The emergent property of inhibitory control: implications of intermittent network-based fNIRS neurofeedback training

Lingwei Zeng, L L Gai, Kewei Sun, Yixuan Yuan, Gao Yun-tao, Hui Wang, X Wang, Zhi‐Hong WenFrontiers in Human Neuroscience 2025 · DOI

This study, published in Frontiers in Human Neuroscience, explores the intersection of graph theory and cognitive training. Researchers investigated whether humans can be trained to modulate the "small-worldness" of their brain networks to improve inhibitory control—the ability to suppress impulsive thoughts or actions.

TL;DR

Researchers used intermittent fNIRS-based neurofeedback to train adults to increase the "small-worldness" of their frontal brain networks. This network-level training significantly improved inhibitory control (measured via the Stroop task) and reduced the brain's "cognitive load" during tasks, with effects lasting at least one week.

What Problem Does This Paper Address?

The traditional "modular" view of the brain suggests that specific cognitive functions, like inhibitory control (IC), are localized in dedicated regions like the dorsolateral prefrontal cortex (dlPFC). However, a growing "network" perspective argues that IC is an emergent property—a complex function arising from the efficient interaction of many brain regions.

While previous studies have shown that "small-world" network properties (characterized by high local clustering and short paths between distant regions) are associated with better cognition, it remained unclear if we could intentionally train these global topological properties. This paper tests whether modulating resting-state network topology via neurofeedback can directly improve cognitive performance.

Methods

The researchers recruited 50 healthy adults and split them into a real feedback group (fNIRS-NF) and a sham (placebo) group.

  • Imaging Technique: Functional near-infrared spectroscopy (fNIRS) was used to monitor hemoglobin concentrations across 35 channels in the prefrontal cortex.
  • The Feedback Target: Unlike typical neurofeedback that targets a single region's activity, this study targeted Small-Worldness ($\sigma$). This was calculated in real-time by creating a functional connectivity matrix from the 35 channels and applying graph theory metrics.
  • Intermittent Training: Participants underwent a five-day training protocol. They received "intermittent" feedback—short bursts of real-time data followed by periods of self-regulation. This approach accounts for the inherent delay in blood-oxygen levels (hemodynamics) and prevents cognitive overload.
  • Behavioral Testing: The color-word Stroop task (identifying the color of a word when the word itself spells a different color) was used to measure inhibitory control before training, one day after, and one week later.

Datasets

The study relied on a newly collected dataset of 45 valid participants (21 sham, 24 fNIRS-NF). No external or public benchmarks were used, as the focus was on the physiological and behavioral changes within this specific cohort.

Key Insights and Ideas

  • Efficiency over Effort: The fNIRS-NF group showed reduced hemodynamic activation (lower $\beta$ values) during the Stroop task after training. This suggests that by optimizing the network's "small-world" topology, the brain became more efficient, achieving better performance with less metabolic effort—a concept known as "neural economy."
  • Asymmetric Plasticity: Even though the feedback algorithm treated all 35 nodes equally, the brain didn't rewire uniformly. The right dlPFC showed significantly greater increases in internal connectivity compared to the left dlPFC. This suggests that certain regions are naturally more "plastic" or central to inhibitory control networks.
  • Causal Link: The study found a significant correlation ($r = -0.32$) between the increase in resting-state small-worldness and the improvement in Stroop task performance. This provides evidence that network topology is not just a correlate, but a driver of cognitive ability.

Key Results

  • Improved Inhibitory Control: The fNIRS-NF group showed a significant reduction in the "Stroop effect" (the delay caused by conflicting information) compared to the sham group.
  • Sustained Effects: The cognitive improvements and network changes were still present at the one-week follow-up, suggesting that neurofeedback can induce lasting neuroplasticity.
  • Network Regularization: Training successfully shifted the brain from a more "random" or disorganized state toward a more "regularized" small-world state.
  • Statistical Significance: The increase in feedback scores was significant by Day 4 ($p = 0.017$) and Day 5 ($p = 0.027$).

Limitations and Future Work

  • Participant Demographics: The study only involved healthy young adults. The authors note that the results might differ in populations with existing inhibitory control deficits, such as those with ADHD or autism.
  • Training Ceiling: The five-day training period was relatively short. It is unknown if longer training would lead to even greater gains or if there is a "ceiling effect" for network optimization.
  • Spatial Imbalance: The observed asymmetry (right vs. left dlPFC) requires further study to understand why the brain prefers certain rewiring patterns over others.
  • Clinical Application: Future research is needed to determine if this "network-based" neurofeedback can be used as a clinical intervention for cognitive disorders.

Summary updated

Abstract

Intermittent fNIRS neurofeedback effectively modulates resting-state small-world networks and enhances inhibitory control, with effects sustained for at least one week. These findings highlight small-worldness as a novel target for cognitive interventions.
BibTeX
@inproceedings{zeng2025emergent,
  author = {Lingwei Zeng and L L Gai and Kewei Sun and Yixuan Yuan and Gao Yun-tao and Hui Wang and X Wang and Zhi‐Hong Wen},
  title = {{The emergent property of inhibitory control: implications of intermittent network-based fNIRS neurofeedback training}},
  booktitle = {Frontiers in Human Neuroscience},
  year = {2025},
  doi = {10.3389/fnhum.2025.1513304},
}
Figure