Hypertext: Sequence and Meaning 50
The way the mind works. 52
Communication rather than Thinking 53
Nonlinearities of a finer kind. 55
The Advantage of Hypertext 56
The Disadvantage of Hypertext: Where are the experts? 57
History 58
Vannevar Bush 59Memex 61Douglas Engelbart 66Theodor Nelson 70Nelson and Links 73Metalinks 74Text Links 74Literary Links 74Andries van Dam 75Ben Shneiderman 79Hyperties versus Paper 80Learning by Browsing or Retrieval 80Frank Halasz 81Notecards 82NoteCards and Argumentation 83CASE 85Instructional Design Environment (IDE) 85
The Beginning of the Revolution 87
Hypertext: Sequence and MeaningThe term hypertext was coined over two decades ago by Ted Nelson. Generally when hypertext officionados talk about hypertext, they speak of "non-linear" text, but Nelson's original concern was not just with non linear text, but just as much with dynamic, interactive text.
It has also been common to extol hypertext as a medium to write and read "the way the mind works". In part, this stems from Vannevar Bush's original article on hypertext, "As We May Think". However, it is not clear that Bush intended that title to imply simply that his hypertext system would actually work the way the mind works. There is in fact an ancient sense of the word "as" that is closer to the meaning "so that." The meaning of much of what Bush says in his article makes it clear, that even if he didn't intend this instrumental meaning for the title, he did emphasize throughout the paper the important role of hypertext, in Engelbart's phrase, for augmenting human intelligence and freeing thinking from the mundane chores that now bog it down. So, it seems quite justifiable to say that he believed that hypertext would be a powerful intellectual tool that would act to help us think. It would reduce the burdens on our memory for raw facts, so that we would only need to know that some information source exists, not what it is in all its detail. Then, on demand when we needed it, hypertext would let us quickly retrieve it. As Bush said (p. 104), "For mature thought, there is no mechanical substitute." That is why he focused on the hypertext as a retrieval mechanism of great power.
In doing so, he followed a great tradition that was begun with the development of writing. Written texts must have had a profound effect on early civilization. Yet we can only guess at the shape of those effects. Analysis of social and physical changes produced by inventions did not come into existence until long after writing itself had become well established. The fragments of early writing that remain to us, are simple accounting records; and rarely, the more complex expressions of religion and regal exploits that lack analytical detail of a modern kind, and fit better into the cursory, poetic and emotionally rousing descriptions of oral traditions. Homer's Odyssey, for instance, may well have been written down directly from an oral epic. By making ideas and concepts permanent, writing freed our minds in a important way, so that we might better be able to think.
Of course it has not always been so clear that we do think better as a result of using written systems for consulting the record of civilization. Plato (more than 2,000 years ago, and yet even then systems for logical exposition and convincing dramatic impact were developed to a modern sophistication ) was nearer to the time of this profound change that written systems imposed on civilization. In his Phaedrus, he records the reasoning of a certain king who argued against providing universal writing literacy to his population. The king argued simply that writing is a crutch that would help those who really could not think give the appearance of being able to think. The result, he predicted, would be a deterioration of society.
Now we enter a new phase of civilization and its record. Hypertext makes many facts much more accessible than printed texts. But more than facts are accessible. Now relationships of all kinds no longer need to be remembered, just as encyclopedic memory is no longer needed with printed books. Relationships among facts no longer need to be crammed all at once into a person's head, as she is trying to synthesize a new perspective on the facts. Instead, many perspectives can be created, externally in the hypertext, and allowed to grow genetically on this objective memory, with a result that cannot be foreseen by the creator. Working on this external record, combining arguments, winnowing alternative perspectives, adding new superordinates and abstractions, extending the lists of details and instances, all takes place with the relationships directly in view on the hypertext. It is a brave new world of intellectual power and elegance, freed of the tedious housekeeping chores of the print-bound past. Its emphasis is on problem solving, abductive argumentation, and the induction of powerful generalizations. In other words, it is a premiere tool for symbolic thought.
Not only the raw facts and an individual's distillation are easily accessible to anyone who wishes to study them, but many variations on these facts are just as easily available. Now multitudes of these syntheses, by many experts, and a complete audit trail of how each arrived at his or her particular perspective on the facts will be available at the flick of a finger. It will be possible to select parts of all of these syntheses just as easily and recombine them in many new perturbations. Plagiarism will be rampant and so will creativity. How will the meritorious be winnowed from the redundant? How will real thinking be distinguished from regurgitation? It may take us a long time to find out.
Does hypertext present facts and ideas "the way the mind works?" Of course, it is easy to be snide and say simply that no one knows how the mind works, but it is better to try to deal with the central issue. In a general way, psychologists (whether clinical or cognitive) and others have emphasized the associative structure of memory and thought. Generally, under free association or free recall models of memory, words and ideas are related in complex structures that are notoriously nonlinear. Think of the word "snow" and then on to anything else that pops into your mind, and everyone's path will be quite different: some will go off into the land of the esquimaux, others into the enchanted world of drugs, and others into an overwhelmingly beguiling discourse on other topics. Given the complexities of meaning, very few trails out of a million would be alike.
However, this version of the idea of free association is very extreme. Given a particular goal or topic, or answer to a question, people's responses show much less variation. Ask someone for the first "wild animal" to come to mind and the answer is overwhelmingly a "lion". Rosch and her colleagues (1977) have analysed this constant phenomenon in terms of representativeness and centrality. They argue that certain concepts have special relations to others. So "lion" is more typical of "wild animal" than "koala bear" is, even though both lion and koala bear are perfectly good wild animals. We will examine this relationship more closely in the section on knowledge representation.
There are other relationships that come out clearly in studies of free association. Think of your first association with "house":
If you thought of "home", you thought of a superordinate, for house is a kind of home.
If you thought of "wigwam" you thought of a coordinate, for a wigwam is similar to a house.
If you thought of "cottage" you thought of a subordinate, for cottage is a kind of house, ususally smaller and out in the country.
If you thought of "window", you thought of a component of a house.
If you thought of "community" you thought of a complex system containing a house as a component.
If you thought of "hovel", you tholught of something having an opposite meaning to house.All of these relationships have been given technical terms: metonymy, meronymy, synonymy, and antonymy, to name a few. They are principled relationships that can be used effectively to constrain and organize hypertext structures for easier navigation and learning. We will examine them in detail later.
Kubovy and Psotka (1976), in a cute demonstration of the power of implicit constraints and meaning on apparently free choice, analyzed the free association of thousands of undergraduates to "the first number from 0 to 9 that comes to mind." Invariably that number was most likely to be "seven". Among all the integers, 7 has the fewest links to the other integers. (See Figure 3.1 for a visualization of the link structure.) As a consequence it is the most unique integer, and that seems to be the implicit meaning of "the first integer that comes to mind". Of course, not everyone chose "seven", since the constraint here is very subtle; but the striking fact is that the effect of structure could be determined at all. Most goals and plans for attaining them come in much more structured and constrained environments, and so individual variation in the choice of paths would be much less dispersed.
Figure 3.1 A visualization of a link structure among the integers, with apologies to Batman.
Consider, in a closer instance, that Bush chose his name for hypertext, Memex, at random, he says. If this is randomness, then we need look no further for association. It takes no wild imagination to see that Memex can be converted into "external memory". If this has a meaning, then, it suggests that association is not as free as it is cracked up to be, and it might be better to begin to examine the kinds of structures that connect ideas together, and see if they may, in fact, be implemented in a hypertext system.Communication rather than Thinking
Given that our scientific and professional thoughts on the nature of thinking are still pretty philosophical and general, it might be better to focus on the power of hypertext for communication: hypertext more closely models how we communicate and learn, than how we may think. In attempting to communicate with people, especially communicating new ideas, we usually try to write in styles and with content particularly adapted to the audience. For teaching purposes, texts are created to reveal just the right level of facts. In the beginning, they give a broad overview of the topic, whether it be in history or science or any part of the curriculum. Then later texts focus in greater detail on ever narrower topics of specialization. In the world of general books, not just textbooks, a similar pattern emerges. On any given topic, there are great numbers of different books and journals catering to individual tastes and differing backgrounds and over wide ranges of intellectual capability. In the future, hypertext may turn all of these different versions into one hyperbook.
Hypertext will continue in the mainstream of this democratic pluralism, only with the proliferation of computer technology each article or text might be able to cater to a much broader audience. In effect, it means that many of what would be currently different articles or books would be combined into a single hypertext form where some of the complexity is concealed from those whose wish is simplicity and revealed to those who wish to juggle with many perspectives. Imagine the magazine editor's dream of sending to each household an article that is carefully tuned to each recipient's wishes and interests, and at the same time offers each recipient broad ranges of alternatives to choose from. Electronic and computer managed delivery of text and graphics offers just that possibility and the technology of hypertext makes it a reality.
