Dizzying the Arithmetic of Memory: Shakespearean Documents as Text, Image, and Code
University of Western Ontario
Galey, Alan. "Dizzying the Arithmetic of Memory: Shakespearean Source Documents as Text, Image, and Code." Early Modern Literary Studies 9.3 / Special Issue 12 (January, 2004): 4.1-28 <URL: http://purl.oclc.org/emls/09-3/galedizz.htm>.
My title refers to a passage in Hamlet that is uncannily prescient of the complexities of electronic text. In the final act, Osric and Hamlet exchange elaborate courtly descriptions of Laertes in a contest where the language becomes head-spinningly figurative. At one point Hamlet says, in reference to Laertes's attributes, that "to divide him inventorially would dizzy the arithmetic of memory". This phrase is oddly applicable to electronic editing as it has developed through the theories and practices of the Internet Shakespeare Editions (ISE) and many similar projects. The activity of "dividing inventorially" a text into a logical structure of parts that are computable according to a machine-readable "arithmetic" can indeed be a dizzying enterprise. This is especially true if the desire to produce a universally descriptive model tempts designers to lose sight of the ground, as Osric loses the sense of Hamlet's words. The problem of universal models, such as that of the now-standard Text Encoding Initiative (TEI), becomes especially vexing when such a model must account for complex and ambiguous texts such as the received source documents for Shakespeare's plays – the quartos and Folio which, prior to any transcription, already occupy uncertain positions between manuscript, stage, and early modern print. This paper will inquire how the passage of a Shakespeare play through different media might prompt us to revise what is by now an old question: "what is the text?". To borrow Hamlet's terms, how can a text be divided, catalogued, and marked-up according to a universal inventory? What kind of information about a Shakespeare play might an electronic transcription (or any transcription) preserve, omit, or perhaps even add? More abstractly, what is the implicit arithmetic of memory that subtends all reproductions of Shakespearean documents, and how might those texts dizzy our capacity to divide them inventorially?
My interest in the questions stated above emerged in the course of working with the full corpus of the ISE's transcriptions of quarto and Folio texts, which I converted from SGML to XML over several months in 2002 using a combination of Perl scripts and other methods. The process was complicated by our desire to produce versions of the transcriptions that conformed to the TEI standard, but also preserved the accuracy and efficiency of the ISE's original markup scheme. Since the tagging differences between SGML, our XML, and the TEI's XML are structural as well as semantic, we were, in short, teaching a computer to ask "what is the text?" and "what is the document?", and to return meaningful answers. Useful connections between the instability of Shakespearean textuality and the polymorphousness of electronic textuality have already been made in recent criticism, especially in criticism generated in part by the ISE. My intent here is not to restate the argument for an affinity between Shakespearean content and electronic media, but, assuming that argument's validity, to explore its consequences in the practical terms of humanities computing and textual studies. The latter discipline's empirical rigor is largely absent from the poststructuralist school of George Landow, and discussions of "electronic text" have therefore suffered from an undue focus on "hypertext," even though the relation of the two terms is more properly synecdoche than synonymy. Even the now-accepted term "hypertext edition" privileges one especially visible (and admittedly consequential) feature of electronic texts over others, such as searchability, interoperability, and extensibility, which are more technical and don't translate well into diagrams, but which define electronic textuality as much as hypertext – perhaps even more so where large-scale projects are concerned. Matthew Kirschenbaum has recently argued a position that editorial theorists and electronic critics alike would do well to observe, namely that "a bibliographic/textual approach calls upon us to emphasize precisely those aspects of electronic textuality that have thus far been neglected in the critical writing about the medium: platform, interface, data standards, file formats, versioning, and so forth" (27). In other words, we must restock the toolbox of traditional bibliography if we, as textual scholars, are to strip the veil of pixels (to paraphrase Fredson Bowers) from electronic texts and understand how they work, not just how they behave on the screen.
Encoding and the Photographic Message
An ontological issue common to all electronic texts is encoding, which is a matter for both theory and practice. This is a term worth clarifying since it carries a range of meanings. In specific computing terms, encoding refers to the standard character set used by an electronic text (typically a species of ASCII or Unicode). Encoding in this sense is not equivalent to the computer's fonts, but bears a closer analogue to early modern hand-typesetting; the compositor's box either held type for a particular character (such as "a", "a", "§", long s, or half-italic colon), or it did not. The latter two examples cannot be encoded by the electronic compositor's box that supplies the character set for the present document, and therefore I cannot represent them here as electronic text, only as images: "Where the Bee ucks, there uck I" (TLN 2045). Try using your web browser's search function to find this phrase in the present document. Although this infamous bit of The Tempest's original typography makes sense to human eyes, your computer almost certainly does not understand what it says – and there lie the real stakes of encoding electronic texts. Whatever a text like the Folio Tempest may be to us, to a machine it is ultimately a structured series of alphanumeric symbols that most humans never see, and that few would wish to see, but which we rely on machines to manage the drudgery of processing. The challenge for the ISE, and indeed any electronic editing project, is to create texts that can make the journey between encoding states, from machine-readable memory to human-readable screen, efficiently and accurately.
