Miller begins by noting that in the last decade or two there has been a major shift towards cultural studies at the same time that the new electronic media have been changing the nature of scholarship. One sign of the second process, he suggests, is the ease with which writing can now be changed: this "encourages the adept in computer composition to think of what he or she writes as never being in quite finished form" (p. 27). This comment echoes the often repeated claim that hypertext promotes the instability of text. I will suggest that in both of these claims Miller nominates as new processes which have been in view all along, and thus overestimates the meaning of the present turn to hypertext.

The division of the disciplines into carefully compartmentalized departments is a relatively new phenomenon: until the late nineteenth century, for example, psychology was considered a branch of philosophy, a convergence evident in the work of William James; Humphry Davy, the great chemist, saw his study of chemistry as a contribution to the understanding of human nature. From Vico, through Coleridge, to George Steiner, there have been important critics of literature who have also situated their work within an attempt to understand cultural phenomena in the widest sense. The present turn to cultural studies, although it has some distinctive new elements in it contributed in particular by gender and postcolonial theory, is rather a return to an interdisciplinarity taken for granted until recently by the best scholars since the time of Aristotle. While the new cultural focus appears to emancipate us from the narrowing exclusivities of the New Criticism, even this is something of a caricature. Despite their attempt to delimit the grounds of literary studies as a distinct area of scholarship, the New Critics from I. A. Richards to William Wimsatt were quick to draw on knowledge from a range of other disciplines when appropriate, from history to philosophy.

Secondly, the stability of pre-electronic text is also something of an illusion. This can be shown on two rather different grounds. Many authors have treated their texts, whether in print or not, as though they were unstable. Wordsworth, for example, was a notorious reviser. Most of his major poems underwent revisions, sometimes quite extensive, from one published edition to the next. Our tendency to reify the printed text has prevented us from appreciating how mutable much of the literary canon has actually been. If an author had the time, or lived long enough, to see a second edition of a work into print, change has been the rule rather than the exception. In a quite different sense, literary texts -- at the least the most interesting ones -- are also rather unstable for their readers. Although the words on the page may remain the same, responses of the same reader across time with repeated re-readings, or a comparison of different readers' responses to the same text, both show that differences in understanding between readings are once again the rule, not the exception. The implications of this perspective for understanding hypertext are rather significant, and lead in a rather different direction from the argument proposed by Miller.

The reader's freedom to pursue interpretive pathways of her own choosing is practically enforced by many literary texts. To take two obvious examples: Coleridge's Mariner shoots the Albatross without apparent motive. Although no reason is assigned for the act, there appear to be few readers who can rest easy without some attempt at an explanation, whether this is due to Original Sin or the Oedipus Complex, as a survey of the critical literature on the poem will show. Is the governess in James's Turn of the Screw to be praised for her heroism in confronting the ghosts at Bly, or should she be locked up for gross hallucinations? These texts provide only rather dramatic examples of a process that seems likely to occur in response to most literary texts: their structure or style calls for the reader's input if they are to be understood, but how the reader contributes is particularly dependent on her existing beliefs, memories, feelings, or dispositions. The act of reading, in turn, may work to reshape some aspects of the reader's input by qualifying feelings, placing memories within a new perspective, calling beliefs into question. Literary texts matter to us in part because they invite us to reflect on what is most distinctive about our concept or image of ourselves: they speak to what is most personal in us.

In this context, the move to hypertext may not enlarge our freedom as readers. On the contrary, as Miller remarks, thinking of Landow's Intermedia, "The apparent freedom for the student to 'browse' among various hypertext 'links' may hide the imposition of predetermined connections." Hypertext, he adds, may actually be "conservative in its implications" (pp. 27-8). It can also, I would add, distract the reader's attention from that self-reflective, formative dimension of reading: the author's links to a new perspective are unlikely to coincide with what the reader might have been led to consider unaided. While such hypertext linking might provide a valuable resource for study, as Intermedia showed, we should also be careful to distinguish reading for study from reading for aesthetic experience or insight. The freedom of the reader may be more vulnerable in a hypertext environment than in any conceivable print form, partly because the links visible on the screen present such an immediate and tangible invitation to follow another's pattern of suggestion rather than work on clarifying one's own more intuitive promptings.

Despite the observation I have cited, Miller goes on to characterize the issues raised by print vs. electronic text in a somewhat misleading fashion: he does so by consistently focusing on the contexts in which we encounter a literary text instead of considering the different reading processes which the two mediums may invoke. Miller elaborates the circumstances though which a reprint of Ayala's Angel, a Trollope novel that he owns, came into being under the imprimatur of Oxford University Press, with all its connotations of authority and conferral of canonical status. He then compares this to an electronic version of the same novel which comes from the Oxford Text Archive, an electronic text respository; but in this case it comes out of a quite heterogenous collection and with no claim to canonical status, and vulnerable to change in a way that a book can never be. The comparison, however, turns out to be a superficial one.

