This article reports on two experiments where undergraduates read five documents on a scientific topic and afterwards answered comprehension questions and wrote either summaries or argument essays on the topic. In the first experiment, students who were instructed to work with the documents for the purpose of summarizing their contents displayed better comprehension and integration of document contents than did students instructed to construct arguments from the documents. In the second experiment, focusing on whether the effects of task instructions on multiple-documents comprehension and integration could be moderated by students' prior knowledge, it was found that only students with high prior knowledge were able to take advantage of instructions to construct arguments while reading, whereas low-knowledge readers seemed to be more hindered than helped by such task instructions. Theoretical as well as educational implications of these findings are discussed. (Contains 10 tables and 5 figures.)