However, electronic transcriptions offer a derivative benefit in that their intermediate status as marked-up text provides an encoding state that is readable by both humans and machines. To use the example of Branagh's 1993 film version of Much Ado About Nothing (discussed in greater detail below), an electronic transcription is most like the 35mm film reels used in theatrical projection, whose frames may be held against a light source to reveal recognizable images. The same is not true of the surface of a DVD of Much Ado, which reveals only one's own reflection. The information on a film reel of Much Ado is nonetheless encoded since it requires the projection apparatus to translate and amplify it, but the content also exists on the reel in a form that both illustrates the operation of the decoding process (through the serial progression of still frames), and divulges otherwise inaccessible information (in the still image within a frame, which freezes motion). Indeed, film's seminal moment in Eadweard Muybridge's 1878 serial photography experiment depends upon just such an intermediate state, since Muybridge's photographic apparatus encoded information about a horse's motion that was undetectable by human eyes. The purpose of the experiment was not to simulate the motion of a horse by decoding the series of images as a prototype film strip, but to look within the encoded state that froze the horse's motion, enabling the experimenters to determine whether a hoof is always in contact with the ground when a horse gallops. The primary use of cinematic film, to represent moving pictures, is a happy accident resulting from an experiment in encoding.
An electronic transcription of Much Ado creates a similar opportunity to freeze the movement of a Shakespeare text as it travels along a chain of transmission through the encoding states of various media. This chain reaches from the indeterminate moments of Shakespeare's foul-paper composition to the electronic encoding and editing projects of the present day. Given the current editorial dictum that no edition is ever definitive and final, we can only ever have these texts in transmission, on their way to becoming something else. But the crucial difference between the Muybridge experiment and Shakespearean textual studies is that in the latter case, there is no horse, so to speak, that constitutes a final state against which truth claims may be verified. The transmission narratives for Shakespeare's plays are as heterogeneous as they are complex, and share common points where texts materialize as physical documents, including Shakespeare's foul papers, scribal transcriptions, playhouse promptbooks, printed quartos, annotated quartos, and the 1623 Folio. Like the surface of a DVD, the original state of the text – if the singular "original state" is even an appropriate term – is not accessible to us directly. All documentary reproductions therefore involve encoding and decoding in some form.
The choice between image- and text-based encodings is not new in documentary editing, and indeed was the subject of debate in the early twentieth century, when new photographic techniques emerged simultaneously with the New Bibliographers' rethinking of the Shakespearean critical apparatus. In a group of short papers read before the Bibliographical Society in 1925, A.W. Pollard, Gilbert R. Redgrave, R.W. Chapman, and W.W. Greg examine the relative merits of image- versus text-based facsimiles in terms of the photographic and typesetting technologies available at the time. Greg emphasizes two points of importance to electronic editors: first, that every reproductive process "in some measure obscures what it reproduces"; and second, that while photographic reproduction is generally more faithful to its source than typographic reproduction, the former is restricted to reproducing a single, possibly flawed original, while the latter allows for a critical apparatus that collates all relevant data (322-3). Where the photograph offers fidelity, the type reprint offers computability; according to Greg, no single technology to date in 1925 could offer both.
The representational status of the photograph complicates questions of documentary reproduction in ways Greg could not have anticipated. The production and reproduction of meaning by any technology is also culturally and semiotically determined. This is especially true in Shakespeare studies, where the paradoxical relationship of books and images is thematized in advance by the very content itself. Any facsimile or multiple text edition of a Shakespeare play necessarily competes with the messages of authorial unity and completeness embodied by the First Folio, which most researchers experience today through Charlton Hinman's complex assembly of photographs in his 1968 Norton Facsimile. It is not surprising that the Folio should present such a complicated site for relations between text and image, since the distinction between the two kinds of content is made on the first pages of the Folio itself. The Droeshout portrait of Shakespeare that appears in the first pages is such a familiar icon in popular culture today that its original strangeness is easily lost on modern readers. Unlike similar title-page engravings, such as Ben Jonson's from his own 1616 collected works, Shakespeare's portrait is unadorned by non-diegetic figures of muses, laurels, or other classical ornaments. The image is, however, annotated by Jonson's facing-page poem which itself calls attention to the juxtaposition of text and image:
O, could he but haue drawne his wit
As well in brasse, as he hath hit
His face; the Print would then surpasse
All, that was euer writ in brasse.
But, since he cannot, Reader, looke
Not on his Picture, but his Booke. (TLN 7-12) 
For readers of a facsimile of the Folio, this injunction to "looke / Not on [Shakespeare's] Picture, but his Booke" becomes impossible since Shakespeare's "booke" is conspicuously absent when the facsimile is present. Readers can thus look only on pictures, and the collection of pictures that makes up a Folio facsimile often functions in popular reception as a step toward the real Shakespeare, even as a fetishized substitute for direct inheritance the author's legacy.
Readers in the twentieth century might not share Jonson's skepticism about the ability of images to capture the essence of an object, since photography supposedly presents the world as it really is – even in excess of the normal powers of human vision. In his essay "The Photographic Message," Roland Barthes explores the function of the press photograph in modern newspapers and magazines, but his conclusions have applicability to photographs of books and documents. Press photographs and facsimile photographs share the same primary quality of being non-artistic, or of presenting their subjects with no accompanying signifiers that could be read as part of an artistic code. Like a press photograph, one cannot describe or paraphrase a page of the Hinman Folio facsimile without recoding it: "In front of a photograph, the feeling of 'denotation,' or, if one prefers, of analogical plenitude, is so great that the description of a photograph is literally impossible; to describe consists precisely in joining to the denotated message a [...] second-order message," which inevitably results in an added, connotative message (18; emphasis in original). In order for a facsimile of a text to function adequately as a tool to improve analyses and editions based on that text, which are themselves new texts, the facsimile must function as a perfect photograph. For example, the value of Anthony James West's study of Folio facsimiles is established primarily through his attention to the accuracy of past facsimiles. He writes in order to answer a bibliographic question: is a given facsimile accurate enough to be useful? The question differs slightly between photographic and photolithographic contexts, however, since the latter leaves open the possibility of editorial intervention between the photographic stage and the final impression of ink on paper. Twentieth-century bibliography's skepticism toward Victorian lithography – which West notes on the part of Charlton Hinman (152) – signals a shift away from imagining facsimiles as lithographic objects that could be both images and manipulable texts. The photograph made a new kind of status available to the preserved document, that of pure denotation.
