Diverse Frontoparietal Connectivity Supports Semantic Prediction and Integration in Sentence Comprehension

Yaji He, Ximing Shao, Chang Liu, Fan Chen, Elizabeth Jefferies, Meichao Zhang, Xiaoqing LiJournal of Neuroscience 2024 · DOI

TL;DR

This study reveals how our brains dynamically adjust frontoparietal network connectivity to predict and integrate meaning during sentence comprehension. It shows that strong predictions rely on increased connectivity with memory and control systems, while integrating unpredictable information strengthens connections with sensory areas.

What Problem Does This Paper Address?

Understanding how we comprehend language efficiently is a major goal in cognitive neuroscience. A key aspect of this is "predictive processing," where our brains constantly anticipate upcoming words and meanings based on context. While we know that regions like the temporoparietal junction (TPJ) and inferior frontal gyrus (IFG) are crucial for this, it's unclear how these areas dynamically interact with other brain networks to support both predicting future meaning (anticipation) and incorporating new information (integration). This paper aims to bridge that gap by investigating the flexible connectivity patterns of these frontoparietal hubs under varying levels of semantic predictability.

Methods

The researchers used functional magnetic resonance imaging (fMRI) to observe brain activity in 22 participants as they read sentences. They manipulated the "semantic constraint" of the sentences, creating conditions where the next word was highly predictable (strong constraint) or less predictable (weak constraint). They specifically focused on two phases: the "anticipatory phase" (before the critical word appeared) and the "integration phase" (after the critical word appeared).

The core analytical technique was Psychophysiological Interaction (PPI) analysis, specifically a generalized PPI (gPPI). This method allows researchers to examine how the functional connectivity between a "seed" region (in this case, the TPJ and IFG) and the rest of the brain changes across different experimental conditions (here, the processing phase and semantic constraint). They used pre-defined seed regions in the left and right TPJ and IFG, chosen based on prior meta-analyses of the default mode network (DMN) and cognitive control network, respectively. This approach helps to avoid circularity in analysis.

Datasets

The study utilized a published dataset of 22 participants (Shao et al., 2022). No new datasets or benchmarks were introduced.

Key Insights and Ideas

  • Dynamic Connectivity is Key: The brain doesn't use a fixed network for language comprehension. Instead, it flexibly reconfigures connections between frontoparietal regions and other brain systems depending on the predictability of the incoming information and whether we are anticipating or integrating.
  • Prediction vs. Integration Roles:
    • Prediction (Anticipation Phase): When meaning is highly predictable, the brain strengthens connections between frontoparietal areas (TPJ, IFG) and networks involved in memory (DMN) and cognitive control. This likely supports the generation and refinement of top-down predictions.
  • Integration (Integration Phase): When meaning is less predictable, or when integrating newly presented information, connectivity patterns shift. For less predictable contexts, integration relies more heavily on connections with sensory (visual) areas.
  • Visual Cortex's Role in Prediction: Surprisingly, strong visual coupling of the TPJ and IFG was observed during the anticipation phase in highly predictable contexts. This suggests that visual imagery or representations might play a role in generating semantic predictions, even before the word itself is fully processed.
  • Predictive Coding in Action: The findings align with predictive coding theories, where confirmed predictions lead to reduced bottom-up processing demands. The observed decrease in connectivity when predicted content appears supports this idea.

Key Results

  • In strong constraint contexts, during the anticipatory phase, the right TPJ showed stronger connectivity with regions in the DMN, control networks (including IFG), and visual cortex. Similar patterns were seen for left TPJ and bilateral IFG with visual areas.
  • In strong constraint contexts, during the integration phase, these connectivity patterns were weaker compared to the anticipatory phase.
  • Conversely, in weak constraint contexts, during the integration phase, connectivity was stronger, particularly between the IFG and visual cortex.
  • An overlap was found in the left lingual gyrus and fusiform gyrus across the interaction effects for all four seed regions (left/right TPJ, left/right IFG), indicating a common role for these visual areas in supporting both prediction and integration under different constraint conditions.

Limitations and Future Work

  • Temporal Resolution: The authors acknowledge that fMRI has limited temporal resolution, making it difficult to fully disentangle the precise timing and direction of information flow between brain systems during prediction and integration. They suggest future research using magnetoencephalography (MEG) to address this.
  • Category Variability: While they controlled for semantic violations, the study used specific categories (tools, buildings) for strong constraints and a wider range for weak constraints. They suggest future work could explore category variability more broadly.
  • Individual Differences: The paper notes that it would be valuable to investigate whether these identified connectivity patterns can predict individual differences in semantic prediction abilities.

Summary updated

Abstract

Predictive processing in the parietal, temporal, frontal, and sensory cortex allows us to anticipate future meanings to maximize the efficiency of language comprehension, with the temporoparietal junction (TPJ) and inferior frontal gyrus (IFG) thought to be situated toward the top of a predictive hierarchy. Although the regions underpinning this fundamental brain function are well-documented, it remains unclear how they interact to achieve efficient comprehension. To this end, we recorded functional magnetic resonance imaging (fMRI) in 22 participants (11 males) while they comprehended sentences presented part by part, in which we manipulated the constraint provided by sentential contexts on upcoming semantic information. Using this paradigm, we examined the connectivity patterns of bilateral TPJ and IFG during anticipatory phases (i.e., before the onset of targets) and integration phases (i.e., after the onset of targets). When upcoming semantic content was highly predictable in strong constraint contexts, both the left TPJ and bilateral IFG showed stronger visual coupling, while the right TPJ showed stronger connectivity with regions within control, default mode, and visual networks, including the IFG, parahippocampal gyrus, posterior cingulate, and fusiform gyrus. These connectivity patterns were weaker when predicted semantic content appeared, in line with predictive coding theory. Conversely, for less-predictable content, these connectivity patterns were stronger during the integration phase. Overall, these results suggest that both top-down semantic prediction and bottom-up integration during predictive processing are supported by flexible coupling of frontoparietal regions with control, memory, and sensory systems.
BibTeX
@article{he2024diverse,
  author = {Yaji He and Ximing Shao and Chang Liu and Fan Chen and Elizabeth Jefferies and Meichao Zhang and Xiaoqing Li},
  title = {{Diverse Frontoparietal Connectivity Supports Semantic Prediction and Integration in Sentence Comprehension}},
  journal = {Journal of Neuroscience},
  year = {2024},
  doi = {10.1523/jneurosci.1404-24.2024},
}
Figure