This research aims at describing the effect of problem-based learning on attitude and ability to solve problems. It was a quasi experiment of the posttest only control group design. The population of this research is all fifth grade students of Cluster 3 Suralaga Subdistric. The sampling technique used was random sampling. The data were collected using test and questionnaire of attitude. The data were analyzed using Manova. The result of the reearch shows that: (1) there is no effect of problem-based learning on students' attitude (tobs = 0.13, p<0,05), (2) there is an effect of problem-based learning on the ability to solve problems (tobs = 2.021, p<0,05), (3) there is no effect of problem-based learning on students' attitude and the ability to solve problems (Fobs = 13.96, p<0.05). Based on those findings, problem-based learning affects the ability to solve mathematical problems, but it does not affect students' attitude of students following the two learning approach of fifth grade students of Cluster 3 Suralaga in the academic year 2012/2013.