Abstract

In this investigation of the educational potential of CD-ROM as a hypermedia environment for education, technological constrains and design challenges are cited suggesting that the success of CD-ROM as an alternative to paper-based media is not a foregone conclusion. It is shown that advances are needed in storage capacity, bandwidth, retrieval time, portability, and input/ouput technology. Challenges in the design of hypertext systems are identified, such as overcoming many inherent constraints of electronic media, improving modes of information access, filtering unwanted information, allowing users to act as designers, and achieving system compability standardization. (Keywords: CD-ROM, optical storage. hypertext.)

As libraries swell with books and other paper-based documents, there is a quiet revolution occurring within. In large facilities, computer-based systems have already supplanted microfiche and card catalogues as the vehicle for access; but now, the penetration of technology into the culture of learning has a new tack. Recent storage technologies such as the CD-ROM optical disc and new modes of access such as hypertext promise also to deliver the content.

The optical disc delivers quantities of information on scale tens or hundreds of times in excess of magnetic disk technology. (In this article, the spelling "disc" is used for optical and consumer products and "disk" is used for magnetic products). Thanks to the consumer market, small. inexpensive, mass-produced optical disc products (audio CDs) are now available. Of dramatic potential is the infusion of this technology in education. CD-ROMs have the potential to bring information--that is, facts, resources, teaching aids, and so forth--closer to learners than ever before. Access to oceans of information, however, can easily overwhelm learners. Hypertext interfaces address this through personalized browsing and connecting (linking). Information is stored, and hierarchy not present at the physical level (as they are in the book), but added later by designers or users. The content may go beyond and include sound and still or motion images, perhaps from different equipment sources. Although the term "hypermedia" is often used, the term "hypertext" usually implies the possible use of sound and graphics, and is preferred here. See Conklin (1987) and Nielsen (1990a, 1990b) for reviews of hypertext systems.

The view presented in this paper is optimistic yet pragmatic. Beating the book is a formidable challenge and one that surely will not precede the twenty-first century. Achieving the educational potential of CD-ROM and hypertext may be contingent upon meeting several difficult technological and design challenges. We shall meet these following a brief look at the role played by the consumer and education markets.

The Consumer Market

Optical disc technology, invented by the Dutch firm NV Philips, was demonstrated in 1972 and released as a product in 1976. Several formats appeared initially, but only LaserVision has survived. A 12-inch LaserVision disc can deliver 30 minutes of video at 30 frames second for a total of 54,000 frames. More importantly, frames can be frozen and accessed at random. The first products were feature-length movies re-released in the new format and boasting unprecedented picture quality. Access to still images fostered multimedia applications for education and industrial and military training.

Consumer response to the optical video disc was lukewarm. The initial investment of $1,600 for the player combined with industry bickering over a standard format left everyone with a "wait-and-see" attitude. Laub (1986) described the events as follows: "Many bitter battles were fought between 1972 and 1976 over the choice of a standard optical videodisc system. Laser Vision won out by endurance, rather than friendly agreement, leaving much discouragement and the memory of wasted time and money in the minds of manufacturers and users" (p. 53).

Another major factor contributing to the limited acceptance of optical video disc technology was the birth of magnetic Beta and VHS video tapes. The lure of recording one's own movies proved irresistible as these writable and less expensive formats contributed to the demise of the laser movie.

Not to be duped a second time, optical technology re-emerged in 1983 as the compact audio disc (CD) following collaboration by Philips and Sony. The CD was (and remains) an unqualified success; eager consumers snapped up more than 65 million discs during the first two years of sales. High-volume sales meant companies could amortize initial costs. Eventually, mastering costs dropped because more companies entered the business and equipment and expertise proliferated.

The audio CD contains more than 550 megabytes of storage; however, this only translates into 74 minutes music. (That's more than 100 kilobytes per second!) Indeed, consumers demanded the best in fidelity, and this they got. With 16 bits per sample and a 44.1 kHz sampling rate, audio CD has a 96 decibel signal-to-noise ratio, far higher than traditional phonographs or tape players. CD-ROM is the same technology as audio CD. Only the data and file structure differ.

Hypertext, as yet, has no penetration as a consumer product. Apple Computer Inc's HyperCard Trademark, distributed free with every Macintosh computer, is the most popularized hypertext product (albeit numerous other exist; see Conklin, 1987, and Nielsen, 1990b). Although HyperCard is a design environment rather than an application, it incorporates numerous examples of stacks, demonstrating the potential of the medium through easy linking and browsing. Other products have appeared that were developed using HyperCard, and these products make use of Hypercard's simple and consistent user interface.

