In this study, we proposed two different approaches for testing the difference of two normal population means using median ranked set sampling. We investigated the performance of these approaches according to power and type I error rates of tests under normality and non-normality with homogeneity and heterogeneity of variances using Monte Carlo simulation. We also investigated the performance of this test under imperfect ranking case. The simulation results indicate that the derived approaches performed better than its counterparts based on ranked set sampling and simple random sampling in almost every considered case. Besides, real data example was presented to illustrate the test procedure under median ranked set sampling.