The article reports a test of harvesting four selected open archives to determine the feasibility and limitations of their interconnection in a particular scientific field: the field of Information Science. It presents organizational, technical and semantic interoperability issues with a state of the art in that question. The difficult issue of common semantic representation for this scientific field, which is obvious from the step of locating and identifying Information Science open archives in the international Directory OpenDOAR, is then deepened with the detailed study of the field "subject". The article concludes with the need to rethink the modes of production of these devices in accordance with procedures shared within the scientific community and among managers of these archives.