ITH - Centrum för studier av IT ur ett humanvetenskapligt perspektiv
vid Högskolan i Borås

2-3/2001

Forming the Text, Performing the Work
- Aspects of Media, Navigation, and Linking
by Anna Gunder

Contents

Part 1
1. The Concepts of Work and Text
2. Storage Medium, Presentation Medium and Text Access
3. Copy, Impression and Edition

Part 2  
4. Structures and Navigation
5. Hypertext
6. Ergodicity
7. Links and Linking
8. Closing Words
About the Author

References

4. Structures and Navigation

Texts of mainly temporal works are structured in one or several units of textual elements. In his discussion on hyperfiction, Johan Svedjedal introduces the term content space indicating the "windows" containing textual elements that, in combination with links connecting them, create digital hyperfictions like Patchwork Girl and others.¹⁰¹ The notion of content space is useful not only in descriptions of digital hyperfictions, but also, if understood as a more or less distinct assemblage of textual elements, in discussions on textual navigation in general.

It is important to stress that content spaces are not absolute categories. Different types of content spaces can be distinguished depending on the person who defines and delimits the content spaces in question. Thus when defined by the originator, a content space is characterized as an authorial content space. And, when defined by an editor, a publisher or by any other person involved in the publishing process, a content space is an editorial content space.¹⁰² Content spaces may also be analytical categories where the analyzer chooses a certain perspective or makes a certain selection – from one perspective, an anthology may be considered as one content space (the whole work); from another, it may be considered to consist of several content spaces (the individual essays).

An academic essay with footnotes, table of contents and so forth, consists of several authorial content spaces. Published in its entirety in print there are no editorial content spaces since no further divisions have been made. However, if the essay, for one reason or other, is published in three sections on three individual web pages in a web magazine, it is necessary to differentiate between authorial and editorial content spaces. The authorial content spaces are the same as for the print essay, but in addition, the essay has been divided into three editorial content spaces. Furthermore, in this example, the article as well as the whole issue of the magazine (on the Web or in print) could be described as analytical content spaces.¹⁰³

Like a work, a content space may consist of basically any kind of text (audible, pictorial, cinematic, typographic, etc.) in any quantity – a single letter or hundreds of thousands of typographic signs, a simple drawing or a graphically advanced scene in a computer game. Of course, content spaces may hold exclusively one sort of textual elements or several. Containing video clips, pictures, typographic signs, and sound, many web pages are examples of what could be labeled multimedia content spaces while most content spaces of a critical edition of a literary work would be described as typographic content spaces. For this reason, content spaces may be mainly spatial or mainly temporal.

A content space is a delimited unit of textual elements. This means that content spaces "end"; they are provided with borders. Figuratively speaking, moats surrounding content spaces may be easy, or more difficult to cross. Consequently, one knows that one leaves one content space for another when taking a longer or shorter stride over a moat – often, one even has to jump over it. In a dissertation, for example, the main text as well as each single footnote is to be considered as one authorial content space. Clearly, these authorial content spaces have clear-cut endings; having read the footnote reference, the reader must traverse the borders of the content space in order to continue experiencing the work. Similarly, having read the authorial content space [strawberries] in Michael Joyce’s digital hypernovel afternoon, the reader has to "jump" to another by clicking on words or pressing keys.

Authorial content spaces in computer games are of various kinds. What is called a "level" or a "world" corresponds, in many cases, to an authorial content space, where certain tasks must be accomplished before the moat can be traversed, i.e. before one can pass to the next level. In some games, the game mainly takes place within one single authorial content space, but there may be several to choose between. One example is the so-called first-person shooter game (of which the best-known is of course Doom and the various clones deriving from it). In these games, the user is running around killing monsters and other opponents. Strategy games (like Command and Conquer) function similarly in that the user chooses a "map," a playing field (i.e. a specific authorial content space).¹⁰⁴ In other games, an authorial content space is a limited area in the fictional world that the user may not leave until she has found the magic crystal, killed the monster and so forth.¹⁰⁵ In the computer game Robinson (designed for children), the visit to the shipwreck is an illustrative example: the user has to find and collect five items on the ship before she may go back to the shore.

On the World Wide Web each individual web page constitutes a content space. Web sites may be constructed differently, and one major difference is that between sites that do not use frames and those that do. Briefly, in the first case, only one page at a time may be displayed on the screen, whereas in the second case, the frames make it possible to divide the screen and thus to show different pages simultaneously. The Swedish web story Det var en gång ett litet land högt uppe i bergen (Once upon a time there was a small country up in the mountains), described by the authors as an "interactive fairy tale," is an example of a web site without frames.¹⁰⁶ The fairy tale is a story based on alternatives; at the end of each page, the reader is invited to choose between several sequels. So when the authorial content space ends, the reader is expected to click on one of the alternatives to go on reading. In doing so, an entirely new page (a new content space) covering the whole screen is displayed. An example of a web site constructed with frames is the home page of Studies in Bibliography.¹⁰⁷ The screen is divided into two fields, or frames: the narrowest field to the left remains more or less unchanged, whereas the larger field displays whichever of the site’s pages that is requested by the user. In other words, by means of frames, two or more content spaces may be displayed simultaneously. Naturally, these content spaces (the web pages) may, in turn, consist of several content spaces (for example an academic essay).

Texts consisting of several content spaces often, but far from always, contain links, i.e. explicit connections between places in the text. A link is a pointer leading the user from one textual spot to another. Links run between content spaces, but they may also run within them (for example, with an indication of the following type: "see p. 6 below."). A link can be digital (click-through) as well as analog, like the typographic link (that in print, for example, run from footnote numbers, page numbers in indices etc.) and the link from "To be continued . . ." often closing episodes of serials on television or on the radio. Of course, several links may lead from the same content space: for example, from the introductory page of the fairy tale presented above, the user may follow twelve links. Often, in print texts, the footnote links, as one could call the links leading the reader from the footnote numbers in the main text to the actual footnotes, are easily counted since their number in general coincide with the number of footnotes. (For a discussion of different types of links and link structures, see below.)

Texts of mainly temporal works may be structured in different ways. Basically, a distinction can be made between two kinds of texts depending on how the user is expected to navigate. On the one hand, there are texts in which the user is constantly faced with textual crossroads of various complexity at which she is consequently obliged to make a choice; on the other hand, there are texts offering no forking paths or side roads whatsoever.¹⁰⁸ In a printed academic dissertation, for example, the reader must make a choice at every footnote mark. She is supposed to determine whether to follow the link – generally leading to a position at the bottom of the page or at the end of the text – or ignore it. Similarly, a web page reader must decide whether to follow a link by clicking on for example a linked word or picture, or neglect it. The structural composition of a traditional novel does not contain such choices since the textual elements constituting the text are fixed in a certain order. Reading a traditional print narrative, the reader knows that she is supposed to start on the first page and read line by line, page by page until the last sentence of the last chapter. Crossing the novel from one cover to the other along the typographic highway, there is not a turning in sight.

It is important to stress that the present discussion concerns the structure of texts, i.e. how the textual elements are ordered and arranged to form content spaces (by the author/originator, editor, or any other person involved in the process). Moreover, the discussion focuses on how the user is supposed to navigate these structures. Navigation, then, has to do with how the user makes her way through texts, or more precisely, how she is expected to do it.¹⁰⁹ The reader of for instance Madame Bovary in book form is expected to read the chapters in order, but the storage and presentation media do not provide the author with any means to prevent the reader from reading the last chapter first.¹¹⁰

So the issue is not what could be called the artistic structure of a text, nor is it a question of interpretation.¹¹¹ In fact, the textual structure as such is independent of how and if the users understand the text and if they notice and recognize an elaborated artistic structure, like a recurrence of certain metaphors creating an intricate net of allusions and interconnections in a work. Accordingly, when discussing Madame Bovary in terms of choices, the issue is not what the users may choose to associate a "porte-cigares en soie verte" (a green silk cigar-case) with. Instead, the vital issue is whether it is possible to click on the words "porte-cigares en soie verte" and to define the structural implications of that possibility or, absence of such possibility.¹¹²

Gustave Flaubert's novel is an example in which all the textual elements constituting the text of the work are within the frame of a single authorial content space. Consequently, there is no jumping between authorial content spaces, but the user follows the predetermined path of ordered and fixed textual elements. Thus all the users, in this case the readers of Madame Bovary, are faced with the same textual elements prearranged in a sequence that does not change from one reading to another. Therefore, quite logically, texts of this kind are labeled monosequential.¹¹³

However, monosequential texts may also very well be divided into several content spaces, yet still offer only one path, i.e. still be monosequential. Madame Bovary, for instance, constitutes one authorial content space. When first published, though, the novel appeared in the journal Revue de Paris in several installments.¹¹⁴ Today, Madame Bovary also exists in a web published version in which each chapter is separated from the previous and the following by links.¹¹⁵ Both these versions of the novel still constitute only one authorial content space. Yet they are published in several editorial content spaces. This should be compared with the "traditional" publishing of the novel in book form where, in general, there are no editorial content spaces. Another example is a movie on television that is repeatedly interrupted by commercial breaks and hence divided into several editorial content spaces.

It should be noted that the examples above differ from "true" serials intended to be published this way; the work as such is divided into several content spaces, which are thus to be characterized as authorial. Dickens wrote Great Expectations knowing it would be published piecemeal, which is also reflected in the work (several authorial content spaces).¹¹⁶ Similarly, the TV dramatization of Madame Bovary was created to be distributed episode by episode; it consists of several authorial content spaces. Making the user eager to return in order to read, watch or listen to the next episode is of major importance to serials and results in the well-known cliffhanger endings of episodes.  