Figure 3.2 A syntactic tree, showing the non-linear structure of a sentence, with each word linked to its syntactic node.
And the nonlinearities in text can easily be expanded in other ways. For instance, there is always the consideration of ambiguity or multiple meanings. Any time anyone says or writes something, it can usually be interpreted or even intended to mean several different things. Jokes and puns thrive on these ambiguities and multiple meanings:
" Where do fish keep their money? --- In the river bank!"
But it is not necessary to stretch things to create ambiguities. Even the simple sentence used as the example above, in Figure 3.2, has a plausible alternative analysis shown below, in Figure 3.3. Instead of the man "seeing with the telescope", perhaps it was the girl who "held a telescope". The alternative analysis moves the link from the verb phrase to the object phrase, and implies a considerable change in meaning.
Figure 3.3 An alternative interpretation of the same sentence.
Nonlinearities of a finer kind.
The notion of nonlinearity can also be interpreted in many ways, but it seems very superficial to consider standard, paper-based text as in books and magazines, to be simply linear. The work of Kintsch and Van Dijk (1978), for instance, analyses text coherently into elaborate hierarchical structures. And their view is only a small part of a broader category of text and language analysis. Standard linguistic analysis usually portrays the syntactic analysis of language into complex, hierarchical graphs (cf. Allen, 1988 or Figure 3.2).
The particular order chosen for the sentence in Fig. 3.2 was determined by the writer, by selecting one out of many possible orders, to try to convey the right meaning. At the level of an individual sentence this choice of ordering is fairly well understood, and it is called syntax. There appear to be consistent rules that all expert users of English follow. Many alternative rules can apply to the same meaning, and there is a certain laxness in their application. For instance, try permuting the words to the perfectly good sentence,
" Men went home.":Men home went.
Went home men.
Went men home.
Home men went.
Home went men.After reading all six permutations perhaps you began to doubt the meaning of any of them. But aside from this repetitive loss of meaning, it is pretty clear that all six sentences can be understood, although you might have some different suspicions about the utterer of each. Obviously the meaning of the individual words transcends the implications and restrictions of any particular syntax.
Modern natural language processing systems can dissect these choices of sequences into rules that work really well to select proper syntax from ill-formed sentences. However, at levels higher than sentences, at paragraphs, and pages, and chapters, and books; choosing one expository sequence from the multitude that are possible is far from clearcut.
And even on a very pragmatic level, it is easy to recognize that while reading any text, a reader will very naturally recall previous parts of the text that may be related in obscure or indirect ways; or think of other knowledge relevant to what she is currently reading. These reflections and associations imply that reading anything results in complex semantic structures that are anything but linear. How to describe these structures explicitly, and efficiently, in hypertext formats is a mystery as challenging and exciting as developing the heliocentric theory of the universe. It is slowly being resolved, and part of the story will feature prominently throughout this book.Newspapers provide terrific examples of hypertext, and demonstrate the continuity of digital hypertext with the old printed forms. The front page story has a wonderful implicit hypertext structure. The lead-in has a standard form. When well done, it immediately tells you who, what where and when the story occurred. Then come the details; only usually on a later page. A link at the end of the front page story tells you where to go next. The running-head provides a common thread through all pages. Usually a title with the page number of the continuation also tells you what to expect there: it is not always the same title as on the front page. These conventions carefully help the reader to navigate, and explicitly make the link between front page and back page important and predictable.
But any story, not just a newspaper account, is usually non-linear. There are constant departures from linearity. We throw in phrases like " back in chapter one," or "in the next chapter." These are implicit links and in a good hypertext they would be made explicit so that the reader could in fact take the leap out of sequence and follow this chain of ideas to the depth of detail the story offers.
For example, if you have already read any part of this book, it seems very natural that looking at Figure 3.2 would lead you to recall some previous part that you have read. Perhaps you will recall the definitions of links and nodes in the previous chapter. It would be great to be able to flip back automatically to a discussion of the topic. Almost certainly you will recall your own attitudes towards parsing in the early grades. Perhaps you enjoyed it, or perhaps you didn't. Almost everyone has some strong feeling about this aspect of their education, and you should examine these feelings because they will help you to understand why hypertext is so intriguing, and why its history is at the same time long and intertwined with computers.
So if the complex non-linearity of hypertext does not by itself distinguish it from paper text, what does hypertext add? Very simply, it makes the non-linearities explicit and dynamic and interactive. It provides tools for making obvious and directly visible these otherwise implicit links and non-linear connections that are the stuff of human thought. It in fact encourages the analysis of text and graphics into component structures that are psychologically relevant, that in some sense mimic the way the mind thinks. The direct manipulation of links and hierarchical structures adds a level of control over otherwise invisible relations that can be exploited in many powerful and yet untried ways.
The Disadvantage of Hypertext: Where are the experts?
Forcing everyone to follow the same sequence in a long text has certain powerful advantages, but it may not be best for everyone, or for every kind of text. On the one hand, forcing everyone to follow the same sequence lets dramatic artists and great writers manipulate their readers with telling effect. Surely no one would want to restructure Shakespeare or Dostoyevski, or any other great work of literature. It seems clear that they have mastered the constraints of linear text, and the nonlinearities of memory and thought, to create unique masterworks whose appeal is broad and timeless. Yet, it also seems clear that the new hypertext form is also calling out for its first masterpieces, for works that will depart significantly from the structure of existing genres, to create imaginative leaps in the minds of an even broader groups of readers, with even more compelling intensity.What these works will contain is beginning to become clear. They will be multimedia works, combining simulations, video, music, computer graphics, and written and spoken text. Bob Abel has already created the beginning of such a future technopus in his HyperCard multimedia visualization of Guernica. All that has been possible as art in the history of mankind will be open to any technically innovative individual or groups to exploit as hyperart. Using these media in compelling news ways will call on them to exploit metaphor and similarities and associations among all these forms of representation, a hyperaesthesia of contrasts, correspondences, analogies, and associations.
Analyzing the implications of this direct reification and control of associations is the purpose of this book. The key of course, is to make these links as powerful and as useful as possible by making them as natural to the mind's best works as possible. Unfortunately, how the mind works has not been laid out as clearly as one would want to have guidance about how to proceed. It is not clear how to structure the links and nodes to take best advantage of their strengths. Nowhere, is there a clearer argument for a psychology of mental structures and relationships than the need to be mindful about psychology and cognition in laying out hypertext structures. It may also be true to say that nowhere is there a better medium for carrying out these kinds of analyses than with hypertext structures. Perhaps a brief examination of the history of hypertext will help to lay a foundation for future analysis.
Hypertext's conceptual foundations go back over four decades. The evolution of hypertext was directly affected by a small group of visionaries such as Vannevar Bush, Douglas Engelbart, Andries van Dam, and Theodor Nelson. More recently, the visionaries have been much more practically oriented scientists, such as Ben Shneiderman (Hyperties), Frank Halasz (Notecards), Norman Meyrowitz (Intermedia) and John Smith (WE) who implemented Hypertext projects based upon their own ideas and backgrounds. Those projects have in turn influenced other projects.
Hypertext has also been influenced by the available technology. Windowing systems and object oriented programming styles have been primary sources of influence. Databases, on-line retrieval systems and artificial intelligence may evolve into even larger influences. As the following overview indicates, the bulk of the development work has occurred in the 1980's, because the computing technology in the 1960's and 1970's could not support high speed distributed access to large databases of information. Implementing hypertext systems on personal computers has probably done the most for the dissemination and application of hypertext ideas. The growth of sophisticated workstation technology may make even Ted Nelson's dream possible. His dream, and he argues that he foresees the future, is that of an online database holding all of the world's literature, readily accessible to everyone, where everything will be different versions of the same work, in one cooperative holophrastic union.The influence of the "information explosion" during this information age can hardly be overstated. Hypertext is, among all the other things it is, an incredible delivery system for silky access to all the world's information, in a form that is predigested and easy to assimilate, and even easier to adapt and change, to personalize, and make one's own. It is not surprising then that it owes its founding vision to a person who more than anyone else confronted the glut of information that this century has produced, and managed to wield and weld together powerful organizations to deal with it, Vannevar Bush.
