Alessandro Lopopolo

The paper investigates how individuals differ in predicting sentence meaning versus upcoming words using neural network models to explain N400 amplitudes.

This paper reanalyzes N400 data using semantic cloze instead of lexical cloze, finding semantic cloze better predicts N400 amplitude than lexical cloze or LLM-derived probabilities.

The paper models N400 amplitudes using a Sentence Gestalt model and compares it to surprisal from a language model, finding both predict aspects of N400 amplitudes.

This study uses a Sentence Gestalt model trained on a large corpus to predict N400 ERP amplitudes during naturalistic sentence processing, supporting the semantic prediction error hypothesis.