Clearly, in distinguishing monosequential works, the number of content spaces (authorial or editorial) is of no importance. The crucial issue is instead the intended mode of navigation. To set up road barriers, where a click with the mouse or the passing of a week is the password to continue along the textual highway, is not in any way comparable to constructing crossroads or circles offering several directions. However, many traditional novels are also available with an abundance of crossroads. In one critical edition of Jane Eyre, for example, an excerpt of the passage in chapter fifteen where Mr. Rochester tells Jane of Céline, the mother of his child, reads:¹¹⁷

Here ensued a pause, filled up by the producing and lighting of a cigar; having placed it to his lips and breathed a trail of Havannah incense on the freezing and sunless air, he went on: –

‘I liked bonbons too, in those days, Miss Eyre, and I was croquant – overlook the barbarism⁴ – croquant chocolate comfits, and smoking alternately, watching meantime the equipages that rolled along the fashionable streets towards the neighbouring opera-house, when in an elegant close carriage drawn by a beautiful pair of English horses, and distinctly seen in the brilliant city-night, I recognized the "voiture"⁵ I had given Céline."

⁴ This use of croquent (the present participle of the French verb meaning "to crunch") as a "barbarism" may be (1) a syntactic inexactitude, (2) eating candy and smoking at the same time, or (3) a colloquial French expression, a croquer (!) (which Brontë uses later, on page 216). When one says in French that a child or a girl is a croquer (!), it means that she is literally "good enough to eat." (The original meaning, "pretty as a picture," based on croquer, meaning to "sketch," has changed through confusion with the first meaning cited.) Therefore, could Brontë, using croquant as an adjective rather than as a participle, be punning by implying that Rochester considered himself one to devour those a croquer (!)?
⁵ Carriage.

Here the reader has to navigate in the true sense of the word creating her way through the text by choosing between alternative routes. Some people will read both footnotes, some will only read one of them, whereas others will not read any of them. In other words, the sequence of textual elements varies from reading to reading (reading is here to be understood as the result produced from the act of following lines of typographic signs and assimilating their content). Depending on which links (in the form of footnote indications in the example above) the reader chooses to follow, the order of the parts of the text changes and is therefore to be described as multisequential.¹¹⁸ Multisequential texts consist of several content spaces which are intended to appear, and to be experienced, in different order and/or in selection.

Based on how the content spaces are structurally organized and interrelated, multisequential texts fall into three main categories. George P. Landow distinguishes between axially structured and network-structured text. To these I would like to add laterally structured multisequential text.¹¹⁹ It must be remembered, though, that this is a theoretical model and that most existing texts are not pure in terms of structure. Therefore, it is more accurate to describe texts as having mainly axial, network or lateral structure.

Characteristic of axially structured texts is that they always have links and one principal, superordinated monosequential text that may be experienced irrespective of the links.¹²⁰ In this kind of multisequential text, there is a main track to which the user will have to go back if she chooses to follow a link; the turnings in an axially structured text will sooner or later turn out to be cul-de-sacs. The critical edition of Jane Eyre cited above is one example of a mainly axially structured text: comments of a certain type have been attached to the novel. The reader may leave the main track in order to check what further information the footnote can provide. But having read it, she is inexorably redirected towards the main text to continue along the main track. An academic dissertation is another example of a mainly axially structured text. Here, as in the previous example, the footnote often contains a reference to a specific passage in a work by another scholar. The dedicated readers will naturally obtain a copy of this work to look up the paragraphs referred to before returning to the main text and argument. So regardless of whether the reader follows one or several links after having left the main track of an axially structured text, she is sooner or later expected to return to it.

In network-structured texts users do not cruise along any clearly defined main track. On the contrary, they wander around a net-like structure of content spaces. As in axially structured texts, the link is obligatory in network-structured texts. However, in the latter case, links function differently (cf. the section on links and linking below), resulting in, among other things, a number of alternatives for the user to choose between at each crossroads or, perhaps a better metaphor in this case, at each circle. Digital hypernovels and web sites are examples of mainly network-structured texts (and works).

The navigation rhythm in network structures is different since the predominating course of direction when following links is onward, onward and onward, which should be compared to the onward and backward rhythm of the axial structure.¹²¹ Surfing the Internet is one example of how it is easy to get caught up in the onward current; we click on something interesting on one web page, then on something intriguing on the next, and so on. Naturally, in this case it is generally possible to backtrack via the browser’s "back" function. However, in a network structure, going back to the previous content space is often just one among many options, while in an axial structure, this is the natural and required movement. In other words, a pure network-structured text could be consumed without any backward movements (although the user may visit the same content space several times) whereas, in a mainly axially structured text, the user must jump back (provided a link is followed in the first place).

The third type of multisequential text, the laterally structured text, is characterized by the absence of links between content spaces.¹²² Moreover, in laterally structured texts, as in network-structured texts, there is generally no main track to which the user always has to go back. At the same time, content spaces in lateral structures are usually less dependent on each other than content spaces in network structures. Typical of lateral structure is also what Aarseth calls random access, which means that all content spaces are accessible at any time.¹²³ A newspaper, an anthology, a magazine issue, and an encyclopedia with relatively long articles are some examples of mainly laterally structured texts (and works). It should be stressed that the concept of lateral structure applies to content spaces deliberately arranged to form a work and intended to function as a whole. The lateral structure has to do with the presentation medium and the presentation signs since several content spaces are presented simultaneously and in parallel. How and where these are stored is of no relevance here.

Content spaces in laterally structured text are generally arranged in a way that facilitates the user’s navigation, making it easier for her to find the things that interest her. Content spaces in lateral texts may for instance be organized in alphabetical order; in chronological order (like in literary history); or arranged according to subject, theme, or importance. In a newspaper, these different principles of arrangement of content spaces are combined in order to capture and keep the reader’s attention. In fact, the newspaper layout is a sophisticated play with these forces requiring careful consideration in order to achieve the desired effect.

The three types of multisequential structures presented above, axial structure, network structure and lateral structure, may be described schematically:¹²⁴

axial structure

network structure

lateral structure

Figure 1. Axial, network and lateral structure.

Finally, take for example a work of digital hyperfiction such as Stuart Moulthrop’s Victory Garden.¹²⁵ Victory Garden consists of several authorial content spaces and is intended to be presented on and read from a computer screen. The user navigates among the content spaces mainly by clicking on words and graphics or by pressing keys. The click determines the next content space – a click on the word "plot" in [Buildup] displays [Lucy Reasons], while the link from "not" in the same content space leads to [Anonymous]. When pressing "enter", the content space following [Buildup] is [Games]. Unquestionably, this is a multisequential, mainly network-structured work.

In order to illustrate the structures discussed in this section, I permit myself to hypothetically move Victory Garden from its natural and intended storage medium and presentation medium. First, I make a printout of every single content space and then I organize them in a certain order in a binder. This version of Victory Garden is monosequential since it is meant to be read from cover to cover in the order presented. Naturally this monosequential version of the work can be stored and presented in digital media as well as in print media. Note that the numerous authorial content spaces are here presented as one editorial content space.

Then I make new printouts. But this time I arrange the authorial content spaces in a collage. Now there is no given trail to follow as in the binder-copy. On the contrary, the spectator/reader may shift her gaze to any of the content spaces and read them in any order. In this version of Victory Garden, the structure is mainly lateral and as in the monosequential binder copy, the links have been removed. Furthermore, it is also possible to transform Victory Garden into a mainly axially structured version; in this case certain content spaces will make up the monosequential main text (with links) while others will function similarly to footnotes, linked to this main text. This mainly axially structured version can be realized both in print and in digital form. Contrary to the monosequential binder version, these multisequential versions of Victory Garden have no editorial content spaces.

[To the top]

5. Hypertext

In a paper presented at a conference in August 1965, Ted Nelson states:

Systems of paper have grave limitations for either organizing or presenting ideas. A book is never perfectly suited to the reader; one reader is bored, another confused by the same pages. No system of paper – book or programmed text – can adapt very far to the interests or needs of a particular reader or student.

However, with the computer-driven display and mass memory, it has become possible to create a new, readable medium, for education and enjoyment, that will let the reader find his level, suit his taste, and find the parts that take on special meaning for him, as instructions or entertainment.

Let me introduce the word "hypertext" to mean a body of written or pictorial material interconnected in such a complex way that it could not conveniently be presented or represented on paper.¹²⁶

Questioning the sequential fixity of traditional print text as the most suitable format for storing and presenting ideas produced by the human mind, Nelson calls for texts functioning in a way more similar to that of the human brain. Similarly to the law of association governing the human thinking, texts ought to be structured according to the same principles of nonsequentiality and association. We think in hypertext, we speak hypertext, so why don’t we write hypertext? Traditional writing, Nelson claims, is "a process of making the tree of thought into a picket fence."¹²⁷

The basic ideas of Nelson’s revolt against the textual sequence, it should be noted, were not quite as new as the term ‘hypertext’, invented to designate a nonsequential, interconnected structure of typographic text and, to some extent, illustrations.¹²⁸ Twenty years earlier, engineer Vannevar Bush had published the visionary (and in retrospect also partly prophetic) article "As We May Think." This classic article is best known for its description of an analog machine with striking resemblance to a modern computer. The memex, as Bush labels this memory machine based on microfilms, is described as a device for storing texts, records and communications and it is furnished with a screen, buttons and a keyboard. But primarily, being "an enlarged intimate supplement to . . . memory," this memex would operate similarly to the human brain:

The human mind . . . operates by association. With one item in its grasp, it snaps instantly to the next that is suggested by the association of thoughts, in accordance with some intricate web of trails carried by the cells of the brain. . . .

