Seminal
The magazine and atlas of EU-funded researchA new publication from the web and AI team at Scott Charnley

Atlas / Netherlands

Marie Skłodowska-Curie Action · Running

UNCERTAIN How the Brain Predicts under Uncertainty in Language Comprehension

Coordinator
Stichting Radboud Universiteit, Netherlands
Runs
1 October 2026 to 30 September 2028
Status
Running
EU funding
€233k
Total cost
—
Partners
1 organisations in 1 countries
Type
Marie Skłodowska-Curie Action
Grant agreement
101276919

What the project is doing

In the project's own words, from its summary on CORDIS.

The human brain constantly predicts input. In language comprehension, such predictions are shaped by contextual uncertainty: listeners rely on context when input is unreliable. Yet, it remains unclear how the brain adapts predictive strategies when semantic context is uncertain. This project investigates whether language comprehension under uncertainty follows the principle of precision-weighted prediction errors, a hallmark of predictive coding, whereby neural prediction errors are dampened in uncertain contexts.

To test this, I will use advanced MEG/EEG neuroimaging and computational modeling across two studies that complement the causal power of a controlled experiment with the ecological validity of naturalistic comprehension. Study 1 employs a semantic volatility paradigm to test whether neural signals of prediction errors are dynamically modulated by contextual stability. Study 2 examines naturalistic story comprehension, using entropy estimates from a large language model to determine whether word-by-word prediction error signals are scaled by contextual uncertainty.

Together, these studies will identify the neural markers of precision-weighted semantic prediction errors and establish whether predictive gain modulation under uncertainty shapes language processing. Moreover, findings will disentangle the role of two key types of uncertainty – volatility and entropy – reflecting distinct computational challenges. The project advances predictive coding theory by testing its domain-generality, bridges cognitive neuroscience, psycholinguistics, and artificial intelligence, and advances research on language processing.

By clarifying how predictive mechanisms adapt to uncertainty in language, the project strengthens models of human communication and may provide a conceptual basis for future diagnostic and technological applications.

Topics

Who is involved

OrganisationCityCountryRole
Stichting Radboud UniversiteitNijmegenNetherlandsCoordinator

Source: CORDIS, European Commission. Data of 5 Nov 2026, reused under the Commission's reuse policy. View on CORDIS · Grant DOI