However, Barthes identifies a paradox that complicates the ability of the photographic facsimile to contain the real document as an uncoded message. The very "analogical plenitude" that produces the documentary photograph as a pure denotation, or as a perfect medium that does not interpolate codes of meaning, also connotes a message that "develops on the basis of a message without a code" (19; emphasis in original). In other words, if connotation is "the imposition of second meaning on the photographic message proper," (20) the photograph's supposed absence of imposed signs upon the photographic message is itself a second-order imposition. A photographic facsimile thus generates its own truthfulness, not through its analogical fidelity to the document, but through its functioning in a larger semiotic field of texts and images.
As both Greg and Hinman understood, this larger field in Shakespeare studies is bounded by the critical apparatus. In a facsimile such as Hinman's, text is present both in apposition to the image, and as the image's message. If the press photograph appears to communicate its content in the absence of a code, the facsimile page similarly effaces its own encoded status as the product of an editor, but must communicate some of the most over-coded content in the English language. The analogical plenitude of the facsimile page, or its realistic fidelity that precludes the need for annotation, exists in a paradoxical conflict with the multiplicity of linguistic, typographic, poetic, orthographic, and other kinds of codes whose complexity has kept scholars busy for centuries. The encyclopedic complexity of the apparatus, seen and unseen, that attends upon any text labelled as "Shakespeare" generates a semiotic influence of its own that competes with the photograph. The facsimile cannot be annotated for many reasons (mostly practical), but in semiotic terms the presence of a single annotation would suggest the potential weight of the variorum apparatus, which would corrupt the clarity of the photographic representation by turning its content into pure data. Even in the absence of electronic text, an annotation can be thought to make a text machine-readable by establishing the addressability of any part of that text. If a note can append a piece of information to a specifically indexable point in a text, the rest of the text must potentially be recodable within a larger system. The necessary antithesis of the clean, facsimile page is therefore the riotous variorum page where perhaps only a single line of primary text remains. The annotated text on such a page could no longer be called primary, and would be subsumed into a field of connotation.
The Photolithographic Message
In "The Work of Art in the Age of Mechanical Reproduction," Walter Benjamin notes that lithography was the fleeting but crucial conjunction of type and image in the history of mechanical reproduction. The ability of lithographers to trace a design on stone with a grease pencil meant that the lithographic image, newly freed from the production constraints of wooden engraving and copper etching, "enabled graphic art to illustrate everyday life, and it began to keep pace with printing" (219). Lithography was soon superseded by photography, but for the purposes of documentary reproduction, lithography was not only the first means of pictorially reproducing Shakespeare's texts, it was also the brief technological moment when text and image occupied the same writing space – the lithographer's stone did not distinguish between them. If the matrix of the compositor's box stands behind the typographical page, and the finality of the etched copper plate or wooden block stands behind the reproduced image, we may regard the lithographer's stone as the sign of an imagined textual condition where text and image function interchangeably.
However, by the time of Greg's writing in 1925, lithography and its later development, photolithography, had become an untrustworthy medium: "Lithography is now generally discarded; it often reduces a text to illegibility and its liability to manipulation is infinite" (Pollard, et al 322; emphasis added). Strangely, the latter point is often considered a virtue of electronic text; to Greg, infinite manipulability negates any meaningful connection between a lithographic copy and an original: "The production of a lithographic facsimile would be no evidence whatever that the supposed original ever existed" (322). Even the lithographic reproduction of a photograph could not satisfy the documentary fidelity required by Greg and many other bibliographers. Photolithography, the reproductive technique of choice for many of Frederick Furnivall's New Shakespeare Society facsimiles, involves taking a photograph of the text, fixing the photo to the lithographic stone, and taking paper prints from it. The photolithographic process takes its graphical content from the photograph, a medium with a supposed indexical relation to the real, but the technique still allows the image of a text to be alterable as text. Even after the replacement of photolithography by other methods of reproduction, editors have imagined reproductions of texts which, like lithographic reproduction, allow typographic text and photographic image to function interchangeably while still retaining their respective virtues as apparatus and archive.
The paradox of taking photographs of books lies in the photograph's connoted unity with its subject, or its dependence on truthfulness, as a quality that is difficult to reconcile with the inherently multiple nature of books and texts. The problem is best described by F.W. Bateson's question, "If the Mona Lisa is in the Louvre, where are Hamlet and Lycidas?" (qtd. in Greetham 342). The provisional answer is that, at this moment, Hamlet is mostly in the Folger library, though bits of it are in the British Library, the Bodleian, the Huntington, and other libraries as well as private collections. But to discuss Hamlet in relation to a Benjaminian aura achieved through mechanical reproduction is difficult, since any Hamlet also subdivides into multiple entities: the document(s) behind the editions; the performance(s) behind the documents. Of course, the point of the question is that any answer can only be provisional. Similarly, the Folio itself only becomes a critically useable object when it is subsumed within an apparatus that selects pieces of it from amongst surviving documentary traces, producing a virtual unity.
