We aim to describe the mediation language between users and indexers in a document retrieval system for a big scientific community intimately related to European Union policies. We assume that this mediation is played by thesauri: sets of indexes apparently coordinating the possible searches by means of term-to-term relations like NT, RT and so on. While persons-to-terms relations are consequent to the use of thesauri for indexing and retrieval, person-to-person relations are embodied into a thesaurus via the implicit repre- sentation of the organisation it serves. In this way, thesauri constitute a network of mediation having historical, social and - because of the scien- tific community served - scientific and technological perspectives. These three perspectives are embedded in time, since changes in organisation change the person-to-person relations, change in retrieval and indexing needs change the person-to-term relations and changes in document type and science change term-to-term relations. In particular, we want to analyse the network originally proposed by the EURATOM thesaurus (1st ed.; European Atomic Energy Community. Information and Documentation Center, Brussels, 1964) and the network of relations - in the three perspectives above - it assumed. Subsequently, we compare the results of this analysis with a more recent thesaurus designed for a community very close to the one originating the EURATOM thesaurus. In doing this, we designed a system that aims the user to browse a path built through the relations. Its in- terface is based on different concepts: Focus+Context and Elastic Grid, which led to the creation of a flexible graphical structure characterised by hierarchically-arranged and scalable information visualisation.