When reading a text, the reader constructs a cognitive representation of the information in the text. This project deals with the interaction between the text and the reader's knowledge in constructing such a representation. The focus is on three topics. First, it is investigated how cognitive categories such as cause, contrast can account for the coherence of the text. The second topic is the development of a discourse processing model that can account for the interaction between text and knowledge and in that sense for inference processes. The third topic is how the text representation depends on the use the reader wants to make of the information in the text.