A single-subject reversal design with counterbalanced phases across two classrooms was used to measure the effects of peer tutoring on the retention and generalization of spelling words in two third-grade general education classrooms. The results revealed that the mean pretest-posttest gain scores during all the peer tutoring phases of the two classes was 31.3, compared to 20.3 in the baseline phases. However, there was no apparent difference between peer tutoring and baseline phases in terms of the percentage correct on retention measures. Generalization test results showed that peer tutoring resulted in 13 to 18 percentage points higher than teacher-led instruction. (Contains 7 tables and 5 figures.)