Relationally Encoded Links and the Rhetoric of Hypertext

George P. Landow

Department of English and the Institute for Research in Information and Scholarship (IRIS) Brown University, Providence, Rhode Island 02912

I. INTRODUCTION

More than two years’ work on designing, writing, editing, and linking documents in Context32 [Land86], the first course employing Intermedia developed at Brown University’s IRIS (Institute for Research in Information and Scholarship), has provided valuable experience of hypertext and hypermedia systems. Context32, which contains more than a thousand text and graphic files joined by approximately 1300 links, appears the most ambitious implementation thus far of a full hypertext and hypermedia system intended for multiple users. Members of the development team at IRIS have previously described various aspects of Intermedia’s object-oriented programming [Meyr86], general design [Meyr85, Yank87a, Yank87b], and educational goals [Land87]. This paper presents conclusions about what works best at each end of a hypertext path or linkway and proposes that, like other forms of discourse, hypertext requires systems of conjunctive and other relational devices.

II. THE NEED FOR A RHETORIC OF LINKING IN HYPERMEDIA

Designers of hypertext and hypermedia materials confront two related, problems, the first of which is how to indicate the destination of links and the second, how to welcome the user on arrival at that destination. Drawing upon the analogy of travel, we can say that the first problem concerns exit or departure information and the second arrival or entrance information. In both cases the designer must decide what users need to know at each end of a hypertext link in order to make use of what they find there. The general issue here is one of interpretation—namely, how much interpretation must the designer-author attach (1) to link pathways and (2) to files at the end of links to permit them to function efficiently in a multi-user system?

At this relatively early stage in the history of hypertext systems, those involved, as one might expect, devote most attention to the simple fact of linking and to the effects upon discourse of such electronically linked text. Although hypertext clearly redefines some of the basic characteristics of page-bound printed discourse, such as the rigidly hierarchical distinction between a main text and its annotation in scholarly works, it still depends upon many of the same organizing principles that make page-bound discourse coherent and even pleasurable to read.

The experience of courseware developers and of student users with Context32 clearly demonstrates that simply linking one text to another in some cases fails to achieve the expected benefits of a hypertext system and even alienates the user. Drawing upon our experience with developing Context32 at Brown, I would like to examine several examples of insufficiently encoded hypermedia materials and then, in later sections, to discuss various approaches to encoding or attaching necessary information. Since graphic images in hypermedia systems so clearly demonstrate the problems created by insufficient encoding, we shall examine them first and then later look at text files.

At an early stage in the development of Context32, some of my graduate-student assistants assumed that placing certain kinds of images that they deemed interesting at the ends of links would suffice to inform students in some undefined way. Portraits of writers and similar documentary or informational use of earlier graphic work, such as a nineteenth-century photograph of homeless boys (Figure 1), exemplify this kind of linked file. Ultimately, I removed most such files or modified them in ways that are described below, but some were left unchanged because I wished to learn how students would react to them (and some because we have not yet had the opportunity to make necessary modifications to the files). As it turned out, students found no educational value either in such links or in the files at their ends, and they resented the time required to call them up, inspect them, and put them away.

Why, then, were such kinds of linking educationally and informationally ineffective? In the first place, these graphic files confused the users, who could not quickly determine why such material had been included. Once confused, they resented the presence of the link. Before discussing several solutions to the problem of efficient handling of informative links in hypertext, let us note the basic assumptions underlying these reactions, the most important of which is that links represent useful, interesting, educationally significant relationships. Such assumptions, we must realize, do not derive from overestimations of hypertext but are intrinsic to such systems. In fact, because links and link-relations play such primary roles in hypertext (and hypermedia), they influence the content they convey and thus exemplify the McLuhanesque principle that the medium is the message [McLu62, McLu64]—or at least that every medium of communication itself communicates an identifiable bias or message. Sometimes the message takes the form of a negative bias created by the limitations of a specific technology. For example, using currently available hardware for the first implementation of Context32, we discovered that both mid-nineteenth-century woodblock illustrations and those that Beardsley created at the end of the century (Figure 2) appear clearly. Photographs and paintings with a wide range of tonality and those depending heavily upon color do not work as well. In fact they reproduce so poorly that we have had to omit most Romantic art—with the result that users of the system might conclude not that our technology has particular limitations that constitute a bias but that there is no Romantic art or that it has little relevance.

