A pre-post experimental design with 6-week delayed posttest was implemented to investigate the effects of reading comprehension strategy (RCS) instruction with and without attribution retraining (AR) on reading outcomes for seventh, eighth, and ninth graders with learning and other mild disabilities. Students were randomly assigned to one of three treatments: RCS+AR, RCS, or a Read Naturally (Ihnot, 1992) comparison condition. Results indicated that compared to the comparison group, both strategy instruction groups performed better on a summarization measure of comprehension after treatment producing large effect sizes. Both groups also performed better after a 6-week delay; however, only the RCS+AR group maintained a large effect size. In addition, RCS+AR students displayed higher attributions for reading success at post- and delayed posttesting. Implications for practice and future research are discussed. (Contains 5 tables and 2 figures.)