Abstract. The motivation for this talk is the view that the ultimate goal of information retrieval (IR) is to support humans to better access information in order to better carry out their tasks. Therefore IR research should provide methods and techniques to improve the retrieval/access process. I have three messages. First I will argue that IR research is not focusing its efforts properly to serve the goal above. The Cranfield paradigm of evaluation tends to lose its power as soon as one looks at human performance instead of system performance. Secondly, I will argue that searchers using IR systems make use of rather unorthodox queries (from the Cranfield point-of-view) and sessions. Their search strategies have not been sufficiently described and cannot therefore be properly understood, supported or evaluated. Thirdly, I will argue that searchers engage in an information seeking process which they have found effective enough based on their searching experiences. They try not to optimize the search result alone but the entire process (and effort) and its expected contribution to their primary task. This can be better understood in terms of the management science theory "incrementalism" than in terms of rationalism.