Reading Hypertext and the Experience of Literature

1 Reading and hypertext

Over the last ten years we have frequently heard the claim that hypertext offers a challenge to existing practices of reading. In particular, hypertext has been used to polarize differences between two kinds of reading: a constrained, linear form determined by the nature of print text, and a decentered, participatory form supposed to be liberated by hypertext. The repressive effect of the book is based on its tangible appearance, as a recent commentator explains: "the physical, stable presence of the text works to deny the intangible, psychological text the reader attempts to construct" (Johnson-Eilola 1997, p. 145). As a result, "books are machines for transmitting authority and disseminations of cultural capital" (p. 136). In contrast, the standard vision of hypertext is that it "obviously creates empowered readers, ones who have more power relative both to the texts they read and to the authors of these texts." Hypertext increases individual freedom because "users are entirely free to follow links wherever they please" (Landow 1997, p. 273).

In other words, the book is dead or dying; hypertext and hypermedia are ensuring fundamental changes in reading and writing. Similarly, radical changes are said to be in prospect for learning: the introduction of the computer will force teachers to rethink their practices, while students will be empowered to learn in new ways (e.g. Landow 1997, pp. 232, 227). Although this is an attractive picture, we will argue in this paper that in other ways it is also misleading. Apart from the wider issue that for many years some teachers and students have been shifting to inquiry-based learning without being driven by the technological imperative (e.g. Boud, 1988), the embrace of hypertext for literature is possible only for those who have paid little attention to the nature of reading. So the issue should perhaps be framed differently. Given what we know about reading and writing, and the psychological processes that support them, how effectively does hypertext facilitate or extend those processes? To what extent does hypertext change the nature of reading, or promote some component process to a more prominent role?

Such questions, however, are not legitimate in the view of hypertext theorists such as Landow. For them, the textual medium determines the nature of response. Not only is the concrete form of the book supposed to drive how we read it; so too the features of hypertext are said to drive its function. To understand hypertext fiction, says Landow, "involves deducing its qualities from the defining characteristics of hypertext" (Landow, p. 183). Similarly, Moulthrop (1993) points to what he calls the hypotext, the underlying structures and specifications of a hypertext: this part, he says, is "arguably the most important" (p. 86). Douglas (2000) refers to the "intentional network" of codes and scripts with which a reader must come to terms, or be frozen out of the reading (p. 133). In other words, the mechanism of hypertext determines reading, rather than the content -- a view which, if true, will profoundly alter the relation of readers to literary texts (Miall 1998). As we show later, our own empirical studies indicate the reality of this alteration.

The structural differences said to exist between book and hypertext have led to a more general claim: a hierarchical model of text deriving from the prestructured nature of the book has been opposed to the so-called topographical model found in hypertext. Since hypertext is non-linear, says Bolter (1992), "In place of hierarchy, we have a writing that is not only topical: we might also call it 'topographic.' . . . Electronic writing is both a visual and verbal description. It is not the writing of a place, but rather a writing with places, spatially realized topics" (p. 25). In this view, compared with the book, hypertext more naturally embraces graphic representations, such as a tree or network diagram, or an image map, and can make them available to interactive linking just like a passage of text. Thus hypertext advocates are drawn to promote the visual over the verbal or abstract order of the book. In fact, the underlying structure driving a hypertext may exist literally as a map: Jackson's (1995) Patchwork Girl provides one example.

The question, then, is how these claims fare in view of what we know about reading. We will first assess the claim that hypertext is a topographical medium; then consider the rhetoric of empowerment in the light of current hypertext design, particularly the role of links in hypertext fiction. The course of this discussion will largely be critical: we will draw attention to discontinuities between hypertext models of reading and much previous understanding of reading. We then present the findings of our own empirical study of readers of literary hypertexts, which, among other things, raise questions about the role of both the topographical and the linking components of hypertext. At the same time, our discussion should not be construed as a dismissal of hypertext as a tool for reading and learning. Our aim, rather, is to show that some current claims about hypertext appear to be misleading. In particular we challenge the claim that now we have hypertext we must accept that the mode of reading appropriate for the printed text is constrained and outmoded. What hypertext is good for is another issue that we will not attempt to consider in this paper (c.f. Miall 1999).

