This paper presents two experiments testing whether an approach to revise a text that fosters the reader's active processing benefits both high and low-knowledge readers. A history text and two alternative revised versions, one fostering the reader's inferential activity and the other reducing it, were employed. Junior high school students with low and high background knowledge about information related to the text topic participated in experiment 1, whereas undergraduate students with low and high knowledge on both the text topic and related information participated in experiment 2. One-third of the students for each condition in both experiments read either the original passage or one of the two revised versions. The revised version, aimed at fostering the reader's inferential activity, benefited low and high-knowledge readers on memory recall and inferences in both experiments, whereas the version aimed at reducing inferential activity had positive effects on recall but not on inferences. Active processing during reading seems to be the crucial factor to explain learning from texts. Consequences for the development of instructional texts are discussed.