Uncovering the Functional Differentiation of Frontoparietal Control Subnetworks
Adaptive behavior relies on a delicate balance between following specific rules that vary across situations and leveraging stable long-term knowledge gained from experience. The frontoparietal control network (FPCN) plays a crucial role in the brain’s ability to navigate this balance.
Here, we investigate how the topographical organization of the cortex supports the flexible deployment of cognitive control within the FPCN. The functional properties of this network might reflect its unique positioning between the dorsal attention network (DAN) and the default mode network (DMN) – two large-scale systems implicated in top-down attention and memory-guided cognition, respectively.
Our study tests the hypothesis that FPCN subnetworks are topographically proximal to the DAN and DMN, and how these differences in spatial positioning relate to functional distinctions. Specifically, the proximity of each FPCN subnetwork is anticipated to play a pivotal role in generating distinct cognitive modes relevant to working memory and long-term memory.
Through a comprehensive analysis of multiple anatomical and functional metrics, we demonstrate that FPCN subsystems share key similarities with their neighboring systems (DAN and DMN). Crucially, this topographical architecture supports divergent interaction patterns that give rise to different functional behaviors.
The FPCN acts as a unified system when long-term knowledge supports behavior, but becomes segregated into discrete subsystems with distinct interaction patterns when long-term memory is less relevant. This suggests that the topographical organization of the FPCN and its connections with distant cortical regions are important influences on how the system supports flexible cognition.
Mapping the Topographical Positioning of FPCN Subsystems
To establish the spatial positioning of FPCN subnetworks, we examined their location on the cortical surface using multiple metrics:
Distance from Sensory-Motor Regions
We calculated the geodesic distance between each FPCN parcel and key landmarks associated with primary visual, auditory, and somatomotor cortices. This revealed that FPCN-A is physically closer to sensory-motor regions, while FPCN-B is more proximal to the DMN (Figure 2A-B).
Anatomical Hierarchy
We used myelin content, a proxy for anatomical hierarchy, to show that FPCN-A has higher myelin values, aligning it more closely with sensory-motor regions, compared to the lower myelin values in FPCN-B (Figure 2C-D).
Functional Hierarchy
Analyzing the principal gradient of intrinsic connectivity, we found that FPCN-A is situated closer to the sensory-motor end of this hierarchy, while FPCN-B is positioned nearer the transmodal end (Figure 2G-H).
Evolutionary Expansion and Cross-Species Similarity
FPCN-A exhibited less cortical expansion from macaque to human and greater cross-species similarity compared to FPCN-B, further indicating its closer ties to sensory-motor processing (Figure 2I-L).
Together, these results confirm that FPCN-A and FPCN-B occupy distinct topographical positions on the cortical surface, with FPCN-A being more proximal to sensory-motor systems and DAN, while FPCN-B is situated closer to the DMN.
Functional Differentiation of FPCN Subsystems
To explore how these topographical differences shape the functional properties of FPCN subsystems, we employed a data-driven classification approach. This allowed us to assess the similarity between network parcels based on their fine-grained temporal features.
The classification analysis revealed that FPCN-A was most often misclassified as DAN-A, while FPCN-B was most likely to be misclassified as DMN-B (Figure 4). This pattern was consistent across task contexts, suggesting that FPCN-A and FPCN-B share greater functional similarity with DAN-A and DMN-B, respectively.
We further investigated the interaction patterns of FPCN-A and FPCN-B with their neighboring networks:
Interaction with DAN and DMN
FPCN-A showed stronger redundancy and functional connectivity with DAN-A, while FPCN-B exhibited greater redundancy and functional connectivity with DMN-B (Figure 5).
Task-Dependent Flexibility
The interaction patterns of FPCN-A varied across tasks, becoming more tightly coupled with DAN-A during working memory tasks and with FPCN-B during long-term memory tasks. In contrast, FPCN-B maintained a relatively stable pattern of interaction with both FPCN-A and DMN-B across task contexts (Figure 7).
Activation Profiles
The activation profiles of FPCN-A and DAN-A, as well as FPCN-B and DMN-B, showed similar modulation by task difficulty, with FPCN-A and DAN-A increasing activation in more challenging conditions, while FPCN-B and DMN-B typically deactivated (Figure 6).
These findings suggest that the topographical positioning of FPCN-A, situated between DAN and FPCN-B, allows it to dynamically shift its interaction patterns to support either external attention or internal memory retrieval, depending on task demands. In contrast, FPCN-B maintains more stable interactions with both FPCN-A and DMN-B, reflecting its role in integrating information from long-term knowledge to guide behavior.
Implications for Flexible Cognitive Control
Our study reveals how the brain’s topographical organization provides an architectural framework that enables flexible changes in neural function across situations. This landscape supports the deployment of distinct cognitive modes, with FPCN subsystems playing a crucial role.
FPCN-A and FPCN-B are positioned and anatomically similar to their adjacent systems (DAN and DMN, respectively), allowing them to interact in a context-specific manner. FPCN-A is more engaged with attention regions (particularly DAN-A) during working memory tasks, while it interacts more with FPCN-B during tasks relying on long-term memory. This flexibility challenges the assumption that regions would consistently resemble their most proximal cortical neighbors.
Despite these differences, FPCN-B shows a more stable pattern of interaction, maintaining connections with both FPCN-A and DMN-B across tasks. This suggests that FPCN-B may serve as a “memory control” network, supporting the integration of long-term knowledge to guide behavior.
In contrast, FPCN-A exhibits the characteristics of a highly flexible “multiple-demand” network, capable of dynamically shifting its interaction patterns to support executive control processes across domains. This architecture may allow FPCN to be influenced by both DAN and DMN at different times, even though these networks are sometimes anticorrelated.
Overall, our findings demonstrate how the brain’s topographical organization, in conjunction with underlying anatomy, provides a framework that enables the flexible deployment of distinct cognitive modes. This landscape allows FPCN subsystems to resolve competition and maintain a functional balance between goal-oriented attentional mechanisms and the retrieval of information from long-term memory.
Future research should explore whether similar topographically organized processes support functional flexibility in other large-scale networks, as well as the extent to which these mechanisms predict individual differences in cognitive performance and daily life functioning. Investigating the causal relationship between topography and function remains an intriguing challenge for future studies.