The emphasis on topography in hypertext, or, in Bolter's words, a writing with spaces, consists of two separate claims: first, it proposes the equivalence of visual and textual information as screen images; second, it requires the connection of separate images within a linked structure available to the reader. The first of these claims is the less familiar one, and here we examine it critically, particularly in the writing of Lanham (1993).

2 Thinking with images

Imagery, or the iconic, forms a constant thread in Lanham's presentation, as it does in the writing of several other authors such as Bolter (1992) or Tolva (1996). Lanham suggests that electronic writing changes the balance between alphabet and icon. Whereas in print culture the icon was suppressed, computer-based writing allows us to restore it to the position it held in pre-print texts, such as illuminated manuscripts (or in the work of writers such as Blake or William Morris, who are said to be rare counter-examples from the age of print). While the icon certainly has been repressed, the case is not as general as Lanham supposes. Although for several centuries texts intended for adults have usually been devoid of icons, this overlooks the great flowering of illustrated topographical books from the late eighteenth and early nineteenth centuries; nor has it ever been true of children's texts. It seems likely that iconic texts have come to be associated in western culture with illiteracy or with early literacy. We tolerate children (readers in training) indulging in icons, but we feel compelled to wean adolescents, since one sign of a good education is the ability to read a text without pictures.

However, Lanham (1993) goes on to suggest that the absence of the icon is responsible for the invention of a type of rhetoric:

When the rich vocal and gestural language of oral rhetoric was constricted into writing and then print, the effort to preserve it was concentrated into something classical rhetoricians called ecphrasis, dynamic speaking-pictures in words. Through the infinite resources of digital image recall and manipulation, ecphrasis is once again coming into its own, and the pictures and sounds suppressed into verbal rhetorical figures are now reassuming their native places in the human sensorium. (p. 34)

Thus, says Lanham, "The graphical and typographical tricks to which the electronic surface lends itself make us self-conscious again about our own apparatus of vision." Print has impoverished our understanding, inviting us to look through it to the conceptual universe it addresses, but electronic text with its visual tricks "makes us self-conscious about the bag of neural tricks that create our own vision . . . The perceptual field of the 'reader' becomes considerably richer and more complex in electronic display" (p. 73). As we will show, our empirical study seems to indicate just the reverse, suggesting that hypertext degrades the quality of readers' engagement during reading.

The shift towards the visual is part of a larger argument that Lanham wishes to make about the cultural forms of the future. Based on his conception of rhetoric, and his suggestion that electronic media mark a return to a rhetorical framework of communication and understanding that print culture suppressed, Lanham sees the "art of persuasion" dominating the information economy now emerging. Thus, "Whatever we choose to call it ... the construction and allocation of attention-structures will be a vital activity in our information society" (p. 227; for a recent statement, see Lanham 2000). Attention-shaping techniques are already apparent, for example, in Futurism or, more recently, in Pop Art. As Lanham argues, "Robert Irwin's minimalist painting and environments are all calculated to bring human visual attention to acute self-consciousness" (p. 228).

The changes in contemporary media towards the visual mode of attention that Lanham points out are undeniable. But the inferences he draws, especially as he attempts to reconfigure our educational curricula to bring them into line, are troubling. His arguments can be criticised on two main grounds: first, that he misinterprets the role of self-consciousness, and second, that he fails to understand the role of imagery in reading.

Perhaps the most highly developed technology of imagery, both visual and aural, is to be found in television advertising. When Pop artists such as Warhol or Lichtenstein set out to exploit an icon of advertising, they did so by defamiliarizing it, placing it in unexpected frames or grossly distorting it in size. This, as Lanham remarks correctly, evokes an acute self-consciousness in the viewer (until we become accustomed to these visual tricks, that is). The aim of the native iconic television advertisement, however, is quite different. It does indeed draw upon a bag of visual tricks to obtain and fix the attention of the viewer: currently these include deliberately jerky and degraded film sequences, "in your face" proximity of a target image, rapid jump-cutting, and contexts in which products become figurative. But attention is evoked through these means for a quite different end: not to promote self-consciousness in the viewer, which would usually negate the purpose of the advertisement, but to evoke feelings that alter the self-concept of the viewer in ways that are largely subliminal, out of awareness.

