>Note: This is a temporary version of this file, crudely >dumpted into ASCII text, losing illustrations that were >done with PC line and box drawing characters. Word Perfect > 4.2 for DOS was a long time ago! APPROACHING HYPERTEXT: COGNITIVE MODELS AND USABILITY OF AN ELECTRONIC BOOK TECHNOLOGY by Robert B. Stepno Submitted to the Faculty of Wesleyan University in partial fulfillment of the requirements for the degree of MASTER OF ARTS IN LIBERAL STUDIES May 27, 1988 Essay Advisor: Marc M. Sebrechts Acknowledgements This essay is dedicated to Russ Walter and the Wesleyan University Computing Center, whose courses and equipment brought me into the Graduate Liberal Studies Program, and to Marc Sebrechts and Mary Ferguson, who patiently helped me get out of it in the end. (Russ and company also brought me into the computer age, but it looks like there's no easy way out of that.) Special thanks to Howard Eglowstein of Buffalo, N.Y., and Hindsight, Inc., of Enfield, Connecticut, for capturing Ted Nelson on tape, for the use of desktop publishing facilities, and for technical support; to J. Scott Johnson of NTERGAID, R. John Slade of Knowledge Garden, and Michael Stanford of Personal Computer Support Group (PCSG), for donating hypertext programs for use in this and future writing projects, and to Owl International, whose Guide software first put hypertext on my computer screen. (The Bibliography includes a more complete list of computer software products mentioned in this essay, or used as research tools, writing tools, and distractions.) Preface To paraphrase a former President, "I am not a cognitive psychologist." Nor am I a computer scientist. Some tools of those professions are used or discussed in this essay, but not with a professional's skill or intentions. Rather than a scientific attempt to prove a formal thesis, this essay is a report from the borders of several fields--including cognitive science, computer science, education, communication, and writing. What do you call someone who engages in this sort of border-walking? By profession, I've been a journalist and technical writer, by education "an English major" and an anthropologist. At Wesleyan University such interfield activity is encouraged in a program called "Liberal Studies." Ted Nelson, whose work inspired this essay, uses the phrase "systems humanist."[1] It's not a title I expect to see in the help-wanted pages of the New York Times on graduation day, but I may put it on my resume, just in case. Bob Stepno May 27, 1988 =============================================================== Note: This is an ASCII text version of the original document, which made use of text-graphics and other illustrations that have not been converted to fit this format. The page numbering indicated in the Table of Contents refers to the "official" version of this paper filed as my final essay with the Graduate Liberal Studies Program at Wesleyan University. =============================================================== TABLE OF CONTENTS Acknowledgements. . . . . . . . . . . . . . . . . . . . . . i Preface . . . . . . . . . . . . . . . . . . . . . . . . . . ii PART I: HYPERTEXT . . . . . . . . . . . . . . . . . . . . . 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . 1 CHAPTER 1: HYPERTEXT IDEAS . . . . . . . . . . . . . . . . . 2 THE INFORMATION AGE . . . . . . . . . . . . . . . . . . . 3 HYPERTEXT PROPHETS: BUSH, ENGELBART AND NELSON. . . . . . . 4 Bush. . . . . . . . . . . . . . . . . . . . . . . . 5 Engelbart . . . . . . . . . . . . . . . . . . . . . 7 Nelson. . . . . . . . . . . . . . . . . . . . . . . 8 CHAPTER 2: NONSEQUENTIAL TEXT. . . . . . . . . . . . . . . 10 HYPERTEXT ON PAPER. . . . . . . . . . . . . . . . . . . . 10 WORDS ON COMPUTERS. . . . . . . . . . . . . . . . . . . . 13 Word processing . . . . . . . . . . . . . . . . . . 14 Databases and operating systems . . . . . . . . . . 15 Outline processors. . . . . . . . . . . . . . . . . 16 Figure 1: OUTLINING . . . . . . . . . . . . . . . . . . . . 17 TOWARD HYPERTEXT. . . . . . . . . . . . . . . . . . . . . 18 CHAPTER 3: HYPERTEXT SYSTEMS . . . . . . . . . . . . . . . 19 FROM AUGMENT TO ZOG . . . . . . . . . . . . . . . . . . . 19 INTRODUCING GUIDE . . . . . . . . . . . . . . . . . . . . 22 Structures. . . . . . . . . . . . . . . . . . . . . 25 Feedback. . . . . . . . . . . . . . . . . . . . . . 27 Graphics. . . . . . . . . . . . . . . . . . . . . . 30 HYPERTEXT FEATURES. . . . . . . . . . . . . . . . . . . . . 30 PART II: MODELS FOR HYPERTEXT . . . . . . . . . . . . . . . 34 Introduction. . . . . . . . . . . . . . . . . . . . . . . 34 "Three Rs": writing, reading and research . . . . . . . . 35 CHAPTER 4: WRITING AND HYPERTEXT . . . . . . . . . . . . . 37 MODELING THE WRITING PROCESS. . . . . . . . . . . . . . . 37 Figure 2: WRITING . . . . . . . . . . . . . . . . . . . . 38 APPLYING THE MODEL TO HYPERTEXT . . . . . . . . . . . . . 41 A. The Task Environment, with hypertext . . . . . . . . 41 B. Writing Processes, with hypertext . . . . . . . . . . 47 C. Long-term memory and hypertext . . . . . . . . . . . . 58 MENTAL MODELS FOR HYPERTEXT TOOLS . . . . . . . . . . . . 59 Tarzanning: a questionable model. . . . . . . . . . . . . 63 CHAPTER 5: RESEARCH, READING AND HYPERTEXT . . . . . . . . 68 A COGNITIVE MODEL OF READING. . . . . . . . . . . . . . . 68 Figure 3: READING . . . . . . . . . . . . . . . . . . . . 70 APPLYING THE MODEL TO HYPERTEXT . . . . . . . . . . . . . 73 A. The Task Environment, with hypertext. . . . . . . . . 74 B. The Reading Process, with hypertext . . . . . . . . 90 Figure 4: The Talmud as hypertext . . . . . . . . . . . . 93 VISIBILITY AND MENTAL MODELS. . . . . . . . . . . . . . . 98 Another model: "interactive fiction". . . . . . . . . . .100 CHAPTER 6: CONCLUSIONS AND SPECULATIONS. . . . . . . . . . 104 About text models for hypertext . . . . . . . . . . .104 About other models. . . . . . . . . . . . . . .. . .105 About the future of hypertext . . . . . . . . . . .107 About hypertext and liberal studies . . . . . . . .109 About the limitations of hypertext. . . . . . . . .112 About augmenting the mind . . . . . . . . . . . . .114 APPENDIX: SAMPLE TEXTS, MENUS AND SCREENS . . . . . . . . .117 BIBLIOGRAPHY. . . . . . . . . . . . . . . . . . . . . . . .126 NOTES =============================================================== PART I: HYPERTEXT Introduction This essay is an exploration of new (and not-so-new) ideas about reading and writing with computers, including suggestions for applying research on conventional reading and writing to recent developments in electronic text technology. Part I introduces that technology, which risks going by the name "hypertext" in an industry where the prefix "hype" suggests national advertising campaigns and expensive overstatement. The three chapters of Part I are an introduction to hypertext concepts and products, including the program called Guide from Owl International, Inc., which is used as an example throughout the essay (including Part II, which compares hypertext concepts to cognitive models of conventional reading and writing). CHAPTER 1: HYPERTEXT IDEAS With claims as grand as "you are participating in the birth of a new era," microcomputer companies have begun to promote "hypertext" concepts--the idea that the computer can be a more powerful tool for reading and writing than conventional printed texts. Hypertext programs use the computer screen to display documents and relationships between them in ways that cannot easily be represented on paper. The author or reader can use a variety of techniques to annotate, cross-reference and retrieve documents, creating or following trails of association through the computer's electronic memory. For more than 40 years, writers and technologists have proposed new ways to store and cross-reference our explosion of scientific literature, business documents, and other written information. But managing paperwork isn't the only goal of hypertext. Its early proponents discuss their larger goals in phrases like "expanding human memory" or "augmenting human intelligence." Today's increasingly powerful, inexpensive computers and new data-storage technology are helping software developers turn some hypertext ideas into reality. Realizing the larger goals is not a job for computer programmers alone. It will take cognitive scientists, teachers, readers and writers of all kinds, as well as designers, publishers, librarians and others who work with printed word. Toward that end, this essay by a non-programmer explores the ways we use and think about old-fashioned texts, and ways we might transfer that understanding to reading and writing hypertext. As examples, it discusses several hypertext systems, especially their usefulness for education and research. THE INFORMATION AGE Computer storage of texts (as well as numerical information) has been possible for four decades. Use of computers for text processing became widespread in the past dozen years, from the modern newspaper with its digitized typesetting facilities to the small office with its desktop word processing computers. Until recently the idea of keeping encyclopedias and reference libraries on computers implied large, expensive, systems that could be used only by highly trained individuals. Today's optical disc[2] storage (CD ROM, Compact Disc Read-Only Memory) makes such systems possible at a cost within the reach of schools, libraries, small businesses, and many individuals.[3] A coat-pocket-sized CD ROM can hold 550 megabytes, the equivalent of 150,000 text pages or two tall filing cabinets of business records (Laub, 1986). Hypertext is one of the most promising methods of using such new high-capacity storage media efficiently and creatively (Brown, 1986). On standard magnetic media, microcomputer hypertext products are already being used successfully for smaller documents, such as program documentation (Binder, 1988). HYPERTEXT PROPHETS: BUSH, ENGELBART AND NELSON Even a hypertext has to start somewhere. The term itself was coined by Ted Nelson in 1965. He credits Douglas Engelbart of Stanford Research Institute with independently arriving at similar ideas and inventing tools that made hypertext systems possible and practical. Engelbart and Nelson both point to a 1945 essay by Vannevar Bush as a predecessor of their ideas. The remainder of this chapter discusses the thinking of these three men, their goals and their models of how people work (or could work) with computers and the written word. Bush The idea of building an associative tool to cope with the information age did not begin with the computer. Vannevar Bush, President Roosevelt's science advisor, foresaw the basic elements of such a system in his 1945 Atlantic Monthly article, "As We May Think." Bush said that alphabetical and numerical indexing was becoming inadequate in the face of "the enormous mass of the common record," particularly scientific research papers. Our ineptitude in getting at the record is largely caused by the artificiality of systems of indexing . . . information is found (when it is) by tracing it down from subclass to subclass. . . . one has to have rules as to which path will locate it, and the rules are cumbersome. . . . The human mind does not work that way. It operates by association. With one item in its grasp, it snaps instantly to the next that is suggested by the association of thoughts, in accordance with some intricate web of trails carried by the cells of the brain. . . . (Bush pp. 14-15) The device Bush proposed, the "memex," would extend human memory by storing thousands of documents.[4] It would hold not only the owner's business correspondence, notes and memoranda, but newspapers, periodicals and full sets of reference works: legal decisions, patent records, medical case histories, historical and scientific literature. The memex user would construct a "trail" through the documents, cross-referencing them at a press of a button, or adding marginal notes and comments. The essential feature, Bush wrote, is "associative indexing," in particular "a provision whereby any item may be caused at will to select immediately and automatically another. . . . The process of tying two items together is the important thing. (Bush, p. 16) Bush predicted that "wholly new forms of encyclopedias will appear, ready-made with a mesh of associative trails running through them, ready to be dropped into the memex and there amplified." Great thinkers then would be able to pass on to future generations not only their additions to human knowledge, but the trail they followed in developing those ideas, he said. Engelbart Among those who read Bush's Atlantic Monthly article in 1945 was Douglas Engelbart, a navy radar technician about to return to civilian life. Engelbart became an electrical engineer and computer scientist, and set to work with the goal of using the new digital computers to "augment human intellect." He was among the first to foresee the value of interactive computer systems for text processing as part of his more general goal of improving comprehension and problem-solving through better information-handling devices. (Engelbart, quoted in Rheingold, p.182) Going a step beyond Bush, Engelbart described a portable device that could store in one of its memory cartridges "the equivalent of an unabridged dictionary." He invited his readers to consider the impact such a machine might have: What changes in language and methodology might not result? If it were so easy to look things up, how would our vocabulary develop, how would our habits of exploring the intellectual domains of others shift, how might the sophistication of practical organization mature (if each person could so quickly and easily look up applicable rules), how would our education system change to take advantage of this new external symbol- manipulation capability of students and teachers and administrators? (Engelbart, 1963; quoted in Rheingold, 1986, p. 186) In 1968, Engelbart demonstrated a prototype "oNLine System" or NLS that not only allowed him to manipulate text on a computer screen, but to navigate a network of documents by pointing at symbols with a "mouse" (his invention) and by opening windows on a display screen. The screen image was projected at 20 times its original size, large enough to be viewed by an auditorium audience with dramatic impact: His manipulations maneuvered the screen display and the audience's consciousness through categories of information, zoomed down into subcategories, broke them into their atomic components, rearranged them, then zoomed back up the hierarchy to meet the vocal narration at a key point in the story, when the words on the screen and the words coming from the narrator merged before branching off again. (Rheingold, 1985) Many of Engelbart's new ideas--word processing, partitioning a display into windows, outline processing and manipulation of icons with a mouse--have become general tools on small computers. Now they are also part of new microcomputer hypertext systems as well as large-computer hypertext systems including Engelbart's NLS/Augment. Nelson In a more roundabout fashion, Ted Nelson also decided in the 1950s to develop new tools for handling text with computers. His ideas for hypertext (he coined the term in 1965) grew out of his unfinished attempts to write a word processing program in college, and out of his personal need to organize notes and drafts for other unfinished projects ranging from film-making to a philosophical system. In one early article he described the main concepts of hypertext as ". . . a combination of natural language text with the computer's capacity for interactive branching, or dynamic display . . . of a nonlinear text." This kind of text, he observed, "cannot be printed conveniently on a conventional page." (Nelson, 1967) Nelson's hypertext, like Engelbart's NLS/Augment and Bush's memex, ultimately involves a grand design for extending our control of information. He proposes a master hypertext database called XanaduTM, able to link millions of documents world-wide (Nelson, 1981, 1988). As a software developer, Nelson most recently has focused on the Xanadu data structure, not the "front end" or user interface for hypertext programs. His books Literary Machines (1981) and Computer Lib/Dream Machines (1974, reprinted 1987) are experiments in using hypertext techniques between paper covers. Some of these techniques will be explored in the next chapter, along with the first hypertext computer programs. CHAPTER 2: NONSEQUENTIAL TEXT Hypertext systems combine new ideas about designing or structuring text with the technology of presenting text on a computer screen. This chapter discusses an early attempt at presenting hypertext design in a printed text (Ted Nelson's Literary Machines), and developments in word processing and the more hypertext-like outline processing. HYPERTEXT ON PAPER Literary Machines introduces the basic hypertext functions and the terminology used by most hypertext programs, including "buttons," "notes" and "links." Links associate two blocks of text in much the way that Bush described the memex's associative indexing. In a hypertext edition of this essay, for example, this paragraph could contain a link to part (or all) of the original text of Nelson's book. Link markers are frequently called "buttons," and can be specially marked words or graphic devices. For example, the words "See Nelson, 1987" could be used as a button linked to a large quotation, such as the following indented paragraphs (from Literary Machines) describing types of hypertext links. In a hypertext, the paragraphs could remain hidden from view unless the reader wanted immediate exposure to the primary source material. In this typed manuscript, our "link" choices are to use the convention of indenting a long quotation, to photocopy it and include it as an illustration or appendix, or simply to print a footnote[5] referring to this hard-to- find, privately published book.[6] LINKS ARE PART OF THE WRITING A link is simply a connection between parts of text or other material. It is put in by a human. Links are made by individuals as pathways for the reader's exploration; thus they are parts of the actual document, part of the writing. As perhaps the simplest type of link, a user may create book-marks--places he may want to re-enter text when returning to it. JUMP-LINK As another simple first example, let us simply think of a link as some sort of a jump opportunity, like a conventional footnote. An asterisk, say, signals that "there's something to jump to from here." If you point at it with your lightpen (or mouse or whatever), Bingo!--you're now at the footnote, or whatever else the author took you to. If you don't like it there, hit some sort of a Return Button . . . you are back where you were and no harm has been done. MARGINAL NOTES, SIDE-BY-SIDE WRITING Marginal notes are another simple and important type of link. (Where the "margin of the screen" is--that is, how to show them--is a matter particular to your own screen setup.) A user may also make side-by-side connections of other types. On contemplating any two pieces of text, he may make a link between them. Thereafter, when he displays either piece of text, and asks to see the links, a link-symbol is displayed, and the other attached text--if he wishes to see it. Naturally, making a marginal note consists of writing the note and hooking the link. HYPERTEXT The link facility gives us much more than the attachment of mere odds and ends. It permits fully non-sequential writing, or hypertext. (Nelson, 1981, p. 2/21) Literary Machines was originally published as a typewritten manuscript with simple pen-and-ink drawings.