The models are tested on documents collected, over a two year period, from four users. Results are given which demonstrate that the models can be used to personalise information access tasks. The tracking model is quantitatively tested on collections where the interests are known and retrospective analysis is used to demonstrate the ability of the model to track real user's interests.

As well as containing content information, documents collected from users also contain a great deal of link connectivity information. In order to take advantage of this connectivity information, a broader definition of links is developed and employed. In addition, we discover the characteristics of links within a collection necessary for the effective combination of term and link information; a collection must have a sufficiently high density of high quality links. The utility of links and the benefits of employing a broader definition of links are investigated, using document classification, on a webpage collection, a collection of academic papers and a collection of email-type documents.