The use of iconic media in advertising is, of course, representative of a larger ideological purpose evident in almost all television programming from the apparently neutral news story to the latest soap episode: the invitation to acquiesce in and identify with the cultural forms that serve the interests of the current commercial world view. What is disturbing about Lanham's conflation of artistic and media iconicism is that it effaces the critical role of the artistic. Moreover, it is print artistry that has until now offered the best prophylactic against the ideological pressures of the market place, precisely because it employs words and avoids explicit images.

Words on the printed page probably remain the most potent resource for critical self-reflection. While this might involve self-consciousness as one phase of a broader response process, self-consciousness as Lanham refers to it is quite different from self-reflection. Lanham biases his understanding of art towards the self-conscious, towards art about art, or to reception processes that theorize art. "The digital computer," says Lanham, "seems a machine created for Art-about-Art" (p. 46). As Rau (2000) has put it, such an approach imposes "the idea of hyperfiction as embodiment of literary theories of the active reader." We would oppose this self-reflection, a process in which a reader experiences a potentially self-transforming interaction with a literary text. This seems to require alternating phases in response, as Oatley (in press) has suggested: "We cannot always be moved and think about something in an observational way at the same time. What we can more often do is to move in and out along the continuum of emotional distance, be fully engaged emotionally at one moment, and then in the glow of that emotion, think about the experience in a more distanced way." This view also challenges the notion that our usual mode of fictional reading is that of complete absorption, or a trance-like state (e.g. Birkerts 1994). During this process the commitments of the self can be reorganised: away, perhaps, from the absorption with the narcissistic emphasis on sexual potency or social rank that advertising culture promotes, towards a broader and more resilient sense of the self and its inherent values.

In arguing for the image in reconfiguring literary texts for the computer, Lanham promotes a fundamental shift in the way this new medium will transmit texts for future generations. Just as the experience of seeing a film made from a novel that we have previously read is nearly always slightly disappointing, because the director's images (however excellent in themselves) fail to coincide with our own as readers, so the computer imagery that Lanham envisages will displace the imagery we normally create during reading. It is this imagery that forms an important component of the self-reflective process. Reading (say) a novel is a highly constructive process: as we encounter each character, each new setting, our own resources of imagery are drawn upon to give inner reality to the unfolding story and its feelings and values. Our images are not neutral. We may surmise that they are selected during the response process, with no intervention of the conscious will, because they symbolize our most personal, most deeply-felt values. Most of the time during reading we are hardly aware of this process, yet it is what gives us that sense of engagement during reading, the feeling that something of individual significance is at stake.

Having exposed our individual images to the powers of the literary text, however, the devices of the text can in turn act on those images to recontextualize them. The values they hold for us can be modified or extended in various ways as we watch a particular character's story unfold. We can come to understand better our personal value system and the feelings through which it is expressed and acted on. This is perhaps the most important process that literary texts can perform, and it is critically dependent on making available our own, personal images.

The importance we are attributing to images is supported by several empirical studies of reading. Perhaps the most interesting finding is that images in response to a literary text appear to be similar to those for personal experience. In a study of image production during different kinds of reading, László (1990) found that images evoked by a literary text paralleled experiential images in quality, whereas images generated in response to a newspaper article seemed more similar to those for social categories. Readers' descriptions of their images showed that literary and real life images contained more attributes and "physical-perceptual" components than images generated for categories or reports � in other words, they were both more detailed and richer in sensory aspects. In addition, as measured by reaction time, images for literary and real life experiences were formed significantly more quickly than those for reported characters and situations.

So, to return to hypertext, we suggest that Lanham's system for making literary reading into a multimedia exercise would largely disable this personal level of response. In addition, it seems liable to replicate in a new and dangerously attractive form the imperatives of the standard textbook approach to literary reading in which meaning is predetermined. Multimedia, especially through the use of another's imagery, is liable to impose a set of limited, standardized meanings on a literary text. Far from democratizing literary reading, then, Lanham's proposed system tyrannizes over it, doing so the more insidiously because the visual medium itself has a powerful attractiveness (unlike the conventional school textbooks derided by Lanham, p. 9). It makes gratifyingly immediate what for the book reader forms the beginning of a long-term evolutionary process as a given image grows or is modified over time (including those times when we are not reading: c.f. Birkerts's "shadow-life of reading": 1994, p. 95).