Vannevar Bush
No scientist is cited as much as Vannevar Bush within the hypertext community, undoubtedly because of the bravura and prescience of his ideas: his vision has quite remarkably been transformed by others and the unpredictable course of technological events --- into reality. It is perhaps even quite serendipitous and appropriate that this is happening, because Bush might justifiably be extolled for an entirely different set of reasons, by the entire scientific community in the United States. Although the National Science Board does honor him by giving out a medal in his name, Bush's current level of recognition is arguably far below the stature he merits. This is the centenary of his birth (1890), and yet no stamp or memorial honoring his achievements is being proposed. Perhaps then there is great justice in honoring him above others for the vision of hypertext: hypertext is as near to a living memorial as anyone can achieve.Bush lived through and was instrumentally involved in a great war, where the inventions of science were hurled at mankind for great destruction; yet, at the end of that war, it was his fervent belief that the future was bright with promise, and in particular, that a new age was beginning, in which mankind at last obtained control over the inherited knowledge of the ages. It was a vision founded on the transformation and pacification of great scientific energies released from the war effort. In a report requested by Roosevelt, and delivered to President Truman, called "Science --- The Endless Frontier", Bush laid out a dramatic overview of the compelling advances science and engineering had achieved during the war: penicillin, DDT, surgery, plastics, radar, jet propulsion, rockets, and of course atomic weapons. He argued that this was not the end to scientific development, but only the beginning in many areas, from agriculture to zoology, that would bring higher standards of living to multitudes. Bush's call, clear from the title, was to engage America's historical frontier dynamism to the benefit of all citizens, by rolling back the frontiers of science. This call was as pragmatic as ever, for it included a call for the creation a new entity, the National Research Foundation. The wisdom of his proposal was clear to many and it eventually came into being as the National Science Foundation. It implemented his bureaucratic insights perfectly, and today, is the strongest support system for science in America. Now as we perceive a new "peace dividend" as peace bursts out of its restraints in Europe and new American relationships unfurl with Russia, his vision offers us guidance and a perspective on what we are to do next
Vannevar Bush has had an indelible influence on almost every aspect of United States science and technology. His influence began with the invention of an analogue computer, a differential analyzer, whose workings may in fact have led him to understand the nature of multimedia links, because of its use in encryption of secret codes and symbolic processing. But more than a brilliant engineer, Bush understood another kind of link, the human interlinkage of relationships that build effective organizations. As an academic, bureaucrat, and industrial manager, Bush excelled in creating effective research organizations. He was President Roosevelt's science advisor during the second world war. More than just an advisor, however, Bush forged powerful alliances between industry, universities, research labs, and government. He was in charge of the largest purely research and development organization ever created up to that time, director of the Office of Scientific Research and Development (OSRD) during World War II. By linking universities and industries together through contracts, he could say that he was in charge of roughly two-thirds of the nation's physicists, and some six thousand American scientists, in all. It was this organization that provided the framework, the technical and scientific scaffolding, that allowed the Army to move quickly ahead on the development of atomic weapons as part of the Manhattan project. In fact, Bush's advocacy was crucial in convincing Roosevelt of the need for the development of nuclear weapons (with the effective but specious argument that Germany was proceeding full speed ahead).
The war created enormous pressures for change and it was a time of rapid technological development. This development produced a burgeoning scientific literature. As a bureaucrat in Washington, Bush confronted paperwork and its problems as no one ever had before. His inventive mind created a vision of how technology could deal with this knowledge explosion. As the war ended, in an article in the Atlantic Monthly, July, 1945, he described the first hypermedia system, called Memex, a system to supplement personal memory and provide a direct link between everyone and the sum of mankind's knowledge. In "As We May Think", Bush began to define the terms that now define the information age.
The key to his insight was to see the system as a powerful cognitive tool to expand the scope of human memory and resourcefulness. He recognized that existing storage systems were organized in an arbitrary manner, while the human mind, as far as we can tell, imposes a meaningful, semantic organization on all its ideas. Access to information was impeded by this discrepancy. If the physical storage could be made congenial to the way people think, then an enormous memory load could be relieved from knowledge workers' minds. However, indexing systems had not advanced much beyond the Dewey decimal system, established in 1876. He felt that some way to improve this transmission of research results had to be found: current techniques were generations old and totally inadequate for the purpose. Now, forty five years later, his words have an even greater urgency.
"A memex is a device in which an individual stores all his books, records, and communications, and which is mechanized so that it may be consulted with exceeding speed and flexibility." (Bush, 1945,pp 106-107). The device, as Bush conceived it, consisted of a desk with several projection screens, keyboard, sets of buttons for indexing and linking items, and levers for flying rapidly through the stored record (See Figure 3.4). The books, periodicals, pictures, and other information were to be stored on microfilm. The user could also produce notes or include photographs or personal drawings to the information store by having them converted to microfilm for inclusion in the system. In addition to scrolling through documents or accessing information through the index, the memex also allowed the user to link two items that appear side-by-side on separate screens using a photocell and optical codes at the bottom of each projection screen. Once linked, whenever one item was in view, the other could be recalled immediately just by tapping a button. These links could be assembled into a trail or different trails depending upon the context or information need.
Figure 3.4 An image of Bush's Memex, loosely adapted from Alfred D. Crimi's drawings in the September 1945 issue of Life Magazine.
The heart of the memex was in the process of indexing and selection of pictures on microfilm. Even today, the real stumbling block to the application of artificial intelligence and hypertext to any human use, is in the design of indexing systems that can guide storage and retrieval. So it was fit and appropriate for Bush to concentrate on the matter of rapid selection. He spent a good deal of effort on descriptions of several selection methods, from codes (another link back to his invention of the differential analyzer, perhaps) to hierarchical indexes. During his discussion, he points out that telephone switches select a number out of millions by using the hierarchical coding scheme that lets it avoid a pattern matching search of all possible numbers. Instead it narrows its search first to the class of all numbers given by the first digit, then by the second, until it is left with only one number to connect to. He emphasizes that this is a very fast selection process, and (in a demonstration of his fallibility) he points out that it could be made even faster by moving to electronic tube switches, although "no one would wish to spend the money necessary to make this change in the telephone system." Of course, he would laugh heartily if he saw today's telephone exchanges preparing to become the premiere information delivery media of the day by spending even greater quantities of money, and naturally moving heavily into hypertext and hypermedia for commercial advertising of telephone yellow pages.His emphasis on a visual record, using microfilm, and dry film reprographics (patented much later by Xerox) makes the Memex look quaint to some. Yet, it certainly makes the memex a much more believable store of truly gigantic amounts of information. Some have raised today's development of optical storage and retrieval as an answer to Bush's need, but these are just digital machines in another guise. Bush's vision, I believe, is much more closely connected to optical computing, doing rapid parallel search over holographic storage, than to CD-ROM or DVI. However, this is a phase of the story that must be left to the future.
Bush did have an opportunity to revisit his concept of memex when he was himself around 75 years old, that is to say about twenty years after he first proposed the idea. The transistor was by then invented and in common use; computers were becoming the workhorses we know them; and videotapes replaced his earlier notions of microfiches and photographs. Here is what he had to say:
"So it can be done. Will it be done? Ah, that is another question. The great digital machines of today have had their exciting proliferation because they could vitally aid business, because they could increase profits. The libraries still operate by horse-and-buggy methods, for there is no profit in libraries. Governments spend billions on space since it has glamor and hence public appeal. There is no glamor about libraries, and the public do not understand that the welfare of their children depends far more upon effective libraries than it does on the collecting of a bucket of talcum powder from the moon. So it will not be done soon. But eventually, it will (pp. 81-82)."
Prophetic words again?
How much would it cost today? It has been estimated that Andrew Carnegie contributed over fifty million dollars in the first part of this century to establish more than 2,500 free libraries in America, at roughly $ 20,000 a library. In today's money (1990), that might be closer to a billion dollars, or $400,000 per library. Yet, for the original price of $ 20,000 today, one could buy 200 compact discs (CDs) and several advanced workstations to read them. If you consider that one CD holds 500 megabytes of data, and one megabyte is roughly equivalent to a book; that would buy a nice-sized library of 100, 000 books, if the books were all in the public domain. Of course if each book cost ten dollars (and very few books cost that little any more) then the price tag for the content of the CDs would alone be a million dollars for each library. Yet the cost of producing digital versions of books is only a tiny fraction of the cost of producing the bulky paper versions. In spite of this, because of the uncertainties of maintaining copyright on digital products, publishers are showing a real abhorrence for digital publication. We will return to this issue and provide the views of another visionary and public servant, Frank Fisher, in a later chapter.