Hypertext as a medium also conveys its own positive bias or message, for hypertext’s system of linked files conveys the strong impression that its links signify coherent, purposeful, and, above all, useful relationships. From which follows:

Rule 1: Hypertext links condition the user to expect purposeful, important relationships between linked materials. Such was the capacity of hypertext systems that I originally planned to draw upon when I began to work with software developers at IRIS, and from this capacity one can deduce one of the principles embodied in Context32 that is also a principle of hypertext:

Rule 2: The emphasis upon linking materials in hypertext stimulates and encourages habits of relational thinking in the user. Such intrinsic hypertext emphasis upon interconnectedness (or connectivity) provides a powerful means of teaching sophisticated critical thinking, particularly that which builds upon multi-causal analyses and relation of different kinds of data [Land87]. But one must note a third, cautionary principle:

Rule 3: Since hypertext systems predispose users to expect such significant relationships among files, those files that disappoint such expectations appear particularly incoherent and nonsignificant. When users follow links and encounter materials that do not appear to possess a significant relation to the file from which the link pathway originated, they feel confused and resentful.

As the examples of the author’s portrait and documentary photograph suggest, appending brief texts in the form of titles to the images does not always provide enough information for the user, because titles do not sufficiently establish a relationship between the two linked files. Such inadequate relational encoding or markup does not appear only in hypertext and hypermedia systems, of course, but such systems accentuate the user’s negative reaction. In fact, inadequately presented visual information characterizes many illustrated textbooks, particularly literary anthologies, that include portraits of authors, works of art, and other supposedly relevant visual materials. As studies have shown, students generally ignore such materials. Books permit the student user to avoid apparently nonsignificant and insignificant materials—one simply glances at them and turns the page or, in many cases, simply never glances at them at all—but hypermedia systems, whose linkages suggest that the user will encounter significant relationships between materials, make ignoring such materials more difficult. They force the user to confront relationality—or its absence.

III. Solutions: The Rhetoric of Arrival

As the examples of a writer’s portrait and documentary information about the historical background have shown, conventional titles do not adequately direct the user how to relate a graphic image to other materials. In contrast, “Victorian Design: Medieval Revival” [Figure 3], an illustration of an item exhibited at the 1851 Crystal Palace exhibition, the first world’s fair, exemplifies the kind of encoding or rhetoric required to enable the user to make sense of graphic files—that is, to discern one or more conceptual relations between them and files to which they are linked. The appended text reads: “This curiosity from the Crystal Palace exhibition of 1851 is not a suit of armor but a stove built in the shape of one. What do such bizarre glances back at the past tell us about the Victorian age, which invented the idea of Progress as we know it? Can you find in the poems you have read any examples of thus clothing present purposes in ancient forms?” Students who highlight the link marker and issue a command to follow receive a menu of various Victorian literary works, including Tennyson's “Morte d'Arthur,” a poem set in medieval England, and “Tithonus,” one set in the Greece of ancient myth. Experience with these materials has taught that to be educationally effective they must follow these principles:

Rule 4: Linked graphic materials must appear with appended texts that enable the user to establish a relation between file of departure and that of arrival. The solution we have adopted appears in Figure 3, whose text (a) provides factual information, (b) encourages users to relate that information to a problem on which they are working, and (3) contains links that allow them to pursue various investigations. From this follows two principles:

Rule 5: The entire text accompanying visual material serves as an introduction and not just the opening sentence or so. And:

Rule 6: The accompanying text does not have to specify all relevant information the designer wishes the user to have; rather, emphasizing that a relationship exists at all may be enough. From which follows:

Rule 7: Texts serve not only to provide information but also to reassure the user that the link embodies a significant relationship and to provide some hint, however incomplete, of how that relationship can be formulated by the user.

The visual information, which provides interesting data of an unexpected sort, also enforces the principle that relevance is in the mind of the beholder and that the investigator's function in whatever field is to inquire what connections might exist among various kinds of data and and their relative value might be evaluated.