Lanham thus promotes the image to an inappropriate position in the world of the text. This, of course, is nothing new except for its location in a new medium. It echoes an argument that occurred across the eighteenth century and into the nineteenth. Edmund Burke in 1757 objected to theories of poetry influenced by Horace, who claimed that the best poetry was like painting. Burke (1990/1757) argued that the clearness of an image militated against the passion that words can arouse, and that "so far is a clearness of imagery from being absolutely necessary to an influence upon the passions, that they [the passions] may be considerably operated upon without presenting any image at all, by certain sounds adapted to that purpose" (p. 56). Thus external imagery would interfere with the feelings evoked by literature. Similarly, both Coleridge and Wordsworth objected in strenuous terms to the contemporary fashion for the picturesque, what Coleridge referred to as "the despotism of the eye" (1983/1817, I. 107; c.f. Wordsworth, The Prelude, xii. 127-31). Coleridge elevated the distinction into a principle: we should, he says, reserve images for those objects that admit of clear conceptions, since "deep feeling has a tendency to combine with obscure ideas, in preference to distinct and clear notions" (1969/1808-09, i. 106). He saw deep feeling, in this respect, as the location of our values and as the vehicle of moral growth.

The rhetoric for a computer age that Lanham proposes with its emphasis on the visual thus disregards some long standing principles relating to the limits of imagery; and it invites complicity with prevailing forms of media that are antipathetic to the self-reflective powers of literary response. The computer medium will, as Lanham argues, facilitate some powerfully creative new uses for texts and images. But as a medium for "repurposing" (p. 131) literary texts from the past, Lanham's account only shows that the graphical interface of the computer is poorly designed to represent what readers do. As a tool for studying literary texts the computer may be all that Lanham proposes; but as a tool for the primary act of reading literature Lanham's computer vision seems reductive, even trivializing.

3 Thinking in links

The second implication of the topographical metaphor is the linking of images that is a distinctive property of hypertext. Bolter (1992) points to the key aspect of this:

A hypertext has no canonical order. Every path defines an equally convincing and appropriate reading, and in that simple fact the reader's relationship to the text changes radically. A text as a network has no univocal sense; it is a multiplicity without the imposition of a principle of domination. (p. 25).

It is here that hypertext theory is most problematic. The process of linking was originally conceptualized by Bush (1945) in his proposed memex as a means of registering the connections made by an individual thinker as he or she explored and related concepts during the research process. Bush's underlying premise was that the mind worked associatively. Whether or not this is true, the dominant model of hypertext now rests on claims that it represents the associative nature of all thinking, thus bringing it closer to the working of the mind for the user. As Dryden (1994) puts it: "In its structure of branching links and nodes, hypertext simulates the mind's associative processes, thereby providing an electronic platform for constructing and recording the reader's literate thinking" (p. 285).

Hence Bolter can claim that each possible pathway is an "appropriate reading," and see this as empowering for the reader, a liberation from the linear rigidity of the printed text. In this respect, however, as Rouet and Levonen (1996) suggest, "ideology has often been substituted for scientific investigation" (p. 160) by hypertext theorists. Rather than using what we already know about reading or the mind, "contemporary hypertext thinking," says Dillon (1996), "seems tied to concepts of association and non-linearity of access even as it distorts them" (p. 27).

In itself the concept of linking is not new. It is often pointed out that hypertext users may be accomplishing in a different medium the non-linear reading that we carry out when consulting an encyclopedia or jumping to a footnote (e.g. Johnson-Eilola 1994, p. 201). Cross-referenced or annotated structures of information can be built in a computer environment, but it is far from clear whether readers perceive them or are able to use them in the same way as in a printed text. Readers find it difficult to tell where they are within a group of nodes; they cannot judge whether they have read something essential and tend to give up too early; they find it hard to decide on an appropriate sequence through material (c.f. Charney 1994, p. 249). The physical structure of the printed book, given its familiarity, may give readers looking for information a better framework for a search or learning strategy. It is true that giving readers more familiarity with a particular hypertext allows them to use it more effectively over time, but differences in design principles between hypertexts make it less likely that learning will generalize from one hypertext to another. Learning to navigate around the geography of Brazil doesn't help me when I visit Switzerland.

