The research described in this thesis presents an investigation and evaluation into the Geld of adaptive hypermedia techniques. The aim of this research was to develop, implement and test a new adaptive hypermedia technique i.e. shading fragments of text. The course of research started off by carrying out a study to develop a simple student model 6amework containing both static and dynamic variables that could be used for educational adaptive applications. The first empirical study carried out compared three hypermedia applications based on dating techniques in archaeology, one using adaptive navigation support, one using shading to grey out complex fragments of text. These two adaptive applications were compared against an existing (non-adaptive) Archaeology application that had been used in the Department of Archaeology for the previous three years. An evaluation study was carried out to compare the usability aspects of these applications using pre and post-session questionnaires and a log of interaction was also recorded. This study found that shading fragments of text was an accepted method of adaptive hypermedia. It also showed that some of the static variables previously identified as part of the student model such as educational backgrormd were very important. The second empirical study also compared the usability of three applications based on the same content i.e. dating techniques in archaeology. One supported the shading method of adaptive hypermedia and adaptive navigational support in the form of dynamic link annotation by changing the colour of links depending on their state i.e. ready-to-be-leamed, or not-ready-to-be-leamed. All visited links were underlined. The second application only supported dynamic link aimotation without shading and the third was again the existing (non-adapted) hypermedia application was used as a control application. Again, the results reflected user attitudes towards the shading from the first study i.e. that it was a useful adaptive hypermedia technique especially when combined w i* link annotation. In general users also performed better when using this application, for example they re-visited less documents and visited more glossary terms. This study also revealed that time could be used as a variable in a dynamic student model to predict and assess user performance.