Questions used in school have not been analyzed by the item and only in one representation. This study aims to produce quality multi-representation-based cognitive instruments on Newton's law material for junior high school students in Banjarmasin. This study specifically aims to describe (1) the validity of cognitive instruments, (2) the reliability of cognitive instruments, (3) the level of difficulty of cognitive instruments, and (4) the discrimination power of cognitive instruments based on multi representation. The method used is the Research & Development (R & D) method by using the adaptation of the procedure model from Borg & Gall. Data were analyzed using classic formulas and through Rasch applications. The sample of the study was 204 eight grade students from Public Junior High Schools 14 Banjarmasin, Public Junior High Schools 25 Banjarmasin, and Public Junior High Schools 28 Banjarmasin. The results showed that (1) the validity of cognitive instruments developed is considered valid, (2) the reliability of cognitive instruments developed is relatively reliable, (3) the level of difficulty of cognitive instruments developed is divided into two categories; very difficult and very easy, and (4) the discrimination power of cognitive instruments developed is divided into three categories; repaired, accepted but needs to be repaired, and accepted. It can be concluded that the multi-representation-based cognitive instrument on Newton's law material on eighth grade students of junior high school in Banjarmasin is suitable for assessment of learning outcomes. So that the instruments developed can be used by teachers to assess student learning outcomes in the material of Newton's law.