This study aimed to determine the effectiveness of the use of realia media on science learning outcomes in primary schools. The study was conducted using a quasi-experimental method with non equivalent control group design, by looking at the differences in pre-test and post-test between experimental and control class. The research subjects were students in fourth grade of SDN 2 Pringsewu Timur with class A as the experimental class and class B as the control class selected by cluster random sampling technique. Data analysis was performed through t-test, which was previously tested for normality and homogeneity prerequisites. The results showed that there were significant differences in learning outcomes between the experimental and control classes (significance level p = 0.006). Thus it can be concluded that the use of realia media is effective in improving science learning outcomes of students in fourth grade SDN 2 Pringsewu Timur.