The Education Market

Since development costs are generally too high for entrepreneur developers, current CD-ROM products are targeted at a broad market only indirect links to education. School boards, perhaps innocent beneficiaries in the end, are not the focus of motivation (or inspiration) in the development of these products, at least at the present time. One survey of CD-ROMs cited products for cataloging, acquisitions, reference, indexing, business, medicine, law, government, science, geograpy, and, finally, education (Desmarias, 1989a, 1989b). While only a few of those products surveyed are targeted specially for the classroom, many have obvious potential as reference sources for students (Tanner & Bane, 1989).

Developing educational products is risky business. School boards are viewed as stoic fortresses, and when the infusion of new technology is involved the issue is particularly contentious. As Helsel (1989) pointed out, classroom tradition is strong; and change only occurs when new technologies adapt to existing methodology, rather than the reverse. Teachers feel resentment when outsiders--bureaucrats or technocrats--tell them that something is inevitable or better. Teachers' practice of lecturing and questioning with the students listening, answering, and reading assigned materials is much the same today as ever.

Furthermore, business perceives education as a finally depressed market. Schools, like farms, always appear to be on the brink of disaster with purse strings drawn tight by governments. Although a potentially huge market, education is fragmented, politically sensitive, and extremely hard to penetrate. Large corporations, despite their outward benevolence in the form of fellowships and equipment donations or discounts, are passive. When a product takes off, corporations want to be perceived as prime developer; but in terms of real investment, they remain distant and stingy. The market is too sensitive, too volatile, and too divided.

CD-ROM penetration is already high in libraries. Since the discs are so inexpensive to manufacture after initial product development, they are considered the ideal distribution medium. Sources such as Dialog Information Services (e.g., ERIC) or Silver Platter Information Services (e.g., PSYCH LIT), offer their subscribers quarterly updates on CD-ROM. It remains to be seen whether the same level of service will one day extend to the classroom.

Although access to vast stores of information at the click of a button is potentially valuable, there are serious limitations of CD-ROM (and related technologies) that must be reconciled before its acceptance in the workplace or school as a serious alternative to print-based media. Thanks to the consumer market, the cost constraint has been substantially met. In this section technological constraints are identified. Design challenges, which are more relevant to hypertext, are presented later. Our standard for comparison is the book.

Display Technology and the Human Interface

A distant from CD-ROM per se, but paramount to the viability of electronic documents, is the quality of the output display. High-quality cathode-ray tubes (CRTs) present an image as an X-Y grid of picture elements, or pixels. The EGA standard found on IBM PC/AT systems, for example, has a resolution of 640 by 350 pixels. Comparing this with the typical resolution of 1,200 dots per inch for a typeset page, the quality of the image on an EGA monitor corresponds to a one-half-inch by one-quarter-inch rectangle. If low-end laser printers are the standard for comparison, the EGA format corresponds to a two-inch by one-inch rectangle.

Current state-of-the-art CRTs have resolutions in the range of 2,000 by 2,000 pixels, with a modest palette of colours; however, the tubes are large and expensive. Liquid-crystal displays (LCDs), notable for their economy of scale, are also expensive and are generally monochrome with about half the resolution of CRT displays. Despite impressive advances, the quality, size and expense of output displays must improve before users embrace new technologies such as CD-ROM as a preferred alternative to print-based documents.

Although our prejudices for computer technologies parallel our strong visual bias, we should acknowledge the presence abd bidirectionality of the audio and kinesthetic channels. Computers use audio output at a primitive level (e.g., beeps) but rarely any form of audio input (e.g., speech). The primary input channel is kinesthetic, or haptic (keyboards, mice, etc.), and isaccompanied by parallel ouput via tactile feedback. Improving the visual, audio, and haptic channels is the focus of considerable research in the relatively new multidisciplinary field of human-computer interaction. By comparison, interaction with books is relatively primitive. Books are, for the most part, a medium of graphic output only.

