This paper reports on an experiment and a behavioral evaluation of a text browser called SuperBook. SuperBook takes one or more existing documents in a standard text formatting language as input and presents the text in a multiwindow display with search and navigation enhancements.

The users in the experiment were 20 university students. The experiment consisted of one training session followed by five evaluation sessions. In the training session, each user was given a tutorial introduction to SuperBook. Then they were assigned randomly to work with a selected text in conventional print or SuperBook form.

In the evaluation sessions, a variety of tasks were performed. Structured search tasks required the users to find the answer to a specific question by using the reference text. Open-book essays required the users to look up material in different places in the text and to write about the information. Incidental learning tasks required the users to acquire additional information about a text while performing typical tasks without explicit directions to attend to the additional information. Finally, subjective ratings assessed the users’ reaction to the form (printed or SuperBook) and the content (a text on statistics).

For all four tasks, SuperBook was easier to use than conventional printed material. SuperBook produced more accurate searches (structured search); SuperBook users wrote essays that were judged superior to those written by users of printed material (open-book essays); SuperBook users recalled more chapter headings (incidental learning); and SuperBook users rated the text much easier to use than the printed text users did (subjective ratings).

This is one of the first papers on the comparison of hypertext and paper materials. It not only shows the advantages of a hypertext system but also suggests which properties of the system are important under various conditions. Anyone interested in hypertext should read this paper.