The present study investigated which factors facilitated solving arithmetic word problems in fifth graders by comparing five training techniques: (1) Self controlinterpretation training(SCI) that was to acquire both the self regulated uses of solving skill and strategy and the self control ability of evaluating one's own solving process, correcting it and interpretating it to others; (2) Blind training(BT) expected to enhance the awareness of solving skill and strategy; (3) Error finding training (EF) that was to monitor the other's solving process; (4) Ordinary teaching training(C) used in a public elementary school, and (5) 30-SCI training(30-SCI) that children had been given the SCI training for 30 hours before the basic learning had begun. In Experiment 1, fifth graders were trained under a given condition for three hours and then given four posttests. Group 30-SCI did better performance on each posttest than the 4 other groups. Group SCI also did better performance on posttests 1, 2, and 3 than Group BT. Experiment 2 using four training techniques confirmed the superiority of the SCI technique to the others found in Experiment 1 in third graders.