Retrieval practice has been tested on the factual materials with low interaction of elements. However, in the actual education, students need to learn both the factual materials and the problem-solving materials. Up to now the retrieval practice has not been found in the highly interactive and complex-programmed materials. The success rate of initial retrieval is a strong guarantee for the retrieval practice. For procedural knowledge, the methods of free recall and problem solving are difficult, which cannot guarantee the success rate of initial retrieval, and it is difficult to observe the retrieval practice. Therefore, in order to ensure the effectiveness of initial retrieval, and finally observe the retrieval practice on procedural knowledge, it is assumed in this study that students should be provided with instructional s in the process of learning to help them construct answers. This study examined the effects of s retrieval on procedural knowledge through two experiments. In experiment 1, 150 undergraduates participated in a 4 (learning conditions: worked example-repeat learning group, worked example-problem solving group, worked example-free recall group, worked example-s retrieval group) 2 (test intervals: immediate test < after 5 minutes>, delayed test <after a week>) hybrid experiment. The results show that the main effect of learning times is significant in terms of the beginner’s performance, the test scores of the second study of homogeneous samples are significantly better than the scores of the first learning samples; however, the main effect of learning conditions is not significant, the differences among the worked example-s retrieval group,worked example-problem solving group and the worked example-free recall group are not significant; in the final test stage, the main effect of the test intervals is significant, the scores of the immediate test are significantly higher than the scores of the delayed test. The main effect of learning conditions is not significant, and there is no obvious retrieval practice. According to the fact that the subjects almost got full scores in the tests of initial and final stage, the ceiling effect appeared in the first experiment, and there was no evidence of the positive influence of instructional suggestions on success rate of initial learning, the second experiment was designed. With the more difficult materials, 116 undergraduates participated in the second experiment, which is a 4 (learning conditions: worked example-repeat learning group, worked example-problem solving group, worked example-free recall group, worked example-s retrieval group) 2 (test intervals: immediate test <after 5 minutes>, delayed test <after a week>) hybrid experimental design. The results show that the main effect of learning methods is significant in terms of the beginner’s performance, the test scores of worked example-s retrieval group are significantly higher than both of the test scores of worked example-problem solving group and worked example-free recall group, in the final test stage, the main effect of both immediate test (after 5 minutes) and delayed test (after a week) is significant, the test scores of worked example-s retrieval group are significantly higher than the test scores of other groups, and the significant retrieval practice was found in the experiment. The results of this study have a positive influence on actual instructions. In the process of learning procedural knowledge which is less difficult, students can promote their learning through learning the homogeneous samples repeatedly or directly solving the homogeneous questions. As for the learning of complex procedural knowledge, teachers can provide students with appropriate instructional s in the course of learning to help students promote their learning and solve problems effectively.