Man cannot hope fully to duplicate this mental process artificially, but he certainly ought to be able to learn from it. . . . Selection by association, rather than indexing, may yet be mechanized. One cannot hope thus to equal the speed and flexibility with which the mind follows an associative trail, but it should be possible to beat the mind decisively in regard to the permanence and clarity of the items resurrected from storage.¹²⁹

According to Nelson’s own statement, the term hypertext was invented as early as in 1963, two years before it appeared in print for the fist time.¹³⁰ The concept is presented in the previously cited conference paper given in August 1965 and printed in the conference proceedings. However, this was most likely not the first talk given on the subject; in an issue of the Vassar college newspaper, The Miscellany News, dated 3 February 1965, there is an article titled "Prof. Nelson Talk Analyzes P.R.I.D.E." In this article, it is explained that the P.R.I.D.E system stands for ‘Personalized Retrieval Indexing and Documentary Evolution’ and is a new organizational method invented by Nelson:

In this system passages of material would be translated into machine language and filed in the machine in any sequence. With the proper instructions the machine would print out any sequence the writer wished to try, freeing him from the necessity of keeping the ideas in his head. Mr. Nelson pointed out that we often do not think in linear sequences but rather in "swirls" and in footnotes. He introduced the concept of the hyper-text, which would be a more flexible, more generalized, non-linear presentation of material on a particular subject.¹³¹

Regardless of exactly when Nelson publicly launched the term hypertext, his idea of flexible, interconnected and non-linear networks was clearly outlined in 1965, and had at that time been around for several years. Nelson’s life work, the Xanadu project, had been running since 1960 developing what would later be called The Xanadu Hypertext System. Xanadu, Nelson explains in Literary Machines, is a new form of storage and delivery, or presentation of texts (in the broad sense) that permits "promiscuous linkage." It is a system in which everything is linked and where changes are assimilated and stored.¹³²

In the early definition of hypertext cited above, Nelson stresses that hypertext is a body of interconnected material that is almost impossible to store and present on paper. Instead, the optimal and natural storage and presentation medium for hypertext is the computer. However, in Literary Machines, published in a first version in 1981, the definition is somewhat revised and modified as it explicitly includes also print texts:

By hypertext I simply mean non-sequential writing. A magazine layout, with sequential text and inset illustrations and boxes, is thus hypertext. So is the front page of a newspaper, and so are various programmed books now seen on the drugstore stands (where you make a choice at the end of a page, and are directed to other specific pages).

Computers are not intrinsically involved with the hypertext concept.¹³³

Nelson reserves the term ‘hypertext’ for multisequential (or, in Nelson’s terms, nonsequential) structures (axial, network or lateral) consisting of mainly typographic text, possibly with some illustrations.¹³⁴ Other types of text, such as cinematic and audible text, ordered multisequentially are referred to as hypermedia, a term that was also coined by Nelson in 1965.¹³⁵

Today, the term hypertext is integrated in the everyday language and used in numerous ways and in various contexts. As a consequence, there are many different definitions of the term. Often, the term is reserved to designate certain multisequential structures in certain storage and presentation media. For example, hypertext is sometimes to be understood as digitally stored network-structured text with digital links (a web site, a work of digital hyperfiction, etc.). At other times, the term also includes print media and axially structured text (a dissertation, an academic essay, etc.) whereas it excludes texts with a lateral structure and without predefined, explicit links (a newspaper page, an art exhibition, etc.).¹³⁶ In its broadest sense, hypertext denotes all kinds of texts arranged multisequentially.

Regardless of which definition of hypertext one subscribes to, it is possible to make a distinction between works that are intended to be presented as hypertext and those that are not. Based on the distinction between work and text, Johan Svedjedal has distinguished between hyperworks and non-hyperworks, depending on whether the work in question is intended to appear in hypertextual form or in non-hypertextual form.¹³⁷ Digital hyperfictions such as Shelly Jackson’s Patchwork Girl and Linda Carroli and Josephine Wilson’s *water always writes in *plural, a web site, and a printed dissertation, are examples of hyperworks, i.e. works created to appear in hypertextual form. Non-hyperworks, on the other hand, such as a traditional novel, a film, or a poem, are intended to be presented in non-hypertextual form, that is, they are structured monosequentially. But non-hyperworks may also be represented as hypertexts. This is the case, for example, with critical editions of literary works and DVD versions of films. In the same way, a hyperwork may be presented in non-hypertextual form. Take for example a version of an academic essay with a table of contents and footnotes in which the table of contents has been omitted and every comment and bibliographical reference has been moved into the main text. In this case, the text is no longer a multisequential hypertext, but a monosequential non-hypertext. Another example of a hyperwork presented in non-hypertextual form is the binder version of Victory Garden described in the previous section. These are new and non-hypertextual versions of what was originally hypertextual hyperworks.

[To the top]

6. Ergodicity

Espen Aarseth introduced the term ergodic to describe texts through which paths are "produced by a non-trivial element of work" and he differentiates between ergodic literature and nonergodic literature.¹³⁸ Aarseth’s understanding of the concept of literature, it should be noted, reaches far beyond the usual, prevalent definition, and includes not only traditional novels like Moby Dick but also, for example, computer games and so-called MUD games. In ergodic literature, the user must actively participate and create a path through the text by choosing between different alternatives. In other words, the user must work herself through the text.¹³⁹ One might argue that reading Proust’s À la Recherche du temps perdu is hard work; and there are indeed multiple ways of interpreting the work. However, the issue here is not user efforts of this kind. Instead, the discussion concerns the intended mode of navigation which, in ergodic text, always includes a higher degree of participation on the part of the user than in nonergodic texts.

In ergodic literature, on the one hand, a non-trivial effort is required to traverse the text. The user must choose in which direction to run and which monster to shoot first in Doom, she must decide which person to ask for the kidnapped child in Gabriel Knight and whether to click on "a woman," "stands on a street corner," "waiting" or "for a stranger" in *water always writes in *plural.¹⁴⁰ In nonergodic literature, on the other hand, the user may choose to read a footnote and move the eyes to the bottom of the page, or decide to continue reading the headline article and follow the link to page 12. In nonergodic literature, following a link – provided, of course, that there is one – is rather uncomplicated and automatic; it is done without much thinking and only a trivial effort is required.¹⁴¹ In ergodic literature, however, the decision of what to do next is of great importance and often has far-reaching consequences. The user is aware of this and knows that she might have missed something, that a problem might have been confronted in another way and that she may "die" if she chooses the wrong way or speaks to the wrong person. In contrast, reading a footnote or not in a nonergodic text (like a traditional dissertation or article) generates no similar consequences and the choice is not decisive in the same way. The footnote does not disappear if the reader does not read it; reading a footnote does not literally change the main text, and the reader does not have to start from chapter one if she fails to read footnote thirteen etc. However, in an ergodic text, this could be the case.

The examples above illustrate the difference between the non-trivial effort required of the reader of the ergodic text and the trivial effort of the nonergodic text. It is important to stress that the distinction does not run along the border between print and digitally stored text, even though the occurrence of ergodic texts is more common in digital media than in print. The Chinese work I Ching (ca. 1000 B.C.), Raymond Queneau’s Cent mille milliards de poèmes (1961), and Julio Cortázar’s novel Hopscotch (1966) are some examples of ergodic texts in print pointed out by Aarseth.¹⁴²

Non-trivial efforts required by ergodic texts, however, vary in complexity and character. Sometimes making a choice is not only choosing between alternatives but also, in effect, creating alternatives (which is also, somehow, a choice). Michael Joyce has introduced the terms exploratory hypertext and constructive hypertext which describe these phenomena in hypertext:

By exploratory use, I mean to describe the increasingly familiar use of hypertext as a delivery or presentational technology. . . . Exploratory hypertexts encourage and enable an audience . . . to control the transformation of a body of information to meet its needs and interests. . . .

By constructive use, I mean to describe a much less familiar use of hypertext as an invention or analytic tool. . . . More than with exploratory hypertexts, constructive hypertexts require a capability to act: to create, change, and recover particular encounters within the developing body of knowledge.¹⁴³

In exploratory hypertexts, the user is not allowed to – or at least not supposed to – touch or change anything. Figuratively speaking, it is like being a ten-year-old on an educational visit at the local industry: keep your hands on your back, your eyes open, and your mouth shut. In constructive hypertexts, on the contrary, you must still keep your eyes open. But you are also expected to make changes and speak up.

Joyce’s distinction between exploratory and constructive hypertext is a sufficient basic dividing line. However, it should be noted that these categories could be subdivided. For example, Aarseth’s distinction between configurative and textonic user functions can be viewed as a distinction between two kinds of constructive texts.¹⁴⁴ In texts with a configurative function, the user may not only metaphorically, as in exploratory hypertext, but also literally, choose and create her way through the text. The textonic user function means that the user may add permanent features to the text. In text-based gaming environments like MUDs and MOOs, for example, users may modify and create objects:

Since objects in a text-based MUD are made out of programming code and words, there is no limit to what can be called into being within the virtual world. An expert MUDder might have his own private castle, with hidden pathways and working drawbridges; he could recruit other people to come live in it and swear fealty to him, or he could amuse "newbie" visitors [i.e. beginners] with puzzle rooms or frighten them off with ferocious trolls. Even a very uncertain programmer can create objects with personal resonance, like a Chinese dancing fan that only looks graceful in the hands of its creator.¹⁴⁵

In these environments, the user may create objects and specify their appearance, but what is more, it is also possible to specify their behavior. Objects may be created procedurally, to use Janet H. Murray’s terms.¹⁴⁶ Such works, open to the manipulation and interaction of the user, are closely related to the kind of works that Espen Aarseth labels cybertexts, i.e. texts that contain "some kind of information feedback loop."¹⁴⁷

[To the top]

7. Links and Linking

Links are explicit connections between works and parts of works. As demonstrated, they can appear both in monosequential and multisequential texts. Metaphorically speaking, a link is a thread with a beginning and an end, which the user may grab and follow like a guiding line. To follow a link, then, is to be guided from one place to another; from the sentence in the main text to a specific footnote, from one web page to another, from the first page of a newspaper to a certain page in the paper etc.

