Dialogue and Interpretation at the Interface of Man and Machine. Reflections on Textuality and a Proposal for an Experiment in Machine

Jerome McGann

Introduction

The project began as a response to the difficulties realized in implementing SGML-based textual models. These difficulties are especially acute for non-informational texts like poetry. For one thing, the "visible language" of text of all kinds does not yield to an SGML approach. Likewise, the radically self-reflexive organization of poetical works - where every textual unit not only might engage multiple signifieds but positively solicits such multiplicities - presents great difficulties for the hierarchized formalities of the SGML-based text. We shall see how both sets of problems arise because SGML models fail to engage a foundational aspect of textuality.

Let me invoke here an observation in G. Spencer Brown's Introduction to that remarkable book Laws of Form (1967). <1> Brown says that in writing his book "I found it easier to acquire an access to the laws [deliberated through the work] than to determine a satisfactory way of communicating them"(xxii). The comment seems at once, and paradoxically, both modest and outrageous, particularly for a book that takes as its point of departure and central subject "self-referential paradoxes". That modest/outrageous statement is as much a self-referential paradox as the famous one Brown cites in the Preface to the American edition of his book: "This statement is false".

Brown's observation about his expressive difficulty cuts to the center of the Laws of Form his book seeks to elucidate, as I hope to show in a later section of this paper. Laws of Form, it turns out, are expressive (trans)forms and are reflected - and reflexive - as such. And certain kinds of text - Brown calls these "injunctive" texts - reveal why it is difficult to communicate the laws they themselves realize.

This paper will thus reconsider the issues taken up in Laws of Form. My interests are, however, more narrow and more practical, not to say pedestrian. I want to elucidate some key but neglected formalities of textual documents and to meditate satisfactory ways of communicating those formalities. I am particularly interested in documents that have been indispensable for traditional humanities disciplines: language and indispensable for traditional humanities disciplines: language and literature, history, philosophy, art.<2> Realizing the need to develop a reliable system for representing such documents in a form that lays them open to the power of digital analysis, humanities computing during the past 20 years worked hard to develop a model for text markup with general applicability. Despite the problems its own implementation exposed, the model of TEI was developed and has become a disciplinary standard.<3>But with increasing numbers of humanities scholars using digital tools in their research work, the realization is growing that TEI's problems are not technical but systemic. To address them properly we have to step back and think not about TEI but about "text" itself.<4>

What is text? I am not so naive as to imagine that question could ever be finally settled. Asking such a question is like asking "How long is the coast of England?". But now we have to ask it again because when the question was re-posed by our digital culture, the humanities response proved inadequate: on one hand a reactionary refusal to admit that this new culture had any right to ask such a question (Sven Birkerts); on the other, the emergence of TEI and the proposal that its view of text would serve the interests of humanities scholars for digital culture. Much is at stake here. Even now we are beginning the process of re-editing -- of representing -- in digital form the entirety of our received textual and documentary archive. How successful this effort will be depends on how clearly we understand the materials we have to work with. Digital tools often appear strange and wondrous, especially because they spawn and mutate so quickly into rich and surprising possibilities. Books and documents

In that general context, then, let me begin trying to think a way back into the problem of textuality itself. To do this I find it useful to begin rethinking the problem through the example of the thinking of Dante, whose grasp of the issues was acute.

I

In the Vita Nuova, Dante regularly attaches explanatory prose descriptions to the poems he imbeds in his famous autobiography. <5> These "divisions", as he calls them, are "made to open the meaning of the thing divided". Some of the divisions are so brief as to seem perfunctory. Others appear so simple and transparent that we wonder what use they might serve. Then again, in the case of one sonnet -- an especially cryptic one as Dante himself acknowledges - no division is supplied because, Dante says, "a division is only made to open the meaning of the thing divided; and this, as it is sufficiently manifest through the reasons given has no need of division." The "reasons given" are the words in the text that explain "the occasion of this sonnet". <6>

I shall have to pass without remark much in these "divisions" that could be usefully taken up, even in the present context. My focus will be on a pair of related matters: the fact that dividing the poems into parts should seem to Dante a way of opening up their meaning; and the fact that the divisions Dante makes seem arbitrary. For example, after quoting the sonnet "Tutti li miei penser parlan d'Amore" Dante says that "This sonnet may be divided into two four parts" ("Questo sonetto in quattro parte si puÚ dividere") and he then proceeds to do two things: first, he restates in schematic terms the prose sense of each of the designated parts; second, he then indicates where in the poem each part falls. Although the exegesis may in any particular case be more or less elaborate, this is the general double form that it always takes.