The principles of making effective links to files containing largely visual information also apply when linking text to text. Here, too, one must employ devices that enforce hypertext capacity to establish intellectual relations. A user who activates the link marker in “Victorian Design: Medieval Revival” receives a choice of links to files about the poetry of Tennyson and Browning, both authors who, like the designer of the stove, use old forms to solve contemporary problems. The first of these files, “Tennyson’s ‘Morte d'Arthur’” (Figure 4), shows how one can emphasize the relationality of a text file by a combination of information, questions, and links. This text file begins by offering the student information about the poem's publication history and about its indebtedness to Sir Thomas Malory's Arthurian work, after which it continues with a series of alternating factual statements and questions that ask the student to apply the new information contained in them to draw broad conclusions about medievalism, the relation of the poem to Tennyson's own life, and its connection to Carlyle, a passage from whose work is included. The file also contains link markers indicating the existence of paths to files on the poet's biography and Carlyle. Planned additions include an essay on medievalism and links to it. The point here is that one must employ the same structure of information, questions, and link markers in both graphic and text files.

IV. Solutions: The Rhetoric of Departure

We have examined methods of encoding points of arrival before examining the encoding of link markers, which are points of departure, because the basic need for such encoding appears so clearly when users first confront files new to them. Experience with Context32 has shown that links and link markers also require similar kinds of encoding to be used effectively, and this implementation of hypermedia employs at least six different forms, three of which are internal to the file containing the link marker and three external.

1. Internal:

(a) Link marker apparently independent of accompanying text.

(b) Link marker whose spatial proximity to text indicates probable nature of link destination.

2. External:

(a) Link descriptions

(b) Menu called up by link marker at site of multiple links.

An instance of 1.(a), a link marker that appears independently of accompanying text, occasionally appears in overview files, such as that for Tennyson (Figure 5). Overview files, which are graphic directories, play an important role in Context32 since they simultaneously inform the user in a general way about the kinds of information available in relation to a particular subject and also enforce the point, a central theme of the course, that individual phenomena, such as an author (e.g., Tennyson, Pope), aesthetic category or periodization (e.g., Victorianism [Figure 6], Neoclassicism), or other topic (e.g., religion in England, Darwinism), relate to a range of subjects and approaches. An example of such naked or unaccompanied link markers 1.(a), which are relatively rare in Context32, appears in the center of the Tennyson OV (overview) file near the poet's name. Although this marker lies adjacent to a text specifying the main subject of this graphic directory file, that text does not suggest any link destination since the user is already within the Tennyson directory file. This marker simply represents the destination of another link (since linking creates link markers within each linked file) to an index file, and as such it represents something of an anomaly and should be replaced by form 1.(b). Each of the separate boxes or texts specifies to varying degrees what the user can expect to find on arrival at the link destination. The markers situated near the titles of individual works of Tennyson and the boxes labeled “Biography” and “[biographical] Timeline” clearly indicate their destinations. That labeled “Literary Relations” leads to another graphic directory file [Figure 7], which takes a standard form, and after using Context32 to study a few authors or works, the student can expect to encounter this kind of graphic representation of literary relationships. Since more variety exists in relation to other materials organized by the overview file, the user does not know in advance what to expect from the label. Thus “Cultural Context,” which here links to the Victorianism OV (Figure 6), also links to an essay on “Tennyson and Victorianism,” but authors less important in the course often do not have such additional linked documents. Similarly, there is an even wider variation in the number and nature of documents linked to “Religion” and “Science and Technology.”

One approach to informing the user more about the nature of a link destination appears in those files, chiefly literary relations directories, that exemplify 1.(b) because the text their link markers accompany is so specific that it indicates the likely nature of the link destination. Such specifying texts also appear in primary text documents. For example, when a file on a modern author contains a link marker near “Freudianism,” “World War I,” or an aspect of literary technique, such as “Theme,” or [Imagery,” the user can expect to encounter an essay (or diagrammatic presentation of one) on these subjects at the end of a link. These observations on the way that Context32 indicates link destinations in various kinds of graphic and text files lead to a crucial principle of hypermedia:

Rule 8. Any file in a hypermedia (or hypertext) system is a directory file. Although I have been writing as if Context32 only employed those graphic files specifically entitled "OV" (for overview) as directories, users can and do rely on whatever file they find themselves in to organize linked information. From this characteristic of Context32 follows a point at which hypermedia capacities converge with an important emphasis of contemporary literary theory:

Rule 9. Regardless of what kind of directories the authors and designers include in a hypermedia system, users can organize it according to their individual interests. Modern literary theorists argue that literary investigations should be organized not in the traditional manner according to authors or periods but according to individual texts [Scho85]. Hypermedia, which has strong individualistic and democratic potential, permits users to make any file (and any interest) the organizing principle of their investigations.

