The University of Maryland participated in the Ad Hoc task of TREC 2008 Legal Track. The TREC Legal Track is designed to model a real-world challenge known as e-discovery. In e-discovery, given a request for production and its associated complaint, the goal is to return all relevant documents without also returning an unreasonably large number of non-relevant documents. In our experiments, we focused on exploiting metadata using blind relevance feedback, iterative improvement from the reference Boolean run, and the effects of using terms from different topic fields for automatic query formulation.