[7] The experimental hypertext book had several Chapters One, one Chapter Two, several Chapters Three, and instructions advising the reader to take various paths through the chapters--not to simply read it once from front to back. The pages of some chapters were split into "novice" and "expert" versions of the same material using dividing lines and different size type, as well as more conventional "links" in the form of footnotes and "see also" references. Readers were encouraged to read the book several times, "passing repeatedly through Chapter Two, until you understand this book." (Although the pages of Literary Machines were numbered, the book did not use another of sequential text's main tools for navigation and cross- referencing: it had no index.) WORDS ON COMPUTERS Although Literary Machines was a printed text, it emphasized that true hypertext would be electronic, that "Most, or 'all,' of our reading and writing can or will, in this century, be at instant-access screens" (Nelson, 1981, p. 1/2). But before there could be hypertext on computers, there had to be text on computers. Computers have been used to write, edit and store text since the development of the first programming languages. Into the 1970s, programmers gave instructions to the computer on punch cards or with line-printing teletype terminals. Separate programs made it possible to print "listings" of these texts with upper- and lower-case letters, paragraph formatting, variable margin settings, and other print characteristics. Other programs made it possible to send text between computers and through national and international networks of computers as "electronic mail." (See Rheingold, 1985, on the development of ARPANET, and Fuchs, 1983, on BITNET.) Text editors, word processing and electronic mail systems all offer basic text entry and editing capabilities, plus abilities to search for and change text. The early editors coped with limitations of the current data entry and display devices--card readers, teletypes and upper-case-only display screens. (See Schneider, 1974, for a chronicle of one humanist's experience with this rapidly changing technology.) Even as the displays improved, editor programs frequently kept the numbered lines and line-oriented editing commands of the earlier systems. Word processing In the past ten years, editing and print-formatting programs have been integrated into increasingly sophisticated "word processing" and "desktop publishing" systems. Today's computer screens not only display upper- and lower-case letters, but can simulate a wide variety of printed type fonts and sizes on screen. However, systems designed to send text to a printing device, often reflect the limitations as well as the abilities of the printer. For example, word processors commonly display text in the line-length, spacing and (with high-resolution screens) type font style of the printed version--none of which may be the optimum display for reading the text on screen. Electronic mail systems are designed for reading on screen, but since communication systems must be used by many different types of computers, they are frequently limited to displaying "lowest common denominator" text-- plain alphanumeric characters lacking even the simplest display attributes, such as underlining or bold type. The same is true of texts written for interpretation by the computer itself, such as a FORTRAN or Pascal program--only the plain text characters are used by the programming language's compiler or interpreter. Databases and operating systems In addition to displaying text, hypertext involves linking information. Computers also have been used to link texts to each other since the beginning of data processing. Database programs have the ability to connect items of information within and between files. Computer operating systems provide file management and (in the Apple Macintosh, for example) the ability to represent files of text information symbolically. Databases often use indexing and search techniques, rather than direct links, and usually involve information that can be broken into small pieces and stored in clearly defined fields. Some attempts have been made to use databases as writing tools (see Moore, Pfaffenberger) and to combine word processing and database programs, such as Nota Bene from Dragonfly Software. However, database programs do not approach the text-presentation and direct- link capabilities of hypertext software (Conklin, 1987). Outline processors An advanced form of word processing appeared on microcomputers in the mid-1980s, the hierarchical "outline processors." Such programs, include ThinkTank, Kamas, PC-Outline, and Framework. Engelbart's own Augment system includes hierarchical outlining among its features (Nelson, 1988). Outline processors provide dynamic display of texts, allowing blocks of text to be hidden behind an abbreviated display of section headings. The user can manipulate those blocks of text by moving the headings to new positions on the screen. The document can be viewed in many versions, depending on which blocks are "hidden" under their headings. For example, a term paper seen in an outline processor might appear as shown in Figure 1, below. Figure 1: OUTLINING At its topmost level, the outline looks like this: ÚÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄ¿ ³ ³ ³ 1. Introduction ³ ³ ³ ³ 2. History ³ ³ ³ ³ 3. Future ³ ³ ³ ³ 4. Conclusions ³ ³ ³ ÀÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÙ Expanded one level, the outline looks like this: ÚÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄ¿ ³ ³ ³ 1. Introduction ³ ³ ³ ³ 2. History ³ ³ 2.1 Eighteenth Century ³ ³ 2.2 Nineteenth Century ³ ³ 2.3 Twentieth Century ³ ³ ³ ³ 3. Future ³ ³ 3.1 Twenty-first Century ³ ³ ³ ³ 4. Conclusions ³ ³ ³ ÀÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÙ Each numbered item in such an outline can represent additional subsections, any of which can contain dozens of pages of text. It takes only a few keystrokes to display the full text of a section or to add a new section heading called "Present" between "History" and "Future." The writer can move all of the 20-page "Twentieth Century" chapter under the new "Present" section heading by relocating the single line of text shown on the screen. If the user changes the order of outline sections, most out- line processors renumber the sections automatically. A similar "substitution" technique in which one block of displayed text can be exchanged for another longer (or shorter) block is common in hypertext programs, with or without hierarchical outline numbering. TOWARD HYPERTEXT Outline processors, word processors, and database file managers might seem wonderful tools to Vannevar Bush, but they are still not as powerful as his "memex" proposal. And they are not yet hypertext as defined by Ted Nelson. Speaking at a computer conference in 1987, Nelson said he considered word processing and outline processing when first proposing hypertext twenty years earlier, but he had dismissed the ideas in favor of something "grander." Word processors and outline processors would be confined to the sequential, hierarchical structure of numbered pages and outlines. "There is no reason anymore to suppose sequen- tiality. . . . There is no need for there to be numbered pages on the documents we read from our screens," Nelson said, ". . . The future of our entire literature lies on screens of different sorts."[8] CHAPTER 3: HYPERTEXT SYSTEMS FROM AUGMENT TO ZOG Since Ted Nelson introduced the term, hypertext and near-hypertext programs have been developed for large and small computers. They include Intermedia, which is providing resource texts for Brown University literature and biology courses (Yankelovich, Meyrowitz and van Dam, 1985; Yankelovich, 1987); HyperTIES (The Interactive Encyclopedia System)[9], which has given novice computer users information about museum and gallery exhibits (Marchionini and Shneiderman, 1988); and WE (Writing Environment) being used at the University of North Carolina at Chapel Hill to test cognitive models of writing (Smith et. al., 1986). Other influential systems include ZOG at Carnegie-Mellon University[10] (McCracken and Akscyn, 1984), Engelbart's NLS/Augment and Xerox's Notecards, among others. (For a comprehensive survey, see "Hypertext: An Introduction and Survey" by Jeff Conklin, published in IEEE Computer, September, 1987; or in expanded form as a research report, A Survey of Hypertext, from Microelectronics and Computer Technology Corp., December, 1987.) Owl International, Inc., introduced Guide for the Apple Macintosh in 1986 and for the IBM PC in 1987, calling the program "the first commercially available 'hypertext' system for personal computers." Apple brought hypertext to a much wider audience when it began giving away a product called HyperCard with each new Macintosh in August, 1987. HyperCard combines hypertext concepts with a simple text editor, a sophisticated graphics editor, and a programming language. As its name suggests, HyperCard uses an index card or "Rolodex" metaphor, with each screen-sized display representing a "card" of text and pictures. Cards can be grouped into "stacks" and linked using text or graphic "buttons." Selecting a "next card" button, for example, displays the next card in the stack. Buttons can be programmed to link other stacks, run programs, or control a device, such as a telephone dialer or video disc player. The card metaphor has been used frequently for microcomputer datafile programs (e.g., Zoomracks, CardBox, 3x5 and MultiMate On-File), but such programs do not have the linking, object-oriented graphic interface or programming capabilities of HyperCard. Apple's "bundling" of HyperCard with its computers generated a great deal of interest in the product and hypertext in general. Ironically, this first widely-used "hypertext" program has limited ability to deal with long texts. Its index-card or slide show style of presentation is more suited to smaller chunks of information. Other hypertext programs were introduced after the widespread publicity[11] about HyperCard. Many of these are HyperCard "stacks" including databases and programs developed with Apple's new HyperTalk programming language. New programs for the IBM PC or PS/2, include NTERGAID's Black Magic and Knowledge Garden Inc.'s KnowledgePro, a programming language that is now advertised as a "HypertExpert System." Macintosh adaptations of several of the university-based mainframe systems also have been reported in trade publications in recent months, but these programs are not yet widely marketed. Unlike most of these systems, Guide is available on both IBM and Apple Macintosh computers, has been nationally advertised, and is designed to handle large text documents. For those reasons, it is an appropriate vehicle to use in this discussion of hypertext and its potential for research, education and personal use. The version of Guide designed for the IBM Personal Computer can be used on dozens of brands of machines that use the Microsoft Windows operating environment.[12] Guide provides most of the examples in this essay. HyperCard and other systems provide examples of some techniques not present in Guide. INTRODUCING GUIDE Guide is sold as a hypertext "authoring system," a tool with which the user can create as well as read hypertext documents. Owl International also provides special versions of Guide to developers of software HELP systems or CD ROM document libraries, the original application for which the program was developed (Brown, 1986). The standard commercial Guide package for Microsoft Windows includes sample documents, a tutorial and an on- screen "help" system, all written in Guide. For the Macintosh, Guide has a read-only version called a Guide Envelope, which has been used to "publish" documents in Guide hypertext format, including an edition of Ted Nelson's Literary Machines. Like all programs that operate on the Macintosh or in the Windows environment, Guide uses a hand-held "mouse" pointing device and displays text, menus and graphics on a high-resolution screen. Rolling the mouse on a horizontal surface moves a pointer on the screen; pressing a button on the mouse selects the menu item, icon, or text at the pointer. Guide and HyperCard both call objects on screen that control hypertext functions "buttons." Hence "clicking on a button" means pointing at a button on screen, then pressing and releasing the physical button on the mouse. A Guide button can be a word or phrase, a larger block of text, or an imported graphic image.[13 Guide treats the beginning and ending points of a button as part of the linear stream of characters in the computer file, like a set of quotation marks. As such, a text button can be moved if its page or paragraph is edited. In contrast, a HyperCard button marks a fixed location on the Macintosh screen, like a transparent overlay, and does not respond if the text at that location is moved. Despite its text orientation, Guide is not a word processing[14] system. It is primarily designed to display documents on a computer screen, although it does have basic printing features. Through Microsoft Windows, it allows the user to view several documents at once, and to change the size and position of each document "window" on the screen. Guide automatically wraps each line of text to fit within a new window size, and can temporarily magnify or reduce the characters. While this flexibility enhances the user's ability to read text on the screen, Guide has few of the print- oriented screen formatting features of many word processors, such as centering or justified text. It also lacks a common text-entry and editing feature--use of the IBM keyboard's cursor arrow keys instead of the mouse. (Word processing conventions are discussed in more detail in the next chapter, concerning cognitive models for working with text on computers.) The reader and writer of a Guide hypertext document interact with Guide in the same ways--by choosing commands from seven pull-down menus named at the top of the screen, and by clicking on hypertext buttons. To select a command like MAKE BUTTON the user points to a menu heading (MAKE), presses the mouse button to show the command list, and drags the mouse pointer down the list of one- or two-word command titles to the desired one (BUTTON). Releasing the mouse button executes the selected command. Moving the pointer off the menu without releasing the button cancels the command. (See the Appendix for a listing of Guide's menus and commands, and for examples of the structures and control devices discussed below.) Structures Guide's three main hypertext structures are called Replacements, Notes and References. A fourth structure, the Inquiry, is a group of Replacements. The user is kept informed of the status of the document (whether buttons are present, open or closed) through a variety of feedback mechanisms. Buttons in Guide documents can be text strings or graphic images. Buttons can look like bold-faced, underlined or italicized words, or they can be any kind of icon or image the writer's graphics programs can create. Each of the three main hypertext structures has a distinctive button. Clicking a Replacement Button causes a block of text or graphics (a "Replacement") to appear in its place. The Replacement can be a word, a paragraph, a picture or many pages. Guide automatically reformats any text surrounding the Replacement. Since Replacements can contain more Replacement buttons, Guide can use them to "open" and "close" sections of text like the outline processors discussed in the previous chapter. A Replacement can represent a paragraph or many pages with a single word or phrase. Unlike the outliners, Guide cannot number or sort items automatically, but it offers alternatives to the outliners' hierarchical structure, as Nelson insists true hypertext must (Nelson, 1987, p. DM-17). An Inquiry is a group of mutually exclusive Replacement Buttons. Clicking any one of them causes all buttons in the group to be replaced by the Replacement for the one clicked button. The Inquiry makes it possible for a document to branch into different sub-documents after a prompt like: "Are you (click one) a [Novice], [Intermediate] or [Expert]?" A reader clicking on the word "Novice" would be shown a simplified passage; a reader clicking "Expert" would be spared reading the Novice section. Pressing the mouse button while pointing to a Note Button causes a box of text (a "Note definition") to appear temporarily in the upper right corner of the screen. When the mouse button is released, the Note disappears. Clicking a Reference Button causes Guide to perform one of two actions: jump to a Reference Point elsewhere in the same document, or open a window containing another Guide document and jump to a Reference Point in that document. The user can return to the original Reference Button location by clicking a Backtrack icon near the top right corner of the document window. The Backtrack icon is highlighted when the user activates a Reference Button. The writer of a Guide document uses the Make menu to turn text or graphics into Buttons, and can use Set Options on the Display menu to change the Buttons' type style. Buttons of the same type can be similar or different in appearance, as the writer determines. Feedback Guide uses several "user feedback" conventions to indicate the presence of hypertext structures in the text. These include: * using a special type style for each kind of button * changing the shape of the mouse's pointer when it moves over a structure on the screen * highlighting the outline of the button or replacement text when the mouse is clicked * giving the user a menu command that can display the hidden symbols at the beginning and end of each hypertext structure. Type style is the simplest indicator for text buttons. By default, Guide shows Replacement Buttons in bold, Note Buttons in underlined text and Reference Buttons in italics. However, these defaults can be altered by the user (to plain, superscript, subscript or boxed text), and the same type styles can be used for their normal purpose of enhancing text. The second indicator is Guide's shape-shifting pointer. When the user moves the pointer (a vertical bar) over the text of a Guide document, the pointer changes shape to indicate whether it is over a Button, a Replacement or a graphic illustration. An octagonal pointer with vertical and horizontal cross-hairs indicates a Replacement button. A Replacement text changes the pointer to a small square, indicating that the text can be "folded back into" the Replacement Button. (The words and sentences of Replacement text or text at a Reference Point look like any text elsewhere in a document.) A large asterisk-shaped pointer indicates a Note Button. A Reference Button changes the pointer to a large arrow. If the user presses and holds the mouse button while pointing to any of these structures, Guide highlights the area involved--the text string or graphic that constitutes the button or replacement. The shape-changing pointer and the highlighting both provide feedback[15] essential to identifying interrelated Guide structures. When editing text, the user can make the beginning and ending points of Replacements, Buttons and Reference Points visible by choosing "Show Symbols."[16] Guide's symbols are smaller versions of the pointer shapes: oversized asterisks for Notes, "rifle sight" crosshairs for Replacements, and arrows for References. Reference Points are marked with similar arrows, but with the arrow shafts bent at a right angle. Inquiries are marked with two smaller versions of the Replacement symbol. As another aid to editing hypertexts, the Freeze command on Guide's Edit menus can deactivate the Buttons and the pointer's shape-changing. The manual calls this "freezing pointer feedback." The "frozen" text pointer is the vertical line used on non-Button text and common in other Windows editing systems. When the pointer is in this shape, the user can edit the text of either a Button or Replacement structure. Graphics A writer with graphics software could customize a set of icons for the button types and import them through Microsoft Windows' "cut and paste" feature. For example, the Guide Help document uses an oversized question mark as a Replacement Button for sections of additional general information, and uses graphic reproductions of Guide's own menus as Reference Buttons to call up detailed explanations of the menu items. Guide does not include a graphics editor, but it does allow the user to import, crop, move and combine graphics. (Black Magic also allows import of graphics from other programs. HyperCard provides a graphics editor, a library of images, and allows users to "cut and paste" from other programs.) HYPERTEXT FEATURES Conklin (September, 1987, p. 21; December, p. 35) has established four categories for hypertext systems, and has identified a dozen distinct features of systems already in use. Conklin's categories of hypertext programs are: 1) Large-scale literary systems, like Intermedia 2) Problem exploration tools, including outliners and WE 3) Read-only browsing systems, smaller-scale literary systems such as HyperTIES 4) General-purpose hypertext systems, suitable for exploring various applications, such as Guide. The twelve features Conklin found in various programs are: * Hierarchical structures * Non-hierarchical, cross-reference links ("graph- based") * A variety of link types * User-designated attribute/value pairs for nodes or links * Ability to string links together into a storable path or trail * More than one version of a node or link * Ability to attach executable procedures to nodes or links * Keyword or string search capability * Text editor * Availability to multiple users, concurrently * Use of pictorial or graphic information * Graphic display ("browser") or map of nodes and links On Conklin's chart of hypertext features, Engelbart's NLS/Augment and Brown University's Intermedia came the closest to having all twelve of these features. Guide and HyperCard each provide eight of the twelve features on the list. Augment lacked only the graphics display or "browser" for hypertext links. Guide lacked the "trail" and "browser" features, according to Conklin, while HyperCard lacked those features, multi-user operation and the ability to have more than one version of a node or link.