Research on the development of image digital libraries has so far yielded little knowledge about the actual uses of rich visual content. This paper presents initial findings of a study on the decision making strategies that expert users employ in a large scale image digital library to choose and evaluate digitized photographs for specific projects. The paper establishes the foundation for the research in the literature on representation and remediation and describes the methodology of the research project, which includes two-stage semi-structured interviews with seven expert users across the spectrum of factors motivating the sophisticated, project-based use of digitized photographs. Findings presented here highlight the nature of visual expertise and the indicators of a new model of learning in visual collections.