Hypertext navigation tools as mechanisms for the investigation of hyperspace properties: Spatial and conceptual relations, metric space and mental representation

This thesis is divided into two distinct but related inquiries. First, to investigate the impact of different information in hypertext navigation tools, four groups of users were each given a different navigation tool to complete an information search task through a 22-page hypertext document. Each tool contained a preponderance of spatial or hierarchical information, a combination of both or the absence of both. Measures were recorded of the overall task time, number of pages accessed and memory measures for the page titles in the document. As well, a distance-like ratings task was employed to tap into users' mental representations of the document. The results indicated that providing spatial or hierarchical or spatial/hierarchical information combined improves users' abilities to efficiently move through a document but does not enhance recall of page titles over that of users given an alphabetical list. The ratings task indicated that mental representations of the document were comparable across all four groups. However, relationships between the representations and the actual path patterns of users were only evident for users given spatial information.; The second line of inquiry in this thesis focuses on investigating behaviour/representation discrepancies through the use of artificial neural networks. An in-depth analysis of two neural networks trained to complete a spatial task revealed that different underlying representations can produce similar behaviours. In this case, a metric and a non-metric representation both produced behaviours that adhered to metric principles. A coarse coding scheme centred around virtual, optimal locations for each of the processing units in the networks appears responsible for the network's ability to process both types of information and produce accurate spatial judgments.