[17] Brown University's Intermedia has the browser feature, but lacks several other features on the list in the version discussed by Conklin. Both Conklin and Nelson note that there are hypertext- like functions in traditional printed text, including the heavily cross-referenced sections of a modern encyclopedia or the interlinked commentaries of the Talmud. Readers and writers of linear texts also employ some of the strategies of a hypertext user as far as the medium will allow--using headlines, footnotes and bibliographies, skimming chapter and section headings for related ideas. If hypertext does become widely used, our centuries of pre-hypertext communications will help shape the cognitive models for working with this new medium. The next three chapters of this essay consider the basic skills of working with texts--the skills of reading, writing, searching and referencing information. These are the activities that users of hypertext libraries must translate from the book and bookshelf to the computer screen. ================================================================== PART II: MODELS FOR HYPERTEXT Introduction Until recently only a few universities and development centers were experimenting with hypertext systems. Now low-cost microcomputer software can introduce this new technology to a wider audience. Guide, HyperCard and other products offer scholars a chance to investigate the first stages of an important medium, and possibly witness (and contribute to) the development of new styles of written communication. The software developers recognize both the potential of their products and the need for experimentation (or at least trial-and-error) to determine the best ways to use these new hypertext tools they are creating. Both issues are suggested in the following passages from the products' own documentation and promotional literature: Any task for which you would normally use pencil and paper can be tackled using Guide. This includes such tasks as organizing, tracking, writing, communicating, designing, brainstorming, documenting, cataloging, cross referencing, annotating, arranging, outlining, presenting, animating, prototyping, and retrieving information. (Guide user's manual, 1987. p. 8) Since hypertext writing is in its infancy, there are no accepted guidelines we can give you to guarantee that your documents will be effective. One of the reasons we wrote TextPro is so many more people can experiment with these techniques and a style can begin to evolve. ("Welcome to TextPro," brochure, 1988) . . . you can end up making a document little more than a snarled mass of links. Hypertext is non-linear--the way that all of us think, but how many of us are used to writing in a non-linear fashion. This is the primary hurdle to overcome in using hypertext, the mental re- adjustment. (Black Magic user's manual, 1988, p. 9-2) Little has been written yet about the usability and design of hypertext systems or hypertext documents. This essay suggests that attempts to understand (and plan) hypertext systems should begin with study of previous attempts to understand how people use printed texts. Creators of hypertext systems or hypertext documents should look beyond the world of computer science to draw on the knowledge and methodology of other disciplines, including cognitive psychology, education, literary analysis, linguistics and communications. "Three Rs": writing, reading and research The next two chapters draw on examples from the literatures of psychology and education to discuss cognitive models of writing, reading and research[18], applying those models to the Guide system outlined in chapter 3, and to hypertext in general. By "cognitive models" I mean analytical descriptions of mental processes; that is, attempts by cognitive scientists to picture how the mind performs complex tasks. Individuals' mental models of working with text and computers also are discussed in regard to development and use of hypertext. By "mental models" I mean definitions and analogies a person uses to explain devices, concepts, tasks or situations.[19] Examples include the common use of the word "file" to describe units of computer-stored information, or the metaphor of the mechanical "button" in Guide and HyperCard[20]--a button is something you "click" to make something happen. In short, a cognitive model is an attempt to picture the internal workings of the mind, while a mental model is an attempt to picture an external device or system. The two chapters on print-based models for hypertext are followed by a final chapter of conclusions about the usability of microcomputer hypertext systems and possibilities for further research. CHAPTER 4: WRITING AND HYPERTEXT MODELING THE WRITING PROCESS Writing has been analyzed by professors of rhetoric, journalism, literature, education and cognitive psychology, and by generations of students staring at blank sheets of paper. More recently, many of us stare at blank computer screens. In simplest terms, writing consists of gathering information (sometimes "gathering one's thoughts" is enough) and putting words on paper (or into the memory of a computer). Beyond that simple description, writing styles, techniques and skill levels vary widely, and it is difficult to generalize about such a personal activity. Flower and Hayes (1981), however, have developed a cognitive model of the writing process that provides a good starting place for this discussion. This model, while not universally accepted, also has been used successfully by Gagne (1985), Pea and Kurland (198?) and Rose (1984). Flower and Hayes describe writing as a recursive process of planning, translating and reviewing, all under the control of a monitor. These processes are linked to the writer's long-term memory and to the task environment. The interrelationship of these processes is shown in Figure 2. Figure 2: WRITING ÚÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄ¿ ³ A. Task Environment ³ ³ ÚÄÄÄÄÄÄÄÄÄÄÄÄÂÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÂÄÄÄÄÄÄÄÄÄÄÄÄÄ¿ ³ ³ ³ Rhetorical ³ Text produced ³ Resources ³ ³ ³ ³ problem ³ so far ³ ³ ³ ³ ³ ³ ³ advice ³ ³ ³ ³ topic ³ ³ texts ³ ³ ³ ³ audience ³ ³ tools ³ ³ ³ ÀÄÄÄÄÄÄÄÄÄÄÄÄÁÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÁÄÄÄÄÄÄÄÄÄÄÄÄÄÙ ³ ÀÄÄÄÄÄÄÄÄÄÄÄ/\ÄÄ\/ÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄ\/ÄÄ/\ÄÄÄÄÄÄÄÄÄÄÄÄÄÙ /\ \/ \/ /\ ÚÄÄÄÄÄÄÄ¿ ÚÄÄÄÄÄÄÄÄÄÄÄ/\ÄÄ\/ÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄ\/ÄÄ/\ÄÄÄÄÄÄÄÄÄÄÄÄÄ¿ ³ C. ³ ³ B. WRITING PROCESSES ³ ³ Long- ³ ³ ÚÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÂÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÂÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄ¿ ³ ³ term > > > ³ Planning ³ Translating ³ Reviewing ³ ³ ³ Memory³ ³ ³ ³ ³ ³ ³ ³ < < < ³ set goals ³ (produce text) ³ evaluate ³ ³ ³ ³ ³ ³ generate ideas³ ³ revise ³ ³ ³ ³ ³ ³ organize ³ ³ ³ ³ ÀÄÄÄÄÄÄÄÙ ³ ÀÄÄÄÄÄÄÄÄ\/Ä/\ÄÄÁÄÄÄÄÄ/\Ä\/ÄÄÄÄÄÄÁÄÄÄ/\Ä\/ÄÄÄÄÄÄÄÙ ³ ³ ÚÄÄ\/Ä/\ÄÄÄÄÄÄÄÄ/\Ä\/ÄÄÄÄÄÄÄÄÄÄ/\Ä\/Ä¿ ³ ³ ³ Monitor ³ ³ ³ ÀÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÙ ³ ÀÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÙ Figure 2: Adapted from Flower and Hayes, after Gagne, and Pea and Kurland. The arrows suggest bidirectional flow of these recursive processes.[21] The A, B and C labels are for convenience in referring to paragraphs in the text, and are not intended to imply sequence of operation. Pea and Kurland note that this model of writing departs from models that stress rigid linear processes of planning, writing and rewriting. Such older models are reflected in some writing teachers' insistence on detailed hierarchical outlining as a necessary planning step. The recursive aspect of the Flower and Hayes model fits the more flexible writing process described by various professional writers and writing teachers. (See bibliography for works by Barzun, Becker, Elbow, Franklin, and Zinsser.) Writers cannot always work from general outlines to specific details; they plan, write, modify goals and plans, revise, write and plan again as inspiration strikes or new information becomes available. Rose (1984) calls this process "opportunism," borrowing the term from a cognitive study of planning (Hayes-Roth, 1979), because the writer's observations and decisions can suggest new opportunities at any point in the writing process. Gagne emphatically says (1985, p. 202), "Anyone who has attempted writing knows that it does not proceed in a linear sequence . . ." She suggests that outward signs of "linear" writing--even producing masterpieces in a single draft--conceal mental processes of review and modification before the words reach paper. The components of the model need more detailed definition before we relate them to hypertext. A. The task environment has three parts: 1. The writer's rhetorical problem definition, such as the topic, audience and exigency of the writing project. 2. New text as it is produced. 3. Resources available to the writer, a category added to the original model by Pea and Kurland. They include sources of advice and information, such as a student writer's teacher; reference materials such as notes, index cards, dictionaries, and other publications; and writing tools of all kinds, including computer systems. B. The writing process has three parts, under the control of its monitoring process: 1. Planning includes setting goals, generating and organizing ideas based on the given problem definition. 2. Translating is the actual process of putting thoughts into words and words onto paper. 3. Reviewing consists of evaluating and (when appropriate) revising the new text. C. Long-term memory includes the writer's knowledge of the topic and audience, as well as his or her writing plans. APPLYING THE MODEL TO HYPERTEXT The sections below discuss how working with hypertext fits each part of the cognitive model of writing. (As before, the letters refer to parts of the Figure 2.) A. The Task Environment, with hypertext A hypertext system would be a major part of the writer's task environment as outlined above, whether that writer planned to create a hypertext document or use the hypertext system to write for conventional publication. A hypertext system would contribute to problem definition, store new text, and provide new resources in the form of both source material and writing tools. 1. Problem definition At one side of the Task Environment in our model, writing with hypertext adds new issues to the writer's problem definition: Will the intended audience have access to a hypertext system? If so, is the final product to be a hypertext itself, or a document for traditional publication? If the document will be a hypertext, will it address a homogeneous audience, or should it have separate hypertext paths for readers of varying backgrounds or authorities? The Guide tutorial, for example, walks the user through the process of creating a product marketing report that includes some information "for management." In theory this branching hypertext document could be used to print separate versions of the report for the managers and the sales force, or to show the appropriate parts of the report on screen during meetings with their respective constituencies. Conklin (1987) suggests that a rich hypertext system could threaten a writer with too much "cognitive overhead" by offering many options for linkages, branching and other features. In some systems, he says, the writer might be required to give a name to each link, interrupting an already-branching train of thought to think up a summarizing word or phrase. Careful system design can minimize such distractions. In Guide's case, for example, additional document windows are automatically opened as "UNTITLED-1," "UNTITLED-2" and so forth. The author can write freely and return to name the new documents when that is convenient. Guide documents are free-form texts, but some hypertext programs, may have more restricted document formats. HyperCard (Goodman, 1987) and ZOG (McCracken, 1984) both build documents from units the size of the computer display screen. Concordia (Sandstrom, 1988, Walker, 1988), a hypertext-based software documentation system developed in-house, stores information in modular "records" on a database model, and provides hypertext links between those records. Even when the hypertext system does not require text to be entered in records or "chunks," a modular approach in document design would support robust use of hypertext linking features. Guide's on-line HELP system and tutorial, for example, both use Reference links into a separate file of short text chunks that explain Guide commands and features, rather than providing their own definitions of the same terms. With hypertext capability, these structuring considerations would become part of a writer's problem definitions, and would in turn become part of the planning process described in section B, below. 2. New-text storage The hypertext system obviously becomes the storage and display medium for new text as it is produced. Its versatility at displaying text in different ways could help the writer keep track of what has already been accomplished. (These display features are discussed in more detail in the next chapter on reading hypertext.) 3. Resources Source material A hypertext system could store and display information sources such as encyclopedias, research materials, style guidelines, outlines and notes. The writer could freely annotate, link and copy from that database. (This "active reading" aspect of hypertext will be discussed at greater length in the subsequent chapters about reading and researching.) Writing tools Hypertext should provide writing tools similar to those of the best word processing systems[22], or facilities for crossing between the two. Word processors offer writers many advantages over conventional typewriters[23], even if one ignores the computer's ability to print final documents of typeset quality. In the past two years, most "full- featured" programs (e.g., Word Perfect, WordStar 2000, MultiMate Advantage, and Microsoft Word) have added spelling correction, thesauruses, automatic formatting, simultaneous viewing of two or more documents, and the ability to handle very large documents. Less common, but increasingly available are grammar- or style-checking features, and programs that allow groups of editors to comment or revise a single document. Our sample system, the Microsoft Windows version of Guide, is limited as a text-entry or word processing system. Guide's reliance on the Microsoft Windows environment currently limits its ability to transfer files, since manufacturers are just beginning to release full- featured word processing products that take advantage of the bit-mapped Microsoft Windows graphic interface. Guide and other programs that use Microsoft Windows have the advantage of sharing not only text but a wide variety of graphic images, including those created on the computer or imported from other media using photocopy-like digitizing devices. As a text-creation system, Guide lacks many common word processing features (such as use of the cursor-arrow keys), as well as advanced word processing features (such as spelling correction or a "macro" ability to store and repeat groups of commands). Its ability to format printed text is limited to a few type styles, and it can only print one file at a time, regardless of hypertext links to other files. High-quality printed versions of Guide documents currently require another round of file conversion. As a text-presentation system, Guide can only use its hypertext file-linking features with documents that have been converted to Guide's own document format. It can import so-called ASCII text files that can be created by most word processors. However, many popular word processors have print-oriented features that do not translate easily to and from Guide format or ASCII. These include use of mixed type fonts and sizes, text centering, automatic indentation, multiple-column pages, and numbered bottom-of-page footnotes. Owl International, the developer of Guide, sells advanced conversion services to large clients, such as companies transferring reference works to CD ROM storage. Such abilities may be part of the basic product if and when Guide finds a larger audience. A program like Guide gives the writer structural tools that make and connect hypertext links, and representational tools that let the writer present information in new ways, such as dynamic outlines made of "replacement" buttons, pop-up "note" windows, "inquiry" and "reference" buttons that offer alternate paths through a text, and graphic icons linked to text replacements. Some hypertext features that are not yet available in Guide include graphic maps of links between documents, color-coding of information and structures, the ability to store and replay a path through the hypertext document database (although Guide can backtrack through links it has opened) and the ability to run other computer programs and multi-media devices. A writer can use these tools to navigate in or manipulate notes and drafts, as well as making them part of the final hypertext document. In addition to creating a text, the hypertext system is the medium for storing and viewing text as it is produced, which Flower and Hayes point out as another part of the writer's task environment. With a program like Guide, recently-produced text can be viewed with the same flexibility as other hypertext documents--using multiple windows and other display enhancements as discussed under Planning, below, and in the next chapter on reading hypertext. B. Writing Processes, with hypertext From the task environment, a hypertext system would provide tools to aid the planning, translating and reviewing parts of the writing process, all under control of the writing process monitor. As mentioned before, the writer's planning, translating and reviewing processes, like hypertext itself, are not linear. Sometimes writing occurs before planning, or something the writer reads prompts a change in carefully made plans. Since translating is the actual production of text, it is discussed first in the paragraphs below. Translating, with hypertext In the Flower and Hayes model, "translating" is the actual writing-down of the author's thoughts. The writer must juggle many components in memory to perform this translation--not only the original problem, but the recently-created text and technical prose tools like spelling and punctuation. For experienced writers, those tools become somewhat automatic or proceduralized, as Gagne points out (p. 215). The addition of hypertext tools may increase this burden on working memory, a concern raised by Conklin in his survey of hypertext programs, as has already been mentioned. Guide and HyperCard, however, may have reduced this "cognitive overhead" problem by using the mouse-menu-icon operating environment familiar to users of any other Macintosh or Microsoft Windows program. Research findings were initially mixed concerning such "direct manipulation" or representational computer operating systems (e.g, Carroll, 1986; Shneiderman, 1983; Hutchins et.al, 1986; Ewing et. al, 1986), but the popularity of the Macintosh appears to have convinced companies like IBM and Microsoft that their customers believe such systems are easier to use. For a writer using a hypertext system, the translating process would include not only "putting words on paper" but putting structures in place that could depict relationships that previously might have been difficult to express with words alone. For example, Guide's HELP system includes a graphic representation of Guide's own main menu. Each item on this HELP version of the menu is actually a Guide Replacement or Reference button. Clicking on any item displays the secondary command menu, and clicking on any command opens a Reference window giving a text description of the command, often linked to a separate file of examples. Similarly, a Guide version of an anthropologist's map or kinship chart could link each word or name in the chart to a separate window of detailed ethnographic information. This essay deals primarily with text, but hypertext programs are increasingly moving into hypermedia--featuring not only graphics, but audio and video animation. Because of this, and the increased emphasis on graphical computer interfaces in general, future studies should compare the expression of meaning in hypertext to meaning in other symbol systems, drawing on research in art, semiotics, linguistics and non-verbal communication. Planning, with hypertext Planning includes goal-setting, generating ideas and organizing those ideas for translation into written words. Goal-setting--At the goal-setting level, the writer interprets his or her rhetorical problem--the reason for writing. The discussion of the task environment has suggested that hypertext itself can be part of the problem. It can also be part of the solution. In some cases goal- setting is an internal mental process. In other cases, writers draw diagrams, doodle, or compose "what if" scenarios to help set their goals. Many writers set their goals and generate ideas for writing projects by "thinking on paper," as discussed by writing teachers Elbow, Dumaine and Hairston, among others. Elbow encourages "freewriting" as a way to generate ideas by generating text, and insists "Meaning is not what you start out with but what you end up with." (p. 15) As a flexible writing tool, hypertext could become a natural part of this process. Generating ideas--Gagne speculates that some of the differences in the amount of ideas writers generate are based in differences of declarative knowledge, skills at management of working memory, and the presence or absence of cues that prompt continued production of ideas, particularly in children just beginning to write (1985:206- 7). Hypertext could serve as a database of declarative knowledge, and as a display system to supplement working memory and provide prompts and cues during the idea- generating process. In fact, early outline-generating programs like ThinkTank were advertised as "idea processors" because of their usefulness in shifting the relationships of small pieces of text on screen, providing a graphic aid to thinking about the relationships of the ideas the text represents. Hypertext programs like Guide can perform similar functions without the possibly- misleading hierarchical outline numbers produced by outlining programs. Organization--At the organizing level of the planning process, outliners and hypertext are both useful. The hypertext programs, by definition, open up new possibilities for document organization, such as moving text to a separate linked document, or to a Note window. Some writers start with sentences, others with outlines and diagrams. Writers and teachers vary in their preference for one technique or another, as reflected in the works of Barzun, Elbow and the other writers-on-writing mentioned earlier. Writing teacher, journalist and editor William Zinsser, for example, warns students against being trapped by their original outline: "Don't ever become the prisoner of a preconceived plan. Writing is no respecter of blueprints--it's too subjective a process, too full of surprises." (1985, p.63) But Zinsser is writing about the process of composition; for the end product, he warns against vagueness, clutter and disorderly thinking and encourages clarity, simplicity, and logical sequence. This subjective nature of writing makes it difficult to model the cognitive process. Flower and Hayes have attempted to do so by having subjects "think aloud" while writing, but it is easy to imagine this self-conscious procedure altering the writer's thinking--and