What constitutes a basic unit of cybermedia will be an important consideration for those who design its interface. A "basic unit" is the smallest fragment within a medium that can be taken out or isolated and still be enjoyed or understood as a complete work; that is, a fragment possessing closure.
The difference between a scene in a film and a single shot illustrates the difficulty of defining a basic unit in any medium: if the film is considered as drama, then anything less than a complete scene is insufficient to convey the essence of the form. But if film is visual art and the emphasis falls on the technology of cameras, film stock, and editing devices, then the single shot is the basic unit, and shots are the essential pieces from which the complete film is built.
So while some people may argue that an article, an essay, or a chapter from a book is the basic unit of a printed work, the characteristics of hypertext urge acceptance of a smaller increment, possibly the paragraph.
The danger in stringing diverse paragraphs (or shots) together is that you lose the transitions. In a fragmented hypertext environment, when users branch off on a new path, they must be able to see immediately why a link exists between where they came from and where they are. If they are dissatisfied with this new path, they must be able to return to the previous place quickly and easily.
A coherent scheme for links in a fragmented space could be seen in such early text-only computer games as Adventure and Zork, in which the basic unit was a room. A description of the room located the player and provided some clues for continuing in the game. The order in which a player visited rooms was usually not important, but the order was constrained to an extent because no one room led to more than a few others. Some rooms had only one entrance/exit; others had several. A complex area could be broken up into separate rooms, e.g. "front of cave" and "back of cave."
The parallel to hypertext environments is apparent. The form requires manageable chunks that can be joined by links. While existing systems link complete documents brought in from print environments, the next step is to break out the chunks that are being linked, the essential chunks, or objects, or rooms -- the basic units. The link structure is important, but not more so than the composition of the disparate fragments. Each time the user stops moving through the medium, the fragment he or she confronts must be able to stand alone, to make sense on its own, even though it offers multiple doorways to other places.
If the chunk does not give the impression that it stands alone, the users will feel as if something is missing, as if the information is inadequate, incomplete; they will have to back up and read an introduction, or trudge on and read more. To make cybermedia work on both deep and shallow levels, it must always allow users to feel free to quit where they are, to create their own ending. (At the same time, they should also feel that there is always more to explore.)
The basic unit in today's online environment is a screen or frame. Within that screen there may be windows, within which different things can exist at the same time. While a screen may be compared to a magazine page (on which various graphical elements can play diverse roles) or even to a full-size newspaper page (translated to the screen, each news story or ad might exist inside a window of its own), there is still a large difference in usability between the print environment and the screen environment. The limited size of the screen and the limited options for movement within and out of that screen make the screen environment inferior to the print environment for reading text.
In print, a big chunk of text can be scanned easily, especially when there are elements of typography and art to aid in the process. The screen environment is not well suited to the kind of scanning we use in print environments. Hypertext does not re-create a print environment on a screen; rather, it offers a wholly different way of moving through and among texts. Because of this, hypertext can vastly improve the usability of texts within a screen (or frame) structure. But because the screen is so different from print, it must be understood that hypertext requires a different handling of text.
Organizing information into an expanded article format (as in the article you are reading now) is similar to an idea expressed by Jakob Nielsen; he describes clusters of nodes arranged in mini-networks. This would mean grouping a number of closely related nodes, or chunks, together in a way that facilitates easy movement among them, with probably a more deliberate effort required for the user to exit from one cluster and go on to another.
The problem inherent in any metaphor from the physical world, like "desktops" and book formats, is that it must limit the user to some extent by constraining thought. The less that new media work like an existing physical form, the more users will be encouraged to think of them as something with new potential. But designers cannot create metaphors without any frame of reference; users will hardly understand a system that is not at all like anything they have known before. Probably the best solution will be one that combines diverse metaphors, such as those from architecture and from film, in a surprising (but not unpredictable) form that will both enable users to understand the corollaries and encourage them to think in new ways.
This is the best book I have read about hypertext. It is now out of print; however, a revised and substantially updated edition is available:
I was fortunate enough to see Nielsen speak at a conference, and afterward I asked him whether it would be worth my while to buy the new edition, since I had read the first version thoroughly. "Well, that depends what your interest is," he said. "About half of it is new, because things have changed so much." I did buy the new edition and have not been disappointed (although I haven't had time to finish it yet).