Storage Capacity

What would be the implications if a computer's graphic output matched the quality of a typeset page? Would the limitations vanish? Perhaps, but at a price. The 100,000 page capacity of CD-ROM represents raw text with each character encoded in one byte. Even ignoring the graphic images often found in books, magazines, or encyclopedias, the text that remains is difficult to represent with traditional character-oriented coding systems (such as ASCII). Boldface, underlining, italics, pitch, point, fonts, margins, columns, proportional spacing,running heads, page numbers--these parametric traits instill graphic cues carrying a lot of information. They are needed at the ouput presentation stage, yet they are troublesome to encode for storage on magnetic or optical media. Add graphics, no matter how subtle or simplistic, and the magnitude of the problem explodes. At issue is not the technical problem of encoding (for the diversity of images can be dealt with a number of ways), but that the 100,000-page storage capacity shrinks rapidly as more elaborate representations emerge. In the extreme, a page richly endowed with the above attributes is stored as a bit-image. What is the cost? At typesetting resolutions and without data compression, an 8.5- by 11-inch bit-image gobbles up 17 megabytes of storage. This translates into 32 pages per CD-ROM! The question arises: How much data is enough? Computer-based learning environments must encompass not only core information but also possibilities for exploration. McClintock (1988) put it this way:

           Two megabytes is the typical verbal content for a
           not-too-demanding survey text and the use of
           illustrations can quickly make that data content
           expand by a factor of one hundred or more. Even if
           the computer-based courseware had the minimum two
           megabytes of information for the root of required
           reading, what would then happen with the branching
           of the study tree? As things stand, the branching
           would be either extremely shallow and narrow or it
           would cease to be computer-based. (p. 198)

Futher to this point, Vannevar Bush (1945/1986), in his visionary essay "As We May Think," offered an example of a research topic suitable for investigating in a hypermedia environment: "Why was the short Turkish bow superior to the English long bow?" In Bush's scenario, a researcher examines on-line history books and encyclopaedias; but also, acting on whim and intuition, delves into a physics book on elasticity. Elasticity is quite a distant branch on the original study tree! A sufficient repository for such whimsical yet essential explorations would be truly staggering in size. It seems our thirst for information quickly reduces 550 megabytes to a pittance. A report on one commercial installation of optical storage technology notes that the system is a jukebox containing 110 platters (Fisher, 1989). The question "How much is enough?" remains.

Storage Format and Content Inquiry

What about the 54,000 images stored on a Laser Vision disc? Indeed, the quality of the image is first-rate, but the storage format is analog. An intricate graphics-plus-text image is aptly captured and stored, but the content cannot be queried without an explicit and external index. The problem is not so much that the storage format is analog but that the image is captured rather than a representation of the content. This constraint may diminish as scanning and pattern recognition techniques mature (e.g., see Robinson, 1989) but current state-of-the-art technology, at any price, prohibits extraction of nontrivial information from images, whether the storage format is analog or digital. As yet, a computer does not exist that can distinguish the image of a cat from that of the dog.

The need for databases of nontextual or nonnumerical information has spawned a subbranch of research in data engineering. Image database management systems (IDBMS), based on optical storage technology, are beginning to appear in corporations with large and centralized document centres. At present, however, the solution is brute force: documents are captured and stored but retrieval proceeds through primitive indexes. Query-by-content--the forte of text and numerical DBMS--is not possible. A typical client-record IDBMS holds, for example, header information such as name, address, client number, and, more importantly, hand-written reports, diagrams, or photographs. Although the entire document may be captured and stored, retrieval is based on key fields entered manually during conversion.

The content remains a mystery until the document is retrieved and viewed. Also, the conversion is a Herculean, round-the-clock effort. One corporation spent a full year converting 1.5 million documents to optical storage (Fisher, 1989). At another installation, a conversion in-progress will eventually see 10 million documents migrate from paper to image-based technology (Rowan, 1990).

The challenges for IDBMS are formidable. The relational query-by-content paradigm was transformed electronic data processing; but can this evolve beyond the management of data fields to a general class of objects including images? Grosky and Mehrotra (1989) pointed out five levels at which the problem must be approached, and the difficulties grow with each. At the lowest level, Iconic Data contain the images themselves, either in analog or digital format. The problems here lie mainly in choosing standards for symbol encoding and data compression. Next, Image-Related Data describe the storage format of the image with low-level attributes such as registration (alignment), resolution, colours, grey scales, grids, and so forth. Extracted Information, the third level, is interpretative but still low-level. Images are processed through a model-based filter to identify the underlying structure. Topological features, physical components, and simple relationships (e.g., "A" overlaps "B") begin to emerge at this level.