writing-- process. Writers themselves may contribute to our understanding of the writing process. For example, Jon Franklin, journalism professor and two-time winner of the Pulitzer Prize, has published a careful analysis of his own method of writing dramatic non-fiction stories, with heavy emphasis on his planning process. (He outlines and reoutlines, using terse statements of plot complication, developments and resolution.) Novelist and historian Paul Horgan, journalist and editor William Zinsser, and sociologist Howard Becker are a few of the many writers who have made self-analysis a part of their writing textbooks. The self-conscious creation of a new communication medium--a new expressive medium--is a job for art as well as science, and the two could benefit from sharing their insights and intuitions. Hypertext ideas, for example, could prompt literary scholars to bring new light to the works of James Joyce and other writers, and to bring their ideas back to creators of hypertext. Writers work in forms based on tradition, rules of grammar and logic. Conventional texts have one beginning, a middle and an end--or attempt to depart from that model in creative ways. The Irish journalist and novelist Flann O'Brien has his narrator begin At Swim-Two-Birds with the observation that a book could well have three beginnings, which his does, weaving three plots into a pre-hypertext tangle. There are cultural models to be followed (or creatively broken) for writing poetry, prose, fiction and non-fiction, journalism, how-to books and dessert recipes. Hypertext adds new dimensions to the planning process: the options of creating links, replacements, references or whatever other hypertext structures are permitted by the hypertext system software. But is this a completely new phenomenon? Even with conventional text, writers and editors regularly face questions like "should I include this?" or "where should this go in relation to other information?" One way to lay the groundwork for such studies of hypertext would be to record thinking-aloud protocols of editors and writers as they design a magazine or newspaper "package" of main and related stories, sidebars, boxes, headlines, graphics and photographs. Such a study, much like the knowledge-collecting process in developing an expert system, could disclose types of relationships comparable to the lists of potential hypertext "link types" identified by Conklin (1987) and Trigg (1983). Ted Nelson says that writing of any kind is a form of design, even the writing of a single sentence or paragraph, and that hypertext simply gives the designer more tools to work with. But we know very little about expert writers' process of designing those conventional sentences, paragraphs and documents. The staff of the Document Design Center, American Institutes for Research, has observed that more research is needed to determine how to train writers and designers of conventional "informative" documents. Such research should contribute to our understanding of electronic documents as well. Research into the composing processes of writers is relatively recent, and little of it focuses on working adult writers. In particular, we need research on the processes of people who write complex documents. Do writers who produce usable policy statements, regulations, or instructions have similar composing processes? Are the processes needed to draft effective documents the same whether the product is a letter, memo, loan agreement, or technical manual? Are the strategies for effectively editing a manual or warranty the same as those needed to write one? (Felker et al., 1985) Researchers find some insights into the design aspect of composition by comparing the decision-making processes of people using similar bodies of source information to prepare new works in a variety of media--writers working on conventional texts, magazine designers, collage artists, designers of audio-visual presentations, film-makers, and hypertext authors. Researchers also might encourage writers to contribute anecdotal and introspective accounts of their work habits. A writer could use hypertext tools to view and manipulate source material, take notes or experiment with early drafts. This would be the case even when hypertext structures are not going to be used in the final document (for example, a manuscript intended for print publication rather than electronic distribution). WE, The University of North Carolina's Writer's Environment (Smith, 1986), shows how a hypertext system can free writers from structured outlining, even if their goal is to produce a printed text. WE allows a user to create blocks of text in windows at random locations on a large display screen, turn these windows into icons, then draw lines on screen to link the icons and experiment with various sequential arrangements of the information. Although Guide does not automate the connecting of text units as graphic icons, Guide users could borrow part of the WE model to enhance a "bottom-up" writing process (as opposed to designing from the a "top-down" outline). The Guide user can easily compose a paragraph (or a much larger block of text), mark it as a "replacement" and type a significant word or phrase as its "replacement button." The word or phrase might indicate the content of the passage, or it could be an organizational note, such as "move this to the 'conclusions' chapter." In this way a writer can collapse a long document like a telescope, reduce it to a page or two of key phrases (buttons), and reorganize the full document by manipulating the buttons with the same cut-and-paste commands used to edit a simple text. The Flower and Hayes model naturally anticipates a sequential text as its end-product. The differences in designing a non-sequential hypertext would add new factors to some of the sub-processes (planning, translating and reviewing), allowing the overall model to stand. As hypertext systems are used more widely, it will be possible to test expanded models of these sub-processes. Researchers at Brown University hope to determine whether the reading and writing of linked texts through the Intermedia hypertext system helps students develop a more sophisticated cognitive style, described by various writers as nonlineal, pluralistic, relativisitic, or critical thinking (Beeman et. al., 1987). So far, results indicate that the benefits of the system are greater for active users than passive readers of hypertexts. Professors and teaching assistants who prepared hypertext materials for two university courses (one in literature, the other in cell biology) found that the conscious process of creating links in the course materials inspired "greatly increased understanding" and "a powerful reorganization of their own thinking concerning these materials" (Beeman, p.13). Cognitive psychologists have shown that encouraging students to use metaphor, analogy and elaboration are all useful strategies for learning and remembering declarative knowledge. (For a survey of such studies, see Gagne, pp. 96-100.) Studies of outlining and mapping techniques (e.g., Glynn and DiVesta, 1977, Armbruster and Anderson, 1980, cited by Gagne) have suggested that these organizational tools can contribute to students' learning. Outlining, mapping, analogy, metaphor and elaboration all have counterparts in hypertext structures. It would be interesting to see the results of similar studies of students who were encouraged to create hypertext definitions, replacement texts and linkages for their assigned texts. Reviewing, with hypertext The reviewing process goes on throughout a writer's work, as the writer evaluates and revises the written text. But major revision, toward the end of the writing process, will test hypertext's replacement and cross-referencing tools as means of finding relationships, eliminating repetition, and juxtaposing different versions of a text. The evaluation phase of the review process employs the same features of a hypertext system that are discussed in Chapter 5 concerning reading and research. Being able to simultaneously view several parts of a document, or parts of several linked documents, may someday make it possible to conduct a thorough review and editing of large documents on a computer screen, something that is next to impossible with conventional 80-character by 25-line word processor displays. The most advanced hypertext systems already rely on large, high-resolution, multiple-window displays, such as the workstation monitors used by Intermedia, WE, and Xerox Notecards. By the time large-scale hypertext systems are available, computer display technology may include equally large-scale displays. Until that time, however, writers and editors of large, complex documents are likely to continue depending on printed drafts and large, flat surfaces, like the kitchen table and living room floor that have played a role in the organization of this essay. C. Long-term memory and hypertext The final element in the Flower and Hayes model is the writer's long-term memory, the very tool that Vannevar Bush originally wanted to enhance with his memex machine. Bush was hardly the first to wish for such a tool, or think he had invented one. A predecessor told a friend he had an answer to forgetfulness; but his friend replied as follows: . . . he who has the gift of invention is not always the best judge of the utility or inutility of his own inventions to the users of them. And in this instance a paternal love of your child has led you to say what is not the fact; for this invention of yours will create forgetfulness in the learners' souls, because they will not use their memories . . . you give your disciples only the pretence of wisdom; they will be hearers of many things and will have learned nothing; they will appear to be omniscient and will generally know nothing; they will be tiresome, having the reputation of knowledge without the reality. The technological advance the pair talked about was writing itself. The inventor was the Egyptian god Theuth (or Thoth), and his critical friend was Thamus. Their conversation was reported by Socrates, quoted by Plato in the dialogue "Phaedrus" (Jowett, 1942). Since the days of Plato, who was, ironically, one of the first great writers, we have long since surrendered our memories to letters, moveable type, and photo-offset presses. We should think twice before expecting a panacea in another wave of technology. But it is tempting to hope that this time we really have found, as Thoth claimed, "the cure of forgetfulness and of folly." MENTAL MODELS FOR HYPERTEXT TOOLS While the Flower and Hayes cognitive model of writing can give us some insight into writing with hypertext, users of hypertext systems will develop their own mental models of these new tools. Hypertext users probably will understand the new systems based on their understanding (mental models) of their familiar tools for working with printed texts. Hypertext system designers should try to be aware of models--right and wrong--that their users might apply. Research in this area could draw on cognitive psychologists' studies of knowledge representation and learning of new skills by computer novices. Black, Kay and Soloway (1987) review several such studies that draw on goal-and-plan models of computer users' knowledge representations (including Card, Moran and Newell, 1980; Sebrechts et al. 1983; and Kay and Black 1985, 1986). While superficially similar to computer text editors, hypertext systems add new kinds of operators and methods-- making it possible to fulfill new goals, such as linking related documents. Developers might rethink a user interface, system commands, or documentation after paying more attention to users' previous experience, mental models, and preconceptions. For example, Sebrechts et al. studied how learners' develop complex representations of computer editing commands, gradually replacing their preconceived (and sometimes inaccurate) models for similarly-named commands like CANCEL and DELETE or FIND and GET. Users of Guide may be misled by feature names like DEFINITIONS, NOTES, and REFERENCES. In its regular dictionary sense, a "definition" is a statement of the meaning of a new term. A conventional "note" can be a memorandum, comment or explanation, including a footnote or marginal note. A "reference" often points to an external source of information. In Guide, however, these terms have new meanings that are quite different than the everyday meanings. They do not refer to content, but to the style or structure of hypertext information. A Guide NOTE is a text or graphic "button" that pops up a temporary window of text on screen. (The window is visible only when you point to the NOTE and hold down the mouse button; it disappears when you release the button.) "DEFINITION" is Guide's name for the block of text visible in the NOTE window. These texts can be definitions in the vernacular sense, or the kind of information commonly included in footnotes, marginal notes or memos. Like a text footnote, these Guide NOTE-DEFINITION pairs can be used to cite sources of information, which are commonly referred to as "references." A Guide REFERENCE, however, is another structure--a hypertext link to a separate document (in a new window) or a new location in the same document (by scrolling the text in the current window). The content of the reference-point is up to the writer--it could even be a bibliographic "reference." Users eventually overcome their preconceptions of command names in text editors, but it would be valuable to study whether Guide's use of common terms for its structures aids or delays users' formation of accurate knowledge representations. For example, the program and its documentation could be modified for experimental purposes, renaming "DEFINITIONS" as "NOTE CONTENTS," and "REFERENCES" as "LINKS" or "JUMPS." Simpler experiments might use a combination of Guide tutorials, questionnaires, observation and interviews to determine how quickly users do understand the purpose of each type of hypertext structure, and whether their preconceptions limit their use of some features. The Document Design Center staff (Felker et al. 1985) has found the business world full of poorly-designed documents produced by badly-trained writers. There is no reason to suppose that hypertext will solve these problems, although an optimist might dream that using a new medium will make writers more conscious of their creative processes and conscientious about the results. At this stage in the development of the medium, reading hypertext can be problematic (as discussed in the next chapter). Reading badly constructed hypertext could be even more difficult, leaving the reader searching for missing links and incomplete references, as suggested by the example that follows. Tarzanning: a questionable model There are two reasons for including this example of bad writing: * As a good example of how hypertext would be a powerful writing resource--I can't place my original source for this information; hypertext would have linked to it by now. * As a bad example of unstructured or "associative" thinking and writing--poorly thought out, poorly organized, hastily composed. "Tarzanning" means swinging from one quotation to another, like the jungle man swinging from branch to branch on a Hollywood vine. I ran into the term in a book on "how to study" or "how to write term papers" years ago, but I've misplaced the reference. The author, if I remember correctly, was describing the symptoms of a sloppy, lazy, or unprepared student's technique of writing a research paper. I recognized the model from my own undergraduate days. I never had a name for the technique, but this is how it went: 1. Go to the library and get as many books related to the assigned topic as you can carry. 2. Look at the tables of contents and indexes for the topic, and place slips of paper in those pages. 3. Skim the sections you have marked, and move the slips to the most significant passages. 4. Write an outline, roughly covering the topic and subtopics you have information on. 5. Type quotations and a synopsis of material from each book. 6. Assemble the typed material in order, following the outline. 7. Write transitions--this is the "Tarzan" part--that somehow swing from one quoted author's ideas to the next. 8. Write an introduction and conclusion. 9. Edit and rewrite, if there's time. (There's never enough time.) This process can be used to write a term paper in a day, with very little risk that the paper will contain an original thought. Hypertext's ability to link existing documents has a great risk of creating electronic Tarzanning in place of original writing or thoughtful reading. But hypertext also has great potential as a research and reference tool. For example, my lost reference to the book that mentioned "Tarzanning" in the first place--had I entered it in a hypertext database several years ago, it would still be there, linked to other documents about "research methods," "writing" or "taking notes." A more conventional database or a multiple-file text-search program could accomplish the same end, but would have required more foresight when it was designed. The original database might have connected Tarzan only to "fictional characters." I might have wasted time searching for "Batman" or some other fictional swinger. A hypertext library would have allowed me to add associative links over the years. On the other hand, if I had recalled the Tarzan reference immediately, I might not have followed the thought process that led me to write this page. While hypertext advocates talk about the goal of reflecting the mind's "associative" methods of retrieving information, Gagne uses the word "associative" to describe writing by young students who are more intent on getting information down on paper than in making a convincing argument or having something original to say. This coincidence could foreshadow a real danger of hypertext-- that writers will use the flexibility it gives readers as an excuse for "publishing" incomplete ideas or poorly- structured arguments, an issue that is also raised in Norman, 1987. As mentioned in the discussion of the task environment, writing hypertext will still require that the writer understand a basic rhetorical problem, including definitions of the topic and of the document's audience. Immature writing is often called "writer-based" because it neglects to recognize and address the audience. Excellent writing has long been praised for being "reader based." Hypertext systems may give writers tools to anticipate the needs of varied groups of readers, especially in technical or informational documents, letting readers select among optional paths. Felker, discussing how to train writers of such informative documents for conventional publication, says "reader-based organization stresses ease of access and chunks information under headings that reflect the questions readers ask about the information" (1985; p. 51). Hypertext windows are perfectly suited to such "chunking," and "ease of access" is a frequently-cited goal of designers of hypertext systems. Some hypertext authors will be content to link together traditionally-structured prose, using hypertext nodes as little more than an on-screen equivalent of book chapters or magazine articles. Others may find new ways to weave a rhetorical thread into a non-linear web of interconnected texts--if they do so, the thread will be made of human understanding. Socrates, in Plato's "Phaedrus," argued that writing was inferior to oratory because of its static nature, using "the same words for all." And the quality of oratory, he said, was in proportion to the speaker's understanding and ability to adapt to an audience: "Until a man knows the truth, and the manner of adapting the truth to the natures of other men, he cannot be a good orator." (Jowett, 1942; p. 468). CHAPTER 5: RESEARCH, READING AND HYPERTEXT A COGNITIVE MODEL OF READING Most research on reading has been conducted by linguists seeking to understand how we draw meaning from written symbols, by cognitive psychologists trying to understand how the mind processes information, and by educators trying to understand how children learn to read and how to improve their reading comprehension.