Charney (1994) points to some of the basic problems of reading hypertext: it imposes a greater demand on short-term or working memory; readers may find navigation becomes arbitrary through a lack of cues to the meaning of links between nodes; and hypertext may disable the reader's existing knowledge about how texts are structured and about different text genres. If we are to offer hypertext presentations for learning, hypertext information cannot simply be mapped on to long-term memory, as Charney points out (p. 243), even if memory does have a comparable network structure -- but this is, in itself, a contentious proposal. Even in the limited world of hypertexts designed for instruction, readers often become disoriented, which suggests that the topographical metaphor so far represents wishful thinking. Hypertext cannot offer to model the reader's mind. The author's associations are not those of the reader. As Dobrin (1994) puts it, in terms of hypertext linking, "The author's conception of the connection's relevance is not the reader's, and the reader gets lost" (p. 310). Association, in any case, will probably not turn out to be a particularly helpful way of construing the mental processes at play during reading or writing. We have, after all, been this way before: eighteenth-century associationism was challenged by Kant and Coleridge; its twentieth-century version in behaviorism has been challenged by several versions of cognitive theory. It hardly seems a promising basis on which to build a theory of hypertext.

The analysis of Charney is limited to an information processing model of reading and learning, the type of reading which has been most extensively studied and theorized. "I focus," she says, on the application of hypertext for readers "who read to learn, to understand and evaluate the ideas and argument of others, to come to realizations about the subject matter," and to integrate their learning with what they already know. But this, according to Charney, is already a process that will "push hypertext to its logical extreme" (p. 241). If so, the prospect of effective hypertext models for literary reading seems far off (c.f. Douglas 1994). Charney's central and compelling point is that hypertext designers currently work largely in ignorance of what is known about the reading process. She shows that according to discourse theorists such as Kintsch and Van Dijk, Meyer, and others, the reading and memory processes engaged during learning or information seeking "are both strongly conservative forces" (p. 259). Hypertext systems that are premised on creative and imaginative linking, the type celebrated by Bolter or Lanham, thus conflict with the familiar, systematic processes of reading, and would appear to be inherently ineffective (p. 259).

4 Reading hypertext fiction: An empirical study

That readers do indeed experience difficulties with hypertext is shown in an empirical study of hypertext readers that we recently conducted. On neither of the grounds we have just discussed, the role of imagery and the linking machinery of hypertext, was hypertext found to be an effective medium for literary reading. In two studies involving 130 readers, we examined the differences between reading short stories on computer when they were in either a structurally linear form or an apparently hypertext form.

In the first of our studies, 70 participants ranging in age between 17 and 28 read a modernist short story, Elizabeth Bowen's "The Demon Lover" (Bowen 1981). The story was presented as a series of 24 separate nodes. Participants in the control group read it in a structurally linear format, activating a "next" link positioned at the bottom of the screen in order to move from node to node (see Figure 1); those in the hypertext group read the same text in a simulated hypertext form, where in each section they were required to choose one of two or three links in order to continue (see Figure 2).

A shaft of refracted daylight now lay across the hall. She stopped dead and stared at the hall table; on this lay a letter addressed to her.  Next

Figure 1. A short segment from "The Demon Lover" shown in linear form

A shaft of refracted daylight now lay across the hall. She stopped dead and stared at the hall table; on this lay a letter addressed to her.

Figure 2. A short segment from "The Demon Lover" shown in hypertext form

Linking words or phrases in the simulation text were selected on the basis of their tendency to promote attention to one of three literary features: plot, character or foregrounding. Plot links mirrored as closely as possible the sort of commands evident in text-based electronic game environments; they were generally concrete nouns or prepositional phrases. Character links were names, personal pronouns, or emotions. Foregrounded links were embedded in phrases containing stylistically remarkable elements. Thus, in the node shown in Figure 2, �refracted daylight� is the foregrounded link, �letter,� the plot link, and �her,� the character link. This linking pattern enabled an examination of how story elements may have motivated link choice; for example, it might be postulated that readers who are moved by imagery would be more inclined to select the link that describes the light playing across the hall, that readers who are story driven might wish to discover the contents of the letter, and that readers who tend to read with empathic attention to character might select �her� in the hopes of learning more about the protagonist.(1)