The fourth and fifth levels contain Image-World Relationships and World-Related Data. At this point the problem is highly interpretive and extremely difficult. Progress is slow. The goal is to map low-level information into real-world entities, to establish relationships between these entities, and to provide a textual description, abstracted at the level of the application. We will not dwell on the particular problems facing image database management systems except by noting that research is now able to focus on many of the enabling mechanisms that may one day result in image-based systems with queried-by-content interfaces.

Bandwidth and Retrieval Time

Another task dimension also illustrates that CD-ROM capacity is not so vast. Although 550 megabytes equates to more than 100,000 pages of unembellished text, it represents only 74 minutes of high-quality audio. There's nothing overwhelming about that. In practical terms, high-quality motion video is untenable on CD-ROM because it requires considerably more than audio. The problem lies not only with the quantity of data but also with the rate at which it is retrieved--the bandwidth. CD-ROM has a fixed bandwidth of 175 kilobytes per second. At a rate of 88 kilobytes per second for CD audio, no visual information can be retrieved coincident with audio information (Bruno, 1987). But by reducing the quality of the audio or video signal (mostly like both), a combination of motion and sound is possible, equivalent in quality to an animated cartoon. However, this still consumes CD-ROM data at an astonishing rate.

Access time--the time between asking for data and getting it--is determined for CD-ROMs by mechanical delays in moving the optics to the track (actually, spiral) containing the data. This takes one second on the average. If retrieval requires one access to an index and another to data, this can grow to two seconds. Magnetic disk drives, with access time in the range of 20-40 milli-seconds, are at least 25 times faster than CD-ROM disc drives. After the novelty wears thin, users of CD-ROM applications may tire of the delay. The bandwidth and access time of CD-ROM will improve only slightly because they result from mechanical limitations in the drives.

At present, most CD-ROMs are installed on a dedicated system with one user. This needn't be the case. One promise of the Information Age is global connectivity, offering rapid access to large information stores at remote sites (see Wright, 1990). It has even been suggested that the library of the future may be a computer network service rather than a building (Nielsen, 1990b, p. 74).

New problems loom when the information retrieval goes beyond simple text. Video transmissions are bandwidth-intensive, and when distances exceed arms reach technological constraints emerge. For example, the 10 megabits/second bandwidth typical of local area networks (LANs) can support dozens of users within several kilometers. Although typical LANs perform admirably for text of software traffic, they grind to a halt with two or three video channels. Longhaul video transmission using satellite, fibre, or microwave links is out of the question except for big corporations with big telecommunications budgets.

Portability and Convenience

The personal computer was quickly followed by the portable and laptop computers. Down-sizing has now produced the notebook and palmtop formats. Although the portability and convenience of books cannot be topped, recent technological advances, such as the CD Walkman, foretell of the portable CD-ROM book. Recently, Sony introduced the Discman containing a 3-inch optical disc storing 100,000 pages of text. It seems the Japanese are taking seriously the potential of the electronic book. An early vision of this came from the Learning Research Group of Xerox Palo Alto Research Center. "Dynabook" was portrayed as a dynamic self-contained knowledge manipulator in a portable package the size of a notebook (Kay & Goldberg, 1977). Although portability was never realized, tales of the dynabook have inspired researchers ever since.

Numerous book-like interfaces have been demonstrated (e.g., Benest, 1990; Bolger, 1989; Egan, Remde, Landauer, Lochbaum, & Gomez, 1989), but none has seriously addressed portability. Many design issues (such as browsing and the placement of annotations or bookmarks) are admirably addressed, but the products reside on microcomputers. In a field study of hypertext user, 33% complained that the hardware was not as convenient as paper (Nielsen & LyngbEk, 1990). The ability to read a book at home, in a park, or on a train is a compelling feature--one that presents a formidable challenge to computer and optical technology. There is also substantial evidence that people read about 30% slower from a computer screen than from printed text (e.g., Gould & Grishkowsky, 1984). This is probably related to factors such as resolution and refresh rate, so technological advances may obviate this element/

Despite numerous problems with portability, it is evident that researchers, in anticipation of new technology, are meeting many design challenges. In the next section, we will meet some of these.

The above limitations are purely technical and they can be brushed off, perhaps, by waiting for advances in technology. However, there are other dimensions of the task lacking. These are presented as challenges to developers because their unravelling will come through inspired innovation and cooperative design efforts as opposed to breakthroughs in semiconductor physics, materials science, or algorithms for data compression or pattern recognition.