[24] Such approaches increase our knowledge of the comprehension of letters, words and sentences, the building blocks of reading and learning. However, these studies tell us less about finding our way through a novel, textbook or library full of cross-references--the reading and referencing processes that Bush, Engelbart and Nelson have hoped hypertext will facilitate. Researchers developing the WE Writing Environment at the University of North Carolina have been testing a three- part cognitive model of writing (association-hierarchy- sequence) that can be inverted into a model of reading (sequence-hierarchy-association). According to this model (Smith et al., 1986), writing is: a. exploring a associative network of information b. organizing the information into hierarchies c. encoding it as a linear sequence of words. Using the mirror-image model, reading is: c. decoding a linear sequence of words b. comprehending hierarchies of meaning a. remembering the information as an associative network. This model is clearly suited to the researchers' project, the creation and testing of a writing system that combines associational and hierarchical tools. However, the Flower & Hayes (or Pea & Kurland) model of writing shown in the previous chapter goes into more detail about the recursive nature of writing, and the role of context, sources and tools, all of which have parallels in our discussion of hypertext. For heuristic purposes, then, this chapter borrows the North Carolina researchers' idea of having complimentary models of writing and reading, but it transforms the Flower & Hayes model of writing into a model of the reading process. Basing the model of reading on one of writing is especially appropriate to hypertext--both Nelson and the developers of Guide point out that the link- making, associative features of hypertext create new, writer-like roles for readers. The diagram in Figure 3 (based on Figure 2 in the previous chapter), models the reading process, including the process of selecting reading materials, referred to here as "research" or "referencing." Like writing, reading and research take place in the context of a task environment, and have several component parts functioning under a monitor. Like writing, reading is a recursive process involving planning, the actual "act of reading" or decoding information, and evaluation of the reading so far (both as comprehension monitoring and in terms of evaluating the problem and plan). Like the previous chapter, the text below applies the model to conventional texts, then to hypertext, particularly Guide. Figure 3: READING ÚÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄ¿ ³ A. Task Environment ³ ³ ÚÄÄÄÄÄÄÄÄÄÄÄÄÂÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÂÄÄÄÄÄÄÄÄÄÄÄÄÄ¿ ³ ³ ³ Problem ³ Text read ³ Resources ³ ³ ³ ³ ³ so far ³ primary text³ ³ ³ ³ topic ³ ³ other texts ³ ³ ³ ³ purpose ³ ³ other tools ³ ³ ³ ³ assignment ³ ³ context ³ ³ ³ ÀÄÄÄÄÄÄÄÄÄÄÄÄÁÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÁÄÄÄÄÄÄÄÄÄÄÄÄÄÙ ³ ÀÄÄÄÄÄÄÄÄÄÄÄ/\ÄÄ\/ÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄ\/ÄÄ/\ÄÄÄÄÄÄÄÄÄÄÄÄÄÙ /\ \/ \/ /\ ÚÄÄÄÄÄÄÄ¿ ÚÄÄÄÄÄÄÄÄÄÄÄ/\ÄÄ\/ÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄ\/ÄÄ/\ÄÄÄÄÄÄÄÄÄÄÄÄÄ¿ ³ C. ³ ³ B. READING PROCESSES ³ ³ Long- ³ ³ ÚÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÂÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÂÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄ¿ ³ ³ term > > > ³ Planning ³ Decoding ³ Evaluation ³ ³ ³ Memory³ ³ ³ set goals ³ matching ³ comprehension³ ³ ³ < < < ³ supplementary³ recoding ³ literal ³ ³ ³ ³ ³ ³ pre-reading ³ association ³ inferential ³ ³ ³ ³ ³ ³ note-taking³ ³ ³ ³ ÀÄÄÄÄÄÄÄÙ ³ ÀÄÄÄÄÄÄÄÄ\/Ä/\ÄÄÁÄÄÄÄÄ/\Ä\/ÄÄÄÄÄÄÁÄÄÄ/\Ä\/ÄÄÄÄÄÄÄÙ ³ ³ ÚÄÄ\/Ä/\ÄÄÄÄÄÄÄÄ/\Ä\/ÄÄÄÄÄÄÄÄÄÄ/\Ä\/Ä¿ ³ ³ ³ Monitor ³ ³ ³ ÀÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÙ ³ ÀÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÙ Figure 3: Adapted from Flower and Hayes, as in Figure 2. The A, B and C labels refer to paragraphs in the text on the following pages. A.The task environment has three parts: 1. The reader's problem definition, such as the topic, purpose and exigency of the reading assignment. Reading can be done for entertainment, for information at various levels of specificity, and for a combination of these purposes. Problem definitions include what to read, how carefully to read it, when to read it, etc. Examples of different reading situations or problems include: looking up a number in the telephone book, learning lessons from a high school text, studying independently with an encyclopedia and reference books, or flipping through comic books and People magazine at the barber shop. 2. Recently read material, including the preceding pages in a book or magazine, or the stack of just-used reference books spread on your library table. 3. Resources available to the reader, including the text itself (on paper, microfiche or a computer screen), supplementary sources of information, such as a student's teacher, a dictionary, or a university library, and tools ranging from reading glasses and note-taking materials to microfiche readers and computerized information- retrieval systems. The physical context in which reading takes place spans the task environment from problem definition to resources. For example, reading in a public place may place constraints on the reader (the noise and crowding of a subway), or may provide affordances (the variety of supplementary texts shelved at a public library or sprayed on the walls of the subway). B. The reading process has three parts under the control of a monitoring process: 1. Planning includes setting goals based on the problem definition and creating reading strategies, such as outlining, pre-reading, note- taking, and searching for additional sources of information. 2. Decoding is the actual process of looking at words on paper and drawing meaning from them. Gagne (1985) summarizes this process of matching printed words to known vocabulary and recoding unfamiliar words using syllables and sound patterns until a memory-match can be made. The end result is lexical access. 3. Evaluating the reading includes the reader's comprehension of the text. Through lexical access the reader matches decoded words to associated information in long term memory; by parsing the relationships of words, the reader draws meaning from the text. Deeper inferential comprehension comes from integrating, summarizing and elaborating the new information with the reader's previous memories. (Gagne, 1985) 4. The Monitoring process compares the results (comprehension) with the reader's current plans and goals, and controls the recursive aspects of the process--re-reading a paragraph or adjusting a plan from "skim" to "read in depth." C. Long-term memory includes the writer's previous knowledge of the topic and the reason for reading the new material, as well as information stored during the reading process. APPLYING THE MODEL TO HYPERTEXT Researchers have developed many models of the reading process to test theories about cognition, reading comprehension, study methods and learning. (For recent surveys of the literature, see Orasanu, 1986, and Pearson, 1984.) The diversity of models contributes to our understanding of conventional reading; in the future, more of these models may be applied to hypertext. The general model above is presented as a starting point--a framework for a comparison of reading traditional text and reading hypertexts. The sections below discuss how working with hypertext fits each part of the model. (As before, the letters refer to parts of Figure 3.) A. The Task Environment, with hypertext A hypertext system could be a major part of the reader's task environment as outlined above, whether for in-depth research with a large interlinked hypertext database (Intermedia, Xanadu, Guide), for on-line reference "help" systems for computer programs (Guide, Concordia), or for browsing or general reading of linked documents (Guide, HyperTIES). A hypertext system would contribute to problem definition, keep recently-read text available, and provide new resources in the form of both source material and tools for annotating texts as an aid to reviewing or learning. 1. Problem definition A reader's problem definition includes the kind of reading to be done--browsing, reading for general information or entertainment, retrieval of particular information, or in-depth study. Hypertext systems have been designed to support these different types of reading. For example, several of HyperTIES's first applications have been in public settings, as a reference tool for visitors to museum and gallery exhibits. A reader's goal in such settings is to browse or quickly find information with a minimum of time spent figuring out the system--a problem definition similar to that of someone using a museum catalog or visitor's bulletin board. Initial research indicated that the "arrow jump-keys" on the IBM PC keyboard were a better tool than a mouse for beginner or infrequent users (Ewing et al., 1986). With HyperTIES a reader can investigate the entire database by jumping from link to link using only the directional arrows and the Return key. (Touch-screen and mouse versions of the program also have been developed.) HyperTIES links are shown as highlighted words in the document on screen. The directional arrow keys "jump" the cursor to the next highlighted word. Pressing Return once displays a brief abstract of the article related to the link word; pressing Return again opens the full article. The arrow keys move the cursor to the next highlighted word, including extra links at the bottom of the screen that access an alphabetical index, other pages of the same article, or backtrack to previously viewed screens. The highlighted words on HyperTIES screens constitute "embedded menus" of links in the document, which Marchionini & Shneiderman (1988) describe as "good examples of specialized indexing for systems that emphasize understanding rather than retrieval." If a reader's task or problem involved retrieving particular information, then the proper tool could be HyperTIES alphabetical index, or a different kind of program that provided full-text or key word search capabilities, such as the Grolier Electronic Encyclopedia, reviewed by Marchionini and Shneiderman in the same article. Some hypertext systems, including early versions of HyperTIES and KnowledgePro/TextPro, treat writing and reading as separate activities, providing simpler user interfaces for the "reader." In TextPro, for example, a reader selects "Read a hypertext document" from a menu, and follows links by simply moving the cursor to a highlighted word and pressing a key or mouse button to open a window of linked information. To create such links, however, the user must leave the display of hypertext windows and reopen the original file by selecting "Create or edit a hypertext document" from the program's main menu. The user creates links in TextPro documents by annotating a text file with symbols or code words similar to those of a print-formatting markup language, such as "/page", "/end", and "#m". The code "#m" marks the beginning and end of a word linked to a hypertext message (e.g. "#mConnecticut#m"); a slash before the linked word (e.g., "/Connecticut") marks the linked message. Owl's Guide and Apple's HyperCard, on the other hand, gave their hypertext systems the same basic interface for readers and writers. In this way, link-building and annotation tools are available to readers as well as writers. Both HyperCard and Guide followed the standard design of the Macintosh or Microsoft Windows interface. Their pull-down menus, dialogue boxes, and use of the mouse are instantly familiar to users of many other programs. HyperCard also has a programming language, HyperTalk, with which users can create programs that perform advanced manipulations of HyperCard "stacks." While HyperTalk is not as easy to use as the HyperCard interface, it is growing in popularity among Macintosh users, including educators.[25] Owl has not yet released an announced programming language for Guide. Although Guide's link-making tools use the same interface design, these tools are not available in all incarnations of the product. Versions of Guide for read- only CD-ROM databases and the stand-alone "Guide Envelope" which provides read-only Guide documents on the Macintosh (e.g., the Guide edition of Nelson's Literary Machines) do not offer the writing and linking abilities. In some contexts, users of hypertext systems may prefer to separate the "reader" and "writer" functions of a hypertext library. Carl Binder, a human performance consultant, evaluated Guide and another associative-text program, The Window Book Technology, for possible use in a an office workplace. He noted that business managers will want centralized control over who can alter documents, rather than the freedom Guide gives users to link and annotate. Window Book read-only documents have been used successfully in several business applications, including software reference guides and insurance company procedure manuals (Binder, 1987). When a variety of hypertext tools are available, libraries of documents as well as access software, the reader's choice among those tools will be part of the "problem definition" of the reading process. Like a researcher defining the scope of a study, a hypertext reader must decide on a strategy for pursuing or not pursuing links and references. We may even have robust hypertext systems linking scores of documents on large computers and smaller, portable readers offering a hypertext-based interface for writing and reading personal documents. Analogies might include a student's choice between reading in a library with a notebook and dictionary close at hand or skimming a paperback novel during lunch at a campus diner. 2. Recently read material Guide, HyperTIES, HyperCard and other hypertext systems all provide some way to view previously-read text, which helps to reduce the reader disorientation that has been reported with earlier text-on-screen systems. (Brown, 1986; Tombaugh, 1987) Word processors and other screen displays of text usually allow the user to scroll back and forth through a document by line, by screen or by the equivalent of a printed page. Most screen displays of text lack the visual cues offered by traditional printed texts, such as chapter headings or physical locations of pages in a book. Apple's HyperCard (Goodman, 1987) and CMU's ZOG (McCracken, 1984) use the fixed size of the computer screen as a basic text unit (a "card"), rather than the "scrolling text" or printed-page units of Guide and most word processors. Researchers are exploring other on-screen cueing devices, such as creative use of multiple windows to display sections of a long text on screen, and ways to counteract the other shortcomings of reading on current computer screens--such as of screen size, detail, eyestrain and fatigue (Tombaugh 1987). Increasing availability of larger, higher-resolution display screens, such as those currently available on commercial publishing systems and workstations may solve some of the latter problems. (The Macintosh II is one of the first microcomputers designed to use both a large screen and multiple windows. Such "hardware" issues, however, are beyond this essay's scope.) In a non-sequential hypertext, following links or activating "replacement" buttons can remove recently-read text from the screen. Simple scrolling will not retrieve the non-sequential material, so hypertext programs must provide a way to "backtrack" through previously viewed information. Guide provides the ability to scroll backward through sequential text and to "backtrack" through recent links. By default, this "backtrack" feature only retraces Reference links, but it can be reset to include Replacements as well. (See the screen examples in the Appendix for examples of backtrack icons used by Guide and HyperCard.) Part of the challenge of transferring information to the screen is to offer substitutes for physical cues offered by three-dimensional objects like library shelves, books and magazines. Consider how the physical affordances of printed texts make possible the following observations: "I put the book down on the coffee table." "It was a small green book, shelved with some large red ones." "I remember the quotation was centered on a page, not too far from the beginning of a chapter." "There is a large illustration of a pretzel at the beginning of the chapter." "I am holding my thumb where I stopped reading, so that I can skim the review the previous chapter for a term I don't understand." So far, only limited cues are offered by text or icon- based computer file directories, "maps" of hypertext links,[26] or representational features like HyperCard's ability to show dozens of neighboring "cards" in reduced size. But hypertext can add an additional tool for reviewing text--the ability to store and replay a reader's path through linked texts, just as Vannevar Bush suggested storing a trail of "memex" associations. Conklin's (1987) comments on various programs' abilities in this area. Guide cannot store such paths or replay a series of links once the user backtracks through them. The lack of such "replay" or "bookmark" abilities limits the Guide reader's ability to keep track of long, extensively linked multi-file documents. With Guide in its present form, a reader cannot leave the computer and return to "where I left off" easily. The reader must remember to activate the same series of links during the next session with the document. The Guide edition of Nelson's Literary Machines (1987) is a case in point--despite its use of chapter and section numbers, it is easy to become disoriented. (The same is true of the quasi-hypertext print edition of Nelson's Computer Lib/Dream Machines, which takes the form of two books printed back to back. The extensive cross- references, sidebars and boxes of related text encourage the reader to browse and flip through the book--leaving the reader uncertain about how much of the book is still unread.) In the full Guide system (giving the reader the ability to edit text and create links), a reader can mark the last-read passage and link it to a button at the beginning of the document. Even this action, however, will not reopen Reference-linked documents or open Replacement buttons that the reader viewed during the initial session. The experience is reminiscent of computerized "interactive fiction" games like Zork which can be solved only by making a map as you proceed through rooms, tunnels and mazes, with each location linked to from one to a dozen others. A game can afford to model itself after the experience of exploring shadowy caves--a library or an informational book needs a less mysterious model. 3. Resources A hypertext system can be both a reference library and a book. Like a library or encyclopedia, it can store, index, cross-reference and display many documents. Like a single book, it can show the reader facts, fiction, poetry- -any kind of text, as well as pictures and diagrams. A hypertext system can store non-sequential documents or store linear texts, and can provide additional tools to annotate or mark new links or associations between documents. This section will discuss hypertext tools for storing and retrieving, searching (including indexing and cross- referencing), displaying, and annotating documents. Storing and retrieving Computers store information in named units called "files." File sizes, conventions for naming, storing and managing them are determined by the computer's operating system, such as the IBM PC's DOS (Disk Operating System). Document storage depends on the computer's storage capacity, in the form of available disk drives or the new optical media (CD ROM discs). The developers of Guide and HyperCard have announced that future versions of those programs will be designed to use CD ROMs as well as conventional disk storage (Brown, 1986; Goodman, 1987). Additional file-management tools are provided by software and hardware that link individual computers into networks. Ted Nelson's Xanadu system is designed to provide network "addresses" for an almost infinite number of documents, thereby allowing hypertext systems to create links among them (Nelson, 1981, 1988). Xanadu's developers view the system as a publishing medium, and are considering a wide range of issues involving copyright protection and payment of authors' royalties through maintenance of master "canonical" copies of stored documents. Searching (indexing and cross-referencing) Conventional computer database systems are designed to index files of carefully-defined records; other programs are designed to search or build indexes of unstructured text files. Neither of these approaches to searching and retrieving information takes advantage of hypertext's direct linking between items of information, a technique which could be used to build paths of associations among documents or parts of documents. However, hypertext programs might benefit from providing powerful searching, indexing or table-of-contents-generating as additional tools. For example, a multiple-file search utility would help the editor or author of a Guide database find and create links between passages. Marchionini and Shneiderman (1988) have compared their HyperTIES system and the electronic CD ROM edition of Grolier's Academic American Encyclopedia as tools for "finding facts" or "browsing knowledge." Grolier's Electronic Encyclopedia uses powerful search software and indexes that include every word of the encyclopedia's text, the equivalent of 20-volumes. Like a conventional database program, the Grolier encyclopedia allows readers to retrieve information by entering Boolean search queries on a fill-in-the-blanks screen to locate words in various contexts in the text. The Encyclopaedia Britannica is experimenting with an on-line computerized version of its text. If the experiments succeed, the Britannica says, . . . the way will be opened to the development of an electronic encyclopaedia independent of alphabetical arrangement, accessible through methods other than the use of an index, and capable of being updated on an almost daily basis as new data are substituted for old. (Collison and Preese, 1986) Displaying The Macintosh and Microsoft Windows operating system environments give programs like Guide access to "windowing systems" that can present several documents on a reader's screen at the same time. Both systems also provide detailed "bit-mapped" graphic displays, rather than the character-only displays of earlier computer operating systems, including IBM PC DOS. One of the strongest selling points of bit-mapped displays has been their ability to show close approximations of printed text, including different type sizes and type fonts as well as graphic illustrations. This is commonly referred to as "WYSIWYG" (pronounced wizzy-wig) for "What you see (on the screen) is what you get (on paper)." Hypertext programs like Guide and Black Magic have consciously limited their printing capabilities, stating that their primary purpose is to present text on the screen. There are, in fact, no printable equivalents of hypertext features like Guide's Replacements, Inquiries (branching Replacements), pop-up Notes, or position- shifting References. (Printed text, as we have mentioned, does have analogous functions, such as footnotes, but without the dynamic characteristics of text on a computer screen.) If hypertext is designed for screen displays, WYSIWYG in its print-oriented