It is important to emphasize that these links constituted the only difference between the linear and simulation models. The layout and content of the text was otherwise identical in both conditions, for each of the links in the simulation led to the same subsequent node. For this reason, it was at times difficult to select linking words or phrases with obvious semantic or logical connections to the subsequent paragraph, although this is indeed what was attempted. For example, the short node shown in the above figure leads to a longer node detailing the protagonist's initial annoyance that the letter has not been forwarded to her current address. She carries the letter to an upstairs room where she looks over its contents. The passage follows well from both the character and plot links, but readers who choose the foregrounded link expecting further discussion of the �shaft of refracted daylight� may be disappointed given that the disposition of the light in the hallway does not figure largely in the subsequent node. When considered in view of existing hypertexts, however, the fact that all links in the simulation did not have strong semantic or logical connections to the subsequent material is not necessarily a weakness of the design, for it reflects the randomness of many literary examples of the genre, both online and off (e.g. Moulthrop�s Hegirascope [1997], and Malloy and Marshall's [1996] Forward/Anywhere). One other feature of the design was that a back button was not provided. This was to prevent simulation readers from discovering that each link led to the same node.

We examined readers' behaviour by recording reading times per node, link choices, and by inviting readers to comment aloud about their experience of reading at the end of the story. Not surprisingly, we found that the hypertext readers took over four seconds longer per node on average than the linear readers (linear: M = 38.37 secs; hypertext: M = 42.73 secs), a significant difference, t(20) = 3.16, p < . 01. In contrast to the linear readers, the hypertext readers tended to feel either confused during their reading, or to feel that they must have missed something. Here are some typical comments made by readers after completing the hypertext version of "The Demon Lover":

The story was very jumpy. I don't know if that was caused by the hypertext, but I made choices and all of a sudden it wasn't flowing properly, it just kind of jumped to a new idea I didn't really follow. (S124)

It almost seemed like there were bits of information that were missing; but I was able to put the puzzle together, so to speak, was able to figure out what was going on just from the action and the dialogue in the story. (S113)

In all, 75 per cent of the hypertext readers reported varying degrees of difficulty following the narrative. Only 10 per cent of the linear readers made similar complaints.

We also carried out a detailed analysis in which we correlated reading times per segment with the presence or absence of particular story factors, such as stylistic foregrounding, new arguments, and shifts in time, space and causation (based on Miall and Kuiken 1999). Our results suggest that the attention of the hypertext readers was diverted to the surface features of the text, and that their reading patterns became increasingly fragmented as the story progressed (Dobson and Miall 1998).

To test the validity of these initial findings, we replicated the study with a second text, Sean O'Faolain's "The Trout". (O'Faolain 1980). "The Trout" is about half the length of "The Demon Lover" and is also more accessible to modern readers; while Bowen writes with a complexity of style reminiscent of the early Modernist era, O'Faolain's writing is at once lucid and replete with striking imagery. "The Trout" in particular bears the stylistic markers of the fairy tale. By choosing dissimilar texts, we hoped in particular to discern whether the findings of the first study were peculiar to the text being read, or were prompted by the structure of the hypertext itself.

Readers were 60 students who, like participants in the first study, either read the story in linear form separated into nodes or in a simulated hypertext format. Again, we found that readers of the hypertext version of "The Trout" took longer on average per node to read the text (linear: M = 45.13 secs; hypertext: M = 52.92 secs), a significant difference, t(9) = 12.438, p < .001.

Readers' comments about their reading experiences were obtained on tape, as before, and after transcription were subjected to content analysis, resulting in a range of constituents and constituent categories. Comments ranged from interpretive descriptions of the story to the various aspects of reading on computer. Tags reflecting each feature were embedded in the transcripts to enable electronic searching and tabulating using the text analysis software, Concordance. Following is a short segment of a coded simulation protocol:

[R S327] [foregrounding] [interp-thematic] At the beginning when she's running out of the tunnel, and sort of drinking in the sun, it's, it seems kind of like a rush through maybe childhood into adulthood, too. [interp-symbolic] She's coming through this darkness to this brightness. I don't know whether that sort of symbolizes knowledge, or empowerment, or anything, but it seems to tie in. [foregrounding] The phrase, moon mice on the water, I have no idea what exactly that means. [interp-language] I can't figure out how to tie in mice with the moon unless it's something to do with the moon being full of cheese, but I really like that phrase. [defamiliar] It, it, I guess partially because it was something different--I mean, you don't normally connect mice with the moon. That made me stop at it and go, whoa--[readerpace+] which was odd, especially since that was a section where I was speeding right along trying to find out exactly what would happen.

