Designing information systems to fit work patterns is becoming a recognized design principle (Greenbaum and Kyng, 1991). Studies of individuals doing their regular work in their usual settings is a growing area of research, for example, for understanding computer system design processes (Curtis, Krasner, and Iscoe, 1988; Davies and Nielsen, 1992; Forsythe, 1992, 1993, 1994; Orlikowski, 1989; Rosson, Maass, and Kellogg, 1988; Shneiderman and Carroll, 1988; Soloway et al., 1988; Suchman, 1987). In medicine, system developers and evaluators have been investigating the interpretive nature and complexity of routine work practice. New design methodologies are based on the assumption that system design must be based on actual work routines rather than on abstract models of information processing or formal models of reasoning (Nyce and Timpka, 1993). System developers are beginning to realize that more attention should be paid to how to incorporate implicit forms of professional competence and knowledge into system design (Nyce and Graves, 1990; Nyce and Timpka, 1993). Such studies could help developers improve system acceptance by tailoring systems to users' perceived needs in ways that will fit into their work practice patterns and routines (Graves and Nyce, 1992), match their values and provide them with benefits (Kaplan, 1987). One area ripe for study is the relationship between technologies of image delivery and how they may change the nature of work. Visualization technology, it is claimed, already has significantly changed work in science, engineering, and medicine (Kaufman, 1994). Changing what information is presented, its layout, and its mode of presentation, changes how professionals think about their work (Ruhleder, 1995). There are few formal studies of visualization systems and their relationship with work practices. Among them are studiesof CAD (Computer Assisted Design) systems, (Henderson, forthcoming), CT scanning (Barley 1986), visualization systems in neurology (Nyce and Graves, 1990), and a clinical imaging system (Kaplan and Lundsgaarde, 1994). Barney et al. (1990) express the needto make explicit the modes of understanding involved in accurately translating and using visual information. They advise identifying and describing the forms of reasoning and practice that are important to technological modeling of visualization so that information systems developers can construct environments that support routine practice effectively. For example, Ramey, Rowberg, and Robinson (1992) studied the task domain in diagnostic radiology for the purpose of designing radiology workstations. This paper concerns an evaluation study of a physician's work and how that study led to design suggestions for a clinical imaging system. It reports on how formative evaluation can help fit system design to work practice.