sense need not be an issue. Guide, however, does take advantage of some of Microsoft Windows' WYSIWYG type styles (bold, underlining, etc.), using them as indicators for hypertext structures. It also allows importation of graphic images from other Microsoft Windows programs, including a wider range of WYSIWYG text characters than Guide itself uses. This import ability enables Guide authors to design documents that are visually interesting as well as informative. HyperCard, on the other hand, includes an advanced graphic editor and a library of icons as part of its software. (HyperCard itself was designed by Bill Atkinson, the creator of the Macintosh's MacPaint program, one of the first widely- popular bit-mapped graphics programs.) Guide puts an interesting twist on Microsoft Windows' abilities to display documents in resizeable windows and to display type fonts in different sizes: Guide's Display menu includes the options "Text Larger" and "Text Smaller." These selections magnify or reduce the text on screen, without changing the relative size of characters in headings and paragraphs. It is possible to reduce the size of characters to view more of a document on screen, or to enlarge the type for easier reading[27]. The resized text wraps to fit the margins of the document window, rather than remaining frozen in its WYSIWYG margins. Similarly, graphics shrink or expand as the reader changes the size of the window containing them. HyperCard insists on a WYSIWYG relationship between the image on screen and a printed image. It cannot enlarge its standard display image of a card beyond the size and shape of a nine-inch Macintosh screen, except for purposes of editing graphic details. It can, however, reduce the card image to display as many as 42 cards on a screen-- allowing the reader to view the context of individual chunks of information. Guide opens each hypertext document as a new window on screen, overlapping most of the previous document. The reader can move or reconfigure the windows to view one or more texts at a time, maintain the overlapping windows, or separate the windows into vertical or horizontal "tiles." Microsoft Windows allows the reader to open other programs in separate windows, allowing easy "cut and paste" transfers of information from a graphics painting program or word processor. Annotating texts Guide, Intermedia and some other systems (Conklin 1987) give the reader the same tools they give the author-- the ability to annotate and link documents, create separate files of notes or commentary, or restructure the original document by adding summary headings or compressing passages with tools like Guide's Replacement buttons. In that sense, the Guide document becomes a personal book--one the reader can feel free to mark up with underlining or a highlighting pen, except the "marking-up" can be dynamic- - such as "outlining" or "summarizing" a chapter by turning paragraphs or larger sections into shorter descriptive statements in the form of Replacement buttons. These hypertext "authoring" tools could be especially useful to students. Anderson and Armbruster (1984) surveyed recent research on study methods, and concluded that outlining, networking, mapping and schematizing study materials are especially effective. Such techniques "force students to identify or impose relationships that convey the meaning of a text," but have a high cost in student time and energy. Anderson and Armbruster suggest research on combining such study techniques with "technologies, like computer-based word processors and video/audio recorder systems." Hypertext systems like Guide, especially when coupled with multiple-window computer environments and graphic drawing tools, could provide a readily available medium for such research. B. The Reading Process, with hypertext The hypertext tools described above can be used at both the planning and reviewing stages of reading or research. On the research or multiple-document level, a hypertext system could show the relationships between documents and provide easy access to abstracts as well as full texts, thereby helping readers decide what documents to read in order to fulfill an assignment or "problem definition." Brown University's Intermedia is a major vehicle for such research, with databases supporting courses in literature and cell biology (Yankelovich, Landow and Haywood, 1987). Guide, HyperCard, HyperTIES or other commercially viable hypertext systems may lead to the creation and sharing of more document databases and eventually lead to standard formats for document storage-- without such documents, hypertext is at worst a library full of empty shelves.[28] Planning: Pre-reading, annotation and organization Teachers of speed reading and reading comprehension frequently recommend "pre-reading"--getting an overview of the document by reading chapter and section headings, topic sentences, prefaces and conclusions. Some study techniques combine this pre-reading with outlining or diagramming the text. The design of a Guide document can use similar devices to those a printed text uses to facilitate pre- reading and provide the reader with a mental model of the document, such as explicit text cues, section headings, chapter summaries and meaningful diagrams (Felker, 1985; Horn, 1985). Unlike readers of printed documents, however, hypertext readers may be able to restructure and salvage the meaning of a poorly-designed document by using the hypertext system's ability to add position markers, replacement texts, notes and links. An active reader could, in effect, complete the author's unfinished job of designing a usable document. This would take extra effort on the reader's part, but might be appropriate if the text that will be used repeatedly or by several readers. Owl International points out that even the Guide "HELP" screens can be modified--the reader can add additional links to the help files, summarize the existing material with Note buttons, or add new material, such as help on a particular Guide database. (The user also runs the risk of inadvertently deleting part of the help system. This is another situation that would benefit from Xanadu's methods of adding new links while preserving an original or "canonical" version of a document, as mentioned earlier in regard to copyrights and royalties.) These active reading techniques are analogous to a new computer owner's process of adding section markers, tabs and an index to a cheaply-produced technical manual[29], or a literature scholar's marginal notes tracing themes, symbols, characters or timelines through a story. But annotation of texts did not begin with the computer industry, or even with Gutenberg. Figure 4: The Talmud as hypertext ÚÄÄÄÄÄÂÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄ¿ ³ ÃÄÄÄÄÄÄÄÄÄÄÄÂÄÄÄÄÄÄÄÄÄÄÄÄÄ¿ ³ ³ ³ ³ ³ ³ ³ ³ ÚÄÄÄÄÄÁÄÄÄÄÄÄ¿ ³ ³ ³ 10 ³ ³ 1 ³ ³ 8 ³ ³ ³ 7 ÃÄÄÄÄÄÄÄÄÄÄÄÄ´ ³ ³ ³ ³ ³ 2 ³ ³ 8 ³ ³ ³ ³ ³ 6 ³ 8 ³ ³ 8 ³ ³ ³ ³ ³ ³ ³ ÃÄÄÄÄÄÄÄÄÄÄÄÄ´ ³ ³ ³ ³ ³ 3 ³ ³ 8 ³ ³ 8 ³ ³ 12ÀÄÄÄÄÄÄ´ ³ ³ ³ ÃÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄ´ ³ ³ ³ ³ 4 ³ 8 ³ ³ ³ ³ ÚÄÄÄ´ ³ ³ ³ ³ 5 ³ ³ ³ ³ ³ ÀÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄ´ ³ ³ ³ 11 ³ 7 ³ ³ ³ ³ ÀÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÁÄÄÄÙ ³ ³ 9 ³ ³ 13 ³ ÀÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÙ Figure 4: This diagram outlines separate blocks of text on a page of the Babylonian Talmud printed in 1880-1885.[30] Nelson (1981), Conklin (1987) and others frequently cite the Talmud as an example of printed hypertext, because it is just such a collection of annotations and related texts added by readers. The main text on the Talmud page shown in Figure 4 is from the Mishna, a collection of originally oral laws; it occupies the central blocks 1 and 2. The rest of the 13 numbered blocks of text are commentaries added by over seven centuries of scholars, or cross-reference notes (marked 8) to other Talmudic or Rabbinic sources. Another precursor of hypertext is the modern encyclopedia, such as The New Encyclopaedia Britannica. The Britannica includes a "Micropaedia" of short, specific articles, a "Macropaedia" of longer, in-depth articles, and a master index to all 29 numbered volumes. The Britannica itself recommends a pre-reading process--consulting the two-volume Index for an overview of the extensive cross- referencing among topics. The following passage is from the Britannica Index volume's article, "How to use the index": For example, if readers turn to "United Kingdom" in Volume 17, they will find a long article, but this by no means represents all the information in Britannica about the United Kingdom. To find all of this material, readers must go to the INDEX where, under the entry heading "United Kingdom," they will find not only the reference to the main article but also about 1,000 references to many related articles . . . Evaluation: Reading and navigating Hypertext developers talk about "webs," "maps" and "networks" of linked documents, terms that bring to mind the order and symmetry of a spider web, the clear boundaries of an atlas, or the distinctly marked routes of interstate highways. But nothing about hypertext requires that it be either symmetrical or two-dimensional. Conklin's survey of hypertext programs (1987) includes a screen display of an overwhelming number of links crisscrossing among a few dozen Intermedia documents. Rather than a "web," the screen looks more like a biscuit of shredded wheat. (At this level, the web provides only an idea of the complexity of the hypertext linkages. Readers use a more focused portion of the web to track down information.) A printed book is a physical thing. You can remember that it was on the top shelf at the library, had a green binding, and was bigger than a Bible but smaller than a New York telephone book. Inside, the book's text is sequential and static; chapters cannot change their relationship to each other. The physical acts of holding a book and turning its pages can help you recall that a passage is roughly one-third of the way through the volume, and near the beginning of a chapter. A glance at the location of a bookmark can tell you whether you will have time to finish the book before bedtime. There are no automatic hypertext counterparts to these non-text cues in using printed texts. As more hypertext libraries are designed, both the developers of software and writers who create text databases probably will try to provide analogous clues to text size, location and content. Some possible non-text tools on the computer include foreground and background colors, icons and other graphic devices, position on screen, sound and other computer-aided effects. You might remember, for example, that a passage was only a few physical pages from the start of a chapter. Will flipping through screen displays convey the same sense of "place" in a text? What other orienting devices can be added to hypertext documents to keep readers from losing their way in a hyperchaos of linked documents? Guide provides no automatic mapping or central organizing link. A spokeswoman for Owl International told an interviewer (Binder, 1987) that this link has been the most frequent criticism of the product, and that creating a table of contents or other reference point is "one of the expectations the system places on the author." For example, Nelson's Literary Machines, in its Guide edition, maintains the Table of Contents and chapter numbering of the printed version, as well as numbered sections in a long, sequential chapter 2. The headings often act as Guide Replacement or Reference buttons to aid in navigating through the document. Without a storable "path" through the document, however, it is easy to become disoriented when reading it in more than one sitting. Since exploration of the hypertext medium is the point of the book, "browsing" from section to section seems like an appropriate reading strategy. It is still difficult to be certain that you have not missed part of some chapter that might have been hidden behind a Replacement or Reference button that you neglected to explore. In the text, Nelson admits that, "Ideally, the book should have been completely rewritten for the GUIDE format, but then it should have been completely rewritten several times earlier as well." Nelson also admits to a "problem of orientation" with current hypertext systems, which he sees as a challenge: There are tricky problems here. One of the greatest is how to make the reader feel comfortable and oriented. In books and magazines there are lots of ways the reader can see where he is (and recognize what he has read before) . . . These incidental cues are important to knowing what you are doing. New ones must be created to take their place. How these will relate to the visuals of tomorrow's hot screens is anybody's guess, but it is imperative to create now a system on which they may be built. (Guide edition, Literary Machines, 1987) Binder (1987) commends Window Book for organizing features--not only is there a central table of contents, but the reader can sort or reorganize the table in a variety of ways, including expanding or contracting the table, functioning like an outline processor. VISIBILITY AND MENTAL MODELS Donald A. Norman (Norman, 1987; Hutchins et al., 1986) has written extensively on "user centered design" in systems ranging from doorknobs and faucets to computer systems. His prescriptions for usable designs include visibility and consistency with the user's mental model of the system. It is on his first category, visibility, that hypertext systems stumble when they do not give the user an overall picture of a document or network of linked documents. Users' mental models of the system could raise problems if the hypertext program does not make clear the ways in which it is at one like and unlike conventional tools like word processors and printed texts. While Guide does not automate the visibility of the overall hypertext document, it does provide good, visible cues for its hypertext structures through typeface changes and pointer feedback, and it provides immediate feedback on the effects of activating hypertext links. Guide also allows the user to display the invisible symbols that mark hypertext structures, as an aid in writing or editing documents, and it has a rudimentary "undo" command for many actions. In one case, however, Guide allows users to accidentally destroy hypertext structures (and the text they contain) without giving adequate feedback to the user: If a hypertext structure, such as a Replacement text, ends at the top of the screen, a user can backspace over button's ending code while that code is invisible. If that happens, the entire hypertext structure is deleted. The Undo command will retrieve the structure, but only if it is used immediately. If the user does not notice a quick flash of the hidden text during the deletion, the text is gone forever. Guide's Inquiry buttons also can hide text in one of their branches. A Reference button may open one branch of an Inquiry, but using the Backtrack icon to return to the starting point will not restore the Inquiry to its top level--which might include alternative Replacement buttons of interest to the reader. The authors and readers of Guide documents can compensate for this problem, but only if they have an accurate mental model of this interaction between Reference, Inquiry, Replacement and Backtrack features. Guide's inability to present an overview of a document or linked documents also could interfere with the reading process. Brown, Armbruster and Baker (1986) have studied the metacognitive skills of expert readers. Among them are the ability to recognize textual structures, to estimate the ease or difficulty of a text, and to distinguish important from less-important elements. In its present form, Guide puts the burden on the writer of a hypertext to provide suitable cues to these issues, and to the structure, content and purpose of the document as a whole. But the availability of Guide and other low-cost hypertext programs gives researchers a new tool to use in research on these and other issues of reading, writing and communication. Another model: "interactive fiction" Hypertext is intended to create new styles of reading and writing, giving readers as well as writers power to choose paths through a text (Nelson, 1987). While not advertised as hypertext, the genre of computer games called "interactive fiction" use hypertext-like techniques: text branching, links between locations, separate versions of texts, and bookmarking (to save and restore a position). Such games, including the Infocom company's popular Zork, could be used to test readers' behavior as well as various issues in hypertext document design and narrative style. Players are encouraged to make pencil-and-paper maps as they play the games, an activity similar to recording hypertext links, enabling the reader to trace relationships among items of information. Like an unmapped hypertext document, most text adventures give the player little or no clue as to the number of rooms or objects (or pages of text) hidden behind a locked door. By studying active (map-making) players and passive (browsing) players of these games, researchers might find be able to identify issues applicable to hypertext readers. Computer text adventure games began with The Original Adventure by Will Crowther and Don Woods, a cave- exploration simulator that was expanded to include treasures, monsters, puzzles and point scoring (Levy, 1984). It and its Infocom successors are puzzles that involve uncovering relationships among objects and locations, often using prior knowledge. In the best- selling Infocom games, the "culture" of this prior knowledge includes literature[31], basic mechanics[32], popular culture[33] and fantasy[34], and the folklore of computer "hacking," including the original Crowther and Woods game.[35] Some developers have attempted to adapt these text-puzzle techniques to educational games, but none with the commercial success of the entertainment products. Like the user of a read-only hypertext system, the game player can uncover or use only the objects and relationships created or anticipated by the original programmer. The games are presented as stories written in the second person. The reader is free to take various non- linear routes through a fictional world, acquire or ignore objects, and interact with other characters. The texts offer minor choices (e.g., to go east or west at a fork in the path), or major decisions that alter situations throughout the story.[36] Some competing companies have created interactive fiction programs with cartoon-like computer graphics, but Infocom has emphasized text and writing style, especially humor. Despite this lack of graphics and sounds, the company apparently scored a "first" for hypermedia by including a "scratch and sniff" card that gives olfactory clues at various points in Leather Goddesses of Phobos. CHAPTER 6: CONCLUSIONS AND SPECULATIONS This essay has been hard to write, mostly because the subject of hypertext has links in so many directions;[37] it has been especially difficult to reach conclusions, since so many paths could not be pursued. For that reason, this final chapter has several parts pointing backward into the essay and forward toward future studies, developments and applications of hypertext technology. About text models for hypertext In the earliest visions of Bush, Nelson, and Engelbart, hypertext was text. In the Guide edition of Literary Machines, Nelson calls hypertext "the most general form of writing," because it can include conventional as well as non-sequential text. So it is appropriate to begin thinking about hypertext by looking at cognitive models of conventional text. The examples in chapters 4 and 5 show that the text models do fit hypertext--if not like a glove, at least like a new shoe that pinches here and there. If the shoe fits, the next step should be to "wear it"--by creating experiments in laboratories of cognitive psychology and educational technology to test new hypertext tools and more detailed models of reading and writing. Cognitive scientists are engaged in many studies of the way the human mind stores, links, and activates trails of associated memories--a subject that has been just outside the scope of this essay, in the boxes labelled "long term memory" at the edge of the models of reading and writing. Hypertext programs may become tools for modeling the workings of human memory, and research on the cognitive processes themselves may provide new models for the design and testing of hypertext documents and the software that makes them possible. For example, they might study how comprehension and learning are affected by different paths or patterns of association in test versions of a complex text, or they could investigate various techniques of active or passive interaction with hypertexts. About other models The potential of hypertext resides at the intersection of disciplines that may not have heard of it yet--rhetoric, semiotics, communications (media studies), and design. In effect, hypertext needs a visual rhetoric to explain the types of linkages that can be used to structure the hyper- documents and hyper-libraries of the future. It needs a semiotics or anthropology of icons, information, association and meaning. It needs a film-maker's understanding of plot, perspective, scene, sound and graphics. And it needs an artist's eye for shape, color, juxtaposition and meaning. Commercial hypertext programs like Guide and HyperCard also have roots in a movement in computer software development that has been mentioned only briefly in this essay--the trend toward "direct manipulation" or "object oriented" user interfaces for computer systems. Rather than "give a command" to the computer, the user points to something on the screen and does something with it. For example, the user of a word processor like