As is evident here, the tags were simply markers within the text. They did not bracket information (i.e., there were no tags demarcating the end of particular features). In some cases they were also applied multiply to a single statement. Two tags, for example, are assigned to the first sentence of this passage: foregrounding and thematic interpretation (interp-thematic). The first marks a direct citation of the following highly foregrounded passage from the story: "she emerged [from the Dark Walk] gasping, clasping her hands, laughing, drinking in the sun" (p. 383); the second marks her interpretation of the protagonist's actions as a rite of passage. The second-to-last feature, defamiliarization (defamiliar), signals that this is a transformative reading moment, a moment during which the reader is prompted to reconceptualize her understanding in response to stylistically foregrounded language.

The frequency with which comments were made by each group of readers was counted. This enabled us to perform analyses of a number of types of comment, showing that on a number of dimensions the experiences of the two groups of readers were significantly different. Differences were tested by Chi-square (categories where expected cell frequencies were less than 5 were omitted from the analysis).(2) The results are shown in Table 1. Numbers shown after each category refer to the frequency of comments by linear and simulation readers, respectively.

Table 1. Frequency of linear and simulation readers' comments classified by group

Group	Constituent categories	Chi-square
Style	imagery (22/14); visual (10/6); foregrounding (4/11); defamiliarization (3/12)	X2 (3) = 11.28, p < .02
Self of reader	identification (23/13) ; reader emotion (19/12); autobiographical, general (16/28); involvement+ (5/6)	X2(3) = 7.60, p < .1
Story	story is confusing (4/11); story fails to flow (2/15); segments of the story appear to be missing (0/13); story, dislike (5/6)	X2(3) = 8.96, p < .05
Computer reading	computer reading, enjoy (11/2); computer inhibiting (12/9); computer reading, dislike (12/21); computer distracting (7/13)	X2 (3) = 10.82, p < .02
Linking	link choice enjoyed (5/6); links promote attention to text (11/2); links promote control (8/13); links distract attention from text (9/8); link choice disliked (7/15)	X2(4) = 10.31, p < .05

Bearing in mind that, apart from the pattern of links in each node, the linear and simulation readers responded to the same narrative, the differences shown in Table 1 are entirely due to the presence or absence of the hypertext format. Interestingly, in view of the claims made by Lanham and Bolter for the visual mode, hypertext readers made significantly fewer comments on the imagery evoked by the fiction. In keeping with our expectations, the hypertext version also tended to evoke only generalized comments of involvement with the story, rather than the more specific and emotionally engaged comments of the linear readers. The hypertext readers also tended to find the story confusing or incomplete, some feeling that they may have missed clicking the links that would have made the story achieve coherence. Evidence for the linking mechanism was ambiguous: hypertext readers made more comments on the mechanics of reading on computer and on being required to operate links, but there was also a greater sense of control through linking among these readers, which we discuss further below. The trends in the data are also indicated in the following figure, which combines several of the subsidiary categories and is organized to show on the left those features mentioned more frequently by linear readers.

Readers often made quite revealing comments about their experiences. As with the hypertext readers in the first study, several simulation readers reported difficulties in following the narrative. Participant S305's comments in particular suggest how hypertext structures may change the way in which readers engage with a story:

Reading this story off the computer was kind of confusing at first because as I went to different screens I realized the story wasn't in order. So I had to get pieces, bits and pieces of information from, in different order. And it was kind of confusing at first, but then I just adapted, and learned just to take the information as it was coming, and then I pieced it all together at the end.

This reader's sense of the story as being "out of order" is undoubtedly a consequence of the hypertext structure. Ultimately, she reported that she enjoyed the "The Trout" not because she found it interesting or emotionally engaging, but simply because she was able to piece together the plot. Her comments about choosing between links demonstrate her reading strategy:

I tried to pick something that would let me know what was going on, and then as I started to kind of get a pattern of what I was picking, I kind of, at the end I knew, like, what word to pick to resolve the problem. Eventually I felt the same as [if] I [were] reading through another normal chronological story.

In total, 55 per cent of the hypertext readers reported that they made link choices in order to advance the story, often observing that they did so because, in one reader's words, they "didn't want it to get, you know, too descriptive." While a story-driven approach (i.e. paying attention to how the plot develops: Vipond and Hunt 1984) is commonly recognized as being prevalent among readers of all ages and particularly among younger readers, it seems likely that in this instance the hypertext format further encouraged this mode of reading by bringing into question the sequence of events. Thus readers who were unsettled by the possibility of multiple plots struggled to engage with what they were actually reading because they were troubled about what they might be missing:

Choosing between the links was also very frustrating . . . because I had this feeling all the way through that there was something else going on, other stories, or other details, or other information that I didn't get to read, and I don't know how that would have . . . affected the story, or my perception of the story. (S327)

This reader, in other words, found that the links failed to correspond with her expectations. Or, put another way, her associative process during reading could not be mapped to the available links.