The Nature of the Medium

A whole genre of problems lies in the differences between the look and feel of an electronic document and the look and feel of a paper document. Storage capabilities, access times, and encoding methods are technological problems; closure proximity cues, linking, and presentation are design and implementation problems. Inherent in print media is the feeling of what is near or far, what is important or unimportant, and the sense of completion. Pages are flipped from beginning to end on impulse, just to get the feel of the document. The content is scanned in an instant. Oren (1988) calls this "serendipity"--the ability to browse and explore a document quickly using a single keystroke or mouse click. The time costs must be low. A delay of several seconds will suppress the whimsical browsing of side issues.

Closure is the feeling acquired when finishing a book, a chapter in a book, or any level of the presentation. In hypertext environment, perhaps based on CD-ROM, it is a challenge to impart the same feeling. The system should provide simple statistics, perhaps graphically, or what has been covered and what remains. Is a particular item in a subsection or sub-subsection? What is its hierarchical status in the overall organization of the document? How easy is it to review an item examined earlier? Disorientation is a major and oft-cited problem when browsing in large databases (e.g., Conklin 1987; Heller, 1990). In one survey of hypertext users, 56% admitted often feeling confused about "where I was" (Nielsen, 1990a). Representing topical proximity is a challenge that must be met.

Another detail is the efficiency of presentation. Designers of paper-based products take liberties, exercise generosity, and generally waste much of the information potential of a page. This is necessary though. Consistent and spacious layouts elicit order and structure and facilitate the assimilation of content. The information theorist's term for this is "redundancy"--a measure of the inherent order in a system. For example, the English language is about 50% redundant, exemplified by the observation that a message with half the words removed is still largely intelligible (Weaver, 1949). Thus, we need not read every word in a message: we ingest words in phrases or chunks and quickly proceed. If random sequences of words were as probable as highly structured word sequences, there would be zero redundancy. Such a language would quickly fatigue readers because each word would need to be considered in isolation from surrounding words. A message of random words would be indistinguishable from an intended message. Redundancy is desirable and arguably vital. In the language of information theorists, redundancy combats noise.

Redundancy exists in a graphic sense as well. Presentations with high information utilization (that is, low redundancy) offer too much--the delivery is too busy. Often a design goal is to exploit the technical possibilities of the medium. As a result, the layout becomes dense, the subtlety is lost, the message obscured. For example, novice publishers, armed with sophisticated word processors, page layout software, and laser printers, are often guilty of producing output that is fancy and cluttered--all the possibilities are explored and used. This is also common for presentations on a CRT display. With too low a resolution to waste on spacious layout showing structure and order, information is often packed in at the expense of aesthetic quality.

Herein lies a major challenge for designers: to assemble and manipulate the presentation addressing structure and order rather than pushing the information potential of the technology. With the added dimensions of audio and motion video (albeit, primitive at present), CD-ROM has the potential to surpass paper documents, but the presentation must be addressed with care.

Filters and Amplifiers

The presence of vast stores of information on high-performance machines does not guarantee the demise of the status quo. Of paramount importance is the ability to amplify or filter unembellished information. Hypertext systems promise to amplify through their capability to order and reorder items even though the underlying structure is nonlinear. Books, with hard-wired structure, offer little here.

Several levels of sophistication can be identified (Byers, 1987). The lowest is basic text management, which is similar to querying systems in traditional databases. The keyword, Boolean search strategy used in the ERIC CD-ROM application is an instance of this (MacKenzie, 1989).

The next level--the starting point for hypertext systems--provides static access to textual or graphic information through links built in to the database by the developers. Possibilities are plentiful, but they are limited and cannot be implemented of the user or systems integrator. Most hypertext systems retain a history of the user's path and allow back-tracking or review of past activities. The history is volatile, however, and is not retained for subsequent explorations. In a study comparing equivalent hypertext and paper-based documents, Egan et a. (1989) found empirical evidence that searching in hypertext is quicker and more accurate. Users have also shown a subjective preference toward hypertext documents (Marchionini & Shneiderman, 1988).

Dynamic linking, the most sophisticated method of access, allows the user to add links. Natural and obvious links exists as cited above, but users can forge new ones while browsing through the database investigating a topic. The links are stored for subsequent investigations on the same topic.

