This report details a study of information contained in high-energy (MeV), dual-energy radiography data for the purpose of materials discrimination, using computational methods and analysis of real radiography data. We use a computational method, the Cramer-Rao lower bound (CRLB) to measure information content in a dual-energy radiograph and bound uncertainties on material discrimination for a wide range of radiography system and object geometry realizations. We conduct computational studies of an inverse algorithm to determine material composition using simulated data. and assess the impact of systematic effects such as scattered radiation. Finally, we apply the methods to experimental data, benchmarking the CRLB to material discrimination analysis of high-energy, dual-energy radiography data and exploring a method to make an effective dual-energy scan with a single X-ray shot by using a patterned grid in the X-ray image. We find that although material discrimination is fundamentally limited, broad discrimination between high and low atomic number materials is possible. The Cramer-Rao lower bound on uncertainty tracks well with more complex approaches and with measured data, and may provide a useful tool for system design and optimization. Finally, the use of a patterned optic may enable single shot dual energy imaging when acquiring two shots is not feasible.