A study examined students' ability to comprehend consistent or inconsistent text when activating relevant or irrelevant background knowledge. Fifty-five grade six students reading at or above grade level were randomly assigned to one of four treatment groups: activated/consistent, nonactivated/consistent, activated/inconsistent, and nonactivated/inconsistent. Activated subjects wrote down background knowledge related to a science textbook passage, while nonactivated subjects wrote down background knowledge for an unrelated subject. Subjects then read consistent or inconsistent forms of the science passage, and completed a multiple-choice comprehension test. Recall protocols were analyzed for gist recall, and multiple choice tests were scored according to text type. The results indicated that subjects who activated prior knowledge and then read consistent text comprehended more text information than subjects in other treatments, including those who activated relevant knowledge but read inconsistent text. The activation of relevant preexisting knowledge did not, however, facilitate learning from text that contained information incompatible with learners' prior knowledge. Subjects who did not activate relevant knowledge and yet read inconsistent text appeared to be more willing to accept textual incongruity, that is, were more text dependent, than those who had activated relevant knowledge. (HTH)