Abstract
The study of automatically generated link networks in hypertext repositories yields valuable insights into the algorithms that produced them. In this investigation of these link networks, the phenomena of centrality and its opposite, dispersion, are identified as issues to be dealt with. Different algorithms produce link networks with different degrees of centrality. We investigate this using the MultiBrowser system as a testbed and applying these algorithms to repositories consisting of documents downloaded based on search engine return lists. Once automatically generated hyperlinks are added to this repository, making documents accessible from other documents in the repository, the centrality properties of the linked repository can be investigated.