In a truly dynamic hypertext system, a user can add entire documents to the system. This requires a document-capture facility and a writable and portable storage medium, such as write-once read-many (WORM) optical discs (Rash, 1988).

In the long run, the ability to filter or exclude information is as important as having access. Global connectivity and massive storage systems promise to drown us in information. What secretaries and consultants presently do for executive's daily planning and decision making (i.e., screening, summarizing). must be paralleled in the context of optical discs and on-line access. Information filtering need not preoccupy a human being, though. "Knowbots" are software daemons that screen incoming information on a personalized basis (Wright, 1990). Content inquiry could benefit from user profiles. An inquiry into the physics of elasticity, for example, should return different information for a primary school student than for older students. Profile-guided searches could encompass numerous other student traits, such as histories of previous courses taken or personal interest.

The User as Designer

The above discussion has employed the word "user" in a broad sense. Bush (1945) called them "trail blazers":

          There is a new profession of trail blazers, those who
          find delight in the task of establishing useful trails
          through the enormous mass of the common record. The
          inheritance from the master becomes, not only his
          addictions to the world's record, but for his disciples
          the entire scaffolding by which they were erected.
          (p. 108)

For this discussion, users or trail blazers are courseware developers, educational consultants, teachers, or students. The success of dynamic hypertext systems depends on the extent to which designers can demystify the database and put customizing in the hands of end users. The further down the hierarchy--developers, consultants, teachers, students--this power can be passed, the better. A dynamic hypertext system that a teacher could easily tailor for a unique classroom setting would be a powerful tool. If a student can customize it, so much the better. The challenge for designers, then, is to provide a simple interface with powerful tools that allows a teacher (or student) to design. Links and documents could be added by a teacher and offered to students for their own exploration of a subject. Or, a teacher could assign to students a project to create their own concept of a subject by linking and adding as they see fit.

The user interface offered by Apple's HyperCard, for example, is demonstrably easy to use (MacKenzie, 1989), even for children (Nicol, in press). However, when the underlying database is extremely large the potential for disorientation and cognitive overload explodes (Heller, 1990). User-initiated design can spell chaos unless some major issues in the design of hypertext systems can be addressed.

Standards and Compatibility

A recent review of five WORM optical disc drives from three manufacturers pointed out that the discs for each manufacturer's drive are not interchangeable (Rash, 1988). This is an ominous situation. Indeed, lack of standardization and incompatibility are chronic problems of computer and communications systems and the situation is not likely to change. The term "standard" is sometimes taken as an industry joke: "There's no shortage of standards: Everybody has one" (Rowan, 1989, p. B4).

With the recent drive toward global connectivity, the need for compatibility between and within generations of software and hardware has never been more important. However, standards in themselves are not the solution. Often there is a subtle push and shove as industry moguls promise standardization as a selling feature while delivering uniqueness in order to gain control and limit third-party access.

At issue is cooperation, and its role in the political and economic climate within and between nations. Japan, with one-tenth the number of lawyers per capita as Canada or the United States, is an industry leader in cooperation (at least within its borders). From computer architecture to data communications, there is more standardization and connectivity in Japan than in any other country. Large corporations in the United States sometimes collaborate (e.g., Xerox, Intel, and Digital jointly developed the Ethernet LAN standard), but a greater cooperative effort may be needed before the Information Age swells into homes and classrooms.

CD-ROMs have recently arrived in libraries at many universities and colleges. There is tremendous potential for students to benefit through access to such stores of knowledge, particularly if audio and (motion) graphic information are added. Although installations may be limited to libraries initially, more direct access is forthcoming in the classroom and at home using inexpensive, dedicated systems or through network access to large databases.

The paper-based document has proven its strength for many centuries. Now it is up to designers to meet the demands of critical users before new technologies will be adopted for common use. Advances in optical storage media, such as CD-ROM, suggest that this challenge can be met in the near future it accompanied with parallel advances in the design of hypertext systems.

Benest, I. D. (1990). Computer-assisted learning using dynamic electronic books. Computers and Education, 15, 195-203.

Bolger, J. (1989). CD-ROM delivery: Breaking new ground. Optical Information Systems, 9, 214-216.

Bruno, R. (1987). Making compact disks interactive. IEEE Spectrum, 42(11), 40-45.

