Modelling N400 amplitudes as change in a probabilistic representation of meaning

Please bwaldon [at] stanford.edu (email) for the Zoom link.

The N400 brain potential has aroused much interest as it is thought to provide an online measure of meaning processing in the brain, and has been used in a large number of empirical studies. However, despite this richness of data, the cognitive processes underlying N400 amplitudes are still unclear and actively debated. In the talk, I will present a computationally explicit account of these processes and the emerging representation of sentence meaning. We simulate N400 amplitudes as the change in a probabilistic representation of sentence meaning captured in the hidden unit activation pattern in a neural network model of sentence comprehension, the Sentence Gestalt (SG) model. The model accounts for 17 distinct and diverse established empirical effects from the N400 literature. Its ability to account for such a wide range of empirical findings suggests that the principles of representation and processing it embodies may capture some essential aspect of human language comprehension.