Abstrak This article presents findings from a posttest experimental control group design conducted to analyze the influence of problem based learning approach and students’ previous mathematics ability (PMA) on students’ mathematical communication (MCA) and disposition (MD). Subjects of this study were 67 students of elementary school teacher training programs of two campuses of a state University in Bandung. This study involved two kinds of instruments namely a set mathematical communication ability, and two sets of disposition scale. Data were analyzed by using one and two path ANOVA, Mann Whitney test, and chi square test. Study found that: (1) All in all entirely students and in each level of students’ PMA, MCA of students of PBL class were classified as fairly good and it was higher than MCA of students of conventional teaching (CT) which was classified as low; (2) students of PBL and CT exhibited positive disposition on learning approaches; and (5) There was no interaction between learning approach and PMA on students’ MCA.