A schematic diagram illustrating differences in the structure and state of reader’s hierarchical generative networks across experiments and conditions. Dotted arrows represent anticipatory information that has been passed down through the hierarchy prior to word onset, while solid, upward arrows represent bottom-up information being passed from the lexical input through higher levels of the hierarchy. Increased N400 responses are thought to arise at the semantic feature level, as new unpredicted semantic information is accessed/decoded. The two late positivities are thought to arise when new information leads to large updates in the prior situation model (late frontal positivity, red), or when information conflicts with the constraints of the communication model, triggering re-analysis/repair (late posterior positivity/P600, blue).