A model of the software evolution process, called the Abstraction Refinement Model, is described which builds on the algebraic influence of the Laws of Programming and the transformational Draco Paradigm. The result is an algebraic structure consisting of the states of the software product (system descriptions) ordered by a relation of relative correctness with transformations defined between the system descriptions. This structure is interpreted as the software evolution space, with the intended semantics given by a model combining axiomatic semantics and the Lindenbaum algebra of a first-order logic. Using this interpretation, software evolution can be represented as a sequence of transformations on system descriptions. The primary contributions of the characterization of software evolution are to the understanding of maintenance and its relationship to development. The influence of development on maintenance is shown as the transfer of a "descriptive context" for the software system. This context is used as an information source during maintenance, and is progressively modified by maintenance activities. These activities are characterized by balanced forward and reverse transformations. The use of reverse transformations explaining the role of reverse engineering in maintenance for information gathering and document reconstruction. Additionally, the form of maintenance affects the performance of the activity, with adaptive maintenance differing from corrective, perfective and preventive maintenance. These factors contribute to the descriptive nature and utility of the Abstraction Refinement Model in defining methodologies for maintenance. / Master of Science