At the user interface, two lists of terms are visible: the ``current'' query and a ``suggested'' word list obtained by searching the thesaurus to find matches or partial matches with the initial query-string. Behind the scenes various items of data are stored about the terms; for example their current weight, and an indication of their source (the user, the thesaurus, or relevant documents). These characteristics are used by the interface layer to determine how to display them. We now consider some of the things the query layer must do to create and maintain these lists.