Such unity is entirely absent from the typographic antithesis of the late-Victorian facsimile. In the midst of the New Shakespeare Society's ambitious facsimile projects emerged the most complex presentation of any Shakespeare play ever to appear in print. Teena Rochfort Smith's Four-Text Hamlet exists now only as a plan of work and a three-scene trial version, which survives in an extremely rare copy in the Folger Library (and which is discussed in detail, along with the story of Rochfort Smith herself, by Ann Thompson). Visual complexity is a vexing issue in the editing of Hamlet, whose popularity and canonical status jars with its multiple, incommensurable texts. This play about the persistence of memory – in which so much, like the Ghost, is forgotten – is a continual reminder to Shakespeareans of the vicissitudes of textual transmission through time; like Horatio on the battlements, editors charge the texts to speak but receive no clear answer. The discovery of the 1603 quarto (now Q1) in 1821, and the nagging possibility that some form of Hamlet existed ten years prior to its first publication, added up to a presentational challenge for documentary-minded scholars like Furnivall and his New Shakespeare Society. Present designers of electronic editions will notice familiar problems in the user interface Rochfort Smith planned for her edition. Instead of a critical edition or annotated type facsimile, the Four-Text Hamlet was to omit editorial footnotes in favour of systematic markers that allow the texts to annotate each other. With diplomatic transcriptions of Q1, Q2, F, a "Revized Text" that conflates Q2 and F, and an intricate annotational system to coordinate these four texts, Rochfort Smith's polyphonic edition would, in Ann Thompson's words, "have been the most complex presentation of the texts of Hamlet ever attempted" (131). Given the play in question, and the period, a corollary statement might be that Smith's project was the most textually complex Shakespeare edition imagined up to that point.
However, for the present discussion of Shakespearean documents as texts and images, it is important to note that Rochfort Smith's interface could only have been imagined typographically. An image-based facsimile of the three texts would have forestalled her system of visual markup. Printers – then, as now – could not economically provide her imagined interface with a semiotic vocabulary of colour and modified type. But the pages reprinted by Thompson suggest Rochfort Smith thought of her interface not as a graphic imitation of a document, but as a recoding of the content as manipulable information. Rochfort Smith's explanatory section, "the way in which the differences of text are markt," is essentially a map of the kinds of metadata she supplies in the transcriptions. Her range of typefaces, used to mark variants in each text and to distinguish between kinds of variation, includes the Clarendon font, italic small capitals, Roman small capitals, gothic blackletter, and of course the normal Roman font. Rochfort Smith also recruits an intimidating number of printer's symbols, including the paragraphus (¶) and the section mark (§) which are arbitrarily repurposed in her scheme. Her recoding of the texts as annotation and metadata anticipates – even exceeds – the most complex electronic interfaces for the play, such as Bernice Kliman's Enfolded Hamlet. In print, the result of typographic markup is a noisy page, which itself dramatizes the mystery of a real Hamlet as either a self-confirming or self-cancelling work hiding behind the antic dispositions of documents. On the Four-Text Hamlet page reprinted by Thompson, both Hamlet and Hamlet find themselves multiply related in unsettling ways to kin and usurpers. Even the Folio text, which might assert the primacy of its bibliographic codes as a claim to the throne of a unified kingdom, can only respond, "Not o ^ my Lord, I am too much i'th'Sun" (F.1.2.67).
Editorial Models and Encoding Cruxes
An intentionalist editor of the Greg-Bowers-Tanselle line might level a charge at Rochfort Smith which applies equally to some electronic designers: that an excessive dependence upon – even delight in – the complexity of the visual presentation merely disguises a reluctance to get down to the business of editing. However, such a criticism assumes a particular kind of editorial model, one that would see the singular work through the problematic multiple documents. One form of response to this problem is the hypothesizing and systematic modelling of manuscript states that became a bibliographic science through decades of work by Greg and the New Bibliography. Their central bibliographic tenants have been challenged by revisionist textual scholars in the past two decades, whose most effective critiques of the New Bibliography have questioned the validity of truth-claims based on hypothesized models of absent manuscript documents in place of the documents themselves. Barbara Mowat's 1998 article, "The Problem of Shakespeare's Text(s)," exemplifies this skepticism when she argues,
it is this clinging to illusions about "prompt-books" and "authorial manuscripts" that blinds us to the possibility that there may have been a large flow of manuscript copies of Shakespeare's plays, copies marked by the idiosyncrasies of manuscript transmission, idiosyncrasies that would inevitably have made their way into the printed copies. (136)
To pursue Fredson Bowers's goal of stripping the veil of print would result not in a newly visible document, present in itself as a revelatory physical object, but rather in a detailed hypothesis that reflects its originating model. Mowat's argument about New Bibliographical manuscript theories is similar to that of the present paper, namely that excessively abstract document models inevitably contain blind spots that are themselves visible through encoding "idiosyncrasies." As an alternative to the copy-text approach of the New Bibliography, we might follow the Oxford Shakespeare's rationale, under which the originating moment of a play would be its first performances, where Shakespeare's scripts reached their true embodiment via the collaborative energies of the playhouse. In this light, modern editors and directors alike would look to the transcription to recover, or decode, information about a play's original performance conditions. I will not digress into a discussion as to which of these narratives is the more accurate or desirable, since both embody virtues and flaws that continue to fuel debates beyond the scope of this paper. Instead, I would emphasize that these and other kinds of critical enterprises, however legitimate, may be forestalled by certain models for Shakespearean source documents.
The "idiosyncrasies of manuscript transmission" that Mowat identifies as destabilizing presences in New Bibliographic textual models also have their counterparts in electronic models, which we could term encoding cruxes. An encoding crux could be defined as a situation in which the original source document may – or may not – record important information, but a given transcription's textual model and corresponding markup system lacks the vocabulary or syntax to preserve that information, even in an ambiguous state. The clearest examples of encoding cruxes in Shakespeare texts often involve silence. As Philip McGuire's 1985 study Speechless Dialect: Shakespeare's Open Silences demonstrates, Shakespeare's plays are full of moments where silence serves a significant dramatic function, but we cannot always determine what that function is. McGuire's concept of the open silence is useful in tracing the gaps that appear in a playtext's passage from page to stage, as his own definition illustrates: "An open silence is one whose precise meanings and effects, because they cannot be determined by analysis of the words of the playtext, must be established by nonverbal, extratextual features of the play that emerge only in performance" (xv). One of the most interpretively loaded such moments occurs at the end of Measure for Measure, when the Duke proposes marriage to Isabella but the received texts record no indication of her reaction. We see a similar moment in The Tempest as Prospero forgives Antonio and Sebastian, who are given no lines in response to indicate whether or not they accept forgiveness. In both cases, we can safely assume that something happens; these characters must have had some form of reaction in the original performances, and even if those reactions were neutral they would still impart meaning.
