Four fifth-grade students were presented with frustration-level math probes while three performance dimensions were measured (i.e., percent intervals on-task, percent correct digits, and digits correct per minute (DCM)). Using a multiple baseline design across participants, students were trained to self-monitor time on-task, accuracy, and productivity in sequence and were given their choice of preferred rewards for increases in each performance dimension. The effects of self-monitoring productivity plus rewards prior to and after reaching on-task and accuracy criteria were also evaluated on a probe basis using an embedded multielement design. Results showed that three of the four students increased DCM when self-monitoring productivity plus rewards, but only after meeting on-task and accuracy criteria. Effects of self monitoring accuracy were mixed, and all students showed high levels of on-task behavior throughout the study. The implications of these results for increasing the effectiveness of self-monitoring through a sequenced approach to academic skill training are discussed.