mediated) suggests that value-based decisions involve valuation-specific processes beyond the motor-control-related cortico-cerebellar sequence.

The authors maintain a modest tone regarding causality, noting that mediation analyses do not establish temporal ordering or causal mechanisms, and that the observed effect sizes are modest population-level associations rather than strong individual predictive markers. They conclude that cognitive control should be understood through a nested, systems-level perspective where low-dimensional cortico-cerebellar connectivity organization is linked to behavioral variation across multiple control levels.