The open silences found in Measure for Measure and The Tempest could also be called encoding cruxes because, even if Shakespeare or his company members had recorded instructions at these points in their various manuscript incarnations of the plays, the typography of early modern print could not have recorded that information, except perhaps as an unusually explicit stage direction. Unlike rest symbols in musical notation or the "beat" indicator in screenplays, printed early modern drama has no graphical mark to indicate silence, and even some kind of rest symbol would not tell us what we need in these cases. No traditional print transcription could provide this information, even though it is an integral part of the play; only an editorial intervention in the form of an annotation or other metadata could do this. Encoding cruxes are therefore not purely electronic phenomena, since all documents employ an encoding and markup scheme, such as the present essay's punctuation marks and spaces between words and line-breaks. However, just as textual cruxes become most visible in the process of critical editing (particularly copy-text editing), so do encoding cruxes become most vexing when they trouble the blind-spots of universally scoped electronic models.
Tagging the Bastard
A more complex example of an encoding crux may be found in Much Ado about Nothing, whose principal villain, Don John, is a bastard. The problem is that, strictly speaking, we do not learn he is a bastard until fairly late into the play, and at a moment that may or may not convey a certain dramatic effect. The unknowing reader of the 1600 quarto or 1623 Folio text would learn soon into the play that Don Pedro's brother is a bastard, since stage directions frequently refer to him as "Iohn the bastard," (in F, TLNs 92, 197, 344, 1278, and 1657; in Q, TLNs 84, 174, 312, 1205, and 1566) and his speech prefix ranges from "Iohn" to "Bastard" and its various abbreviations. So, when the first mention of Don John's bastardy appears in actual dialogue in act 4, scene 1, no new information is being divulged by the text. But what is the text in this case? If we take the text to be either a record of past performance or a script for potential performance, we would need to strip it of all but dialogue and the paratextual actions spelled out by stage directions. In such a performance text, John only becomes a bastard in act 4, scene 1, which may be a significant moment.
After the three princes have exited, following their bitter confrontation at the chapel, Hero is devastated and her father swears bloody murder. Reflecting on the suddenness of the princes' accusation, the Friar notes "There is some strange misprison in the princes" (185). To this, Benedick voices his suspicion that
Two of them have the very bent of honour;
And if their wisdoms be misled in this,
The practice of it lives in John the bastard,
Whose spirits toil in frame of villainies. (186-9)
To an unknowing audience, Don John's status as an illegitimate child emerges at the moment when the very unchastity that produces bastards is suspected in another woman. John's plot creates the figure of his own mother's fault in Hero. Benedick thus recognizes Don John's scheming by its apparent reproduction of the extra-marital sex that first produced John. Moreover, John's plot and Benedick's hunch both depend upon the shared codes of an anxious misogyny, according to which bastardy is traceable to female sexual weakness, which remains to be detected and regulated by men. Benedick suspects Don John not just because some villainy is at work, but specifically because that villainy operates on an all-too-ready fear of unregulated female sexuality. If Don John suddenly becomes a bastard as the appearance of female unchastity threatens the social fabric of sunny Messina, the effect requires the active suppression, or passive absence, of certain information about the character of Don John up to that moment.
Is this really a major character point? Could Don John's status as a bastard not be signalled somehow in performance? Or, to take a step back, what do questions like these mean to an editor, or to an electronic encoder? My intention here is not to advance an interpretation of the function of bastardy in Much Ado, but to take the above critical scenario as a test case for electronic editing. Even if the responsibilities of the editor in new media are up for debate, those of the transcriber are not: transcriptions in any media must not add, distort, or suppress information that may be of use to what was once called "higher criticism." Therefore, the reader need not accept the plausibility of the above interpretation, but merely its possibility, which is enough to make trouble for an electronic transcription that must make decisions about speech prefixes and the textual entities they represent. It would indeed be strange for Shakespeare to withhold vital information from his audience (excepting Hermione's preservation in The Winter's Tale), but therefore as a stranger we should consider giving this odd crux welcome; transcribers should at least stay out of its way. My concern is how this encoding crux modifies the question, "what is the text?" or in this case, "what is the text of Don John?". If he is a series of graphic marks on the printed page, then he is a bastard from his first entrance. If he is an effect of transcribed stage dialogue, his own and that of other characters, then he is not a bastard until the precise moment when his crime is suspected. But if he is an electronic object, what then?