Miller misleads himself by failing to make adequate distinctions between the functions to which he appeals. First, he confuses the issue of whether a machine can "read" a text in any sense approaching that of human reading. The possibility that "haunts" us, he says, is "that the human brain is no more than an extraordinarily powerful, complex, and compact computer" (p. 30). He notes that the "machine readable" version of the Trollope novel, because it is electronic, "can move so fast it is almost ubiquitous" without pausing to reflect that this has no bearing on the process of human reading. The meanings of the word "read" are quite different. The effect is to confuse information with knowledge at the most basic level. While it is commonplace and uncontroversial to speak of a computer containing information, it is quite another matter to speak of what we gain from literary reading as information. Only the least important aspects of Ayala's Angel are amenable to being stored in the computer or "read" by it, as Van Peer's critique of literary computing showed.

Yet the burden of Miller's discussion focuses not on human acts of reading, but on the different associations of the novel given by its print or electronic versions. While the Oxford printed text, in the World Classics series comes with a specific weight, colour, and smell, and a carefully inscribed point of origin on the title page that establishes its canonical status, the electronic text has no such environment but is surrounded rather with other electronic patterns, such as the internet route by which it was accessed at the Oxford text archive. The "familiar historic embedding," Miller suggests, has disappeared and been replaced by cyberspace (p. 31). Miller points to the differences with some elegance and wit, but in doing so he avoids discussing what it means to read the novel. It remains the case, whatever associations surround it, that the text of the first paragraph of the novel offers the same words and in the same order whether we read it in printed or electronic form. This constrains what the reader can do rather more significantly than any contingent facts about its origin, binding, or title page.

Miller goes on to point out that the internet as an environment seems to level out its various contents (a part of the democratization that is attributed to it by its enthusiasts). Thus, he writes, "Ayala's Angel jostles side by side in cyberspace with weather maps, satellite images, pornographic bulletin boards . . . " etc. (p. 32). In fact, the internet is much better structured than this comment would suggest. His account makes it seem a promiscuous heap like the city rubbish dump in which any item may occur next to any other item. But for most purposes resources on the internet are only available within well structured and often hierarchical arrays of links or indexes (access to electronic literary texts is a good example of this organization: see, for example, the best current resource The Voice of the Shuttle). It would be impossible to jump directly from an internet text of a Trollope novel to a weather map.

Miller's comment also bypasses the fact that as readers we have in place well developed responses to the different genres of texts we read. We cannot respond to the beginning of a novel in the same way that we do a weather map. Texts themselves contain rather distinctive features that tend to determine how they are to be read, and the appearance of a text in electronic form does not automatically override such features. Indeed, without some prior experience of novels such as those of Trollope, Ayala's Angel might be more or less incomprehensible to the web surfer who was used to dealing primarily with weather maps and pornographic bulletin boards. The internet environment may change many things, but it cannot eliminate the requirement for certain well defined reading processes once we address ourselves to the text of the novel itself.

Miller's subsequent arguments on the changed status of the electronic version of the novel suffer from the same basic inconsequentiality: they bypass what it means to read the novel in favour of such incidental aspects as its "fleeting" appearance when scrolled on screen, or the vulnerability of the text to being changed once it is in our own computer, or our ability to search the text for a given word or group of words. The more pressing question is why Miller, who has for some time been one of the best readers in the literary field, should think that these arguments carry the weight he appears to give them.

Only near the end of the essay does Miller seriously address the question of what it means to read, by referring to what Proust attempted to do through narrative in accessing and relating the memories that form his great novel. His account of this process, however, is instructive for reflecting on the electronic tropes with which Miller has made so free. Miller points out that reading such a novel necessarily involves non-linear aspects: "Marcel treats his memories as though he had a hypertext program for moving around within them. Anywhere you begin will lead ultimately by a series of links everywhere else in that vast storage disk of recollections, but this will not happen according to any predetermined pathways. We readers must do the same." In our reading too we must "do the work a hypertext does for you" (p. 37). This is well put, but it serves to show the great distance between the act of literary reading and the electronic medium Miller has been describing. Human and computer memory are quite dissimilar: we do not access a "vast storage disk of recollections" somewhere in the brain; and links between memories are quite unlike any conceivable hypertext system. It should also be noted that since the links are not "predetermined" as Miller points out, we cannot hope to design an actual hypertext system to instantiate the non-linear structure of such a novel for all readers: my links are almost certainly not the same as yours, and to impose mine on you would disrupt the emergent meaning that for you depends on making your own links.

Miller concludes by pointing to the illusions about reading which will be done away with in the new electronic world of texts. We will no longer be able to think of a novel "as a stable and unmoving organic unity, on the model of a spatial array. Such a fixed text imposed on its readers a single unified meaning generated by a linear reading from the first word through to the end" (p. 38). This is indeed an illusion: as I have suggested, reading was never like this (it is one of the standard straw figures of hypertext advocates). However, Miller goes on to appeal to hypertext as the model of the future, since this will bring into the open our responsibility as readers to choose the pathways we take through a literary text -- "This is the ethics of hypertext. Hypertext brings into the open the way the generation of meaning in the act of reading is a speech act, not a passive cognitive reception" (p. 38).

But since hypertext cannot model the individual reader's choice of links and the interpretive implications of those choices, hypertext will actually limit or destroy the responsibility that Miller urges for the act of reading. Miller gives us that responsibility with one hand -- "The reader is, in the end, responsible for what he or she makes of a text" -- but then takes it away with the other by nominating hypertext as its best representative, a tool that can never be responsive to that "mobile, ungrounded, and unmasterable vibration" that is a text in the process of being read (p. 38).