Despite initial optimism about the feasibility of Machine Translation, it is now accepted as being an extremely different task to implement. This is due in part to our lack of understanding of the human processes involved in language comprehension and production in general, and translation in particular. In addition, the myriad of problems posed by ambiguities caused by structural differences, category options etc , which in most cases are resolved subconsciously by humans, have slowed down the development of a Fully Automatic, High-Quality Machine Translation System, and have convinced many people that this goal is completely unattainable. This thesis is an investigation of the suitability of Head-Driven Phrase Structure Grammar (HPSG, Pollard and Sag, 1987, 1994) for use in a transfer-based translation environment. It provides an account of some of the problems tackled by such a system, as well as the reasons behind the decisions to chose HPSG and a transfer approach Moreover, some of the possible inadequacies of HPSG’s current semantic framework are addressed and some potential alternatives are suggested, namely the incorporation of case grammars and semantic features to guide lexical selection in the target language. The evaluation of these ideas is based on an implementation of these proposals in a system for translation between German and English, using the Attribute Logic Engine (ALE, Carpenter, 1992) for the purposes of monolingual analysis.