The answer depends upon whose model for an electronic text we use. Following the guidelines of the dominant electronic model, that of the TEI, John is always already a bastard. The TEI makes available a well-documented tagset for dramatic texts, but does not sufficiently distinguish between modern playtexts, like a Stoppard play one might find on the shelf of a bookstore, and early print documents that contain dramatic texts whose encodings of information are complex and sometimes ambiguous. It would of course be unfair to blame the TEI for not handling in detail the proper encoding of Shakespearean source documents, let alone all early modern dramatic texts, since that is the job of specialist projects like the ISE. (That the TEI exists at all should be recognized as an achievement.) But the problem that causes John to be a special kind of bastard in the TEI model is that the core TEI documentation does offer a standardized approach to resolving issues of variable speech prefixes within plays, and it even uses Q1 Hamlet – a transcription, not a critically edited text – as its example. The TEI solution is sensible: record the speech prefix as printed, whatever its variation, but give the tag an attribute, or machine-readable link, that ties it back to a single definition of the character that appears in a header at the beginning of the play. For example, Don John's first line in the play would be encoded as
<p>I thanke you, I am not of many wordes, but I thanke you.</p>
(In this example and the one below, the corresponding references to the object "john" are emphasized in boldface.) In other words, the fluctuating print entity known as the Bastard in Much Ado would, through every speech tag, bind itself to a stable definition of his character, abstracted from the sequential progress of the play over time, and equally referenceable from all points in the text. This entity might take the following form in the transcription's metadata:
<role id="john">Don John</role>
<roleDesc>Don Pedro's bastard brother.</roleDesc>
To create such a list of character definitions in the metadata of a transcription, as the TEI guidelines would seem to imply, results essentially in the kind of dramatis personae list we would expect from modern print editions, the kind that would almost certainly tell us John was a bastard right up front.
To add a dramatis personae to a transcription of Much Ado – whether that list is metadata or print – is to add something to the text that was not there before. The dramatis personae list abstracts characters away from their progressive development in the play, and turns them into data objects with stable properties, like being a bastard, that exist outside of time or sequence. The fundamental problem with the TEI model is that, in the interests of regularity, it presumes to extract a logical structure of characters from the transcription text, and to convert that list of characters into an extraneous data object that itself merely reproduces the conventions of print. Its model for dramatic texts silently embodies interpretative assumptions that have nothing to do with that model's function in humanities computing or textual studies: in the Don John example, the assumptions that characters ontologically precede texts, and that transcriptions of early modern dramatic documents are masses of irregular data that can be silently regularized on a non-editorial, machine-readable level. To encode a dramatic text by the strict rules of the TEI is to produce a kind of critical apparatus by default, contrary perhaps to an electronic text's stated editorial methods.
Fortunately, the TEI is only one document model among many in the broader field of textual studies. Even though the TEI is the dominant – and too often unquestioned – model for electronic transcription, it is also detailed to such an extent in C.M. Sperberg-McQueen and Lou Burnard's P4 Guidelines that its status as a model is anything but disingenuous – in other words, one can argue with it. Indeed, my disagreements with the TEI stem not so much from the guidelines themselves as from scholars who seek to apply them comfortably across a universal field of texts. But in the world of Shakespearean cultural production and consumption, the models we use to satisfy the "what-is-the-text?" imperative are far more implicit, provisional, and subject to revision. Perhaps the most revealing place to look at textual models in action is not a specific model for a specific medium, but rather the economy of variables that governs a play as it passes through various media.
We can see the set of possibilities that encompasses Much Ado by returning to the bastard question and tracing its progress through Branagh's film and published screenplay versions of the play. It is worth noting that the latter is not a screenplay, strictly speaking, since it is not simply a shooting script but a glossy souvenir of the Branagh film. Consequently it includes considerable paratextual content, including written content by Branagh, production stills, frame enlargements, publicity photographs, and a synopsis and character notes by the film's dramaturge, Russell Jackson. For all its added interpretive content, such as music, mise-en-scène, cinematography, and acting, the medium of film removes interpretive apparatus such as dramatis personae lists and stage directions. Branagh's film can therefore provide space for the bastard effect, as Don John's familial status is not revealed until Benedick's revelation in the chapel scene. Although the film does not overtly exploit this effect – there is no corresponding cue in the musical score – the significance of Don John's delayed bastardy is available for interpretation by those who would argue the point. Indeed, the film's goal of reaching a broad audience presupposes an absence of paratextual knowledge of the play, including the facts about Don John.
The published screenplay, however, makes different assumptions. Russell Jackson's character notes function as dramatis personae for this cinematic performance, and link production stills of the characters (posing clearly in-character) to descriptions such as this one for Don John: "The bastard brother of Don Pedro, and a malcontent. Although before now he has taken sides against Don Pedro, he is now reconciled with him, but Don Pedro's generosity irks Don John, and he is especially galled by the favours shown to Claudio" (89). It is difficult to situate this description temporally in relation to the film. It refers to none of the story events that occur in the film, so the "now" of the caption must refer to the state prior to the beginning of the film. Yet all of the character information mentioned here, even beyond the bastard reference, is only revealed to the audience through the process of watching the film. The blurb therefore gives us the results of audience interpretation, but locates itself in advance of the events of the film. (Indeed, another "now," in the "now a major motion picture" blurb on the front cover, invites even more confusing speculation as to the book's status. What was it prior to "now"?) In a paradox oddly like that of Shakespeare's source documents, it is impossible to regard the screenplay as purely a record of a past performance or a script for one yet to happen. The set of possibilities at work in the film is therefore incompatible with those offered by the film's print counterpart, since the latter, like the TEI, places characters in an atemporal state exclusive of the sequentiality of performance. Don John's semi-diegetic eye-contact with the camera in his production still provokes us to question, in relation to Don John, not only "what is the text?" but also "when is the text?".