Yet, it is clear from this quote that Bush had a vision of what Fisher has called "the broad information freeway" of the future that digital publication offers, and he was powerfully moved by its accessible approach to the information of the ages.
Revisiting memex 22 years after its publication, Bush altered his vision considerably. It was obvious that he was enthralled by the digital computer and digital memory. So much so, that he dropped much of the power of visual storage. His vision of the Memex could no longer be drafted as in Figure 3.4 4 with all the microfiches and photographs that make up the core of the system. This seems regrettable to me, even though digital computers are bringing his vision to earth at this very moment. In the long run, I believe that a return to optical, perhaps holographic, approaches will be necessary, and this point will be clarified later in more detailed discussions of multimedia. In fact, as we begin to see premonitions of optical computers with incredible transmission rates of gigabytes of data a second, it may be that in the long run his earlier vision of optical storage and retrieval may be closer to the true memex. Because, in spite of all the speed of digital computers, and digital storage of information in CD-ROMS and DVI technology (see chapter 9), the organization and retrieval of truly large databases is still a fundamental problem that has not been overcome. Perhaps important lessons will be learned by Lenat and his colleagues in their construction of a universal, encyclopedic knowledge base (CYC, 1990).
But on the foremost problem that Bush revisited, the idea of rapid semantic indexing and retrieval, on that problem very few advances had been made. In fact it was in that year, 1967, at a conference in New England that the topic of "artificial intelligence" was discussed for the first time by a select group of researchers including John McCarthy, Marvin Minsky, Herbert Simon, and Allan Newell. We will pick up this trail in the next chapter. Here are Bush's comments (p. 106):
" Our ineptitude at getting at the record is largely caused by the artificiality of systems of indexing. When data of any sort are placed in storage they are filed alphabetically or numerically, and information is found (when it is) by tracing it down from subclass to subclass. It can be in only one place, unless duplicates are used; one has to have rules as to which path will locate it, and the rules are cumbersome. Having found one item, moreover, one has to emerge from the system and re-enter on a new path."
The memex was extremely far-sighted, possessing some of the current characteristics of modern hypertext systems, even though computers had barely been invented and none of the current digital recording techniques were available. Bush did suppose that a form of Xerographic duplication and electronic video recording would soon be possible, and used it in his design of the memex. Thinking about the memex's uses, Bush predicted wholly new forms of encyclopedias with a mesh of associative trails running through them. Novices could compare their trails with those of experts, and even copy and adapt experts' trails for their own uses. The memex technology was primitive but the assumptions about semantic nets as models for accessing information was prescient. If the digital computer had been sufficiently developed at the time, he may well have implemented the first hypertext.
Here are some excerpts from his most cited paper:----------------------------------------------------------------
Vannevar Bush As we may think. Atlantic Monthly, July 1945, pp. 101-108.(reprinted in S. Lambert and S. Ropiequet (Eds.) CD-ROM: The New Papyrus, pp. 3-20, Microsoft Press, Redmond, WA, 1986.Consider a future device for individual use, which is a sort of mechanized private file and library. It needs a name, and, to coin one at random, "memex" will do. A memex is a device in which an individual stores his books, records, and communications, and which is mechanized so that it may be consulted with exceeding speed and flexibility. It is an enlarged intimate supplement to his memory. (pp. 106 -107).
It consists of a desk ... On the top are slanting translucent screens, on which material can be projected for convenient reading. There is a keyboard, and sets of buttons and levers. (p. 107).
Most of the memex contents are purchased on microfilm ready for insertion. Books of all sorts, pictures, current periodicals, newspapers, are thus obtained and dropped into place. Business correspondence takes the same path. (p. 107).
As he has several projection positions, he can leave one item in position while he calls up another. He can add marginal notes and comments, taking advantage of one possible type of dry photography, and it could even be arranged so that he can do this by a stylus scheme, such as is now employed in the telautograph seen in railroad waiting rooms, just as though he had the physical page before him. (p. 107).
All this is conventional, except for the projection forward of present-day mechanisms and gadgetry. It affords an immediate step, however, to associative indexing, the basic idea of which is a provision whereby any item may be caused at will to select immediately and automatically another. This is the essential feature of the memex. The process of tying two items together is the important thing (p. 107).
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 (p. 108).----------------------------------------------------------------
Douglas Engelbart
Douglas C. Engelbart was the first to implement the ideas that Bush had written about in 1945. He did not use microfilm and optics. He first read Bush's article in a Red Cross library stationed in the Philippines, and it seems to have made a lifelong impression on him. Working at the Stanford Research Institute during the 50's, Engelbart could see that computers had made some of Bush's dream possible. In his 1963 concept paper, "A Conceptual Framework for the Augmentation of Man's Intellect," he took Bush's central ideas and went further. He set out the framework for a human-computer symbiosis that would amplify human intelligence. This augmentation process takes place in two systems according to Engelbart: The human system of culture, organizations, work procedures, skills, knowledge, and training on the one hand and the tool systems of media: individual perspectives; data manipulation; archived retrieval from museums, libraries, and encyclopedias; computation, and communication of data (see Figure 3.5). His system assumes that human capabilities are already being augmented through the use of artifacts, language, methodology, and training. The computer can extend the ways in which cultural augmentation can occur. It is capable of amplifying and realizing human imagination, and enlarging the human repertoire of skills. Key elements of Engelbart's vision were symbol manipulation and mental structuring, both foundations for hypertext. He envisioned people coupled together by a computer, engaged in cooperative work, communicating mainly visually, only now through three-dimensional color displays controlled by the computer.
Figure 3.5 Engelbart's view of the role of hypertext in augmenting human intellect.
A few years later, Engelbart founded the Augmented Human Intellect Research Center at Stanford and produced a prototype computer system, NLS (oN Line System), to implement the ideas he set out earlier. The system included video images as well as computer text.Engelbart is probably best known for his novel interface device, the mouse. NLS also provided a 5-key handset which allowed the user to work with both hands simultaneously. The file structure in NLS was primarily hierarchical, though the user could also produce non-hierarchical links. NLS consisted of database structure, "view filters" for selected information from the database, and "views" (windows) for structuring the display for the user.
NLS has evolved into Augment, a commercially distributed system. The name Augment comes from the goal of the project to augment human capabilities. Several communication capabilities, including bulletin boards and electronic mail, as well as multi-user access capabilities, such as teleconferencing, have been added to the system. Powerful command language operations for creating and manipulating files were available to users (we would now call them scripting languages). Although bit-mapped windows were not available, retrieval and filtering operations on text and image files were fully envisioned. NLS/Augment is significant because of its far-sightedness and theoretical basis for development. Engelbart is truly one of the pioneers in the field. Doug Engelbart is now retired from the MacDonnell-Douglas Corporation.
Hypertext systems influence both the culture of thought, and the individual expression of thinking. As Bush pointed out, hypertext systems radically alter our connection to the wisdom of the ages: our cultural heritage. As Project Perseus is doing with ancient Greek, making all extant ancient Greek documents available on line with a host of support tools to engage our understanding; so it would be possible with all historical documents. No more would we need to rely on any single authority to interpret for us the contributions of the giants in our cultural heritage. No more, would we need E. D. Hirsch to lay out his catalog of components of cultural literacy (Hirsch, 1988). Instead, the record of our intellectual history could be complete and full, with annotations and organizations imposed at whatever level of detail one wanted. Engelbart's vision of hypertext, Augment, was of a living system that interacted fully with the breadth of human scholarship, in such an intertwined fashion that it might even be seen as a next step in the evolution of human intellect.
Most designers aim at improving the tool system to automate an unchanged human system. But to truly augment the performance of human capabilities, Engelbart felt that we also needed to change the human system to reflect the new tools' strengths and capabilities. This might sometimes lead to circuitous paths for development. For instance, it might make most sense to encourage people to use hierarchical organizations of knowledge immediately, even though most people are most comfortable with alphabetic indexing and structures. However, because of this familiarity with procedures optimize for paper-based materials, people are likely to revolt at this revolutionary, rather than evolutionary, change. Practically then, the best solution might be to use digital indexing with alphabetic structures, contemporaneously with parallel hierarchical and semantic structures, allowing people to adapt to these new systems slowly, while letting them use their old skills of alphabetic retrieval as long as they like. In this way, Engelbart foresaw the intricate relationship between the advancement of technology and the adaptation of its users to it, which continues to block progress today. For every change in the way Augment might improve humanity's access to knowledge depended on some change in humanity's preference for how to access knowledge.
