Given what we know about human episodic and semantic memory (Tulving & Thompson, 1972), we can leverage memory for "events" or episodes in browser design, as well as memory for related facts and meanings about those events. A few have studied the contribution of spatial location memory to document management (e.g., Jones & Dumais, 1986), although those results were somewhat inconclusive in how generablizable they are to electronic worlds.  Applying human memory theory to the study of memory for favorite web pages, we predicted that many kinds of  information about a web page might be stored in long-term memory, including memory for spatial location. We hypothesized that users may remember where they stored a page, what the page looked like, the overall gist of the page, or the title of the page. Users may also know what they were doing when a web page was stored, or the context in which a web page was studied. This is research actively being addressed by Hightower, Ring, Helfman, Bederson, & Hollan (1998), and Wexelblat (1998). They describe the increased efficiency and user satisfaction that can be attained by augmenting browsers with additional usage and image information. For example, Hightower et al. (1998) added a graphical history-map to a browser as a companion tool to aid users as they navigated during search tasks. With "PadPrints", as the tool was called, users visited reliably fewer web pages during search, found their search targets reliably faster, and rated the user interface as more satisfying along a number of dimensions. Since the tool included graphical, temporal and hierarchical detail, it is still unclear which dimension was contributing most strongly to improved user performance and preference. Therefore, we decided to explore more closely what users remember after a brief encounter with a web page in a variety of simulated browsing sessions.

We observed users interacting with Microsoft’s Internet Explorer v. 4.0’s (IE4) Favorites text-based system, as well as several new, 2D and 3D visualizations of Favorites lists and browsers.  Favorites mechanisms allow users to store personally interesting web page addresses locally on the user’s machine. It was hoped that, by examining users’ navigation performance and memories for these web pages, we would be able to better prioritize which cues might be added to next generation browsers to support efficient search and navigation.

When users go to select a page from Favorites they may only have limited information in short-term memory about the contents and context with which that page was viewed and stored. Users may remember where they stored the page, the name of the page, what the page was about, or perhaps what the page looked like to some degree, but probably cannot recite back the entire contents of a web page. Therefore, we have carried out studies examining what specific attributes of a full web page are efficient retrieval cues. We were also interested in the number and types of categories users would come up with when storing these pages—would there be deep, narrow structures or broad, shallow hierarchies in the resultant organizations? Also, how much intra-subject consistency would there be in terms of the category labels assigned to these structures? Do 3D visual cues benefit navigation and subsequent retrieval performance?  Do automatic categorization hints help?  The answers to these questions could be used to improve future versions of web browsers.

The combined findings from several of these on-going studies have told us much about what features to add to next generation browsers to benefit users' navigation and retrieval capabilities. For instance, the title and the image of the web pages were both effective retrieval cues, as were other subjects' categorization schemes.  Allowing subjects to manually lay out their web pages in a 2D or 3D space is effective for both navigation and later retrieval, even up to six weeks later.  It is our belief that users' spatial location memory benefits from this manual layout of the web pages.  We found that supporting a combination of explicit/implicit query + navigation in the user interface was a potent addition during information storage and retrieval.   Interestingly, there is a large degree of overlap across subjects and their categorization schemes, suggesting to us that a default categorization scheme could benefit navigation in 2D or 3D space.  We have experimented with a variety of 3D perceptual cues in our browser research, many of which have contributed significantly to navigation performance.  For instance, 3D spatialized audio was a rich cue for efficient navigation and storage of information in 3D space.

To improve web browsers, we have shown that facilities need to be added to allow the user to retrieve pages on the basis of summary information and/or visual cues (adding "scent" to an existing navigational structure). A common problem for retrieving a page on the basis of summary information was that the hierarchy the user created was not cross-referenced. A specific example of this was when one user classified the "Wall Street Journal" web page into a financial category. At retrieval time the user expected to find the page in the news category. One way of recovering from this type of error would be to add an implicit query feature to the Favorites mechanism. When the user selects a page, all the pages that are semantically similar to the selected page would become highlighted. Figure 1 is an example of how this could be done. In the figure the NASA Home Page was the selected item, and it revealed US News and World Report as a related page. In such a mechanism it would be likely that all of the pages belonging to the same section of the hierarchy would be marked as relevant, in addition to relevant pages that happen to have been placed in other, more distantly related parts of the hierarchy.