An adequate model for a Shakespeare play would need to account for the multiplicity of performance variables, even though it could never account for all the possible variables themselves. The detailed economy of performance variables that Alan Dessen deploys in Rescripting Shakespeare constitutes an impressive start along these lines, yet Dessen's approach does not extend to the complications introduced by different media, or the kinds of cross-format translations that produce encoding cruxes like those described in this paper. A more focused attempt along these lines is made by Peter Holland, who applies the analytical tools of Shakespearean textual studies and editorial theory to films, including Branagh's Hamlet. As Holland argues, "The movement from page to script to film to screenplay is not linear. The intertwining of the textual witnesses of film reflects back on the nature of the process of editing the Shakespeare text but also on the process of creating an edited film out of the multiplicity of different shots" (296). Holland leaves out "stage" from his list, but his concern is film and print rather than theatrical performance. Nevertheless, the complexity (and impossibility) of accounting for all of a Shakespeare text's variables and potential media transformations, in all their nonlinearity, requires careful attention to conceptual blind spots, and skepticism of the universal applicability of any model. If the shortfalls of the TEI are any indication, Shakespearean dramatic texts are a category of special cases, whose textual idiosyncrasies, open silences, and encoding cruxes will ultimately exceed the capacity of any one model to delineate them, dizzying the arithmetic of human and machine memory alike.
In the meantime, the designer of new interfaces of Shakespearean text and image is, figuratively speaking, always staging Hamlet. Behind the complexity of the texts – which, as Rochfort Smith has shown, support radical new stagings of their own – stands a question that most readers and audiences ask of the work and its title character, and which forms the first words of the play: "who's there?". It is a question that the guards on the battlements of Elsinore ask of each other, but also that Hamlet asks of the ghost, which is not unlike a photographic message itself in its uncanny, spectral repetition of something that is absent, yet present enough to convey information. As editors and designers continue to remake and refigure Shakespearean documents in new media, it is worth remembering that the original 1623 Folio was designed in part as a necessarily incomplete answer to the question, "who's there?", or perhaps "what's left?". Shakespeare's memorializers are still the watch on the battlements, urgently questioning the ghost that Shakespeare himself may have played in life.
 The second quarto (1604) provides the only source for this passage. It reads, "to deuide him inuentorially, would dazzie th'arithmaticke of memory" (TLN 3610+7 - 3610+8), but "dizzy" is a generally accepted modernization of "dazzie." An alternative modernization is "dozy," used by Harold Jenkins in his Arden 2 edition (5.2.114). The ISE's electronic transcription of the second quarto is available at <http://web.uvic.ca/shakespeare/Annex/DraftTxt/Ham/Ham_Q2/Ham_Q2Pages/Ham_Q2N2v.html>.
 See the previous special issue of EMLS on the ISE: "The Internet Shakespeare: Opportunities in a New Medium" (January 1998), <http://www.shu.ac.uk/emls/si-02/si-02toc.html>.
 "Searchability" refers to the electronic text's support of both text searches and logical searches based on tagging. "Interoperability" is an electronic object's availability for use by software and platforms other than the ones that originated it, usually achieved by adherence to a third-party standard such as the TEI. "Extensibility" is an encoding system's capacity for modification and extension by its users in ways the designers had not envisioned (hence eXtensibile Markup Language).
 It should be noted that Kirschenbaum's object is to lay foundations for a descriptive bibliography of first-generation electronic objects; i.e. electronic texts which, like Shelly Jackson's Patchwork Girl, were authored for the electronic medium and have no existence outside of it, unlike a Shakespeare play. Kirschenbaum's article is among the best work to date on electronic editing, but its definition of first-generation (G1) electronic objects might not be adequate to the work of the ISE. The transcriptions and electronic facsimiles are certainly not G1 objects, since they represent non-electronic documents, but what of the forthcoming critical editions which will exist as G1 electronic texts? The editors of the ISE edit Shakespeare's plays, but author their own editions within an "ontological horizon [that] is exclusively digital" (20). A distinction between G1 and G2 objects (which Kirschenbaum leaves the reader to infer) becomes difficult in the case of the ISE, given the blurry ontological horizon of Shakespeare's texts, though Kirschenbaum's term remains a valuable one.
 The ISE's electronic transcription of The Tempest is available at: <http://web.uvic.ca/shakespeare/Annex/DraftTxt/Tmp/Tmp_FPages/Tmp_FB3.html#TLN2044>. The long s () shown here is merely a tiny screenshot of a 12pt Times New Roman f with half of the crossbar removed. In addition to its violation of the searchability imperative, the inadequacy of my workaround will immediately be noticed by any user whose browser's default text is set to anything other than 12pt Times New Roman.
 That is, several of these types of documents materialize only within the bibliographic narratives themselves, since we do not have substantial physical evidence for many of them, such as Shakespeare's foul papers. The degree of confidence with which traditional bibliography can hypothesize absent documentary states has been questioned by many textual scholars in recent years; see Paul Werstine, "Narratives About Printed Shakespeare Texts: 'Foul Papers' and 'Bad' Quartos," Shakespeare Quarterly 41 (1990), 65-86.
 Greg goes on to articulate the specific encoding problem faced by the ISE and all other current electronic editing projects: "It comes in the end to this, that to make a reprint that shall be really rigorous in its reproduction of the original, you must possess the fount in which that original was printed, and you must be prepared to follow in all their inconvenience and inconsistency the ways of the original printer. For the sixteenth and early seventeenth centuries we have not got the one, and the other is intolerable" (326); or, as Greg puts it more succinctly, "There is a particular terror lurking in the [Elizabethan half-italic] colon," which remains unprintable in modern character sets. Greg's formulation of the problem echoes the popular Borgesian fable of the map so detailed it became co-extensive with the territory it depicted (see Jorge Luis Borges, "Of Exactitude in Science," A Universal History of Infamy, trans. Norman Thomas di Giovanni [New York: Dutton, 1972], p. 141).
 Current work on the challenges of text and image may be found in a recent special issue of Computers and the Humanities, "Image-Based Humanities Computing" (36.1 ), edited by Kirschenbaum.
