This investigation studies teaching via problem solving with emphasis on Scheonfeld's problem solving theory. The purpose of this study was to determine the effectiveness of training via problem solving based on Schoenfeld's mathematical problem solving theory on improving the student's beliefs in problem solving and increasing the group's activities and student's mathematical problem solving skills. This research has been studied at high school level with two randomly chosen groups of subject one control and one experimental group. The experimental group's teacher had been trained in a period of mathematical problem solving based on Schoenfeld's theory and the control group's teacher had not received any instruction. Pre and post-test was used to measure the progress of students in problem solving. A questionnaire was used to measure their beliefs and group activity before and after the training. The results of a t-test show significant differences between teaching via problem solving based on Schoenfeld's theory and student's group activity (t=3.595, p ) and problem solving ability (t=3.567 , p ). The analysis of u test shows there were significant relationships between teaching via problem solving based on Schoenfeld's theory and student's beliefs (z=-4.349 , p 0.01).