Bush, V. (1986). As we may think. (Reprinted from The Atlantic Monthly, 1945, July, pp. 101-108.) In S. Lambert S. Ropiequet (Eds.), CD-ROM: The new papyrus (pp. 3-20). Redmond, WA: Microsoft Press.

Byers, T. J. (1987,April). Built by association. PC World, pp. 244-251.

Conklin, J. (1987). Hypertext: An introduction and survey. Computer, 20(9), 17-41.

Desmarais, N. (1989a). CD-ROMs proliferate. Part 1: Library/reference discs. Optical Information Systems, 9, 23-29.

Desmarais, N. (1989b). CD-ROMs proliferate. Part 2: Business/science/government CD-ROM discs. Optical Information Systems, 9, 98-105.

Egan, D. E., Remde, J. R., Landauer, T. K., Lochbaum, C. C., & Gomez, L. M. (1989). Behavioral evaluation and analysis of a hypertext browser. In Proceedings of the CHI '89 Conference on Human Factors in Computing Systems (pp. 205-210). New York: ACM.

Fisher, M. J. (1989,April 15). Digging out with image technology. Datamation, pp. 18-27.

Gould, J. D., 7 Grischkowsky, N. (1984). Doing the same work with hard copy and with cathode ray tube (CRT) computer terminals. Human Factors, 26, 323-337.

Grosky, W. I., & Mehrotra, R. (Eds.). (1989). Image database management [Special Issue]. Computer, 22(12), 7-8.

Heller, R. S. (1990). The role of hypermedia in education: A look at the research issues. Journal of Research of Computing in Education, 22, 431-441.

Helsel, S. K. (1989). CD-ROM in general education. Optical Information Systems, 9, 30-37.

Kay, A., & Goldberg, A. (1977,March). Personal dynamic media. Computer, pp. 31-41.

Laub, L. (1986). What is CD-ROM? In S. Lambert & Ropiequet (Eds.), CD-ROM: The new papyrus (pp. 47-71). Redmond, WA: Microsoft Press.

MacKenzie, I. S. (1989). A review of four CD-ROM databases. ECOO Output, 10(2), 14-19.

Marchionini, G., & Shneiderman, B. (1988). Findings facts vs. browsing in hypertext systems. Computer, 21(1), 70-88.

McClintock, R. (1988). Into the starting gate: On computing and the curriculum. Teachers College Record, 88(2), 191-215.

Nicol, A. (in press). Children using HyperCard. In S. Ambron & K. Hooper (Eds.) HyperCard and Education: Education Advisory Council Journal. Cupertino, CA: Apple Computer, Inc.

Nielsen, J. (1990a). The art of navigating through hypertext. Communications of the ACM, 33, 296-310.

Nielsen, J. (1990b). Hypertext and hypermedia. New York: Academic Press.

Nielsen, J., & LyngbEk, U. (1990). Two field studies of hypermedia usability. In R. McAleese & C. Green (Eds.), Hypertext: State of the art. New York: Ablex.

Oren, T. (1988,December). The CD-ROM connection. Byte, pp. 315-320.

Rash, Jr., W. (1988,February). A quintet of WORMs. Byte, pp. 146-151.

Robinson, P. (1989,May). Easy reading. Byte, pp. 203-208.

Rowan, G. (1989,April 17). Personal computers promise dazzling changes. The Globe and Mail, pp. B1, B4.

Rowan, G. (1990,October 23). Paperless office still mirage of industry gurus. The Globe And Mail, pp. C1-C2.

Tanner, D. F., & Bane, R. K. (1989). CD-ROM: A new technology with promises for education. In A. Lathrop (Ed.), Online and CD-ROM databases in school libraries (pp. 234-240). Englewood, CO: Libraries Unlimited.

Weaver, W. (1949). The mathematics of communication. Scientific American, 181(1), 11-15.

Wright, K. (1990). The road to the global village. Scientific American, 262(3), 83-94.

This paper is based upon research conducted at the Ontario Institute for Studies in Education at the University of Toronto.

~~~~~~~~

By I. Scott Mackenzie Seneca College of Applied Arts and Technology

Contributors; Scott MacKenzie is a professor of computer engineering technology at Seneca College of Applied Arts and Technology. His research interests include the role of technology in society, human-computer interfaces, and performance modeling in human-computer dialogues. (Address: Seneca College of Applied Arts and Technology, 1750 Finch Ave. East, North York, Ontario, Canada M2J 2X5. Email:mac@seneca.bitnet.)