This paper presents a framework for analyzing scientists' efforts to construct `do-able' research problems. A problem is `doable' when scientists can align tasks to three levels of work organization - experiment, laboratory, and social world. Articulation is the planning and coordination needed to align levels. The framework is used to analyze data from a study of basic cancer research. Its usefulness is illustrated by a case study of oncogenic antibody research in a biotechnology company. The paper discusses the role of modularity and standardized packages of tasks in facilitating `doability'. The findings suggest that attention to articulation work in the training and work processes of scientists, and to the conditions affecting alignment of levels of work organization, is necessary if doable problems are to be constructed.