Note that in dividing this sonnet Dante chooses to distinguish four parts - that is to say, his partitioning represents a judgment Dante makes about what would be useful for the reader to know. The arbitrariness of his divisions leap to one's mind as soon as we see where the four parts separate from each other. In the case of this sonnet part one comprises only line 1, part two includes lines 2-6, part 3, lines 7-8, and the last part covers the sestet. Looking at his divisions for the other poems in the text, we find that they too have little relation to their highly formalized metrical structures. The divisions cut across those structures in apparently random ways, as we see in the case of this sonnet. The randomness is all the more clear because Dante does not disguise from us the fact that the divisions have as much -- perhaps more -- to do with his purposes towards his readers as they do with the formal structure of the poems themselves.

We notice as well that the exegeses, even the elaborated ones, are spare to a degree. Unlike for instance in the Convivio, Dante gives no thematic or symbolic interpretations. Indeed, wherever the poems seem most obscure he typically retreats even further from explanations, even from the schematic divisioning process that marks his method here. So in the case of the sonnet that he doesn't divide at all, Dante says something remarkable. Though obscure and ambiguous words spring up, like tares among the wheat, in the passages "whereby is shown the occasion of this sonnet" in a clear way, Dante chooses not to provide a divisioning, for the difficulties, he says, can't be solved by anyone who doesn't have a deep insight into the issues in the first place: "And therefore it were not well for me to expound this difficulty, inasmuch as my speaking would be either fruitless or else superfluous".

Dante's method is perhaps most fully revealed in his prose division of the great canzone "Donne ch'avete intelletto d'Amore", the Vita Nuova's central text. "That it may be better understood," Dante says, "I will divide it more subtly than the others". Then, having laid out an intricate set of divisions and subdivisions, he finishes his analysis with a remarkable set of statements:

I say, indeed, that the further to open the meaning of this poem, more minute divisions ought to be used; but nevertheless he who is not of wit enough to understand it by these which have been already made is welcome to leave it alone; for certes, I fear I have communicated its sense to too many by these present divisions. . . .

In setting out these divisions for his poems, Dante recalls us to a crucial primitive level of his work's textuality. The significance of Dante's divisionings can be exposed by considering them in relation Brown's Laws of Form. These laws draw out the structural dynamic implicit in Form as such. The elemental condition or manifestation of Form is the appearance of a mark in an otherwise unmarked space. Brown calls this mark a "distinction" so that the elemental law of form is: a distinction can be drawn. Every conceivable formal world may be traced or tracked back to that elementary law.

Dante's textual divisions illustrate his understanding that the same law underwrites the making of poems. Each poem is a kind of world or universe to itself and any set of poems - for instance, the set chosen for the Vita Nuova - may be conceived and fashioned into a meta-universe. So we might say -- might show, as critics have done for centuries -- that the Vita Nuova is usefully seen as part of other networks and textual universes. Or we might turn the direction of our tracing in the opposite direction, back into the sub-universes concealed, as it were, within the apparitions of the specific poems placed in Dante's autobiography. The latter is Dante's own procedure when he offers us his divisions, and their arbitrariness points to the infinities of order that may be tracked inward of the poem, through the looking glass of its surface(s).
 

II

In that part of the book of my memory before the which little can be read, there is a rubric, saying, Incipit Vita Nova. Under such rubric I find written many things; and among them the words which I purpose to copy into this little book.

Mark, space, direction: if we think of language in linguistic terms, these words as I've been using them appear to us as metaphors, figures of speech. But if we think of language in semiotic terms the same words take on a literality that can be extremely helpful for understanding how language works, and particularly - for my present purposes - how the language of poetic forms works. From that altered scale of attention we may then start to re-imagine, in the graphematic terms realized through digitization, analytic tools for our most complex semiotic devices.