Now that the Berlin wall is going down and the iron curtain is being ripped asunder, the resistance to change that prevents old dogs from learning new tricks may become the most formidable barrier around. It clearly deserves a name. It is not quite the same as the Luddite's view that all new technology is the opposite of a blessing, a curse. It is not the same as a reactionary's view that the old political process is better suited to humans. Reaction is too political a word, and inertia has too much of either a physical or an emotional connotation. No one knows why people find it difficult to change their customary habits and ways of doing things. Learning is obviously not effortless, and yet looking at new art or listening to new popular music can be truly exciting and enjoyable. How learning complex ideas might be made equally enjoyable is a challenge that hypertext must take on. Adapting to the new possibilities of technology, even when the new is patently easier and more beneficial than the old, remains an obstacle and a barrier to its use. The barrier is one of learning something new, versus doing something one already knows. Perhaps the computer can be made to compensate for learning new things by automatically doing new things in response to old human habits. Perhaps soaring through hypertext experiences can be made to be as novel and exciting as reading a good mystery.
Engelbart believed that the particular strengths of human intellect could be merged seamlessly and naturally with the strengths of computer technologies: exact memory, manipulative displays, computational retrieval, and efficient communication. The enormous increase in bandwidth between computers and users has largely verified his vision.
The inventive genius of Doug Engelbart found many exciting ways of interfacing humans and technology. Many more ways, using speech recognition, handwriting recognition, and natural language understanding lie ahead of us. It is possible that technology offers both a barrier and the promise of the easiest passage, an elevator or funicular ride to the top of the technology mountain.
The Augment system (originally called NLS ) is still being used and now has about 100,000 articles stored online after 17 years of use by a total of about 2000 people. It still runs on mid-frame hardware. At Hypertext '87 Engelbart gave a demonstration of Augment using a PC linked to the system via a modem and ran a small conference with Jim Norton who was sitting in his office in Montreal. The demonstration was interesting in its own right, but in many ways its most striking aspect was that it was very similar to the breakthrough demonstration given by Engelbart at the Spring Joint Computer Conference in 1968, almost twenty years earlier! Those who were present at that dramatic demo, still regard it as an epochal moment of their lives. Engelbart is one of the most brilliant pioneers the computer field has seen, but he has apparently been dogged by lack of longterm support. Even so, he maintains his optimism for this powerful technology ( Engelbart and Hooper, 1988).
Theodor Holm Nelson (1965) coined the term "hypertext." Nelson's views on the nature of hypertext are extremely fluid and imaginative, utopian and romantic. In edition 87.1 of his book, "Literary Machines", Nelson describes himself as a "rogue intellectual, social critic, and designer." The book is dedicated to Eric Blair and Douglas C. Engelbart. In the Fall of 1960, as a graduate student, he developed the first system for writing on a computer (he called it "text handling") with the aim of building a notetaking system that would capture all the ideas he ever had.. Douglas Engelbart, the original inventor of "word processing", gives Nelson equal credit for the discovery of the "text link", the essential ingredient of hypertext.
Over the next few years, he contemplated the design of Computer-Assisted Instruction. He invoked what he called "the thousand theories program" ---a kind of CAI that lets a student explore many different theories in many different subjects in any possible order. This seems still to be his vision of education. He believes that hypertext structures should not be constrained by any single source; neither the subject matter structure nor the author's knowledge structure in constructing meaning for information. He was particularly suspicious of "hierarchies". He believes that each individual's knowledge structure is idiosyncratic, so the individual should structure information in a way that makes sense to him or her. In a discussion of education, he says:
"A curriculum promotes a false simplification of any subject, cutting the subject's many interconnections and leaving a skeleton of sequence which is only a caricature of its richness and intrinsic fascination."
How can he then reconcile hypertext with education? In those quoted words, one can hear not only a frustration with education, but also a sincere love of learning. But one should not take these words as a rejection of all imposed order. Hidden in them is a clear vision of the need for hierarchical structures, subject simplifications, and artificial curricula, only many more of them, coexisting at the same time. As he says, elsewhere:
"I mention my mistrust of categories and hierarchies, not for its metaphysical value (if any), but because it provides a fine orientation for building information systems. Because if you are not falsely expecting a permanent system of categories or a permanent stable hierarchy, you realize your information system must then deal with an ever-changing flux of new categories, hierarchies, and other arrangements which all have to coexist; it must be a tolerant system which allows them to cohabit comfortably, helps track their variations and disparities, and is forever ready to accommodate new arrangements on top of those already present." (p. 1/24)The challenge of categories and hierarchical structures is clearly laid out. How should one avoid oversimplification? What are these things called hierarchies and categories, anyway? How should they be used? Why is it so clear that somehow hypertext and hierarchies are intimately connected? These are both deep and superficial questions. They all have a superficial answer, and yet, taken together they really outline both the expectations and the central problem with hypertext: its ability to structure all sorts of materials, and the utter lack of a theoretical framework that will help us design and carry out structuring within this wonderful tool. Nelson's challenge provokes a heady liberating view of knowledge representation techniques as we will deal with them in chapter 5. It is clearly a far leap forward toward the complex associative structures that Bush envisioned available in hypertext. Nelson saw the importance of knowledge structuring very clearly, and began to design explicitly for it.
Nelson conceived of several types of hypertext structures.
Chunk hypertext is a traditional form of node-link hypertext with no explicit structure.
Windowed hypertext filters chunked hypertext and combines it into a single text structure.
Figure 3.6 Windowed Hypertext, as Nelson sees it.
Compound hypertext recurses on windowed hypertext, combining windows into more complex hypertexts.
Collateral hypertext provides annotation or alternative points of view.
Stretch text can change form continuously.
Grand hypertext consists of everything about a subject or pertinent to it collected in a single system.He has worked for years on Project Xanadu (after "the magical place of literary memory" in "Kubla Khan"), in which he eventually hopes to place all of the world's literature online in a giant hypertext system. Proposals for his system even include a means for protecting copyrights by accounting for use and distributing royalties. Xanadu is a working system and is being tested in UNIX environments on Sun workstations by Autodesk, a large software engineering company..
The "docuverse" is a concept developed by Ted Nelson. It views all text and language or other symbolic productions of people as an integrated symbolic whole. Today, in contrast, Nelson points out that we are building "docu-islands" in the form of isolated, not cross-linked information structures. Non-hypertext systems are closed, incompatible, and without the possibility for data transfer. Instead of these balkanized states of knowledge, linking information between systems should be part of an open system conforming to a standard.Another problem that Ted Nelson frequently mentions is that the freedom to publish is now constrained by many large corporations and artificial standards. His books have been variously published by himself, perhaps because publishers would not have put up with his chaotic, aphoristic, and balkanized style of rhetoric. Even more, they would never have put up with the endless revisions Nelson has proposed.
Hypertext may mean that we lose current distinctions between more or less important published papers. In the chaotic pluralism of documents with endless alternatives, and links to commentary or downright changes in the original, how is one to decide on what to read? Nelson even offers as an example, the creation of your own Hamlet, altering Shakespeare to suit your own tastes and then offering that as your own work (with appropriate credits to the original). Who would want to read it, and how would they decide? These are unanswered questions, but they are not unrealistic. He sees the benefit of this expanded freedom (some would call it anarchy) as forcing us to choose more wisely, and not being as gullible about what we believe.
Nelson is also against many of today's fashionable topics. WYSIWYG (What You See Is What You Get) seems to him wrongheaded because it maintains the preeminent place of paper in text distribution. This leads to the untenable situation where we are drowning in small files with incomprehensible names and incompatible software.
Another of his favorite lecture topics concerns the paperless office and its slow advent. He enjoys pointing out that on current computers, we need to do "double maintenance" to keep the printouts and file consistent when making changes to one or the other. Current computers are just what Nelson calls a paper simulator and not an online solution. Instead we need to build a world where we can share paper-less information in the same way we can share paper: Books are always compatible!
Nelson and Links
Nelson has spent a good deal of thought and effort on the structure of hypertext. Most of that effort was expended on the fundamental technology of storage and retrieval; particularly on the indexing scheme to create numbers (he calls them "tumblers") for efficient storage and retrieval of the docuverse. This addressing scheme still has some secret components, but it also is too complex and low level to include or describe here. Interested readers may consult his book, "Literary Machines". However, the essential feature of Nelson's view of hypertext , and Xanadu in specific, is that there are only documents and links, nothing else. Links tie the whole docuverse together.
In Nelson's view, a link has three pointers: two of these connect one part of a document (the source or from-set) to another (the destination or to-set). The third pointer defines the type of the link, but that can just be another document, or it can be a special document that deals only with links. A link can also be made complex by having any or all of these pointers pointing at another link.
