The goal of this research project is to develop an understanding of the documentation needs of maintenance designers for software systems and to build a model of how these maintenance designers interact with knowledge about legacy designs. Managing evolution is one of the fundamental challenges of software engineering. During the design process many design decisions and trade-offs concerning functional and non-functional requirements must be considered. The process of abstracting requirements into models and converting those models into code yields design artifacts where decisions are documented?if at all?informally. To preserve design decisions through maintenance, the current research approach is to annotate architecture models with explicit representations of design decisions as first-class entities. Unfortunately, it is clear that we lack the instruments for evaluating the effectiveness of design decision modeling systems, for comparing these systems to each other, and for using these analyses to drive future research. To accomplish our research goals, we propose a course of empirical study that uses surveys, expert interviews, and talk-through protocol analysis as the basis for formulating a theory of maintenance design. The theory will be tested using experimentation by protocol analysis and validated through ethnography in real-world industrial contexts.