Readers who enjoyed the process of choosing, on the other hand, did so largely because they thought that choosing gave them control over the plot and increased suspense:

The only thing that I really liked is being able to control the hypertext, like I could choose the link that I was going to, and so I just, I had a sense of . . . control of the situation. You know, it was a, a sense of unknown because I didn't know what was behind the link, but I was controlling which link . . . was most interesting to me. (S304)

Either scenario tends to preclude the more personal level of response that we described earlier. Thus hypertext, as a vehicle for literary reading, seems to distance the text from the reader. If readers report enjoyment, their pleasure shifts to the mechanics of plot, the exercise of the intellect in driving the hypertext machinery to see where it will lead. The absorbed and personal mode of reading seems to be discouraged. Moreover, neither imagery, nor the linking process, appears to have operated among these readers in ways consistent with the hypertext theory discussed above. While this conclusion is based on only a limited and somewhat artificial situation for the reader, it is also clear that for readers of the simulation version, attention was directed towards the machinery of the hypertext and its functions rather than to the experience offered by the story. We can see that as a result these readers largely failed to engage with the literary qualities of the text.

It would be premature to generalize on the basis of our findings, since no other empirical studies like ours on literary hypertext reading have been reported, and even individual readings of hypertext fiction are rare, as Douglas (2000, p. 3) has pointed out (for a recent example see Rau 2000). Moreover, our study depended on adapting for hypertext a story originally designed for the printed page. We believe, however, that the problems we have identified are also likely to occur for readers of native hypertext fictions, since hypertext design disrupts some long-standing and central components of literary reading. The studies to investigate this, however, remain to be carried out.

But one conclusion seems clear from this review of current hypertext theories. Despite the prior work of influential commentators such as Bolter or Landow, hypertext is still at a pre-paradigmatic stage; that is, we have no accepted theoretical framework in which to locate it, and no settled body of knowledge on either the nature of hypertext or its appropriate applications. We suggest that the principle cause of this deficiency is that hypertext has not been modeled on what is known about the process of reading. Hypertext undoubtedly has some important applications in the study of literature and for the provision of text and other resources for study. But as a vehicle for the experience of literary reading itself, hypertext appears to promote processes of attention that inhibit the engagement and absorption that are its most characteristic aspects.

5 Postscript: Literary politics

Lanham, whose work is discussed above, first declared his faith in the new electronic environment in 1989. He argued that the electronic medium in itself "democratizes the world of arts and letters," and that "the political direction of the technological force is strong and unmistakable; value structures, markets ideological as well as financial and theoretical, will be reassessed." Thus, he insists, in teaching students of literature "we must accommodate literary study to the electronic world in which that world will increasingly deal" (1993, p. 23). This seems a compelling argument, but at the present moment the market forces that are now set to take over and control the electronic world seem to have little interest in literary study. Much bigger goals are in prospect: interactive, Internet-based television and other high-resolution digitized forms of entertainment. Despite Lanham's attempts at creating a new electronic rhetoric, the reading of literary texts may simply seem rather dull in competition with the multimedia, virtual reality simulations that are now being rushed to market. How will literature fare in an environment where the computer medium has itself largely been appropriated and developed by commercial interests (Golumbia 1996)?

The key issue here will not be how far literature can be made to dance to the multimedia tune in order to seize the attention of the Internet-surfing audience for electronic entertainment, but how far we can establish the distinctive qualities of the literary experience that makes it a clear and significant alternative to what commercial interests are willing to provide. The really urgent issue facing literary scholarship now is to understand what is at stake in literary experience before Lanham and his colleagues convince us to change it into something else. "I hope," Lanham remarks, that technology will be seen not as "driving us where we don't want to go [but] as an opportunity to go where we have never been, and to do things no one has done before" (p. 26). We would agree, but first we need to know where we are, and on that score the claim of hypertext theorists that conventional reading is restrictive and imprisoning has been a disabling error, forestalling serious attention to the relationship of hypertext and reading.