This paper presents the Smoothed Probabilistic Data Association Filter (SmPDAF) that attempts to improve the Gaussian approximations used in the Probabilistic Data Association Filter (PDAF). This is achieved by using information from future measurements. Newer approximations of the densities are obtained by using a combination of expectation propagation, which provides the backward likelihood information from the future measurements, and pruning, which uses these backward likelihoods to reduce the number of components in the Gaussian mixture. Performance comparison between SmPDAF and PDAF shows us that the root mean squared error performance of SmPDAF is significantly better than PDAF under comparable track loss performance.