It is useful to remember that the text we are transacting here and now on this page is not just a vehicle for communicating certain ideas. It is simultaneously a reflection of and on itself, an analysis of itself - to borrow from Dante, a divisioning of itself. The analysis is executed semiotically, that is, through the deployment of the forms of what might be called, after Robert Horne, its Visual Language. This language Horne describes as a composition of words and shapes and/or images. Useful as this definition is, it is not strictly accurate, and the loose feature of Horne's view is important since it pervades, as an assumption, nearly all approaches to textuality. <7>

The point is important and must be pressed. Horne sees Visual Language when he sees wordtext combined with shapes or wordtext combined with images or wordtext combined with both together. But graphically transmitted wordtext is always ab initio "visual" since there can be no wordtext without the presence of what Horne calls "shapes". According to Horne, images are optional to graphic text but shape is not. But in fact whether or not shape enters the pagespace as explicitly drawn boxes or arrows or circles or whatever -- these are the forms Horne sees -- shape is ever present. Graphically transmitted texts, by elementary "laws of form", automatically generate -- perhaps "incarnate" would be the better term -- shape, which emerges along with the primal "mark" that shapes Brown's demonstration.

So far as our common (visual) language is concerned, then, the elementary marks are an alphabet of letters, plus an accompanying set of signs -- explicit or implicit -- for reorganizing the letter marks into different scales and sets of relation. One can distinguish ("si puÚ distinguere") the most common elementary textspace as the page itself (in contradistinction, for instance, to a scroll's textspace or to the nonlinear textspace of, say, a cave wall). This page space is elemental because it replicates at a different scalar level the same kind of distinction marked within the page space by the elementary letter and graphic marks. The relation between the elementary graphic marks and the elemental page space sets the parameters for all types of graphemic directionality. In the spatial conventions of the page regulated at two-dimensions, the normative directions are horizontal left-to-right and vertical top-to-bottom (along with a normative line of directionality from upper left to lower right). So

A page of printed or scripted text should thus be understood as a certain kind of graphic interface. The complexity of the interface varies from a minimal use of the bibliographical codes open to a given paperspace - the text you are now reading is a good instance of such simplicity - to highly elaborated interfaces like those determined as poetic texts. Some of the latter exploit the bibliographical resources of paperspace to an extreme degree -- Pound's Cantos, for example, or Dickinson's various writings -- while others are satisfied to work within a set of basic and commonly used conventions. Whatever the specific differentials, however, a broad heuristic distinction separates informational from imaginative texts. The former aspire to transparency, the latter to noise, redundancy, repetition. One is vehicular, the other, iconic.

Exploiting bibliographical codes does not per se signal that poetic motives are governing a particular text, as the work of Tufte and Horne - as their own books - indicate. Both participate in a long tradition of "knowledge representation" that mixes textual and graphic forms (a tradition importantly advanced in the late work, the existential graphs, of Charles Sanders Peirce). It is crucial to realize that every text may choose, as it were, to engage that self-conscious tradition. We see this in the layout and general book design of Brown's Laws of Form. These apparitional features of the work show Brown's effort to meet the problem of effectively communicating himself. In the event, the rhetorical and design move would elucidate the Laws of Form at a much deeper level.

Figures I and II reproduce the first edition of Chapter 1, "The Form", which begins with a statement of assumptions, an initial presentation of The Form's elemental Definition, and a declaration of The Form's two primary axioms, the Law of Calling and the Law of Crossing.

 
1
The Form
_____________________________________________________________________________


We take as given the idea of distinction and the idea of indication, and that we cannot make an indication without drawing a distinction.  We take, therefore, the form of distinction for the form.


Definition

Distinction is perfect continence.

That is to say, a distinction is drawn by arranging a boundary with separate sides so that a point on one side cannot reach the other side without crossing the boundary.  For example, in a plane space a circle draws a distinction.

Once a distinction is drawn, the spaces, states, or contents on each side of the boundary, being distinct, can be indicated.

There can be no distinction without motive, and there can be no motive unless contents are seen to differ in value.

If a content is of value, a name can be taken to indicate this value.

Thus the calling of the name can be identified with the value of the content.


Axiom 1. The law of calling

The value of a call made again is the value of the call.

That is to say, if a name is called and then is called again, the value indicated by the two calls taken together is the value indicated by one of them.

That is to say, for any name, to recall is to call.
 
Equally, if the content is of value, a motive or an intention or instruction to cross the boundary into the content can be taken to indicate this value.

Thus, also, the crossing of the boundary can be identified with the value of the content.
 