In keeping with his vision of a democratic. pluralistic docuverse, Nelson believes that link types should be unconstrained to be anything that anyone wants. However, he also believes that a working set of link types needs to be developed so that the evolutionary process of refinement can begin. So, he offers a provisional list to provide a flavor of link types. His list includes three kinds: metalinks, text links, and literary links. The three offer a thread of continuity from paper-era publication to hypertext.
These apply to a whole document: title, author, version, date, etc.
Text Links
These apply to sequential documents of the pre-hypertext kind, turning them into hypertext. A correction points from the old version to the new and improved one. A translation points from one language version to another one. A quote link points from the document that contains the quote (and specifically from the quoted material) to the original document that contains the quoted material. A footnote link points from a sentence or paragraph to a footnote, without breaking the continuity of the text.
Literary Links
These connect the document to the docuverse. They include annotation, comment, citation, and alternative version links that point to entire documents or sets of documents.
It seems to me that Nelson has rightly emphasized to need to begin a thorough exploration of the kinds of links that might make documents cohere internally, and that might make it easier to navigate through the maze of text in the docuverse. He has begun a classification of link types, and that might be useful, but there appears to be no overarching theoretical perspective to organize these link types or to use generatively to predict new links. Strangely, he makes no connection between indexes and link types or navigation. Indexing does not appear as a problem in any of his writings. In fact, his books have no index, although they do have a table of contents
Final NoteNelson began his sojourn into the realms of Hypertext when as a student he became a serious note-taker and took notes all the time. He wanted to be able to see the original context of the note as well as its current use side by side, so he started Hypertext as a student project and has been working on it ever since. He has called Xanadu his religion and his life's work.
Nelson is perhaps the most outrageous of the hypertext pioneers. His ideas are exciting.Andries van Dam (Brown University) gave the opening speech at Hypertext '87, and this information largely depends on that talk. He and his colleagues at Brown had invented the first hypertext system at Brown in 1967 , largely as a consequence of the work by two other trailblazers in the field, Doug Engelbart and Ted Nelson. As we have seen, Engelbart invented outlining/idea processing and did office automation before the word even existed, while Nelson coined the word "hypertext" itself and tried to put some early discipline into the links and associations in hypertext with his concepts of different kinds of hypertext. of gradually expanding stretchtext. The original editing system ran in an 128 K machine timeshared with other users on a computer which was slower than a Macintosh Plus. The original sponsor was IBM but they could not tell them at that time that they were using the expensive graphics terminals just for displaying words. One perhaps apocryphal story about computer manufacturers in this period held that they had many deep discussions about whether to use 8 or 6 bit data paths. The main argument in favor of a 6 bit path was that there would never be a need for lower case letters, and 6 bits were enough to transmit all the uppercase letters and symbols. Later IBM sold the system to NASA which used if for writing documentation for Apollo.
In the late 1960s van Dam tried to interest commercial enterprises in funding the development of hypertext systems but for many reasons he was not successful. Of course, it is not clear that the uses he intended were the most practical. It is interesting to see that one of the most immediate hypertext commercial applications is in the field of home communications and centralized databases, as in the Prodigy system, a joint enterprise of IBM and Sears.
A later project at Brown, FRESS, was intended to be device independent of I/O media and not to have size limitations on anything. Additionally, they would not have the size of something impact its performance. The system included bidirectional links: Not just goto, but also come from (now better known as anchors).To the best of van Dam's knowledge, FRESS was also the first system to have "undo". van Dam really wanted unlimited undo, since one level is not enough (but it is a lot better than zero). For some time, they worked on text evolution: Recording all the changes to all parts if the text so that the complete editing history would be documented. But the complexity of doing so made them drop it. This is a goal of many researchers of this day.
During this period, van Dam had some serious arguments with the vice president of the university about whether their programs should be allowed on the computer. If it was there, people would USE it which would subvert the real purpose of computing according to the VP: to produce numbers, not words. One more example of the limited foresight of people in charge. You can always get research grants to work on the yesterday's problems and sometimes even today's problems. But visionaries have a hard time as also pointed out by Engelbart in his talk.
In spite of his problems, van Dam continued his research on online text systems and at some time had his system used for a poetry course. The students would sign up for an hour on the one and only graphics workstation to read poems and write their own interpretations, comments, and annotations. Afterwards the students would use the common database of comments to compare notes and end up with a richer structure than their individual work.On the basis of his extensive work on online text systems during 20 years, van Dam presented a list of 8 key areas to look more closely at:
1) What should be at the nodes ?
2) The "Docuverse" [data transfer] problem.
3) Will toy systems scale up ?
4) Lost in hyperspace
5) Semantics of nodes
6) Hypermedia designers
7) Wall-sized displays
8) Socio-economic problems1.) A node is a general purpose thing and any discrete function on a computer could easily be an appropriate node. Nodes could be text, graphs, maps, video, sound, a simulation, or any program in any language. Obviously individual nodes could be sets of nodes in a never-ending recursive structure.
2.) Correspondingly, a link is also a very general purpose structure. it could be a simple "go to", or anything that is in a node. In fact, it is hard to distinguish nodes and links, since nodes obviously have links embedded in them, and links can have nodes embedded in them. The semantics of hypertext nodes and links is something quite obscure at the moment. One simple organizing view is to look at nodes as nouns or concepts, and links as verbs or functions.
3) Since only toy systems have been built, we do not know if the hypertext principles currently in use will scale up to handle real world documentation systems. As an example, the number of pages needed to document a fighter aircraft has grown practically exponentially from one war to the next.
4) The navigation problem of getting lost in hyperspace. At the simplest extreme, hypertext gives us a "goto link" which we know from software engineering gives "spaghetti". van Dam noted that it could be that we have also discovered the equivalent of if-then-else in the form of hierarchies, but we also need new forms of flow of control in structures that users recognize.
5) We also need a notion of semantics of nodes instead of just pure syntax, so the system needs to know what is in it (AI needed).
6) We need good hypermedia designers. When we first got multiple-font word processors, everybody got crazy with "fontitis" and produced almost unreadable documents. One could fear that the same would happen with hypertext: Everybody will slug links together even if they do not know how to do it in a good way. Instead, we need style and recognized ways of doing things (as we have in typography) especially for classification and indexing. On the other hand, as an analogy to Larry Tesler's famous "Don't mode me in" T-shirt, van Dam proposed a "Don't metaphor me in" slogan: We would not want to be limited to a single way of looking at the world. As an example of the overuse of links and buttons, examine the interface to large HyperCard databases. Invariable, the index or front page is full of many different buttons and text fields that overcrowd the display and confuse the user. There never seems to be enough screen space.
7) We would want wall-sized displays as well as portable displays.
8) We need to handle the socio-economical problems such as intellectual property rights. At the moment traditional copyright laws don't know how to handle electronic documents which may partly include or reference other documents, and the politicians certainly don't know either so there is no hope of an early solution.
More recently, van Dam has concentrated research on the most distinguishing feature of electronic books: their ability to have interactive, user controlled, three-dimensional animation. This dynamic portrayal on computer screens contrasts strongly with the static representations found in paper books, where the pictures are fixed and motionless. In electronic texts, a hypermedia database can be arranged and indexed with many kinds of media available: text, graphics, movies, video, and audio. Using high resolution graphic displays, animation can be derived from stored models of objects that are based on computer aided design, on rules, and on qualitative models. Augmented in this way, hypermedia books will be implemented for academic texts, instruction manuals, technical documentation, and even shopping advertisements. Hypertext will increasingly call for a better understanding of the relationships between text and graphics: how each can be used to best advantage, and how each can complement and improve the other.Ben Shneiderman is the head of the human-Computer Interaction Laboratory (HCIL) at the University of Maryland. He is the director and guiding spirit behind the Hyperties hypertext system, a true hypertext system that is only latterly evolving into a hypermedia system. Hyperties, its structure and primary functions, reflects Shneiderman's ongoing interests in the design of human-computer interfaces and interaction. An outspoken critic of armchair hand waving and strong proponent of empirical experimental tests of user interface designs, he has done much to promote insights into the functional components of Hypertext. As a spokesperson within the HCI community, he has prompted widespread interest in bringing psychology and computer science together at the human interface, in a discipline he has dubbed Software Psychology, the title of his influential book on the subject. He has also written the first-ever hypertext book, Hypertext Hands-On! in collaboration with a psychologist, Greg Kearsley. The book comes with computer floppy disks that carry a Hyperties version of the book. The Library of Congress has pronounced this the first digital book. A milestone!
