The aim of the research was to determine the effect of PISK, STAD and conventional approach to the mathematics achievement viewed from student activities. This research was a quasi experimental using 3x3 factorial design. The populations of the research were all students of senior high school in Sukoharjo Regency. The sampel was 9 classes and it was taken using stratified random sampling technique. The instruments used to collect data were student activities questionnaire and mathematics achievement test. The requirements test for data included the population normality test using Lilliefors method and the population homogenity variance test using Bartlett method. The balance test toward student’s early ability in mathematics data using unbalance one-way analysis of variance and it was concluded that experimental and control classes had balance prior knowledge in  mathematics. The testing of hypotesis using unbalance two-way analysis of variance. The conclusion of this research shows as follows. (1) Students taught by PISK cooperative learning model had better mathematics achievements than students taught by STAD cooperative learning model and conventional model. In addition, STAD cooperative learning model and conventional model gave the same mathematics achievement. (2) Students in high activity had better mathematics achievement than students in middle and low activity; also students in middle activity had better mathematics achievements than students in low activity. (3) There was an interaction between learning model and activity toward mathematics achievement