There is today a growing interest in the use of automation to enhance the effectiveness of a human performing a complex task, to build "tools instead of prostheses for human performance" [Hoc, 1989; Hollnagel et ai., 1985]. Holtzman states that "effective decision systems must concentrate on assisting the decision-maker to gain insight into the decision problem at hand rather than on merely supplying a somehow “right” answer" [Holtzman, 1989]. This latter view has also been emphasized by Woods, who discusses the advantages of a joint humanmachine cognitive system architecture where the "system" is defined as the combination of human and machine (as opposed to the machine alone). He further states that "The challenge for applied cognitive psychology is to provide models, data, and techniques to help designers build an effective configuration of human and machine elements within a joint cognitive system" [Woods, 1986]. The research presented in this paper takes up that challenge in the particular arena of visual problem solving.