 Norton uses these codes to market their Folio facsimile, with the current version encased in a handsome slipcase that bears a full-colour portrait of Shakespeare. For a detailed reading of the Folio's bibliographic codes, see Leah S. Marcus, "The Art of the Unlovely Frontispiece," Puzzling Shakespeare: Local Reading and Its Discontents (Berkeley: U of California P, 1988), 2-25.
 For a broader discussion of the status of photofacsimiles throughout textual scholarship, see G. Thomas Tanselle, "Reproductions and Scholarship," Studies in Bibliography 42 (1989), 25-34.
 For more on the process, see Michael Twyman, Lithography: 1800-1850 (Oxford: Oxford UP, 1970), especially 253.
 An example is the 1886 facsimile of the 1594 quarto The Taming of a Shrew, photolithographed by Charles Praetorius under Furnivall's supervision, in which the facsimile alters some minor readings from the original, unique witness of the quarto. My own forthcoming electronic edition of A Shrew – which provides a linked facsimile, transcription, and modern text – uses the images from the Praetorius facsimile and flags the alterations in the original. An alternative I had briefly considered was to correct the apparent fault in the digital images themselves, which is an option not unlike that offered by the lithographic stone itself.
 Except for the long s, this quotation is not an image, but rather an attempt to reproduce the typography of Rochfort Smith's edition using the basic typographic capabilities of web browsers. Since fonts and character sets are still not standardized – even in 2004 – across browsers and operating systems, this attempt is doomed to fail on some systems. Some users will read the words "i'th'Sun" in a font that approximates Rochfort Smith's original gothic, and some will not.
 Though the two terms are similar, it may be useful to distinguish between open silences and encoding cruxes. McGuire is primarily concerned with the stage, and he does not explore what an open silence might mean in other media and contexts. To interpreters of Shakespeare's texts, whether critics or performers, an open silence is a source of richness, not because, as McGuire says, their meanings "cannot be determined by analysis" (xv) but because they are multiply determinable – as McGuire's own detailed survey of performance possibilities shows. However, to an editor, an open silence is a moment of textual ambiguity that must remain open in the final editorial product, even if the editor offers his or her own interpretation in a footnote. The moment of ambiguity must be preserved and remain readable as such; to decide the ambiguity in an edition would be a form of inaccuracy, though to decide it in performance is valid dramaturgy. The encoding crux is specifically the problem of the editor or transcriber, since it becomes a question of textual accuracy and ambiguity.
 Perhaps the most notable Shakespearean example appears in the Folio text of Coriolanus, where the title character, having been entreated at length by his mother, "Holds her by the hand silent" (TLN 3539). See the ISE transcription: <http://web.uvic.ca/shakespeare/Annex/DraftTxt/Cor/Cor_FPages/Cor_Fcc2v.html>.
 The ISE's transcriptions of the quarto and Folio texts are available at < http://web.uvic.ca/shakespeare/Annex/DraftTxt/Ado/index.html>. As EMLS's anonymous reader pointed out to me, the text of Much Ado is also uncertain as to whether John is a prince or a count, though the distinction may have been more fluid on the early modern stage than on our own. Another parallel might be the name of Hamlet's uncle, "Claudius," which is mentioned in Q2 and F's speech prefixes and stage directions but never spoken in dialogue. To an audience without access to a dramatis personae, the usurping uncle and brother has no name of his own, only his stolen title. Oddly enough, this effect is preserved in the first quarto of Hamlet, in which the King of Denmark is unnamed on both page and stage.
 See 6.11.2, "Core Tags for Drama," <http://www.tei-c.org/P4X/CO.html#CODR>.
 Sperberg-McQueen, an architect of the TEI and of XML itself, offers a similar caveat in his 1996 article, "Textual Criticism and the Text Encoding Initiative":
it is only fair to point out that the TEI guidelines, like any standard notation, may prove dangerous. Like any notation, the TEI guidelines inevitably make it easy to express certain kinds of ideas, and concomitantly harder to express other kinds of ideas, about the texts we encode in electronic form. Any notation carries with it the danger that it must favor certain habits of thought – in the TEI's case, certain approaches to text – at the expense of others. No one should use TEI markup without being aware of this danger – any more than we should use the English language, or any other, without realizing that it favors the expression of certain kinds of ideas, and discourages the expression, and even the conception, of other ideas. (55)
This is a persuasive formulation of the problem of innate bias or "habits of thought." However, it is by no means a solution, since the problem is ultimately not solvable. Instead, according to Sperberg-McQueen, "the danger can be mitigated, or minimized, by facing it clearly and attempting to provide mechanisms for modifying, or extending, a notation to allow it to express new concepts conveniently." Again, the principle of extensibility proves its value for projects which seek a lifespan beyond a few years. Among the several mechanisms of extensibility that Sperberg-McQueen lists, the final one is the most consequential: "Tags now defined in the TEI scheme can be suppressed, renamed, or modified, provided the changes are properly documented" (56). This provision contains a workaround for the bastard tagging problem I raised above, and may even allow encoders to make dissenting decisions about textual ontology. Yet two questions follow from this, which Sperberg-McQueen does not address. How can an encoding scheme make "explicit provision for the views of dissenting scholars" without simply absorbing those views into a self-consistent structure of ideas, ultimately subordinating them to the dominant habit of thought? Or, to push the problem in the opposite direction, at what point does the democratically extensible TEI standard stop being the TEI standard? The ISE has wrestled with both questions in its creation of TEI-conformant quarto and Folio transcriptions. In a more general sense, so has much revisionist textual scholarship in the wake of poststructuralism and the dethroning of the New Bibliography.
Responses to this piece intended for the Readers' Forum may be sent to the Editor at M.Steggle@shu.ac.uk.
© 2004-, Matthew Steggle (Editor, EMLS).