A study was conducted to determine the benefits of an approach to teaching inferential skills that combined an explicit attempt to sensitize children to why and how one should draw inferences to prior knowledge with substantial practice in drawing such inferences during story discussions. Subjects were 20 good and 20 poor fourth grade readers who were randomly assigned to either an experimental or a control group. The experimental treatment consisted of three parts: (1) making students aware of the importance of relating new information to their existing knowledge structures; (2) getting students to speculate, prior to reading, about what they would do in the protagonist's situation and to predict what the protagonist would do; and (3) getting students to answer a number of inferential questions. Students in the control group received literal/inferential questions in a pattern most often reflected in elementary school reading instruction. The results showed that poor readers tended to benefit from the instruction more than did good readers. The poor readers did not differ substantially from good readers in answering the inferential questions. The poor readers also outperformed their control group peers, indicating that the experimental teaching approach was successful. (Examples of teacher-student discussions that occurred during questioning sessions and copies of worksheet questions are appended.) (FL)