Since many studies indicate that young readers have difficulty comprehending many commonly occurring forms of anaphora, an instructional experiment was conducted in which a variety of anaphoric forms was taught, the intervention period was of sufficient duration, and external validity was enhanced so that results would be generalizable to the realm of children and teachers in classrooms. Sixty rural midwestern third-grade students received nine, 40-minute lessons over three consecutive weeks, after which four posttests were given. Students were divided into three groups: (1) the strategy group, in which the strategy was based on successful instructional research found to be effective in teaching sixth-grade students to comprehend main ideas; (2) the basal group, in which the anaphora instruction came from a current popular reader series; and (3) the control group, which did not receive any additional instruction. The results of the study suggest that the implementation of a direct instruction model for teaching children to comprehend anaphoric relationships is superior to basal reader instruction in anaphora and to no instruction at all. Specifically, the findings show that the strategy group outperformed both basal and control groups in ability to find the antecedent for specified anaphoric terms; the basal group did not outperform the control group in anaphora comprehension; the strategy group outperformed both basal and control groups in ability to answer anaphora-specific questions; the overall pattern of results for treatment held true whether students read narrative or expository texts; and there were no treatment-by-achievement level interactions. (EL)