Predictive coding across the left fronto-temporal hierarchy during language comprehension.

Area of Science:

• Neuroscience
• Cognitive Science
• Linguistics

Background:

• Language comprehension involves predicting upcoming words.
• Deviations from predictions (prediction errors) are crucial for understanding how the brain processes language.
• Hierarchical predictive coding offers a framework for understanding brain function.

Purpose of the Study:

• To investigate the neural basis of prediction errors during incremental language comprehension.
• To determine the temporal dynamics and localization of brain activity associated with different types of linguistic prediction errors.
• To evaluate the predictive coding framework's ability to explain these neural responses.

Main Methods:

• Magnetoencephalography (MEG) and event-related potentials (ERPs) were used to measure brain activity.
• Participants comprehended sentences with expected, unexpected plausible, and implausible words.
• Analysis focused on the time-course and spatial distribution of evoked brain activity.

Main Results:

• Between 300-500 ms, increasing prediction errors (expected < plausible < implausible) correlated with greater activity in left temporal cortex (lexico-semantic) and inferior frontal cortex (event model).
• Between 600-1000 ms, unexpected plausible words engaged left inferior frontal and middle temporal cortices (top-down feedback).
• Highly implausible words elicited activity in left inferior frontal cortex, posterior fusiform, and medial temporal cortex, suggesting reprocessing and new learning.

Conclusions:

• The findings support a hierarchical predictive coding model of language comprehension.
• Specific brain regions and temporal dynamics reflect different levels of prediction error processing.
• Predictive coding may unify language processing with broader cognitive functions.