Fifth-grade students, classified as either above average or average readers, were tested on their ability to detect errors in short essays prior to, immediately after, and one week after receiving self-instruction or didactic instruction. In the selfinstruction condition (SI), students were taught a self-verbalization routine composed of task-specific statements. Students in the control didactic condition (CD) received equivalent practice and instruction without active self-verbalization. After training, the performance of the SI above average readers was significantly greater than that of the CD above average readers. The average readers displayed a similar although nonsignificant trend favoring the self-instruction strategy training. The results suggest that self-instruction procedures may effectively promote increased comprehension monitoring during reading. Implications of SI for reading instruction and directions for future research are discussed.