The online reading brain: prediction, segmentation, and integration

This thesis investigates how the brain predicts and processes information during reading. For example, when readers encounter a text such as “I am orange. I am big and thick. You can carve me on Halloween.”, the brain gradually builds expectations about the meaning of the text. These expectations are influenced by prior knowledge, such as knowing in advance that the text is about a pumpkin. The brain also integrates incoming information into meaningful units. To investigate the underlying processes, brain activity was measured using electroencephalography (EEG), which provides high temporal resolution, functional magnetic resonance imaging (fMRI), which provides high spatial resolution, and eye tracking. The thesis includes three studies. The first two studies examined how predictions are generated during natural reading. The results show that the brain continuously predicts upcoming information and updates its interpretation of the text. These predictive processes are reflected in both earlier and later brain responses after word recognition. The third study applied machine learning techniques to demonstrate that the brain organises narratives into short and longer events across distinct reading-related neural networks.