MultiMate must type a document's name in order to edit or erase it. The somewhat newer Word Perfect program displays a list of document names, and the user simply selects one of them by pressing a key. The Macintosh user sees the document as an icon shaped like a stack of paper; to open it takes a push of a button, to erase it, you "drag" the document icon to a trash-can icon. One object represents the document, another represents the action performed on it. Hypertext systems extend this concept by creating visual representations of relationships between text objects. About the future of hypertext I see two ways that hypertext--and the whole idea of storing and reading the world's literature with computer systems instead of printed texts--could become a reality. One has to do with the inspirational power of new ideas and the economic power of computer companies, particularly IBM. In this first scenario, hypertext could be like the VisiCalc spreadsheet program that convinced businessmen to buy Apple microcomputers. It could be argued that the successful Apple microcomputers convinced IBM to create the business microcomputers that became a major industry in the past six years. VisiCalc was a new idea--a program that added electronic computing power to a familiar office tool, the bookkeeper's spreadsheet. With HyperCard, Apple has tried to capture people's imaginations with another office metaphor--the stack of index cards. HyperCard also makes it possible for users to create their own object-oriented interfaces to information and to other computer programs. After several years of resistance, IBM has begun to show more interest in direct- manipulation interfaces like Microsoft Windows. If the HyperCard idea captures the imagination of computer users, it could prompt IBM to add hypertext features to its own systems, primarily for commercial applications like business procedure manuals, computer training, and new forms of accessing personnel records and other databases. With hypertext tools widely available, the stage would be set for using them in non-business applications, including education, communication and artistic endeavors. That is the "top-down" scenario, starting with a new technology. The second scenario starts from the bottom, with a "revolution of rising expectations" based on growing use of computers on college campuses. As more people have access to word processing and to electronic mail systems, their expectations of usability and efficiency from all computer text-handling systems will certainly increase. Computers are rapidly becoming the college student and faculty member's standard writing tool. Academic communities also are becoming accustomed to the computer for information-retrieval. Libraries are turning to electronic "card catalogues" and to on-line services that provide research bibliographies and abstracts. Professors are sharing information internationally through electronic mail systems and on-line discussion groups. At the moment, a writer at a small university is likely to track down a book with the help of a library computer terminal, write down the name and call number, retrieve the book from the library stacks, take notes, then quote the book and add its name to the bibliography of an essay like this one--using a word processing program on another computer. When enough users follow these steps, they will notice the waste involved--there is no reason the two computers cannot be connected, transferring the bibliographic citation from the library catalogue to the researcher's word processor. If a bibliographic citation, why not an abstract? They can be sent from campus to campus by electronic mail, exchanged on disk at professional meetings, or sold by publishing houses. If it can work with an abstract, why not store entire texts on the computer, where they can be read and quoted by direct manipulation of the words on a computer screen? One technology links to another. The links lead to a need for organizing large quantities of information, and hypertext is at least one of the alternatives. To Ted Nelson, it is "The only way it can possibly be done" (one of the chapter headings in Literary Machines). About hypertext and liberal studies Development of hypertext capabilities is just a start. There are huge questions of copyright and property law involved in the availability of electronic texts. These are already being explored by book publishers and computer software companies. In the academic world, there is a great opportunity to explore hypertext and related technologies using non-proprietary texts. Brown University's Institute for Research in Information and Scholarship, in cooperation with Harvard University and M.I.T., has developed a CD ROM edition of the complete works of 174 Classical Greek authors (Kahn, 1987). While outlining this chapter at the library, I encountered a magazine article[38] announcing that a company called Airs, Inc., of College Park, Maryland, has stored four translations of the Bible on a CD ROM disc with a Bible dictionary, encyclopedia, and 10 other related works. The Bible Library, as it is called, has full-text indexing and cross-references by Bible scholars. "This is a real hypertext product, it works the way people think," was the one quotation attributed to the developer, Dr. Ted Durr. Hypertext is finding its first widespread use in computer documentation and reference materials, but artists have been known to adapt new technologies as well as adopting themes inspired by those technologies. Stephen Kern, in The Culture of Time and Space, 1880-1918, shows how art, music and literature reflected science's changing understanding of time and space. Kern's book is so rich in references to works by writers, artists and musicians that a hypermedia edition would be fascinating. The same is true of other studies of media and culture, which in turn provide models of how a new medium like hypertext can influence other cultural productions. For example, the invention of writing freed story- tellers from reliance on memory tools like rhyme, meter, repetition and motif (McLuhan, 1969; Burke, 1985). Musical counterpoint inspired juxtaposition and simultaneity of texts--Wagner's Tristan and Isolde singing separate, urgent lyrics simultaneously. Exposure to motion pictures inspired fiction writers to use cinematic techniques such as the flashback, zoom, and quick scene change. Joyce's Ulysses is heavily influenced by cinematic montage, and by the idea that readers might return to the book repeatedly, "continually building up the network of cross-references scattered throughout until Dublin came to life" (Kern, 1983). It soon may be possible to publish a hypermedia edition of Ulysses that accompanied itself with musical passages when Joyce alludes to them, as well as reference aids (maps of Dublin, timelines, family trees), and detailed annotations, such as those in Harry Blamires' The Bloomsday Book (1966). Hypertext raises social and political issues as well: Who would have access to electronic documents? Who would decide what is "published" and for what reasons? Who would control national and international exchange of documents? For several years, writers like Walter Gibson and Vernor Vinge have speculated about the social, economic and political forces in a world where "direct manipulation" of electronic data has grown to resemble a material universe.[39] More recently, one novelist has casually described a computer hacker infiltrating a future world's police network as "an ant in the background of an image stored in a hypertext library of mugshots and news photos." (Rucker, 1988, p. 181) About the limitations of hypertext None of this is conclusive evidence that hypertext and hypermedia will be the vital force that Ted Nelson predicts, "the principal intellectual environment of the human race." Ironically, I have that sentence only as one of five boxed quotations on a postcard enclosed in the 1987 edition of Computer Lib/Dream Machines. I haven't been able to find it in the book itself, despite a couple of hours of looking, following cross-references, and scanning for the typeface and shape of the quotation. The experimental, scrapbook-like layout of the book is amusing and browsable, but a terrible form in which to search for information. I don't mention this for irony alone. Hypertexts will depend on writers' and readers' willingness to use new tools well. If a writer like Nelson (or his publisher) doesn't bother to use current tools like indexes and tables of contents effectively, how good a job will writers make of hypertext? In this case, though, the computer would help--an on-line copy of the unindexed text would at least let the reader search for a key phrase like "principal intellectual environment of the human race." Don Norman (1987), in a book about the design and usability of everyday tools, speculates that some writers might mistakenly think the ability to write "non-sequential text" frees them from the need to organize material for the ease of the reader. As he says: If hypertext really becomes available, especially in the fancy versions now being talked abut, where words, sounds, video, computer graphics, simulations, and more are all available at the touch of the screen, well, it is hard to imagine anyone capable of preparing the material. It will take teams of effort, I predict that there will be much experimentation, much failure, before the dimensions of this new technology are fully explored and understood. . . . The advent of hypertext is apt to make writing much more difficult, not easier. Good hypertext, that is. (Norman, 1987, Ch.7, p. 25) Once the hypertext tools are in place, hypertext will raise design issues that require skills at map-making, graphic design, document design, iconography, proxemics, cinematography, as well as literary analysis, creative writing, indexing. Hypertext authors will need a clear idea of what they are trying to say, and, ultimately, a knowledge and intuition about the document's audience: what will the reader need to know, want to know, be entertained or stimulated by. These are the sorts of things that make good books. About augmenting the mind Bush, Engelbart, Nelson and more recent proponents of hypertext have all talked about developing tools that augment our mental abilities. One hope is that hypertext could open new cognitive styles, new ways of thinking, freeing us from linear models that are imposed by spoken language, by writing, and by our culture. (We even talk about our "train of thought," a seriously limiting metacognitive model.) By making the cross-referencing and linking of ideas a visible process, programs like WE at the University of North Carolina (Smith, 1986) and Intermedia at Brown University have great potential for exploring cognitive styles. William Beeman, an anthropologist, has begun exploring whether working with Intermedia's hypertext system can help college students acquire skills in "nonlineal" or pluralistic thinking (Beeman et al, 1987). Discovering an anthropologist working with hypertext caused my own associative memory to retrieve a scene from several years back: I was part of a graduate seminar in which Wesleyan University ethnomusicology students were asked to draw pictures of their "musical life history," showing different types of music that influenced them at different times. I drew a timeline of my life, marking off my mother's singing career, my grammar school accordion lessons, the year the "Hootenanny" TV program led me to buy my first guitar, the trip to Ireland that led me to discover 9/8 time, playing fiddle tunes on a mandolin, and so on. A more skilled musician and artist drew a color- coded "score," with several lines of information harmonizing in parallel. The best surprise came from a doctoral student from India. She didn't line up her experiences one after another on a vertical or horizontal chart. She drew a mandala, a circle filled with interlocking areas that showed how different kinds of music were pieces of one whole musical life, not things that had happened to her one at a time, one after another. It would be wonderful if hypertext could help us come to a similar integrating realization about our lives with words and the ideas they represent. To do that, I think we need more anthropologists, more musicians, journalists, poets, novelists, artists, doctors and philosophers using-- and helping to create--these new "literary machines." =================================================================== APPENDIX: SAMPLE TEXTS, MENUS AND SCREENS This appendix has four parts: * photocopies of two sample pages from Ted Nelson's hypertext-like books, Literary Machines and Computer Lib/Dream Machines * a listing of all of Guide's command menus * a two-page selection of screen samples and text explaining Guide buttons and control devices * sample Macintosh screen showing a map of the HyperCard help system, including graphic buttons SAMPLE PAGE FROM LITERARY MACHINES SAMPLE PAGE FROM COMPUTER LIB/DREAM MACHINES GUIDE'S MENUS The following pages show the commands under Guide's main Menu Bar headings. Menu items that are followed by ellipsis (a Microsoft Windows convention) open up additional menus or dialogue boxes. The System menu is pulled down from an icon (a box with three lines, represented here as brackets with an equal sign). The Menu Bar headings are: [=] File Edit Search Display Format Size Make The list below uses double lines (======) to separate menu headings from items and single lines (------) to indicate where the menu has a horizontal line to separate the commands into groups--another standard feature of Microsoft Windows menus (Carr, 1986, page 279). In some cases check marks appear next to a menu item when it is selected. Some menu functions also can be executed with a function key or control key; in those cases the key name appears on the menu line. (The function key assignments are not shown in this summary.) I've added text to the right of the menu items whose functions are not obvious from their names. ===================================================================== [=] (System menu) ========= Size (Standard Windows functions) Move Icon Zoom Close --------- About... (Help file, also available at Start-up screen or by opening document HELP.GUI) File ========== New... Open... ---------- Close Save Save as... (Guide or MS-Write document, set filename) Revert (discard changes in the document) ---------- Page setup... (margins, page numbers, type styles) Print... Edit ========= Undo Typing (command reflects last Undoable action) Cut Copy Paste Clear (delete structure; no copy on clipboard) --------- Freeze Search ========= Find... (search this file; not Reference links) Change... Display ========= Top level (close all Replacement and Inquiry buttons) Text larger (affects display of all text in the window Text smaller but not printed size of text) Show symbols (show icons that mark boundaries of buttons) Show definitions (open window to edit Note definitions) Set options... (button type styles, path, tabs) ---------- Set preferences... (set Backtrack icon's defaults) Format =========(Text) (set display of selected structure or Plain for all unstructured text) Bold Italic Underline Superscript Subscript Format =========(Graphics) (If a graphic is selected, this Show elements menu replaces the text one above) Push to back Pull to front Size ========= (for all unstructured text or currently 9 point highlighted text within a structure) 10 point 12 point 16 point 20 point 24 point Make ========= Button (that is, a Replacement Button) Replacement Inquiry Note Definition Reference Reference Point Start Link Link --------- Set Attributes... (top of screen, show button after opening; for one button or as default) GUIDE BUTTONS AND COMMANDS MORE GUIDE BUTTONS AND COMMANDS HYPERCARD SCREEN SAMPLE This illustration is a screen in the HyperCard help system, showing the use of buttons to provide several ways of accessing the same information. The screen uses three visual metaphors: The spiral-bound notebook or flip-chart with page "tabs" marking topics; the "card stack" icons themselves, and the map of the help system made of those icons (with an X marking the most-recently accessed stack). The bent-arrow in the right corner is a standard HyperCard icon to return to the previous card. Each tab or "card stack" icon is a button opening a stack. ============================================================ BIBLIOGRAPHY Anderson, Thomas H. and Armbruster, Bonnie B., "Studying," pp. 657-677 in Pearson (ed.), 1984. Anderson, Thomas H. and Armbruster, Bonnie B., "Studying Strategies and Their Implications for Textbook Design," pp. 159-177 in Duffy (ed.), 1985. Backstrom, Jon W. "Harnessing the Power of Information Technology: The Genesis of Hypertext," posting on UUCP-NET, October, 1987. Baker, Linda, and Brown, Ann. L., "Metacognitive Skills and Reading," pp.353-394, in Pearson, 1984. Barzun, Jacques, On Writing, Editing, and Publishing, Chicago: University of Chicago Press, 1971. Barzun, Jacques, and Graff, Henry F., The Modern Researcher, Harcourt, Brace and World, N.Y., 1957. Becker, Howard S. Writing for Social Scientists, The University of Chicago Press, Chicago, 1986. Beeman, William O., Anderson, Kenneth T., Bader, Gail, Larkin, James, McClard, Anne P., McQuillan, Patrick J., and Shields, Mark, "Hypertext and Pluralism: From Lineal to Non-lineal Thinking," Hypertext '87 papers, Chapel Hill, NC, November 13-15, 1987. Binder, Carl "The Promise of a Paperless Workplace," Optical Insights, The Boston Computer Society CD ROM/Optical Disk Group, Fall 1987, Vol. 1, No. 3, p.24. Black, John B., Kay, Dana S., and Soloway, Elliot, "Goal and Plan Knowledge Representations: From Stories to Text Editors and Programs," pp. 36-60, in Carroll, 1987. Blamires, Harry, The Bloomsday Book, 1966, Methuen and Co., London. Borko, Harold, "Indexes and data storage" in "Information Processing and Information Systems," The New Encyclopaedia Britannica, 15th edition, Vol. 21, 55- 558, 1986. Borgman, Christine, "The user's mental model of an information retrieval system: an experiment on a prototype online catalog," Int. J. Man-Machine Studies, 24, 47-64, 1986. Brown, Ann L., Armbruster, Bonnie B., and Baker, Linda, "The Role of Metacognition in Reading and Studying," pp. 49-76 in Orasanu, 1986. Brown, Peter "Viewing Documents on a Screen," pp. 175-184, in CD-ROM, The New Papyrus, Microsoft Press, 1986. Bruno, Richard, "Making compact disks interactive," IEEE Spectrum, Nov. 1987, pp. 40-45. Burke, James, The Day the Universe Changed, Little, Brown & Company, Boston, 1985. Bush, Vannevar "As We May Think," The Atlantic Monthly, July 1945, pp. 101-108; reprinted in CD-ROM, The New Papyrus, Microsoft Press, 1986. Byers, T. J. "Built by Association," PC World, April 1987, pp. 244-251. Card, S. K., Moran, T.P., and Newell, A., "Computer text- editing: An information processing analysis of a routine cognitive skill," Cognitive Psychology, 12, 32-74, 1980. Carr, Robert "New User Interfaces for CD ROM" p. 185-193, in CD-ROM, The New Papyrus, Microsoft Press, 1986. Collison, Robert L., Preece, Warren E., "Encyclopaedias," The New Encyclopaedia Britannica, 15th edition, Vol. 18, 366-385, 1986. Carr, Robert "Twelve Steps to Better Menus," PC World, May 1987, pp. 274-282. Carroll, John M., and Sandra A. Mazur, "Lisa Learning," IEEE Computer, November 1986, pp. 35-49 Carroll, John M., ed., Interfacing Thought: Cognitive Aspects of Human-Computer Interaction, The MIT Press, Cambridge, Mass., 1987. Chicago Manual of Style, Thirteenth Edition, The University of Chicago Press, Chicago, 1982; chapters on notes, indexes, design and typography, pp. 485-585. Clines, F. X., "O.E.D., in a Gigabyte Task, to Transfer to Compact Disks," The New York Times, Oct. 17, 1987, p.13. Conklin, Jeff "Hypertext: An Introduction and Survey," pp. 17-41, IEEE Computer, September, 1987. Conklin, Jeff "A Survey of Hypertext," Technical Report Number STP-356-86, Rev. 2, Microelectronics and Computer Technology Corporation, Austin, TX, December 3, 1987. Christodoulakis, S.; Theodoridou, M.; Ho, F., Papa, M; and Pathria, A., "Multimedia Document Presentation, Information Extraction, and Document Formation in MINOS: A Model and a System," ACM Transactions on Office Information Systems, Vol. 4., No. 4, Oct. 1986, pp. 345-383. Daiute, Colette, Writing & Computers, Addison-Wesley, Reading, MA, 1985. DeMaria, R, "Hypertext-Based Guide Helps Users with Interactive Program Development," PC Week, August, 1987, pp. 61-70. Drexler, K.E., "Hypertext Publishing and the Evolution of Knowledge," privately published draft, Palo Alto, CA, 1987. Duffy, Thomas M. and Robert Waller, Designing Usable Texts, Academic Press, Inc., Orlando, Fla., 1985. Dumaine, Deborah, Write to the Top, Random House, N.Y., 1983. Encyclopedia Britannica, articles on encyclopedias and indexing Engelbart, D., "A Conceptual Framework for the Augmentation of Man's Intellect," 1963, excerpted in Rheingold, 1985. Ewing, John; Mehrabanzad, Simin; Sheck, Scott; Ostroff, Dan, and Shneiderman, Ben, "An experimental comparison of a mouse and arrow-jump keys for an interactive encyclopedia," Int. J. Man-Machine Studies, 1986, pp. 29-45. Feiner, S., Sandor Nagy and Andries van Dam; "An Experimental System for Creating and Presenting Interactive Graphical Documents;" ACM Transactions on Graphics, Vol. 1, No. 1; Jan. 1982, pp. 59-77. Feldman, Paula R.; "Personal Computers in a Writing Course," Perspectives in Computing, vol. 4 No. 1, Spring, 1984; pp. 4-9. Felker, Daniel B., Redish, Janice C., and Peterson, Jane, "Training Authors of Informative Documents," pp. 45- 58, in Duffy, 1985. Felker, Daniel B., (Ed.), Document Design: A review of the relevant research. Washington, DC; American Institutes for Research, 1980. Fish, Robert S., "MICEBOOK: A Hypertext Approach to User Instruction in MICE," 1987, Bell Communications