Axiom 2. The law of crossing

The value of a crossing made again is not the value of the crossing.

That is to say, if it is intended to cross a boundary and then it is intended to cross it again, the value indicated by the two intentions taken together is the value indicated by none of them.

That is to say, for any boundary, to recross is not to cross.

That structure in Brown's book is extremely significant for understanding how Laws of Form operate in a textual horizon. The law of crossing governs reflexive functions, and in Brown's book reflexivity does not explicitly begin until the final chapter, whence it continues through the set of chapter notes and appendices that follow. How scrupulous Brown has been in constructing his text along the lines of his laws is underscored by the way those late chapter notes are (not) marked: Brown provides no links to them from within the chapter texts. The chapters draw and then draw out further and further distinctions until, at the conclusion of Chapter 11, the process itself reveals "that the account may be continued endlessly" (68). At that point an implicit injunction becomes explicit in and as Chapter 12, where we are called to reflection, that is to say, where we are called to cross back to Chapter 1: "we return for a last look at the agreement with which the account was opened". In point of textual fa

In the reflexive note to Chapter 2 of his book Brown explains that "the primary form of mathematical communication is not description, but injunction. In this respect it is comparable to practical art forms like cookery [and] music" (77). Although Brown does not include text production among these practical art forms, he might and indeed should have done so, as his own book admirably demonstrates. Clearly Dante regards writing, including poetical writing, as injunctive. Dante's word would not be injunction, however, it would be -- it was - rhetoric.<11>

As Brown's book shows, a primary textual injunction is to make and elaborate distinctions. If these distinctions are rigorously pursued they produce the realization "that the world we know is constructed in order (and thus in such a way as to be able) to see itself":

and so on, and so on you will eventually construct the universe, in every detail and potentiality, as you know it now; but then, again, what you construct will not be all, for by the time you have reached what now is, the universe will have expanded into a new order to contain what will then be. (106)

III

Not many works of philosophy demonstrate that paradox as elegantly as Laws of Form -- perhaps Nagarjuna's treatise on "Emptiness" makes an apt comparison.<13> Some do it with greater force -- Montaigne, Kierkegaard, Wittgenstein (especially in that relentless work known to us as the Philosophical Investigations). How to "edit" such writers presents a constant challenge and adventure, and hence a recurrent opportunity to address anew the unanswerable questions they raise.

Works of imagination, however -- let us say henceforth "poetry" -- make the discourse of paradox and contradiction the ground of their semiosis. In terms of Brown's Laws of Form, this means that distinctions in poetic texts are elaborated in a space that (so to speak) has no intention to maintain original integrity. As textual boundaries are defined and crossed, the marks of distinction constituting the boundaries are canceled or threatened with erasure: because new marks of distinction turn out to be phenomenal illusions, closely akin to mathematical transformations. New distinctions conceal algorithms - hidden injunctions - to cancel the same distinctions by recrossing the boundary initiated when the distinction first appeared to view.

Whereas everyone knows this about poetical texts, we are less clear about how and why this network of recursions unfolds. Yet clarity on the matter is particularly important in a digital horizon if we are to have any hope of building adequate electronic re-presentations of our received textual archive.

Modern linguistic analysis from Saussure to Hjelmslev to Segre develops a four-part analysis of the sign.<14> The signifier and the signified -- the elemental dismantling and reconstruction of the ancient distinction of form from content -- are each shown to replicate in themselves the form/content distinction because in any case both signifier and signified, in order to be recognized as such, have to be separately marked. The marking transaction creates, by the law of form, a new distinction - signifier v. signified - that dissolves a mistaken implication drawn from the earlier distinction of form/content. In the reconstructed sign both signifier and signified are not only "content constitutive", they are so precisely because of their "form function", because they have been marked.

What this otherwise useful analysis does not indicate - what it positively obscures, in fact - is the injunctive or rhetorical character of textuality. The structure of any text overgoes its own internal (signifier/signified) coherencies and/or contradictions. In Jakobsonian terms, this overplus is comprehended under the concept of "reference". The concept functions reasonably well in analyses focussed on informational and non-poetic texts, but its analytic force dissipates when directed toward poetry. This happens because a modern aesthetic understanding shapes our thought about "the literary text". Since poetical works are conceived as "communication sui generis" (or "language [oriented] toward the message in itself") (Segre, 28-29), neither affirming nor denying anything beyond their internal relations, "reference" in the literary text turns (virtually) virtual. They are, as marxists used to say, "not among the ideologies".