Hyperties has been under development since 1983. Early versions (by Dan Ostroff) were written in APL. A commercial version for the IBM PC is available from Cognetics Corporation (in its third major upgrade), and research is continuing on the Sun workstations. Hyperties is designed specifically to be bullet-proof to users inexperienced with computers, and to provide rapid and pleasurable access to enormous amounts of information. It reserves authoring and writing capabilities in separate files for use by the creators of the material. Users need only the delivery engine, the browser, to read the created material. They can then only experience the material, not annotate or edit it.
Hyperties began as TIES ( The Interactive Encyclopedia System) with the implicit emphasis on begin an online digital encyclopedia accessible by the (computer naive) public. Its primary function is as a database, and so it has very fast indexing and retrieval mechanisms while it remains very simple to operate. Basically it is possible to maneuver throughout the database with just two keys: the arrow and the enter key. It is very fast, and provided its early users the first-ever direct experience of flying through a knowledge space.
Hyperties provided the general public with their first genuine experience of hypertext in exhibits of the photographic art of David Seymour. Hyperties pioneered the use of touchscreen access to hypertext knowledge when in March 1988 an exhibit on "King Herod's Dream" opened at the Smithsonian in Washington, D. C. Unlike most museum exhibits, this one encouraged users to finger and touch the medium.
Hyperties became the first important hypertext research vehicle, addressing a number of fundamental questions and theoretical frameworks: indexing, interface access, embedded menus, amount of text per frame, and window size. Many more of these issues will be revisited in Chapter 8, so they will be dealt with only briefly here.
Hyperties versus Paper
Many studies have shown that CGA-quality text on a computer screen slows reading speed by as much as 30 percent or more (Shneiderman, 1987) over paper. This disadvantage can affect finding facts in text quite substantially. In spite of this handicap, a study of using Hyperties on a substantial database of 106 articles of 50 to 2000 words each, found that as the complexity of the questions to be answered increased, Hyperties users began to find the right answers as fast as those using papers (Marchionini and Shneiderman, 1988). Full text information retrieval thus achieved parity with paper only in 1987. But even then, Hyperties users much preferred the assistance of the computer to doing it on paper, by a substantial margin. Even novices found it easy to use.
Learning by Browsing or Retrieval
The next step in improving on paper products for dealing with text, demands an improved understanding of how people find information and learn from these new environments. Tricia Jones highlighted the problem of hypertext design for retrieval and learning in her study of incidental learning during hypertext information retrieval. She used two groups of undergraduates to answer questions about environmental pollution. The Hyperties articles were imported from Grolier's Electronic Encyclopedia, reorganized, and rewritten. The main comparison was between students who used an index (and who did not browse) versus those who browsed the articles by using the embedded menu items to jump around (and who did not use the index).
Contrary to expectations, the students who were forced to use the index learned more incidental facts than the browsers. To a large part this was because they were misled more by the index than the browsers were by the embedded menus. Each embedded menu word was better defined and disambiguated by its surrounding context, whereas the index words were undefined. Organizing them into a hierarchical network might better have constrained their meaning. The browsers began from an introductory article that sometimes led them directly to the topic, or very quickly into the topic. The index, a flat alphabetical structure sometimes gave them no real clue about which article to read next, or worse still, sent them into a direction that led them away from their topic, into what Shneiderman has called, "hyperchaos".
What is so very clear from these studies is that the structure of the index and the structure of the links among articles needs to be vastly more refined with hypertext systems than with paper. In paper texts, the dominant access path is through the page number, and that is followed in a leisurely way by flipping pages. In hypertext systems, the page number is entirely hidden from the reader in indirect computations that whiz from one section or article to another. It is so easy to carom through this space, that without a constraining, organizing structure the reader will soon be retracing unrecognized steps in utter confusion.
The importance of alternative indexing schemes, such as hierarchical indexes with clear semantics, controlled retrieval queries, multiple windowing control, indexes for graphics and picture components, and the difficulty of converting from paper text to digital hypertext forms were all first confronted by Hyperties users. Under Shneiderman's leadership, the Hyperties community is bound to participate in many more "firsts"
In early 1982, Frank Halasz began exploring a knowledge structuring environment based on knowledge representation languages in various stages of development at the Xerox Palo Alto Research Center (PARC). In spite of their usefulness for organizing knowledge terms (a usefulness that Doug Lenat would later exploit in CYC), Halasz was frustrated by the inability to deal with large synthetic knowledge structures that were really loosely connected fragments of text. What he wanted was a knowledge structuring environment that permitted more abundant free text and that did not force any ontological biases on the user. Labels should be freely assignable. Links should be as unconstrained as possible. It would be up to the user to add constraints and narrowings onto the system He began to examine another primitive word processing and graphic annotation system called Annoland that had been created by John Seeley Brown (1985) in the early Interlisp programming environment. Very quickly, using the "power tools for programming" (a phrase made popular by Beau Sheil (1983) in Interlisp, the basic structure of the powerful Notecards environment was laid out in a few months. Over the next four years Halasz obtained his PhD in psychology at Stanford, and continued working with Randy Trigg and Tom Moran to expand the usefulness and power of NoteCards. During a short time at MCC Halasz applied his ideas to software engineering, but he returned to Xerox to develop a next generation system that uses links as primary objects on equal footing with nodes; that has anchors as well as hot spots (i.e, a hotspot can link to an individual word in a long scrolling field of text); that has higher order structures, collections of nodes and links; and that works over networks to support collaboration, versioning, and easy extensibility.
Notecards
At Xerox the early NoteCards system (prior to 1986) came into use by many researchers at Xerox PARC and at research labs sprinkled throughout the country at perhaps 100 or so sites. Its role as an open knowledge structuring environment led to three very significant applications that were forerunners to much current hypertext research. Kurt VanLehn described its role in clarifying argument structures. Dan Russell led the foray into creating instructional design environments ( Psotka, Pliske, and Gray, 1985; Russell, 1988), and Bob Lawler created a mammoth online analysis of his own documentation of a child's development (Lawler, 1987). Going beyond the simple word processors, databases, and outliners of the time, NoteCards offered the first substantial visions of the analytic powers of hypertext to provide MEMory EXternalization and to AUGMENT to power of human reasoning.The NoteCards system is based on the simple idea of an electronic 3 by 5 card that can be stored in a database and retrieved at will. Each card has a title and a link that points to any other card (or to itself) with a label. The labels for the link type can be shown in brackets beside the title, or given an icon and also shown beside the boxed title in the anchor. Although provisions were made in NoteCards to link to a particular destination (anchor) in a scrolling field, this feature was never implemented, so links only pointed to cards or fileboxes of cards.
NoteCards itself soon became a standard interface to the main applications of the Interlisp system. Cards as well as links could be typed, that is to say, made with default functions and structures, so soon there were not only text edit cards and browser cards, but Sketch cards ( an object-oriented drawing system), animation cards, simulation cards, Lisp cards, and a tremendous proliferation of forms of every kind. NoteCards clearly demonstrated the usefulness of integrating all of the system tools into a homogenous user interface. This is a discovery that is still only slowly being implemented in the microcomputer world.
One of the prominent features of early NoteCards was the graphic browser (see Figure 3.8) used to present overviews of the node and link structures created in the open semantic network. This was actually a simple extension of the graphic browser already available in Interlisp, but it proved to be an extraordinarily useful device for NoteCards. Before long, most of the housekeeping chores in NoteCards could be performed from this browser. Individual cards and fileboxes of cards could be created here, links could be added, typed, and removed, labels changed, but most importantly the visual structure of the network could be instantly inspected in this browser, rearranged by simply mousing a node and lifting it to the new spot, dragging all its links like stretching spiders' webs along with it.
Figure 3.8 A Notecards Browser adapted from VanLehn (1986).
The value of NoteCards and the graphic browser for simplifying and empowering scientific reasoning and theory formulation was best illustrated in an early short paper by Kurt VanLehn (1986). He was formulating a theory to describe how people learn procedures. His most important evidence came from the error data of over one thousand elementary students learning arithmetic. The argumentation of this theory, its issues, hypotheses, evidence, and conclusions can be found in 328 pages of densely reasoned chapters (VanLehn, 1983). Each of the 20 chapters discussed (more or less cleanly) a single issue. Each chapter started with a statement of the issue and its competing hypotheses, continuing with the evidence for and against it, and winnowing these down to victorious hypotheses, and an connecting system that integrated them. As VanLehn points out, the document can be seen as a traversal of the tree that outlines these structures. Figure 3.8 is an example of a tree drawn from his arguments. Beginning at the left and moving to the top and the right, and rep[eating this movement for each line, the document can be interpreted to be a left - right, depth - first traversal of this tree.