It should be noted that the following discussion on links mainly deals with how different types of links and linking function structurally. Thus other interesting issues such as the implications of links in terms of aesthetic effects etc., are beyond the scope of this discussion.¹⁴⁸ The aim is instead to provide a new terminology for the description and analysis of links, to present the nomenclature for a "linkology."¹⁴⁹

Link Source, Link Destination and Anchor

The text or the part of a text from which a link departs is referred to as a link source while the text or the part of a text to which a link leads is called a link destination. The exact spot in a link source or a link destination to which links are attached has been called an anchor.¹⁵⁰ In typographic texts, the anchor is usually a fairly small part of the source or the destination; often, the anchor is a word, a footnote number or perhaps a sentence.

Link sources and destinations vary in size and scope; theoretically, they can consist of anything from a single alphanumeric sign to an entire work or even several works. In fact, link sources and destinations in typographic texts could be said to form a continuum of the following type:

alphanumeric sign – word – sentence – paragraph – chapter – part of work – work – several works

Naturally, only a few of the possible positions on the scale are indicated here, and the scale may thus be refined as required. Sources or destinations consisting of, for instance, four words but still not a complete sentence, or of several paragraphs yet not an entire chapter, therefore also fit into the continuum. It is important to remember that link sources and destinations often are part of another, larger link source/destination and, at the same time, consist of smaller sources and destinations. In a novel, for example, a paragraph, a chapter as well as the entire work may serve as link destination and/or link source.¹⁵¹

Digital and Analog Links

It has already been stated that there are both digital links and analog links. Although digital and analog links can, to a considerable extent, be described in the same terms, it is vital to remember that, due to physical and technological qualities, there are some fundamental differences between, for example, a typographic link and a digital link. The most vital difference is that analog links are processed by humans, while digital links are processed by machines (e.g. a computer). In other words, these links use instructions for their realization that are written either for man (analog links) or for machines (for example digital links).

With many digital links, the user, by pressing a key or clicking on a word, etc., signals that she wants to follow a certain link. The link destination is then promptly activated and presented to the user. For technical reasons this is not possible in print, and the user is therefore obliged to actively follow instructions provided (explicitly or conventionally) by the anchor in the link source. An ordinary footnote is a good example: a superscript four after a sentence tells the reader that if she is interested in reading a comment on this section, she will find it after the superscript four at the bottom of the page (or at the end of the essay). In a digital text, clicking on the superscript four may instantly make the comment appear on the screen (as in Word documents.) One could say that whereas digital link destinations always come to the user, typographic links mostly require the user to come to the link destination (by moving her gaze, turning pages, getting another book, etc.).¹⁵² In a way, then, typographic links place a greater responsibility on the users who must find the link destination on their own. Evidently, this leads to an increased risk for mistakes. How many readers have not, by mistake, read footnote number four of chapter five instead of footnote number four of chapter six when working with a book with end notes ordered chapter by chapter?

Typographic links also exist in digital media. Text published in pdf format, for example, often contains exclusively typographic links. Thus in order to get to read the footnotes in an essay published as a pdf file, like for instance Toru Sasaki’s "Ghosts in A Christmas Carol: A Japanese View," the reader must scroll down to the footnote section. In other words, the footnote text is not brought to the reader. As in print, the reader herself must find the link destination.¹⁵³

Text in digital media often has both digital and typographic links. Take for example Tina Young Choi’s web published essay "Completing the Circle: The Victorian Sanitary Movement, Our Mutual Friend, and Narrative Closure" containing nine footnote references.¹⁵⁴ If interested in a certain footnote, the user simply clicks on the superscript footnote number in the main text and the footnote is displayed on the screen. When clicking on the superscript eight, for example, footnote number eight is presented to the reader: "8. Charles Dickens, Our Mutual Friend (London and New York: Penguin Books, 1985), 849. BACK." To return to the main text, the reader has three options: (1) to click on "BACK" in the footnote, (2) to use the back function in the browser, or (3) to scroll back. The third alternative implies that the user follows a typographic, analog link. In addition, the links that lead from the bibliographical references in the footnotes to the works in question are typographic – page 849 in Our Mutual Friend will not appear on the screen when the reader clicks on the reference in the footnote.

Thus in digital media, a typographic link may serve as an alternative to a digital link. For obvious reasons, though, the digital link generally predominates over the typographic link – why scroll when it is possible to click? However, the typographic links are excellent stand-ins when the digital links do not work, for some reason. This is the case when a digital link malfunctions and, perhaps even more so when a digitally stored text migrates to another, analog storage medium. For instance, when printed, the digital links in Choi’s essay will "die" and the typographic links will come into power. In this particular essay, it is also interesting to notice that the digital links take advantage of the user’s knowledge of the functionality of typographic links; a click on "BACK" in footnote six, and the link will lead the user to an anchor in the link destination that is situated before the paragraph containing footnote number six. As a result, the user has to activate her competence as to analog, typographic links in order to navigate from the beginning of the paragraph to the footnote number.

In order to function, both analog and digital links depend on the user having the knowledge of how to deal with these features. In order for the user to know what to do when running into a blue, underlined word on a web page or a superscript four in a dissertation, the knowledge of how to interpret and handle links must be an integral part of her navigational competence.¹⁵⁵ One could say that the user must have internalized a set of instructions informing her on how to recognize and use links. Of course, this set of instructions is not an absolute package but varies from user to user, who acquire them by experiencing different texts. Once you have read an academic essay, you know what a small superscript number indicates and how footnotes function; once you have surfed the web, you know what underlined words in blue mean and why the cursor sometimes turns into the picture of a small hand.

Internal and External Links

Links have their source and destination within a single work or in different works. Links within a work are called internal links while links running between works are labeled external links.¹⁵⁶ Clearly, these types of links closely depend on the concept of work and the adopted perspective, that is, what is considered a work is crucial when characterizing links as external or internal. From one point of view, a link may be described as external whereas from another point of view, it may be regarded as internal. Take for instance a link from one academic paper to another (in for example a footnote reference.) This is an external link from the writers’ point of view. But if the two papers are collected in the same anthology, the link may be seen as internal (within the book) from the point of view of the reader.

In D. C. Greetham’s Textual Scholarship: An Introduction, as in most academic texts, there are numerous examples of both types of links. Examples of internal links are found in the table of contents, from which readers are directed to specific pages in the main text: to read chapter seven go to page 271, etc. On page 271, there are other examples of internal links in the anchors: "see Chapter 2" and "see Chapter 5." There is also an external link on the same page leading from the anchor, "(Gaskell, New Introduction: 337)," to a destination in the work of another author.¹⁵⁷

If a web site is considered a work, internal links run within the web site (within a page and between pages) while external links lead to other web pages. On the Sherlockian.Net home page, for instance, a table of contents of the site is presented to the left:

Figure 2. "Sherlock Holmes on the Web: The Sherlockian.Net Holmepage."

In the table of contents, the web pages within the site are listed and the user may click, for example, on "Arthur Conan Doyle", "The world of Holmes and Watson" or on "How to write a term paper". The links in the table of contents are all internal. However, on the same page, there are also external links: for example, two leading to photo galleries on the web ("Images of England" and "London Stills") and one leading to the Vancouver Museum. If the user clicks on "Sherlockian resources on the Web", a page is displayed where the external links are far more numerous than the internal links.¹⁵⁸

Unidirectional and Bidirectional Linking and Links

Unidirectional linking and bidirectional linking have to do with the possibility, or lack of such, to go back to the place that one just came from, "to retrace one’s steps."¹⁵⁹ Unidirectional linking is when there is no explicit back link in the link destination leading back to the link source. Links involved in unidirectional linking are labeled unidirectional links. Bibliographical references in footnotes indicating, for instance, a specific page in another work, are almost always examples of unidirectional links – there are rarely any links in the destination text leading back to the footnote in the first text.

Bidirectional linking is when an explicit back link leads from the destination of a link back to its source. On one level, it is sufficient to describe texts as bidirectionally linked, i.e. there is a two-way connection between the two linked texts. However, bidirectional linking may also be studied and described in more detail. Bidirectional linking involves two links, one forward link, so to speak, and one back link. At least two anchors are required in bidirectional linking, but three or even four anchors may be used. Consequently, the forward link and the back link may either use the same anchors or different anchors. Here the term homoancoral is used to designate that a forward link and a back link have the same anchor, while heteroancoral indicates that they use different anchors.

In order to fully describe bidirectional linking between two texts, the anchor issue must be considered in both texts, that is on the one hand, the text in which the forward link has its source and the back link its destination, and on the other hand, the text in which the back link has its source and the forward link its destination. As a result, four types of bidirectional linking are possible:

Figure 3. Types of bidirectional linking.

From now on, the source and the destination are described in the syntax source/destination.

