This thesis examines methods to aid in the non-visual browsing of web based documcnts, primarily using synthetic speech output. The serial nature of speech ensures that it is a difficult medium in which to browse documents. Much of the structure implied in the visual appearance of the content is not available through specch. Only a narrow region in the content is perceivable at any given time, and it can bc difficult to navigate to the important segments of the document. This is in contrast to visual interaction, where cues such as changcs in font or colour establish contextual changes in the content and guide the user accordingly. A number of browsing/navigation strategies are presented to offset these problems. These are implemented through WebTree. This is a highly customisable web browser which rcnders documents through a dynamically expandable tree structural view. This mirrors the arrangement of mark-up elements in the source file. Contextual information about each element is provided as appropriate. By expanding and collapsing individual tree elements, thc user decides how much of the content is rendered at any given time. The user can also dctermine whether a certain element is automatically expanded in the rendcring when cncountercd, or whether it appcars in the rendering at all, effectively allowing for the easy gencration of alternative document views. To speed up navigation the user can move through the document based on the element of their choice. Specialiscd element search functions are also available. Additional navigational functionality is provided to deal with the specific requirements of <table> or <form> elements. The thesis reviews different methods previously employed to offset problems with auditory interfaces and compares these with WebTree. Initial user tests and evaluations of WebTrec are prescnted, which show that the approaches taken provide a viable solution, particularly for thc browsing of large or complex wcb-bascd documcnts, by blind users.