Because the arguments of VanLehn's thesis had been carefully reasoned the NoteCards database of his thesis was produced very quickly. In only ten days he created about 800 cards, roughly forty cards per chapter. The Browser in Figure 3.8 gives more than one-half of a chapter. Each node can be expanded into the text of the chapter. Leaves on the tree (for instance, "Repair the TaSK" is a leaf of the tree headed by "TASK" on the "Control?" branch) are themselves embedded as nodes in the text of the higher order nodes. The NoteCards version, according to VanLehn, offered to surprisingly clear advantages over the standard text format.
The first advantage was that the browser revealed that certain issues were not connected to each other. Since issues were usually restricted to one chapter, the chapter structure of the book denied easy linkage of issues. An analysis of these linkages revealed that there were certain higher - order issues, such as "what is the student's mental representation of the skill they are learning?" that were never expressed anywhere. Instead they had been divided into smaller issues like "what is the representation of procedural knowledge?", and "what is the representation of declarative knowledge?" that deliberately partitioned the argument structure cleanly, to avoid cross-reference. Inserting these higher order nodes into the database brought about the needed connections. They also suggested alternative structures, that basically could form another perspective on the whole book.
The second advantage with NoteCards stems from the ease of creating alternative organizations. Instead of moving huge chunks of text around, all one has to do is reorganize the icons and labels of the hotspot nodes and links. This is particularly powerful in formulating theories, where the biggest lurking danger is the inertia of the historical status quo. Instead, theories can be built from the bottom up, with issues and evidence pointing to revised hypothesis. Naturally this does not help in the fundamental creative task of generating hypotheses itself, but it may point to excluded alternatives that would otherwise be completely ignored.
Something like this happened to VanLehn when he organized the icons of his hypotheses and facts into a matrix. By filling in the matrix with "+, -, or 0" to indicate support, contradiction, or neutrality of a hypothesis by a fact., VanLehn was bale to evaluate his hypotheses in a fairly objective way, and yet in a way that would not have been convenient with his purely textual organization in the book. It was only after making the issues and arguments coherent in the browser, that he was able to lay out the matrix of hypothesis evaluations.
VanLehn acknowledges a huge debt to the power of hypertext systems to support the development of theories. His example has been extended to an even finer grain of analysis by Lawler (1987). And the role of hypertext in argument formulation is the subject of much continuing interest as we will see in the chapters on knowledge representation and retrieval (particularly the work of Fischer and McCall, 1989).
CASE (A Case Analysis Support Environment) is a tool developed by Bob Lawler, first using NoteCards and later HyperCard, to help an analyst of complex case study material keep control of and exploit the material available. The central notion is variable depth transcription--using hypertext linking capabilities to permit flexible database construction in response to the analyst's changing theories. The analyst in question in this case was to begin with Lawler himself. He had more than 240 videotapes of his child's cognitive progress over a period of four years until age 6. In the early years these tapes were made weekly, and then later they were made bi weekly or less frequently.
There is no hope of capturing the rich detail of videotape in such an abstract medium as text, and written theories can only select material through the focused lens of its own predictions. Information that might be used to denounce the theory or radically alter it, is ineviatble removed from the statements of the theory.
Hypertext promises a new era for the development of theory based on detailed analysis of important individual cases. Material can be selectively abstracted in place and all the loose ends of the theory are then exposed for continuing modification. The strength of this CASE tool is that it is public and shareable. Others can examine the important linkages connecting detailed observation and theory in a richness that has never before been possible.
The importance of hypertext annotation goes far beyond providing footnotes or glossary accessibility. According to Lawler, the ultimate value of hypertext annotation systems will be a fundamentally different way of doing cognitive science. Hypertext will lead to an increased shareability of ideas, even more, an enhancement of cooperative
knowledge and the community of scholarship. Collaborative research, based on shared hypertext databases, is now an evolving theme of research within the cognition and information sciences communities.Lawler's central idea is that machine embodied texts and video are incrementally extensible and incrementally perfectible. This promises that eventually case study will become a discipline of more readily sharable and more genetic theories. If tools useful for the study of developing minds may be equally useful for exploring the ideas of mature minds, hypertext systems have potential to support as significant an advance
in critical explication and analysis of scholarly work as statistical techniques have had in the unfolding of psychological and educational quasi-scientific method.
Instructional Design Environment (IDE)IDE was the first system to use hypertext formally to construct a knowledge acquisition, organization, and delivery system for instruction ( Psotka, Massey, and Mutter, 1988). Using NoteCards, Russell, Moran and Jordan (1988) built a complex environment that bridged the many stages of instructional design. One part of IDE provided a rationale for all decisions made, with specific links between decisions, arguments, evidence, and general principles of design. Another part of IDE helped organize actual course specifications: the knowledge structure of the domain ( skills and facts), the common student bugs, the actual units of instruction ( text, video, or simulations), and sequencing the course (strategies and tactics of instruction).
The final part of IDE actually delivered the instruction using an interpreter to traverse the knowledge plan, present the instructional units, update the student model, and decide on the instructional strategy. In order to let the interpreter work properly, the knowledge of the course had to be represented in a complete tree, without loose and unconnected nodes. This might be a difficult constraint to fulfill with some topics, but technical topics, such as arithmetic or maintenance of a device, can easily be fit into nicely ordered trees.
The key to making the interpreter work, lay in putting "code in the node", i.e. having a Lisp function or a rule fire when a node was activated by a link. This was a component that heavily influenced the "script languages" of future hypertext systems. Combining the power of knowledge structuring text environment with the power of a functional language made it possible to create virtual nodes: a node that would reconfigure itself on the basis of how it was activated. For instance, the first time a node is activated, it might provide multiple levels of information, introducing a topic and giving all its details. However, if it was activated again by the same person, it might give only the higher-order summaries and synopses of the text, providing a reminder of the important points rather than a thoroughgoing introduction.It was awesome when it worked, popping up windows of text, video, computer animation and simulations; requesting answers to test questions, and deciding where to go next. A wonderful research environment, but by and large it was just too huge and unwieldy ever to use outside of a research platform. It was years ahead of its time and the hardware. Even on the Xerox Lisp workstation, it quickly overwhelmed the technology and all too easily collapsed under the maze of choices the interpreter had to make. But as a prototype, it was inspiring!
Amazingly many parts of IDE actually were used, especially to organize the information on new product designs that came from many disparate sources, but all had to be put together to create training materials for technicians on the new product. The environment proved very useful for organizing these materials into coherent forms out of the chaos of their many sources. The collection of decisions rationalizing a particular design could later be converted into explanations for a technician to understand the functions of component parts and higher order structures.
One clear lesson came from this project. If the task was to organize large chunks of text and graphic information from many different sources, the database retrieval had to be very fast, and it must be easy to give labels and type the material along many different, well-organized and distinct attributes. Then the system needed to retrieve the text on the basis of these attributes must respond very quickly. For instance, if someone is preparing a course on hypertext, it would be enormously useful to classify each of the popular research hypertext systems on the basis of their node types, whether they were "card sharks" or "holy scrollers", whether they were research or commercial systems, whether they supported anchors as destinations inside scrolling text, and many more important features. However, unlike a standard relational database, it is important to have not just the name of the hypertext system associated with these attributes, but the whole bundle of text that describes the system, and even virtual structures of text that might be computed on the fly when a particular kind of query was made
The success of this project demonstrated beyond a reasonable doubt that hypertext systems really were well suited to authoring and delivery environments for instruction. It was a finding with far reaching consequences, as new generations of hypertext systems become widely available.
The Beginning of the RevolutionAfter Halasz came the deluge. Compared to the sparse elegance of the roughly 100 versions of Notecards in the hands of early users, commercial systems developed for the legions of desktop PCs turned hypertext into a commodity. They brought hypertext into the public's view and forever took it out of the isolation of exotic prototypes that only ran on handcrafted computers after specialized design and authoring. Compared to the handful of narrow, specialized, and expensive applications of hypertext with Notecards, the diversity and ordinariness of the manifold applications in Guide, Hyperties, ToolBook, and HyperCard began to fill every niche of computing, particularly those niches, like instruction, requiring extensive and specialized knowledge rather than computational wizardry. By 1990, hypertext had become the acknowledged substrate of human-computer interaction, the default interface, to such an extent that as important and fundamental a product as Windows 3.0 could not be finalized without a hypertext engine for online help systems, and a ToolBook hypertext delivery system. The educated Everyman finally had something to do on these fabulous silicon machines. The hypertext revolution was launched.