The term bidirectional link is here reserved for bidirectional linking that is homoancoral/homoancoral.¹⁶⁰ This is the case, for example, with an ordinary footnote in print where the forward link leads from the footnote number in the main text to the footnote number in the footnote, and the back link leads from the footnote number in the footnote to the footnote number in the main text.

Examples of heteroancoral/heteroancoral bidirectional linking are found at the Sherlockian.Net home page, where the links from the table of contents lead to destinations from which it is possible to return to the source page. For example, when clicking on "Individual Sherlockians," the top of a web page titled "Sherlockian.Net: Individual Sherlockians" appears on the screen. Wanting to go back, I click on "Back to Sherlockian.Net’s Homepage" at the end of the page and the top of the Sherlockian.Net home page is immediately displayed. Thus the forward link departs from the anchor "Individual Sherlockians" and leads to the default anchor (cf. definition below) at the Individual Sherlockians page. The back link leads from the anchor "Back to Sherlockian.Net’s Homepage" to the default anchor at the Sherlockian.Net home page. In this case, four different anchors are involved in the bidirectional linking.

It is important to bear in mind that bidirectional linking involves two links and that each of these may constitute a pair, and form bidirectional linking, with several links. In digital texts, for example, it is all but rare that a word like "back" or something similar has been added in the footnotes serving as anchors for back links. As mentioned, this is the case in Choi’s essay, in which each footnote contains a "BACK" alternative. Thus, there are two links leading back to the main text: the typographic link from the footnote number and the digital link from "BACK". In other words, the same forward link may be used whereas the back links may differ.

Although both unidirectional and bidirectional linking may easily be found on the Web, unidirectional linking seems to be more common. In most browsers, however, there is a return function that generally allows the user to jump back to the previously visited page or spot on a particular page. This creates the effect of bidirectional linking. However, it should be remembered that retracing one’s steps by using the browser has nothing to do with links of works.¹⁶¹ The browser keeps track of the visited URLs; when the return function is used the browser simply displays the previous page. These pages may be interlinked (like when surfing within a web site), which reinforces the effect of bidirectional linking. But it may also be the case that the pages have nothing to do with each other, for example when the URLs were typed manually or selected from the URL list in the browser, or when the pages were retrieved through a search engine.¹⁶² Even though this allows the user to move smoothly between web pages, it is different from using links that are part of the work.

Unconditional and Conditional Links

Unconditional links can be followed under any circumstances at any time; the only thing needed is the accurate navigational competence. Most typographic links are unconditional for all practical purposes; the link from the footnote number anchor in the main text to the actual footnote is accessible regardless of whether the user has read the chapter in question and irrespective of factors such as time and the number of users having read the reference before. Similarly, many digital links in web published texts are unconditional and can be followed by anyone at any time. In most cases, all links from a web page are unconditional.

In contrast to an unconditional link, a conditional link can be followed only if certain conditions are fulfilled.¹⁶³ A suitable example from the everyday life is the password – we are not allowed to follow the link unless we know and can state the correct word. Conditional links may regulate access to a content space making it accessible for instance only thirty-five seconds at a time or only on nights when there is a full moon. With conditions, the access to links can be controlled in detail. If, for instance, links A, B and C have their source in X, link B may be accessible only if the user has visited certain content spaces while the C link may only be followed if all the content spaces have been visited. Link A would then be activated when the conditions governing access to B and C are not fulfilled.¹⁶⁴ This means that when the conditions are fulfilled, the A link is, figuratively speaking, replaced by the B link, which in turn may be replaced by the C link.

In these examples, the user does not explicitly and deliberately choose one alternative or the other – the content space is only available after exactly thirty-five seconds and not before nor after that. In other words, these are examples of non-ergodic conditional links. Ergodic conditional links, on the other hand, require the user to make a non-trivial choice of which link to follow, for example, by choosing between objects displayed on the screen.¹⁶⁵ 

Of course, conditional links in ergodic texts do not always order the user to choose between alternatives; the condition may also be that a certain task must be accomplished. These kinds of conditional links have a key role in many computer games in which the player has to fulfill certain missions in order to get to the next level, the next room, or the next city, and, in the end, to succeed in saving the princess or killing the dragon.

Intricate and complicated conditional linking may, for obvious reasons, be most successfully created in a computer. However, conditional links do exist in other media as well. The most frequent conditional links in analog media are perhaps those attached to anchors like "to be continued…" and others where the condition consists in the passing of a period of time. This type of conditional link that forces the user to wait until the next day, the next week and so fort to gain access to the link destination can be found in print media as well as on television, on the radio and, naturally also in digital media.

Once this real-time related condition has been fulfilled, it generally, especially in print media, looses its power to control access. In 1859, Charles Dickens founded All the Year Round, a publication whose serialization of novels was an important feature. In October 1860, sales were dropping radically, and in an attempt to reverse the trend, Dickens decided to once again feature a story of his own. As a result of this decision, Great Expectations was published in weekly installments in All the Year Round (from December 1860 until August 1861), and as hoped, sales did indeed increase.¹⁶⁶ A skillful combination of the time-related condition and "cliff-hangers", creating suspense until the next installment, proved to be a successful weapon against dropping sales.

However, once the last episode of Great Expectations was published, the readers could access the whole work at any time. The link was still there (the reader had to "jump" from one issue to another), but the effect of the condition had changed since it no longer maintained a certain temporal gap between the two sections. Later, when all the episodes were published together between covers, the gaps, and consequently the links, were removed. In the widespread version of Great Expectations published in book form the stylistic and narratological devices give glimpses of the conditional links (mostly discernible as "cliff-hangers") that played such a crucial role when the narrative was originally published in All the Year Round.

Visible, Invisible and Hidden Links

The existence of a link may be indicated or not. Links that can be identified simply by looking at the text (or more precisely at the presentation signs) are called visible links, while links that cannot be visually detected are labeled invisible links. Links that the user can make visible (by pressing a key or moving the cursor, etc.) are referred to as hidden links. In most cases, it is not the links as such that are visible, invisible or hidden (they are in general all invisible) but their anchors. Anchors are often indicated in some way in order to stand out in the text, making it easier for the user to find the links. This anchor indication is called a cue. A standard cue for anchors on the Web is the blue color in combination with underlining.¹⁶⁷ In print, the cue of footnote number anchors usually involves a smaller font and a superscript position.

Invisible links are practically impossible to achieve in print media since typographic links always contain more or less explicit instructions as to where to find the link destination. In other words, these kinds of links automatically have cues since they must provide instructions regarding the link destination. A good example in print is a page number in an index. Digital links, on the other hand, may be invisible. Truly invisible links, however, seem to be rather unusual and the majority of links are more accurately referred to as hidden links since a certain keystroke or a mouse movement will make them visible. For example, if a web page is not provided with explicit cues, links are usually pointed out, as we move the cursor over the text, by the appearance of the small hand indicating on which parts of the text the user may click. Pressing a certain key may sometimes be another method of revealing links, like in Patchwork Girl, where pressing the control key makes a red box appear around clickable text.

In Michael Joyce’s hypernovel afternoon, a story,¹⁶⁸ however, the links are invisible in that there are no means of making them stand out from the rest of the text; when reading a content space the user cannot tell whether there is what is called a "default link" in the program Storyspace, or which words that are "words that yield," to use the author’s own words, or whether there are any links at all.¹⁶⁹ The fact that the reader who has read the instructions knows that there is supposed to be a default link and that the content space title often gives a hint of at least one of the words that yield, does not change the character of the links. Nor does the fact that it is possible to click on "Links" in the bar to get a list of all the links that lead from the content space. The links are still not distinguished from one another and from the rest of the text in the content space; i.e. the individual anchors cannot be identified.

Uncategorized and Categorized Links

Uncategorized links give no information whatsoever on the character of the link destination. The typical example is the underlined blue anchor in a web document – the user knows that the place is linked, but not to what kind of destination. Many of the links in Linda Carroli’s and Josephine Wilson’s web published work *water always writes in *plural are uncategorized since the user, as a rule, does not know what to expect when clicking on the anchors. For example, when clicking on "her waiting will never end" at one of the web pages, there is nothing that signals to the reader that the link destination contains a picture (a postcard) and a sound. Also, in this particular hypertext, the typography changes – type fonts and sizes vary as do paragraphs and placement of the text – without there being any indications of what the user can expect. When clicking on "Poor service." on a page written in Times font (although in different sizes), there is nothing that indicates that the text at the link destination is written in large, winding letters imitating handwriting.¹⁷⁰

In contrast, categorized links provide the user with more or less information on the link destination. The information provided may, for instance, concern the textual and typographic character of the destination, the relation between the source and the destination or the content of the destination. But differences between categorized links are not only related to what kind of information they give but also to how much information they give. In other words, categorized links have a variable degree of precision.¹⁷¹

Previously, I argued that links leading from blue, underlined text in web documents are uncategorized. This is often true. From the underlined blue word "Rod" one cannot tell whether the link leads to a photo of Rod Stewart, a drawing of my dog Rod or a Rod Stewart sound track. However, the URL displayed when holding the cursor over the anchor Rod could reveal some more information, like anna/pets/dogs/rod.htm. Yet it gives no information on the actual character of the link destination – is it an essay, a picture, a song?¹⁷² The issue here is thus whether the interpretation of words, pictures, etc. serving as anchors – the associations tied to certain words – may create a categorized link. Are the blue underlined words "The New York Times" in a poem a categorized or uncategorized link? I would say that such links feign categorization; the categorization relies on a pure assumption on the part of the reader. I assume that what I associate with the words "The New York Times" is also what the link leads to. I think that the link destination is The New York Times home page but instead I am presented with a poem titled "The New York Times." Of course this is not clear-cut: if the words were found in italics in an article on newspapers, the link would probably be described as categorized.

