This research aimed to reveal the effectiveness of learning with a scientific approach in the STAD type cooperative learning model based on lesson study on mathematics students learning activities. This is a quasi-experimental research where the population was the junior high school students in Rokan Hilir regency which divided into three school levels, namely; high level, middle level and low level. The subject of this research was chosen by applying clustering random sampling method. The second grader of each school was chosen to be the subject of this research by applying purposive sampling. There were experiment and control classes in each subject. The instruments that were used in this research, namely learning media (syllabus, lesson plan and students learning sheet activity) and for the data collection media (question sheet and observation sheet). The design of hypothesis testing of student learning activities in the form of data analysis of student learning activities through statistical tests. Statistical tests used the Kolmogorov-Smirnov data normality test, the Levene variance homogeneity test, the Mann Whitney U difference test, the Independent T-Test, the one-way ANOVA test and continued with the Scheffe test. The result of this research revealed that: students learning activities in the experiment class was better compared to the control class, for every level, there were differences of students learning activities reviewed from school level; and learning with scientific approach in STAD type cooperative learning based o lesson study was effective reviewed from students learning activities. In conclusion, it is proven that learning with a scientific approach in the STAD type cooperative learning model based on lesson study is effective to improve students’ learning activities