In a previous study, the researchers report a study of the short-term effects of using different types of dynamic assessments as remediation models for children with mathematics learning difficulties (Lu, Lin, & Lu, 2004). The results showed that using dynamic assessments as remediation models had effects on mathematics low achievers right after remediation sessions. In the presenting study, the researchers want to explore how the remediation sessions influent those low achievers' mathematic performances in the long run. The concepts of dynamic assessments were based on the notion of Zone of Proximal Development, which contend that all children can achieve the highest level of development with adequate instructions (Vygotsky, 1978/1997). There are three major models of dynamic assessments: scaffolding, coaching, and strategies trainings (Carlson & Wiedl, 1992; Guthke, 1992; Swanson, 2001). In the scaffolding model, the examinees are given a series of hints during the tasks. In the coaching model, the examiners coach the examinees task-specific strategies. In the strategy-training model, the examiners give the examinees general strategies training, demonstrate the strategies, and feedbacks. Among those models, Swanson (2001) found that scaffolding models had best effects, followed by the strategy. In this study, the researchers applied the model of scaffolding and strategies trainings for the content of mathematics problems and cognitive skills. A total of 69 children identified by their classroom teachers as mathematic low achievers from 2 schools took the dynamic assessment remediation models. The participants were randomly assigned to 4 treatment group: Scaffolding-Math, Scaffolding-Cognitive, Strategy-Math, and Strategy-Cognitive. Children took mathematics achievement tests before treatment, right after treatment, and 2 month after the treatment. Another group of 77 children who were not nominated by their classroom teachers as low achievers but scored one standard deviation below the school mean scores in the pre-test served as the controlled group in this study. The controlled group did not take the first post-test. There were two comparative forms of mathematics achievement tests, and the pilot study did not find significant differences for the difficulties. However, the researchers applied Latin-square design, that children at School A took Form A in the pre-test and Form B in the post-test, while children at School B took Form B in the pre-test and Form A in the post-test. Children at school A took the pre-test in the end of the second semester when they were first graders and took the post-test in the beginning of the first semester when they were second graders. Children at School B took the pre-test in the beginning of the first semester when they were second graders and take the post- test 2 and a half month later. That is, children at School B took the tests about 2 month later than children at School A. As a result, children's performances in the tests were transformed into z-scores. Comparing students' z-scores in pre-test and post- test, the results showed that the treatment group did progress significantly in the test of numerical concepts. In the other subtests, the treatment group's performances were not different from their performances in the pre-test. No school by treatment interaction effects were found, although the non-treatment group did perform better than the treatment group in identifying geometric shapes. The implication and contribution of this study to the theories of mathematics learning and the education for children with mathematic learning difficulties will be discussed further.