In hypertext theory, links that "signal to the reader what the relationship between the link’s anchor (or departure point) and target are," are often referred to as "typed links."¹⁷³ Unlike the term categorized links, however, the term typed links is generally restricted to digital links. Categorized links is thus a wider concept including the so-called typed links. The reason why the definition of typed link is not simply modified to include also analog links is that the word "typed" may easily be misinterpreted in contexts where links in print media are discussed.

As to categorized links, it is generally the anchor from which the link leads that carries the information; a traditional footnote number anchor like the one following this sentence is interpreted by means of a set of typographic conventions, which tell you that the link leads to a footnote, which is a comment of some kind, often printed in smaller font and placed at the bottom of the page or at the end of the chapter.¹⁷⁴ This anchor provides relatively little information, however, compared to bibliographical references in footnotes. Of course, bibliographical references also vary in complexity and precision concerning the link destination. In fact, typographic links are categorized by definition since they always, as I have explained, depend on instructions directing the user to the link destination. By indicating works, pages, lines, footnotes, chapters, etc., these instructions inevitably give more or less information on the link destinations.

On web pages, one method of creating categorized links is to make a comment describing the link destination appear in a small textual window when the cursor is placed over the anchor. Another method is the use of icons giving information on the destination. For example, on one of the many Sherlock Holmes web pages, namely Michel Sherman’s page "221B Baker Street: Sherlock Holmes," a small Palm pilot icon signals that the link leads to a Palm DOC format version of a certain text whereas a tiny Acrobat icon tells the user that the link leads to a pdf file.¹⁷⁵

Links may also be categorized by means of different link cues, and a certain color, for instance, may indicate a specific link type. An example of categorization of links by the use of colors is found in Adrian Miles’ web published essay "Hypertext Syntagmas: Cinematic Narration with Links," in which the links are categorized according to the following color system: "Once within the essay, links to the canonical text are blue, links to quotations are green, links to additional commentary are red and links to the references page are black."¹⁷⁶ This is an example of categorization created for one specific work. In fact, explanations like the one quoted above are very often an indication of a work-specific categorization system. Established and traditional categorized links (such as footnote numbers, page numbers in indexes, etc.), on the other hand, need no clarification. As in Miles’ essay, individual categorizations are often additions to these categorization systems that could almost be described as universal.

Ancoral Text

Link sources and link destinations may contain one or several anchors. The terms used here to designate the number of anchors are uniancoral and multiancoral, where the former indicates that there is only one anchor and the latter that there are several.¹⁷⁷ Consider, for example, a paragraph in an academic essay that ends with a small footnote number in superscript. The link source, that is the paragraph including the footnote number, is uniancoral since the superscript footnote number is the only anchor. On the other hand, a footnote containing, for example, one or several references to other works and parts of works is a multiancoral link source since there are links from several anchors.

Another example of a multiancoral link source is the content space named [title page] in Shelley Jackson’s digital hypertext Patchwork Girl. It reads:

 
PATCHWORK GIRL;
OR,
A MODERN MONSTER
BY MARY/SHELLEY, & HERSELF
a graveyard,
a journal,
a quilt,
a story,
& broken accent
(sources)

In this link source there are six anchors, one in each of the following clickable words: "graveyard", "journal", "quilt", "story", "broken accent" and "sources". As already mentioned, the number of anchors in each of the content spaces in this particular hypertext may easily be checked – when pressing the control key, a red box appears around each anchor. There are several link sources of this kind in Patchwork Girl. In the content space [female trouble], for instance, there are two anchors, one of them surprisingly long: "I have a crazy wish! I wish that I had cut off a part of me, something Percy would not miss, but something dear to me, and given it to be a part of her. I would live on in her, and she would know me as I know myself."¹⁷⁸

Above, link sources have been described as uniancoral or multiancoral. Although the majority of links do depart from one anchor within a link source, links sometimes lead from an entire link source. This might be the case with mainly spatial link sources, like a photo or a picture. Link sources of this kind, and destinations, for that matter, are described as omniancoral since the link uses an omnianchor.¹⁷⁹ An omnianchor may coexist with other anchors. If that is the case, the source (or destination) cannot be characterized as omniancoral, but must be described as multiancoral. An example would be a picture from which it is possible to follow a link by pressing "enter" (omnianchor is used), but where parts of the picture are also clickable. The term omniancoral must not be confounded with the term nonancoral designating the complete absence of anchors and, consequently, of links. In other words, a nonancoral text is a text without links. For obvious reasons, nonancoral texts are not particularly interesting in the present discussion on link sources and link destinations, that is ancoral text.

Only mainly spatial link sources and link destinations may have omnianchors. This is because mainly temporal sources and destinations always have a beginning and an end, which serve as anchors. Anchors of this type are somewhat similar to default anchors (cf. below), and they are employed when no other anchor is used. This has to do with the convention for reading and using typographic texts. Links from bibliographical references such as "see Charles Dickens, Great Expectations (London: Penguin, 1996), pp. 123–126" and the like are one example. Here the link leads to an anchor situated to the left at the top of page 123.¹⁸⁰ Similarly, a link to "chapter two" in the same novel leads to an anchor placed at the very beginning of the chapter and, in reality, to the words "chapter two".

On the Web, many link destinations are uniancoral or multiancoral. This is because there is usually an anchor at the top of each web page, even when they are mainly spatial. The default anchor, as this particular kind of anchor could be called, is generally activated automatically in case no other anchor is used.¹⁸¹ Default anchors are frequently used in link destinations, and the top of the destination web page is often the part of the page where the user ends up when following a link. For example, when clicking on "Next chapter" in one of the web published versions of Madame Bovary, the user is presented with the beginning, and not the middle, of the page.¹⁸² With large link destinations, like long typographic texts as in the example above, the default anchor is made manifest to the user (by the top left position of the first line displayed) even though she will probably not take any notice of it. The default anchor, however, is also used with smaller link destinations that easily fit into a screen-sized window. But the users notice these even less than the ones in the larger destinations.

The default anchors described above are all for incoming links. Yet, there are also default anchors for outgoing links. A digitally published collection of poems, for example, could be arranged so that one poem at a time appears on the screen and, to shift poems, the user must press "Enter". Although there is no explicit anchor after the poem, one could say that the link leads from a spot at the end of the poem (default anchor) to the top of the page on which the following poem is displayed (default anchor). Naturally, the existence of default anchors does not necessarily mean that there are no other anchors. In Michael Joyce’s hypernovel afternoon, for example, many content spaces contain two default anchors (one at the top and another at the end of the text) and several anchors (clickable words).

Yet what is perhaps the most interesting aspect of anchors in digital link destinations is when anchors other than the default anchors are used.¹⁸³ These anchors, as well as the default anchors, are generally invisible to the users, who only experience the effect of their existence when they are presented with text situated elsewhere than at the top of a page. If, for instance, I want to explain the word "browser" by means of an explanation in "Matisse’s Glossary of Internet Terms," I would not create a link to the default anchor at the beginning of the glossary, but to the anchor bringing the user to the word "browser" and to the comment belonging to it: "A Client program (software) that is used to look at various kinds of Internet resources. See Also: Client, URL, WWW, Netscape, Mosaic, Home Page (or Homepage)."¹⁸⁴

Figure 4. "Matisse’s Glossary of Internet Terms."

The web page "Matisse’s Glossary of Internet Terms" is a multiancoral link destination (and source), but there are also uniancoral digital link destinations. These are web pages with no other anchor than the default anchor at the top of the page. An example of a uniancoral link destination in print is the following footnote: "² For purely practical reasons, I refer to the user (the reader, the listener etc.) as ‘she’ throughout the essay." Another example is when an article in a magazine has been split into two parts, for instance, one on page three and the other on page five. The first half of the article on page three is the link source whereas the second half on page five is the link destination. The link is attached to an anchor, "cont. p. 5" in the source, while the words "cont. from p. 3" serve as the anchor for the back link. The destination may indeed be uniancoral in this case.

"Matisse’s Glossary of Internet Terms" illustrates the relativity of the term ‘"link destination". The entire document with the glossary (the default anchor is used) as well as parts of it (the "word-anchors" are used) may serve as link destination. Here, it is possible to draw a parallel to typographic links where an entire article as well as a section, a paragraph or a single line, etc. within it may serve as link destination. If the entire web page with the glossary of Internet terms is considered a link destination, it is an example of a multiancoral link destination. Likewise, a print article can be considered as a multiancoral link destination since it includes several anchors (footnote numbers, etc.). But also smaller link destinations may be multiancoral; in the example with Matisse’s Glossary, the comment quoted above on the term "browser" could be described as a multiancoral link destination.

To sum up, texts can be ancoral or nonancoral. Furthermore, a distinction can be made between uniancoral, multiancoral and omniancoral link sources and link destinations.

Linkarium, Exlinks and Adlinks

So far, texts or parts of texts have been described primarily either as link sources or link destinations. In reality, though, many link sources also serve as link destinations and vice versa. A practical term for the analysis of ancoral texts is therefore linkarium, which denotes a specific text or part of a text that is ancoral, i.e. that links lead to and/or from.¹⁸⁵ Linkarium is an analytical tool and the concept refers to the textual section, or selection, that one has chosen to study linkologically. For natural reasons, a linkarium often coincides with content spaces (editorial, authorial or other) or other divisions of works (in chapters, verses etc.). But in theory, ten footnotes, ten lines or ten chapters of a text may as well be considered linkaria. Thus a linkarium can serve as link source and/or link destination in its entirety and/or in parts, by which follows that linkaria may include linkaria. As a consequence, the web page "Matisse’s Glossary of Internet Terms" is a linkarium, but each of the explanations in the glossary could also be considered individual linkaria and analyzed as such.