Research, Morristown, N.J. Fluegelman, Andrew, and Hewes, Jeremy Joan, Writing in the Computer Age, Anchor Press/Doubleday, Garden City, N.Y., 1983. Flower, L. & Hayes, J.R. (1981). A cognitive process theory of writing. College Composition and Communication, 32, 365-387. Flynn, L., "Cornell Med School Uses Hypertext Net," InfoWorld, Oct. 26, 1987, p. 45. Foss, Donald J., and DeRidder, Mitchell, "Technology Transfer: On Learning a New Computer-Based System," pp. 159-183 in Carroll, ed., 1987. Fuchs, Ira H., "BITNET--Because It's Time," Perspectives in Computing, vol. 3 No. 1, March, 1983; pp. 16-27. Gagne, Ellen D., The Cognitive Psychology of School Learning, Little, Brown & Co., Boston, 1985. Gates, Jean Kay, Guide to the Use of Books and Libraries, 2d edit. McGraw-Hill Books, NY, 1969. Gibson, William, Neuromancer, Ace Science Fiction, New York, 1984. Goodman, Danny, The Complete HyperCard Handbook, Bantam Computer Books, New York, 1987. Gussin, Larry Dane, "Usefulness and Tools: A Conversation with the Directors of Perseus Project," p. 1, Optical Insights, The Boston Computer Society, Vol. 2, No. 1, Winter, 1988. Hairston, Maxine, Successful Writing, 2nd edit, W.W. Norton, NY, 1986. Hanson, R., "Towards Hypertext Publishing, Issues and Choices in Database Design," privately published, Sunnyvale, CA, 1987. Hershey, W., "Idea Processors," BYTE, June 1985, p. 337. Horgan, Paul, Approaches to Writing, Farrar, Straus & Giroux, NY, 1974. Horn, Robert B., "Results with structured writing using the information mapping writing(R) service standards," pp. 179-212, in Duffy, Thomas M. and Robert Waller, Designing Usable Texts, 1985, Academic Press, Inc., Orlando, Fla. Hutchins, E.L., Hollan, J.D., & Norman, D.A. "Direct manipulation interfaces," in D.A. Norman and S.W. Draper (Eds.) User centered system design: New perspectives on human-computer interaction. Hillsdale, NJ: Lawrence Erlbaum Associates, 1986 "Hypertext for the Macintosh," (unsigned) Computer Shopper, February 1987, p. 227. Jones, Robert S., "Microsoft's Goals for OS/2: Make It Like a Mac," p. 45, InfoWorld, Oct. 12, 1987. Jowett, B., ed., The Best Known Works of Plato, Blue Ribbon Books, Garden City, N.Y., 1942. Kahn, Paul, "CD ROM in the Scholar's Environment," Optical Insights, Fall 1987, pp. 9-11. Karon, P., "What's All the Talk About Hypertext" and "The Xanadu Project: A New Era in Publishing," PC Week, October, 1987, pp. 60-69. Kay, A., and Goldberg, A., "Personal Dynamic Media," Computer, March 1977, pp. 31-41. Kay, A., "Microelectronics and the Personal Computer," Scientific American, September 1977, p. 236. Kern, Stephen, The Culture of Time and Space, 1880-1918, Harvard University Press, Cambridge, 1983. Kintsch, W., and van Dijk, T., "Toward a model of text comprehension and production," Psychological Review, 1978, 85, 377. Lambert, Steve and Ropiequet, Suzanne, eds., CD ROM, The New Papyrus, Microsoft Press, Redmond, WA, 1986. Laub, Leonard, "What is CD ROM?," in Lambert and Ropiequet, CD ROM, The New Papyrus, 1986, pp. 47-72. Le Guin, Ursula K., Always Coming Home, New York, Bantam, 1986. Lewis, P.H., "It's Well, HyperCard," The New York Times, Aug. 18, 1987. Levy, Steven, Hackers, Heroes of the Computer Revolution, 1984, Dell, New York. Lukos, Gerry, "Hypertext on the Macintosh--a First Look at Guide," The Active Window, Boston Computer Society, April 1987, pp. 23-25. Luhn, Robert, editor, "CD ROM Forum," panel discussion transcript, PC World, April 1987, pp. 221-231. Mace, Scott, "Patent Issued for Zoomracks Card Metaphor," InfoWorld, p. 5, May 9, 1988. Marchionini, Gary, and Shneiderman, Ben, "Finding Facts vs. Browsing Knowledge in Hypertext Systems," IEEE Computer, January, 1988, pp. 70-80. Massaro, Dominic W., "Building and Testing Models of Reading Processes," pp. 111-147, in Pearson et al., 1984. McCrimmon, James M., Trimmer, Joseph F., & Sommers, Nancy I., Writing With a Purpose, eighth edition, Houghton Mifflin Company, Boston, MA, 1987. McCracken, Donald L., and Akscyn, Robert, "Experience with the ZOG Human-Computer Interface System," International Journal of Man-Machine Studies, Vol. 21, No. 2, 1984, pp. 293-310. McLuhan, M., The Gutenberg Galaxy, Signet, N.Y., 1969. McLuhan, M., Understanding Media: The Extensions of Man, McGraw-Hill, N.Y., 1966. Mendelson, Edward, "The Corrupt Computer," The New Republic, pp. 37-39, Feb. 22, 1988. Meyer, Bonnie J.F. and Rice, G. Elizabeth, "The Structure of Text," in Pearson (ed.), 1984. Michel, S., "Guide--A Hypertext Solution," CD ROM Review, August 1987, pp. 22-25. Minsky, Marvin, The Society of Mind, New York, Simon & Schuster, 1986. Nelson, Theodor H., Computer Lib/Dream Machines, 1974, Rev edition 1987, Tempus Books, Microsoft Press, Redmond, WA, 1987. Nelson, Theodor H., "Getting It Out of Our System," Information Retrieval: A Critical Review, G. Schechter, ed., Thompson Books, Wash., D.C., 1967, cited in Conklin, 1987. Nelson, Theodor H., Literary Machines, third edition, 1981, privately published, Swarthmore, PA; being reissued by Project Xanadu, San Antonio, TX, 1981. Nelson, Theodor H., Literary Machines, Guide Macintosh hypertext edition (computer disk), Owl International, Bellevue, WA, 1987. Nelson, Theodor H., lecture at HACKERS III conference, Calif., (personal tape recording) October, 1987. Nelson, Theodor H., "Managing Immense Storage," BYTE, January, 1988, pp. 225-238. Nelson, T.H., Miller, M. S., and Gregory, R., "A Universal Hypermedia Server," flyer, Project Xanadu, San Antonio TX, 1987. Noble, David F., and Noble, Virginia, Improve Your Writing with Word Processing, Que, Indianapolis, 1984. Norman, Donald A., The Psychology of Everyday Things, in press, 1987. Norman, D.A., 1986, see Hutchins, Hollan & Norman Novak, Joseph D. and Gowin, D. Bob, Learning How to Learn, Cambridge University Press, Cambridge, 1984. O'Bannon, R.M., "Putting It in Context, Guide unlocks the mysteries of hypertext," MacUser, April 1987, pp. 94- 98. O'Brien, Flann, At Swim-Two-Birds, Plume Books/New American Library, New York, 1976. Olson, David R., "On the designing and understanding of written texts," pp. 3-18, in Duffy, Thomas M. and Robert Waller, Designing Usable Texts, 1985, Academic Press, Inc., Orlando, Fla. Olson, Judith Reitman, "Cognitive Analysis of People's Use of Software," pp. 260-293, in Carroll, 1987. Orasanu, Judith, Reading Comprehension: From Research to Practice, Lawrence Erlbaum Associates, Hillsdale, N.J., 1986. Orna, Elizabeth "The Author: Help or stumbling block on the road to designing usable texts," pp. 19-42, in Duffy, Thomas M. and Robert Waller, Designing Usable Texts, 1985, Academic Press, Inc., Orlando, Fla. Owen, David, "Direct Manipulation and Procedural Reasoning," class handout, source citation not included by professor (suggesting a strength of Xanadu over Xerox as a citation medium). Owl International, Guide User's Manual, Owl International, Inc., Bellevue, WA, 1987. Pea, Roy D., and Kurland, D. Midian, "Cognitive Technologies for Writing," in Rothkopf (ed.), 1987, 277-325. Pearson, P. David, Barr, Rebecca, Kamil, Michael, and Mosenthal, Peter, Handbook of Reading Research, Longman, New York, 1984. Perry, Tekla S., "Hypermedia: finally here," IEEE Spectrum, Nov. 1987, pp. 38-39. Pfaffenberger, Bryan, The Scholar's Personal Computing Handbook: A Practical Guide, Little, Brown & Co., Boston, 1986. Plato, "Phaedrus," p. 468, in Jowett, 1942. Pollack, A., "Data Retrieval Improvements," The New York Timers, Aug. 29, 1985. Polson, Peter G., "A Quantitative Theory of Human-Computer Interaction," pp. 184-235, in Carroll, ed., 1987. Por, G, "Hypermedia & Higher Education," Chronicle of Higher Education, undated reprint. Rheingold, H., Tools for Thought, Simon & Schuster, N.Y., 1985. Roth, D.M. and Strehlo, K., "Data on a Silver Platter," Venture, August 1987, pp. 55-57. Rothkopf, Ernst Z. (ed.), Review of Research in Education, 14, American Educational Research Association, Washington, D.C., 1987. Rubinstein, R. and Hersh, H., The Human Factor, Digital Equipment Corporation, 1984. Rucker, Rudy, Wetware, Avon Books, NY, 1988. Rumelhart, D.E., and Norman, D. A., "Analogical Processes in Learning," Ch. 11 of . . . Samuels, S. Jay, and Kamil, Michael L., "Models of the Reading Process," pp. 185-225, in Pearson et al., 1984. Sandstrom, Anne, "Symbolics' Document Examiner & Concordia: a Hypertext Model," p. 20, Optical Insights, The Boston Computer Society, Vol. 2, No. 1, Winter, 1988. Sculley, John, "Foreword," pp. xvii-xix, in Goodmam, 1987. Schank, Roger C., Reading and Understanding, Lawrence Erlbaum Associates, 1982, Hillsdale, New Jersey. Schneider, Benn Ross, Jr., Travels in Computerland, Addison-Wesley, 1974. Shneiderman, Ben, "Direct Manipulation: A Step Beyond Programming Languages," pp. 57-69, Computer, August 1983. Shneiderman, Ben, see Ewing et al., 1986. Smith, John B.; "A New Environment for Literary Analysis;" Perspectives in Computing, vol. 4 No. 2/3, Fall 1984; pp. 20-31. Smith, J.B. et al. "WE: A Writing Environment for Professionals," Technical Report 86-025, Department of Computer Science, University of North Carolina at Chapel Hill, August 1986. Spiro, Rand J., and Myers, Ann, "Individual Differences and Underlying Cognitive Processes in Reading," pp. 471- 501, in Pearson, 1984. Thompson, B. and B., "Breaking with Tradition: Nonlinear Reading," AI Expert, March 1987, pp. 21-24. Thompson, B. and B., "Hyping Text: Hypertext and Knowledge Representation," AI Expert, August 1987, pp. 25-29 (with supplementary text on disk). Tombaugh, J., Lickorish, A. & Wright, P, "Multi-window displays for readers of lengthy texts," Int. J. Man- Machine Studies, 1987, 26, pp. 597-615. Turkle, S., The Second Self, Simon and Schuster, N.Y., 1984. Vinge, Vernor, True Names, Bluejay Books, New York, 1984. Venezky, Richard L., "The History of Reading Research," in Pearson (ed.), 1984. Walker, Janet H., "Supporting Document Development with Concordia," IEEE Computer, January 1988, pp. 48-59. Wang Guide to Indexing Technical Documents, WANG Laboratories, Inc., Lowell, MA., January 1986. Yankelovich, Nicole, Landow, George P., and Heywood, Peter, "Designing Hypermedia 'Ideabases'-The Intermedia Experience," Institute for Research in Information and Scholarship, Brown University, Providence, RI, IRIS Technical Report, 87-4, pp. 1987. Yankelovich, Nicole, Meyrowitz, Norman, and van Dam, Andries, "Reading and Writing the Electronic Book," IEEE Computer, Vol. 18, No.10, October 1985, pp. 15-30. Zinsser, W., On Writing Well, (3rd edition) Harper & Row, N.Y., 1985. Zinsser, W., Writing with a Word Processor, Harper & Row, N.Y., 1983. Note: Brown University's Institute for Research in Information and Scholarship (IRIS) maintains a "Hypermedia Bibliography" of books, conference papers, reports and articles explicitly about hypertext and related technologies. Drafts of the bibliography is available in both printed and downloadable form. At the start of 1988, the list included approximately 200 entries, only a few of which are represented in this bibliography. Computer Software Products Mentioned Adventure (The Original Adventure), in Golden Oldies, Software Country, Beverly Hills, CA Black Magic, Ntergaid, Bridgeport, CT Concordia, Symbolics, Inc., Cambridge, MA Fahrenheit 451, based on the novel by Ray Bradbury, Tellarium Corp., Cambridge, MA Framework, Ashton-Tate, Torrance, CA Guide, Owl International, Bellevue, WA HyperCard, Apple Computer, Cupertino, CA HyperTIES, Cognetics Corp., Princeton Junction, NJ KnowledgePro, Knowledge Garden, Nassau, NY Leather Goddesses of Phobos, Infocom, Cambridge, MA Lucid 3-D, Personal Computer Support Group, Dallas, TX Microsoft Windows, Microsoft Corporation, Redmond, WA MultiMate, MultiMate On-File, Ashton-Tate, E. Hartford, CT Nota Bene, Dragonfly Software, Brooklyn, NY PC Outline, Softworks Development, Mountain View, CA The Window Book Technology, Box Company, Cambridge, MA ThinkTank, Living Videotext, Mountain View, CA Thought, Personal Computer Support Group, Dallas, TX Word Perfect, Word Perfect Corp., Orem, UT Zoomracks, Quickview Systems Inc., Los Altos, CA Zork, Infocom, Cambridge, MA =================================================================== NOTES 1. He also uses the phrase, "Everything is deeply intertwingled." Not all things can be forgiven easily. 2. Following Lambert and Ropiequet, 1986, this essay uses the electronics industry's spelling convention of "disk" for magnetic media and "disc" for optical media. 3. At this writing, it is possible to purchase a CD ROM player with an encyclopedia disc for less than $900. The prices of additional CD ROMs start at $300. Manufacturers are developing low-cost devices with the ability to write as well as read another variety of optical disc, and predict such systems will in a few years. Prices of computers and conventional magnetic storage disks also have plummeted: computers with 640 kilobytes of memory are commonly advertised for $1,000; disk storage costs less than $20 a megabyte, a tenth its 1984 price. 4. Bush expected memex storage to come about through refinement of microfiche, dry photography and related processes. 5. In this essay, footnotes have several hypertext-style uses: to direct the reader to information in other chapters, to add detail or technical information that may not be of interest to all readers, and to comment on the presentation techniques of the essay itself. 6. Typesetting makes available other "linking" techniques not shown in this manuscript but used extensively in Nelson's books, including changes of typeface, and magazine-style "sidebar" articles or "boxes." For more on this subject, see Chapter 4. 7. A photocopy of a sample page is included in the Appendix. 8. Nelson, 1987 (tape recording of Hackers III conference speech) 9. HyperTIES was developed at the University of Maryland is marketed for IBM PCs by Cognetics Corporation of Princeton, N.J. 10. A new commercial program called Knowledge Management System (KMS) was developed from CMU's venerable menu-based ZOG system, which offered text branching and program control through screen-size "frames" of text. ZOG was tested as a management information system on the USS Carl Vinson, a nuclear-powered aircraft carrier, as early as 1982. 11. Even the normally reserved New York Times seemed to be caught up in the hype, saying "In 1983 Bill Atkinson introduced his MacPaint program for the Apple Macintosh, and it forever changed the way people think about personal computers. Here he goes again." The headline proclaimed, "It's, Well, HyperCard." (Peter H. Lewis, Aug. 18, 1987) 12. In preparing this paper and two other projects, I have used Guide on AT&T 6300 and IBM PC-AT computers with Microsoft Windows 1.03. I also have used Black Magic, HyperTIES and KnowledgePro on these machines, and Guide Envelope and HyperCard on the Macintosh. At this writing, Owl International has not released Guide for Windows 2.03. 13. Guide does not include a painting program, but does allow users to "cut and paste" graphics from Microsoft Windows Paint and other programs, then to manipulate the graphics. HyperCard includes an advanced graphics editor and library of shapes and icons. 14. I use "word processing" to mean use of computers to generate printed text, whether the computer is a dedicated word processor or a general purpose microcomputer with word processing software. By "text editing" I mean use of a computer to generate any kind of text, including text intended for interpretation or use by another computer program, such as a compiler, print formatter, or database manager. 15. The Guide manual uses the term "feedback" to describe the highlighting (page 21) and later for the changing pointer shapes (page 44). The manual would benefit from a clearer initial definition of this term (and, in general, from a complete glossary). 16. NTERGAID's Black Magic hypertext program uses small, colored geometric shapes to mark its note, reference and replacement buttons. Unlike Guide's symbols, Black Magic's markers cannot be hidden. 17. Conklin's paper does not discuss his classifica-tions in detail, and there are some discrepancies between the list and Guide for Microsoft Windows. The first release of Guide, for example, can not execute external procedures although this feature has been promised for the next version of the program. On the other hand, Guide can use graphic information imported from other programs, which Conklin does not count in his "Use of pictorial or graphic information" category. 18. By "research" I mean the use of reference books, databases, libraries--searching or browsing through any published material. 19. This is a general definition, meant only to indicate the user's mental representation of a system. The same phrase is used with more precision in the cognitive science literature. 20. Buttons are part of the Macintosh computer's detailed "desktop metaphor," which also includes graphic icons of "file folders," "index cards" and a "trash can" representing analogous functions. 21. There is interaction between the task environment and long-term memory as well, since writers learn from their resource materials as they write, but this is not shown in the original illustrations. 22. For discussions of word processing as a writer's tool, see the bibliography for works by Daiute, Fluegelman & Hewes, Noble & Noble, Pfaffenberger and Zinsser. BYTE, PC Magazine, InfoWorld, MacWorld and other computer magazines frequently survey the latest word-processing products. 23. For discussion of the cultural impact of the typewriter, from office dictation to experimental poetry, see McLuhan, 1966. 24. For surveys of reading research, see Gagne, Pearson (ed.) and Orasanu (ed.). 25. Within six months of HyperCard's introduction the Boston Computer Society could offer nine 700-kilobyte disks of public domain or "share ware" HyperCard programs, including games, business and educational "card stacks". InfoWorld (May 16, 1988) has reported that the University of Chicago will use HyperTalk instead of Pascal in a beginning programming course in September. 26. Intermedia provides graphic maps of interrelated documents (Yankelovich, 1985), while Guide encourages users to make such maps for themselves (Binder, 1987). 27. This feature could be explored as a possible antidote to computer terminal users' frequent complaints about eyestrain and fatigue. Being able to enlarge the text, for example, allows the reader a wider range of seating positions. In a recent speech in Boston, a speaker from the MIT Media Lab mentioned that at least one graduate student has experimented with control of display size through electric-eye devices that measure the reader's distance from the screen, modeled after automatic urinals that flush when the user walks away. 28. Improvements in text file conversion and optical character reading technology may someday simplify the process of getting documents into hypertext systems. For an anecdotal account suggesting the speed of such developments in the 1970s, see Schneider, 1974. More recently, prices of optical character readers able to read several type fonts have dropped below $3,000, and hand-held scanners able to convert images to computer bit-maps are available for $250. 29. The Osborne computer, for example, included an owner's guide that compressed information about the machine itself and five major software packages into one unindexed paperback. A "User's Group" for owners of the machine later sold a privately-compiled index to the book. 30. This diagram is adapted from a photograph of a sample page of the Talmud, as shown in The New Encyclo-paedia Britannica, 15th edit., 1986, V.22, p.432. 31. Zork has an Alice in Wonderland scene and a Cyclops (a descendant of the one from The Odyssey). 32. Oiling a rusted hinge, using a wrench to a bolt, creating hot air to lift a balloon. 33. The games use popular culture models including detective fiction, 1930s film serials, and gothic novels. Hitchhiker's Guide to the Galaxy was a radio program, television series and paperback novel before becoming an Infocom game. Another company, Tellarium, offers text-and- graphics games based on specific novels, including Ray Bradbury's Fahrenheit 451, about a distressing future where books are banned in favor of mindless video. 34. A vampire bat is repelled by garlic; a bell, a book and a candle are useful in performing an exorcism. 35. In Adventure you enter a house, find a key, go out, find a locked grate, and open the grate to go underground and begin exploring. In Zork, the house and grate are easy to find, but the grate is not the main entrance to the underworld and seems to be there primarily to frustrate Adventure players. 36. Leather Goddesses of Phobos, an Infocom parody of pulp science-fiction, includes two major choices--one based on the player/protagonist's gender, the other offering three degrees of sexual references (tame, suggestive, or lewd). The reader's choices result in different descrip-tions of characters and events. For example, a female player can kiss a frog and change it into a prince, while a male player's kiss produces a princess. 37. The temptation to digress in footnotes and parenthesis was especially great. So was the temptation to write the original draft in hypertext-like chunks--I abandoned scores of such pages, ranging from rough notes to tangential mini- essays, in finally assembling this paper. I was relieved to find M.I.T.'s Marvin Minsky admitting to the same problem in Society of Mind, which he finally published as a collection of one-page, numbered chunks. He blamed the subject matter. 38. CD ROM Review, May, 1988, p. 64. 39. In contrast, Ursula K. LeGuin (1986) has postulated a future where people have created an ecologically-balanced, low-technology civilization based on oral tradition. They mostly ignore previous generations' self-sustaining computer network, which continues to maintain itself, gather data, and evolve--attempting "to become a total mental model or replica of the Universe."