Modern user interfaces make extensive use of navigation, a metaphor based o
n wayfinding in a physical space. Navigation can be an effective solution f
or many problems in understanding and manipulating a complex information sp
ace. Unfortunately, current research toward domain-independent, intelligent
assistance for navigation lacks some important conceptual and practical to
ols for evaluation. This article describes an empirical study of assisted n
avigation that investigates the relationship between the quality of the ass
istance and overall performance. Our definition of quality is based on the
information retrieval concepts of precision and recall. We use the limitati
ons of the study results to motivate the development of a general-purpose n
avigation testbed. By separating the concerns of retrieving navigational da
ta, maintaining and managing the resulting information space, and providing
different views into the space, the testbed facilitates the development of
navigational mechanisms and supports their evaluation. The testbed will ac
t as a foundation for future work toward the automated generation of intell
igent tools for navigational assistance. (C) 1998 Elsevier Science B.V. All
rights reserved.