Creativity is not only the creation of completely new, innovative solutions to a problem. It also comprises the adequate use of existing knowledge in new situations. In thinking a solution to a problem, the designer has a vague image of the form that will embody the solution. Creating collages, sketches and other types of (external) visual representations are used to help in shaping and establishing this image. For this, designers make extensive use of design precedents. In design, precedents provide the frame of reference for the development of new solution principles and product forms. They act as sources of knowledge to generate an image of the possible solution space, to get an impression of modes, styles, trends, applications of materials and production/assembly techniques, etc. Finding the right image can be a difficult task. Many of the existing computer tools to support designing with precedents suffer from serious drawbacks: firstly, to be useful, a database with precedents has to have a significant amount of images. Describing and indexing each image is a time consuming, labor intensive, expensive task. The second drawback is related to the way images are understood and interpreted. Attribution of meaning is very personal, subjective and situational and can hardly be determined beforehand by the editor of a collection. These shortcomings could very well be the reason why none of these initiatives have left the research environment to become a successful commercial product. The high costs associated with the indexing of the data and the amount of maintenance required to keep it up to date, are major obstacles for the assimilation of these technologies in industry. In this paper, we present an approach that eliminates the human mediated description, indexing and organizing of large collections of design precedents. It explores both, theoretical and technological aspects of the use and handling of design precedents. On the theoretical side, it discusses questions related to how to represent design precedents in such a way that they can be effectively used in design education and in design practice. On the technological side, it shows how to implement such representations in a Query By Example (QBE) computer program using a Content Based Retrieval Approach (CBIR) so that it eliminates the problems associated with human mediated indexing and description. It is done by three empirical studies whose main objectives are to: • explore different ways of (automatically) representing design precedents • test the suitability of the QBE approach to handle large collections of design precedents • learn about the criteria used by the designers to assess the relevance of design precedents used during the design process • design a proof-of-concept system for research and demonstration purposes. One of the main results of these studies was the possibility to observe more closely, by studying their way of using precedents with protocol analysis, the form creation process. These observations suggest that the designers move, very early in the design process, from a discourse of the function of the artifact to be designed, to a discourse where the both, the potential users and the product to be designed are described by means of other elements that will be present in the product’s use. That is, the concern shifts from the function of the product to the context of use. In the same way, there is a third moment in which the main concern is materials and geometry: Form. These results are useful in what they provide clues on how to better design computer tools to support this process. This paper discusses the practical implications of the results on the development of a system to handle design precedents. One of the motivations to use Content Based Image Retrieval System (CBIRS) was the promise of results without having to describe, organize and index each image, as is necessary in current systems to handle design precedents. This allows escaping the subjectivity of the interpretations, escaping the imprecisions of language and avoiding differences in opinion between users. However promising, as was proved in the tests presented in this paper, this approach falls short of fulfilling all the designer’s needs for visual information. The reason is that the algorithms available cannot recognize what the image contains (in semantic terms) but humans can, and with great facility. This ability was reflected in the searching process of the designers in our studies. It is very natural for them to expect living room furniture if using a sofa and a lamp as seeds for a query, because a user can understand that these two are related, and that the common aspect is that they are both elements of a living room. To the system, they are geometrically so different that the results are completely incoherent. Image recognition is useful, but not enough! This paper concludes proposing a strategy to solve this problem in future systems to handle design precedents.