Now, to return to the problem of encoding link destinations within files. The most specific textual specification of a link destination occurs in 1.(c), when specific directions inform users of a link destination, usually by inviting them to make their way there. For example, the file entitled “Point of View in ‘The Prussian Officer’” begins: “ Point of view [link to definition],” and in the file entitled “Neoclassical Couplet” the following sentence appears next to an example of Pope’s satire on women: “Follow to see a woman's view: Lady Mary Wortley Montague’s couplets.”

In addition to such internal means of indication link destinations from within the file, Intermedia provides three external ones—2.(a) link descriptions, 2.(b) menus of link descriptions, and 2.(c) local maps. After choosing a link marker, the user can put down a menu and select a link description rather than follow a link without further information (Figure 8). Many, though not all, link descriptions inform the reader if the linked files contain primarily graphic or text materials. During the first implementation of Intermedia and Context32 users do not seem to have employed this capacity very much, perhaps because invoking it involves additional delay. A second external means of indicating link destinations involves automatically generated menus that appear when the command is issued to follow a link marker to which two or more files are linked (Figure 9). A third external indicator takes the form of Intermedia's automatically generated local map (Figure 10), which contains the names of all files linked to the file currently active. Although local maps for files containing relatively few links are easy to use and helpful as a way of discovering link destinations, those for files with many links are difficult to use.

V Acknowledgements

Intermedia is the culmination of two years of intense effort by a large team of developers led by Norman Meyrowitz, Scholar's Workstation Group Manager. I would especially like to thank him and Nicole Yankelovich, our Project Coordinator, for their continual resourcefulness, tireless effort, and unfailing good humor. I would also like to thank Helen DeAndrade, Tim Catlin, Page Elmore, Charlie Evett, Matt Evett, Ed Grossman, Nan Garret, Karen Smith, Tom Stambaugh, and Ken Utting for their contributions to the Intermedia system and David Cody, Glenn Everett, Suzanne Keen Morley, Kathryn Stoekton, and Robert Sullivan for their contributions to Context32.

The work described in this paper was sponsored in part by a grant from the Annenberg/CPB Project and a joint study-contract with IBM.

VI References

[Conk86] J. Conklin. “A Survey of Hypertext.” MCC Technical report Number STP-356-86, October 23, 1986.

[Garr86] L. Garrett and K. Smith. “Building a Timeline Editor from Prefab Parts: The Architecture of an Object-Oriented Application.” OOPSLA '86 Proceedings. Portland, Oregon: 1986.

[Land86] G. Landow, D. Cody, G. Everett, K. Stockton, and R. Sullivan. Context32: A Web of English Literature. Providence, Rhode Island: Institute for Research in Information and Scholarship: Brown University, 1986.

[Land87] G. Landow. “Context32: Using Hypermedia to Teach Literature.” Proceedings of the 1987 IBM Academic Information Systems University AEP Conference. Milford, Connecticut: IBM Academic Information Systems, 1987.

[Lars86] J. Larson. "A Visual Approach to Browsing in a Database Environment." IEEE Computer, June 1986.

[McLu62] M. McLuhan. The Gutenberg Galaxy: The Making of Typographic Man. Toronto: University of Toronto Press, 1962.

[McLu64] M. McLuhan. Understanding Media: The Extensions of Man. New York: McGraw-Hill, 1964.

[Meyr85] N. Meyrowitz. “The Intermedia System: Requirements.” Providence, Rhode Island: Institute for Research in Information and Scholarship, Brown University, September 1985.

[Meyr86] N. Meyrowitz. “Intermedia: The Architecture and Construction of an Object-Oriented Hypermedia System and Applications Framework.” OOPSLA '86 Proceedings. Portland, Oregon, 1986.

[Ong82] W. Ong. Orality and Literacy: The Technologizing of the Word. London: Methuen, 1982.

[Scho85] R. Scholes. Textual Power: Literary Theory and the Teaching of English. New Haven: Yale University Press, 1985.

[Yank85] N. Yankelovich, N. Meyrowitz, and A. van Dam. “Reading and Writing the Electronic Book.” IEEE Computer, October 1985.

[Yank87] N. Yankelovich, G. Landow, and D. Cody. “Creating Hypermedia Materials for English Literature Students.” SIGCUE OUTLOOK, September 1987.

[Yank87b] N. Yankelovich, B. Haan, and S. Drucker. “Connections in Context: The Intermedia System.” Providence, Rhode Island: Institute for Research in Information and Scholarship, Brown University, 1987.