Naturally, the terms uniancoral, multiancoral and omniancoral can be applied to linkaria and it is possible to distinguish between uniancoral linkaria, multiancoral linkaria and omniancoral linkaria (only an omnianchor). (Nonancoral texts are of course not linkaria.) In addition to the aspect of anchor, it is also necessary to account for links leading from or/and to a linkarium. In the following, links leading from a linkarium are referred to as exlinks, while incoming links are labeled adlinks. It is important to remember that the number of exlinks and/or adlinks to a linkarium does not necessarily correspond to the number of anchors. The terms used here to indicate the existence of exlinks is exiteral, where "ex" (Lat. out of, from) has to do with the direction of the link, whereas "iter" is the Latin word for journey, or way. Furthermore, by adding "uni" (Lat. unus = one) or "multi" (Lat. multus = many), the number of exlinks can be given. Thus the uniexiteral linkarium has only one exlink while the multiexiteral linkarium has several.

By simply replacing the part of the words specifying the direction of the link, that is "ex" with "ad" (Lat. toward, to), the terms will instead concern adlinks: aditeral signals that there are adlinks, uniaditeral that there is one incoming link, and multiaditeral that there are several incoming links. It might be possible to establish the exact number of adlinks by examining links within a single work or a selection of works. However, the task is almost insuperable when dealing, for instance with a considerable book collection or the World Wide Web. The reason is that it is often impossible to track down or even to notice a link the back way. For instance, how many links have their destination in Madame Bovary, Collection folio (Paris: Editions Gallimard, 1972), page 23? On the Web, search engines provide means of searching for web pages linked to a certain web page, i.e. adlinks. This gives an idea of the number of adlinks to a certain page, but the result must be considered as inexact and approximate. It should also be remembered that adlinks listed by the search engines are exclusively those leading from other web pages. This means that adlinks from, for example, an article in print are not included. What is interesting regarding the listing of adlinks in print, are the so-called citation indexes, which list not only exlinks, but also adlinks leading to a work, from essays, dissertations, etc. published in certain journals.

In contrast to link sources, which can only be described as uniexiteral or multiexiteral, and to link destinations, which can be either uniaditeral or multiaditeral, linkaria may be characterized as uniexiteral or multiexiteral and uniaditeral or multiaditeral. However, seeing that a linkarium is not necessarily both exiteral and aditeral, terms describing the absence of either exlinks or adlinks are needed. When there are no exlinks from a linkarium, I choose to call it nonexiteral and, in analogy, nonaditeral if there are no adlinks. A web page within a web site that has adlinks but yet no links to other pages, or a dead end in a digital hypernovel, are examples of nonexiteral linkaria. Tables of contents, on the other hand, are often nonaditeral since there are usually links leading from them but not to them.

There are nine possible constellations when combining these six terms to describe the number of exlinks and adlinks, or absence of such. However, the combination of nonaditeral and nonexiteral is theoretically impossible since this would imply that the linkarium has no links and would therefore, per definition, not be a linkarium. In the figure below, the terms and the different types of linkaria are schematically depicted (three arrows indicate multiple links and not specifically three links):

Figure 5. Types of linkaria.

A uniaditeral multiexiteral linkarium could for example be a footnote of the following type: "¹⁷ Cf. Smith 1998, p. 166 and Johnson 1842, pp. 69-125." It is uniaditeral because there is one internal link (from the footnote number in the main text) leading to the linkarium, and multiexiteral because there are two external links leading from the bibliographical references, in addition to which there is an internal back link via the footnote number (from the footnote number to the main text).

The figure above, on the other hand, is an example of a multiaditeral nonexiteral linkarium since it is referred to several times in this study, i.e. there are several internal links leading to the linkarium but no links leading from it. The footnote to which the link from the anchor at the end of this paragraph leads is a uniaditeral uniexiteral linkarium since the link is bidirectional and there are no other links from the footnote.¹⁸⁶

Linkaria may be characterized either as uniancoral, multiancoral or omniancoral.¹⁸⁷ Labeling a linkarium uniancoral, multiancoral or omniancoral might sometimes seem sufficient. However, considering the fact that linkaria often are both aditeral and exiteral, it is useful, and sometimes necessary, to distinguish between aditeral anchors and exiteral anchors. It might also be interesting to indicate the number of anchors that links lead to (aditeral anchor) and from (exiteral anchor).

In order to include information as to whether the anchor is exiteral, aditeral or both, the letters "e" (exiteral) and "a" (aditeral) are added as a subscript to "A" (anchor). Hence, A_a indicates an aditeral anchor whereas A_e indicates an exiteral anchor. In combination, the A_a/e indication depicts an anchor that is both aditeral and exiteral.¹⁸⁸ Moreover, linkaria may, as already mentioned, have, or be, what could be called an omnianchor, {A}. This type of anchor may also, of course be aditeral, exiteral, or both: {A}_a, {A}_e and {A}_a/e.

A more precise identification of anchors in a linkarium is therefore to simply indicate the type of anchor (A_e, A_a, A_a/e,{A}_a, etc.), and (if not an omnianchor) the number of each type (¹A_e, ²A_a, ⁵A_a/e, etc.). For instance, a linkarium with five aditeral anchors, eight exiteral anchors and two aditeral-exiteral anchors would be depicted as follows¹⁸⁹:

Figure 6. A multiaditeral multiexiteral linkarium.

However, two issues remain unaccounted for in this figure: the number, and the character of the links leading to and/or from the individual anchors (and thus to and from the linkarium). Regarding the number of links, the terminology used to describe whether there is one, several or no adlinks and/or exlinks leading to/from a linkarium may be used.¹⁹⁰ For example, a traditional footnote number in print generally has one incoming link and one outgoing link and would therefore be characterized as a uniaditeral uniexiteral anchor. Clickable words on a web page are often nonaditeral uniexiteral anchors since there is no adlink explicitly attached to the anchor and seeing that there is only one exlink. Generally, default anchors at the top of web pages are multiaditeral, because their primary task is to serve as link dock for incoming links, and nonexiteral, because they lack exlinks.

A Method for the Description of Links

The schematized figure of a linkarium and its anchors above gives information on the absence of adlinks (nonaditeral) and exlinks (nonexiteral) to anchors but not on the number of links that lead to/from each anchor. It is evident, for example, that the five exiteral anchors (⁵A_e) are nonaditeral since they would otherwise be indicated as A_a/e instead of A_e. Whether these five nonaditeral anchors are uniexiteral or multiexiteral, however, cannot be concluded from this figure. In order to describe the different exlinks, the five anchors must be analysed separately. Only then is it possible to show that three of the anchors are nonaditeral uniexiteral and two are nonaditeral multiexiteral.

In a detailed analysis of anchors and links, it is useful not only to characterize anchors as uniaditeral/multiaditeral and/or uniexiteral /multiexiteral, but also to indicate the exact number of adlinks and exlinks. Hereafter, the number of links will be marked in Roman numerals after the subscript letters "a" and/or "e": A_eI, A_aVI etc. An identification number is also necessary to facilitate discussions by individualizing structurally identical anchors. This ID number is placed in a subscript position before the A. The designations of three uniexiteral anchors (³A_eI) would then be the following: ₁A_eI; ₂A_eI ; ₃A_eI. With anchors that are both aditeral and exiteral (A_a/e), the ID number and the number of links are indicated similarly: ₁A_aX/eI ; ₂A_aX/eI etc.

In order to describe the character of the links, the types of links and linking must be added to the description of anchors and linkaria. As in the previous schematizations, a link is depicted as an arrow. The different link characteristics are indicated by means of abbreviations, symbols, or a combination of these two. Abbreviations are used for the following concept pairs: digital (dig.) – analog (ana.), internal (int.) – external (ext.), and categorized (cat.) – uncategorized (uncat.) The unbroken line symbolizes a visible link, the dashed line a hidden link and the dotted line an invisible link. Furthermore, the conditional link is marked with the letter "c" whereas the unconditional link lacks this "c".

Finally, as for unidirectional and bidirectional linking, links are depicted as single directed arrows if unidirectional. As discussed above, two links are involved in bidirectional linking. They may either use different anchors in the linkaria (heteroancoral bidirectional linking) or the same anchor (homoancoral bidirectional linking). Homoancoral bidirectional linking, i.e. bidirectional linking where the forward link and the back link use the same anchor in the linkarium, is indicated as follows: "<═>". The use of two lines in the symbol shows that there are two links involved, making it possible to characterize them separately (one link may, for example, be digital and the other analog, typographic.) The upper line symbolizes the back link whereas the line below symbolizes the forward link. Heteroancoral linking is depicted with two arrows, one indicating which anchor in the linkarium that the forward link uses, and the other to which anchor the back link is leading. Because the attention, when analyzing individual linkaria, is focused on that particular linkarium, unidirectional and bidirectional linking will only be discussed in relation to exlinks. As a result, adlinks are always described as unidirectional in descriptions of individual linkaria. Moreover, it should be noted that if a link runs within a linkarium, the anchor from which, as well as to which, the link leads is indicated.¹⁹¹

Thus the symbols that are to be used in schematical descriptions of linkaria are the following:

Anchor¹⁹²

A = anchor
DA = default anchor
{A} = omnianchor

Links and linking

Figure 7. Symbols for schematic descriptions of linkaria.