Figure 1: Implicit query incorporated into the Favorites mechanism

Figure 2: Small thumbnails incorporated into the Favorites mechanism

Figure 3: Larger, interleaved thumbnails incorporated into the Favorites mechanism

Abrams, D. (1997). Human Factors of Personal Web Information Spaces, http://virtual.inesc.pt/rct/rct.51.html, M.S. Thesis, University of Toronto, 1997.

Hightower, R.R., Ring, L.T., Helfman, J.I., Bederson, B.B., & Hollan, J.D. (1998). Graphical mulcale web histories: A study of PadPrints. In HyperText ’98, The Proceedings of the Ninth ACM Conference on Hypertext and Hypermedia, Pittsburgh, pp. 58-65.

Jones, W., & Dumais, S. (1986). The spatial metaphor for user interfaces: Experimental tests of reference by location versus name. ACM Transactions of Office Information Systems, 4, pp. 42-63.

Pitkow, J., & Recker, M. (1996). "The GVU 5th WWW user survey" http://www.cc.gatech.edu/gvu/user_surveys/.

Tauscher, L., & Greenberg, S. (1997). Revisitation Patterns in World Wide Web Navigation. ACM SIGCHI'97 Conference on Human Factors in Computing Systems. Atlanta, Georgia, March 22-27, ACM Press.

Wexelblat, A. (1998). Paper can be downloaded at http://wex.www.media.mit.edu/people/wex/Footprints/footprints1.html

University of Washington Psychology Department (1996-on).   Affiliate assistant professor of psychology. Actively involved in departmental research through professor/student contracts and internships. Guest lecturer on usability engineering for various other departments on campus.

Compaq Computer Corporation (1990-1995).  Research lead, software human factors design and test. Managed and contributed to all software user interface design and evaluation research. Conducted cognitive and task analysis modeling procedures to benefit user interface design. Worked on user interfaces to the web, speech-driven voicemail applications, systems management products and consumer software applications, both in-house and with third parties.

Rice University, Department of Psychology (1990-1995).  Adjunct faculty member in Psychology. Instructed graduate course in Information Processing and Attention with an applied emphasis; Mentored Human Factors students.

Lockheed Engineering and Sciences Corporation and NASA-JSC (1989-1990).   Applied research lead, Human-Computer Interaction Laboratory. Directed all applied research in the human-computer interaction laboratory at Johnson Space Center, as well as conducted own avenues of research for user interfaces to intelligent systems and multitasking environments. Published and reviewed technical articles, wrote grant proposals and prepared project schedules.

Bell Communications Research (1988-1989).  Member of technical staff, Human Factors Engineer. Designed graphical user interfaces and wrote technical requirements for a packet-switched planning network. Aided in the preparation and dissemination of corporate user interface design guidelines.

Ball State University (1978-1981) Muncie, Indiana.  M.A. in Experimental Psychology; B.A. in Psychology, Minor in French, Honors College graduate, Summa Cum Laude.

Chen, C. & Czerwinski, M.P. (1998).  From latent semantics to spatial hypertext--An integrated approach.  In Proceedings of Hypertext '98, June, Pittsburgh, PA.

Czerwinski, M.P., Larson, K. and Robbins, D. (1998). Designing for navigating personal web information:  Retrieval cues.   Paper to appear   in the Proceedings to the Human Factors and Erogonomics Society's 42nd AnnualMeeting, Chicago, IL,  October, 1998.

Robertson, G.G., Czerwinski, M.P. & van Dantizich, M. (1997).   Quantifying immersion in desktop virtual reality.  In proceedings of ACM UIST '97 Symposium on User Interface Software & Technology, October, Banff, Canada.

Robertson, G. , Czerwinski, M., Larson, K., Robbins, D., Thiel, D. & van Dantzich, M. (1998).  Data mountain: Using spatial memory for document management, Paper to appear in Proceedings of ACM UIST'98 Symposium on User Interface Software & Technology.