Texts seen in this light turn dark and passive. They seem not to address us but rather to lie down and await examination, like corpses under autopsy or treasured secrets. That textual structure was fashioned in the rhetoric of romanticism, where a textuality was sought that would not appear to have "a palpable design upon" the reader. But all texts are generated through algorithms; romantic texts are coded with special instructions to obscure the design codes, or rather to make those codes appear not as reading instructions -- not as marked text -- but as pure character data.

The signifier/signified/referent structure implicitly poses two (related) questions to a text: "what is it saying?" and "what is it doing in saying what it says?". This second question points toward the injunctive feature that is open in every text and in every part of every text. Ultimately one would want to be able to describe and analyze what literary texts are doing, "sui generis", in saying what they say - how they function in society and culture at large as they are literary works. To construct that kind of comprehensive analysis, however, we will have to undertake a thorough re-examination of their rhetorical and injunctive character at micro levels. The history of any text's emergence is both a record and an index of how it has been used, what it "meant". Those records should be recovered for a programmatic analysis of their injunctive features, which is to say for their Dantean "divisiones" and their Brownian "laws of form".
 

(In)Conclusion

In such circumstances what is needed is a dynamic engagement with text and not a program aimed at discovering the objectively constitutive features of what a text "is". That dynamic requirement follows from the laws of form themselves, as Brown's work shows. But what equally follows is that the analysis must be applied to the text as it is performative. We begin with an understanding that text is always the marked or materially distinguished text - the text as image and/or audition - and that the textual analysis is itself part of the marking processes that governs the object of study.

The problem at this point becomes at once more clear and more difficult to address. One is perhaps reminded of Brown's observation that "I found it easier to acquire an access to the laws [deliberated through the work] than to determine a satisfactory way of communicating them". This difficulty arises because the act of communication promotes and re-energizes the original (historical) ambiguity of the textual signifiers. They are arbitrary forms, open to an indefinite range of significations. The more complex the form of the signifier the more deeply run the ambiguous options of meaning. In the visible state of language they scale up from letters and diacriticals to wondrous scriptural and bibliographical creatures.

In this situation the limitations of determinately marked forms can be exploited for more dynamic operations. The proposal I imagine here is directed at visible text only and involves approaching the text not in terms of its semantic "content" but as a physically shaped construction. We do this on the assumption that the physical arrangement of the text amounts to a reflection or interpretation - a marking - of its semantic meaning. The reflection will inevitably introduce a "deformance" of the work, and thus will appear in one or another perspective as what Joyce once called "a cracked mirror": because even at the purely bibliographical level the semiosis of any specific set of signifiers will lie open to different possible readings. The text's non-self identity extends itself through all marked levels precisely because it is the operation of marking that divides the text from itself.

The project imagined here attends only to the text's bibliographical codes in order to begin with a relatively simple set of rules for marking or interpreting textuality. We want to teach the computer a set of rules for reading texts. Trying to teach it higher order rules presents enormous difficulties. It seems possible, however, to develop an initial set of rules for bibliographical coding options and forms. Part of the programmatic operation is to implement these rules in order to expose and generate a more complex set of rules extending to higher orders of textual form.

The ultimate event in this program will be a dialogue between the computer and the human beings who are teaching it how to read. We want to study the bibliographical formations that appear out of the computerized readings. These readings will, we believe, inevitably constitute a set of (de)formations full of surprises for the rule-givers. What those surprising readings will be cannot be predicted, but that they will come is, we think, as certain as the fact that no text is commensurate with itself.

The more sophisticated we are the more we normalize textual incommensurates. We have internalized an immensely complicated, many-levelled set of semiotic rules and signs, and we control the contradictions of actual textual circumstances by various normalizing operations. We can hope to expose these normalizations - which are themselves deformative acts - by opening the conversation here being proposed between analogue and digital readers. We begin by implementing what we think we known about the rules of bibliographical codes. The conversation should force us to see - finally, to imagine - what we don't know that we know about texts and textuality.

At that point, perhaps, we may begin setting philology - "the knowledge of what is known", as it used to be called - on a new footing.

Endnotes