As an example of how these symbols can be combined and used in descriptions of linkaria, we could, for instance, focus our attention on a hypothetical multiancoral linkarium with seven anchors and, in all, sixteen adlinks and exlinks. Seven of the sixteen links are adlinks, while nine are exlinks. One anchor is an aditeral default anchor with four adlinks (DA_aIV): two unidirectional, visible, unconditional, digital, internal, and categorized links, one unidirectional, invisible, conditional, digital, internal, and uncategorized link, and one unidirectional, hidden, unconditional, digital, internal, and uncategorized link. One anchor is uniaditeral (A_aI) and has one unidirectional, hidden, conditional, digital, internal, and uncategorized link.

Three anchors are exiteral. Two of these have three exlinks (₁A_eIII; ₂A_eIII), while one of them has only one exlink (A_eI). One of the anchors with three exlinks has two unidirectional, hidden, conditional, digital, external, and categorized links. The third link from this anchor is involved in heteroancoral bidirectional linking. The forward link that leads from the anchor in this linkarium, is hidden, conditional, digital, internal, and categorized, while the back link is hidden, unconditional, digital, internal, and uncategorized (cf. adlink to DA). The other exiteral anchor with three exlinks has three unidirectional, visible, unconditional, analog (typographic), external, and categorized links. The single link from the uniexiteral anchor is unidirectional and internal (within the linkarium). Furthermore, the link is visible, unconditional, digital, and categorized. Note also that this exlink is an adlink of the default anchor. This is an example of how large linkaria include smaller ones.

Two anchors are both aditeral and exiteral. As they have one adlink and one exlink each, they are to be characterized as uniaditeral uniexiteral (₁A_aI/eI; ₂A_aI/eI). In both anchors the links are involved in homoancoral bidirectional linking. In one case, the forward link, as well as the back link, are visible, unconditional, digital, internal and categorized. The links to/from the other uniaditeral uniexiteral anchor differ from one another: the forward link is visible, unconditional, digital, internal, and categorized, while the back link is hidden, conditional, digital, internal, and categorized.

Schematically, the multiancoral linkarium that I have described could be illustrated in the following way:

Figure 8. Schematic description of a multiancoral linkarium.

Examples

Finally, three authentic linkaria will be analyzed. The first one is the second footnote of chapter four in Robert L. Pattens’ Charles Dickens and his Publishers:¹⁹³

Two major areas of friction can be distinguished. First, Dickens found Bentley’s interference in the editorial policy of the Miscellany annoying. Dickens’s duties as editor were defined by the second of nine Agreements with Bentley² made on 4 November 1836. . . .

²The agreements are printed in P [The Pilgrim Edition of The letters of Charles Dickens, ed. Madeline House, Graham Storey, and (vol. iii) Kathleen Tillotson (Oxford, Clarendon Press), 1965-.] i. 648-51, 654-5, 662-4, 666-80; ii. 471-5.

This linkarium is uniaditeral multiexiteral since it only has one adlink, the one that comes from ‘²‘ in the main text to ‘²‘ in the footnote, but several exlinks (one from each of the bibliographical references and one from the footnote number.) The linkarium is also multiancoral since there are six anchors in all (the footnote number + each bibliographical reference). One anchor is both aditeral and exiteral (the footnote number) while five are exclusively exiteral (the bibliographical references). Thus the footnote can be described as uniaditeral multiexiteral multiancoral [¹A_a/e ; ⁵A_e].

Additionally, considering the exact number of links, the aditeral exiteral anchor, A_a/e, may be characterized as uniaditeral uniexiteral, A_aI/eI, seeing that there is one link leading to it and another one leading from it. With only one exlink, the five exiteral anchors are all nonaditeral uniexiteral, A_eI. The links that lead from the nonaditeral uniexiteral anchors are all typographic (not clickable), external (the link destination is in another work), categorized (there is certain information provided on the link destination), visible, unconditional, and unidirectional (there is no link leading back from the link destination). The link leading from the uniaditeral uniexiteral anchor is also typographic, categorized, visible, and unconditional but internal (the link destination is within the work) and bidirectionally linked. Seeing that the forward link (the exlink) and the back link (the adlink) are attached to the same anchor, the linking is homoancoral. In fact, since the forward link and the back link also have the same anchor in the main text, this could be described as a bidirectional link (as the linking is homoancoral/homoancoral.)¹⁹⁴ Schematically, the footnote can be depicted as follows:

Figure 9. Schematic description of the second footnote of chapter four in Robert L. Pattens’ Charles Dickens and his Publishers.

The second linkarium to be analyzed is a web page titled "Mrs. Watson’s Baker Street."¹⁹⁵ The page pictures an open fireplace functioning as the table of contents of a fairly small web site on Sherlock Holmes. On the mantelpiece, there are clickable items (books, a candle, a note, a photograph, a syringe, and a globe) that may bring the user to other pages of the site. There is also a clickable spider web in one corner of the fireplace.

Figure 10. "Mrs. Watson’s Baker Street."

The link from the globe leads to the page "Mrs. Watson’s Baker Street: An Atlas," on which a silhouette of Sherlock, placed in the upper left corner, serves as the anchor of a link leading back to the index page. This silhouette, indicating a link to the index page, is found on all the web pages within the site. Thus, after a few clicks around the site, it is possible to characterize the web page above as a multiaditeral multiexiteral and multiancoral linkarium.

The multiancoral linkarium contains eight anchors – the objects and a default anchor at the top of the page. The seven "object-anchors" so to speak, are nonaditeral uniexiteral (⁷A_eI) since all the adlinks to the linkarium lead to the default anchor. The default anchor is multiaditeral nonexiteral and has no less than ten adlinks (DA_aX).¹⁹⁶ These adlinks are all unconditional, digital, internal, categorized (the silhouette of Sherlock gives an idea of the link destination) and hidden (it is not possible to tell whether the silhouette is clickable just by looking at the page.) The links from the nonaditeral uniexiteral anchors are similar to the adlinks. However, because they are exlinks, it must also be indicated whether they are unidirectional links or involved in bidirectional linking. As a matter of fact, in five cases there are links in the destinations leading back to the default anchor in the link source (heteroancoral bidirectional linking.) These links are the ones that lead from the books (₁A_eI), from the photograph (₄A_eI), from the globe (₅A_eI), from the syringe (₆A_eI), and from the spider web (₇A_eI). On the other hand, the links from the candle, ₂A_eI, and from the note, ₃A_eI, are unidirectional since their destinations (which are JavaScripts) are not provided with back links. Schematically, the web page "Mrs. Watson’s Baker Street" would be described in the following way:

Figure 11. Schematic description of the web page "Mrs. Watson’s Baker Street."

Lastly, a linkarium in a computer game will be examined. The linkarium in question is from a children’s computer game named Robinson in which one of the tasks is to find useful things on a shipwreck.¹⁹⁷ As soon as the task of building a raft has been accomplished the link to the shipwreck will be activated and Robinson will suddenly appear inside the ship in some sort of hallway. There are several rooms to investigate and five items to be collected before Robinson may go back to the shore. If the user tries to make Robinson leave the ship without all the five things in his pockets, the link is unavailable and Robinson says: "– There are still things to be found."

There is only one way into the ship and only one way out, which implies that the linkarium is uniaditeral uniexiteral. Seeing that Robinson leaves the ship from the same spot to which he arrived, the linkarium could be described as uniancoral. The designation of the only anchor in the linkarium is A_aI/eI. The adlink and the exlink are unidirectional, hidden, conditional, digital, internal, and categorized. The adlink is conditional because the user must first find a way to build the raft and then to make it possible for Robinson to sail on it. The exlink is conditional since it requires that Robinson has collected five things. Described schematically, the shipwreck linkarium in Robinson looks as follows:

Figure 12. Schematic description of the shipwreck linkarium in the game Robinson.

[To the top]

8. Closing Words

This study, which aims to clarify and develop some theoretical standpoints, has sought to provide a framework for descriptions and analyses of the way in which texts and works are stored, presented to the user, and navigated. Storage, presentation and reproduction of works are some aspects discussed. The structure of works and texts, as well as the navigation related to these structures, has also been examined. Furthermore, the study is concluded with a comprehensive discussion on links and linking.

The framework presented in this article aims to facilitate broader analyses of works and their manifestations. Apart from dealing with the purely artistic devices, these analyses also embrace what could be called media structures. What is now needed are in-depth analyses of individual works of art, in which studies of these media structures are combined with explorations of their narrative structures. I hope to contribute to the forming and performing of such analyses.

[To the top]

About the Author

Anna Gunder is a doctoral student at the Department of Literature at Uppsala University. This article is part of her doctoral dissertation Hyperlitterärt berättande – narrativa strukturer och ergodicitet i ett antal skönlitterära hyperverk [Hyperliterary Narration: Narrative Structures and Ergodics in some Hyperworks] (prel. title). Part of Anna’s dissertation is also the previously published article "Berättelsens Spel" ["The Game of Narration"], dealing with the narrative technique in Michael Joyce’s digital hypernovel afternoon, a story (Human IT 3/1999). Anna is writing her dissertation within the research project "IT, Narrative Fiction, and the Literary System" run by the Section for the Sociology of Literature at Uppsala University. The research project is funded by The Axel and Margaret Ax:son Johnson Foundation. Professor Johan Svedjedal is the project leader.
E-mail: anna.gunder@telia.com

[To the top]

Notes

[To the top]
[To part 1]
[To references]