A study tested the hypothesis that readers continually evaluate the informational content of text in terms of its relevance to processing goals and congruence with existing knowledge, and that they prioritize this information in terms of its strategic importance to the task at hand, integrating it with existing schemata when it is compatible. Subjects, 180 college students placed in one of three conditions, read versions of a story with differing cues, information types, and themes. After reading, the subjects completed a free recall measure and a 60-item recognition test. Results indicated that when the text contained two distinct and contrasting types of information, the context established at the time of encoding affected what statements the subjects judged to be true. Their inferences appeared to be similarly affected. Manipulation of the retrieval context had little or no effect on subjects' recognition of these statements. In addition, recall of one type of information was maximized when schemata established at encoding and retrieval were consistent, while incongruent contexts served only to muddle the process of reconstruction. (Graphs illustrating the results are included.) (FL)