This study was attempted to investigate the effects of self-instruction and attributional training on mathematics achievement and self-efficacy of learning disabled students in the middle school. For achieving this aim, the following hypotheses were provided : 1. Self-instruction would improve the students' mathematics achievement and self¬efficacy. 2. Attributional training would improve the students' mathematics achievement and self¬efficacy. 3. When self-instruction and attributional training were combined, the students' mathematics achievement and self-efficacy would bring higher improvements than each of the above two cases The subjects involved in this experiment, were 40 female learning disabled students who were met all of the following three criteria. First, the students have at least normal IQ(above 85) and their mathematical achievement scores were ranged from 25 to 50(M=4l). Second, the students were classified as those who should join in a compensatory education program of school. Third, the students who were perceptually or neurologically impaired were excluded from LD group. Students’ achievements were compared through pretest and posttest on both the mathematic achievement and the self-efficacy. Base on the pretest, students were assigned into one of four conditions; ① self-instruction ＋ attributional training, ② self-instruction, ③ attributional training, and ④control group. Ten students were assigned into each condition, all students participated in the mathematics training program during 50-minute sessions on 6 consecutive school days. Training sessions were conducted by the two teachers In the school. After these sessions of the experimental treatments, all students received the posttests of the mathematics achievement and of the self-efficacy. The following major findings were obtained through this study : First, in all conditions, the mathematics achievement and the self-efficacy were shown to be significantly improved(ps＜.01). Second, self-instruction’ s effects on the mathematics achievement were significantly great(F(1,35)=7.465, p=.010), but not on the self-efficacy(F(1, 35)=.343, p＞.05). Third, although attributional training promoted the mathematics achievement and the self-efficacy, its effects were nonsignificant(F(1, 35)=3.519, p＞.05, F(1, 35)=1.771, p＞.05). Finally, no self-instruction × attributional training interactions were obtained on all measures. These results suggest some implications for teaching. Learning disabled students may be improved in their mathematical skills through self-instruction while their solving problems and through receiving attributional training that were linked with their successful problem solving.