Providence, RI 02912
Jerome J. McGann
Department of English
University of Virginia
Charlottesville, VA 22906
The Text Encoding Initiative and other recent markup research projects in humanities computing have fostered lively debate and discussion on the nature of text. If we are to represent text in a form suitable for electronic processing, we need to define what text is and what structures we want to be able to recognize in text. This focus on the structural components of text, pioneered by Allen Renear and his colleagues at Brown University, became the foundation of much of the SGML-based encoding seen in today's electronic text projects. But it also developed at a time when many textual scholars, notably Jerome McGann, began to move away from the linguistic and structural properties of text and to place more emphasis on what McGann calls "bibliographical codes", that is on texts as material phenomena where appearance and rendition are important.
This session will take the form of a debate between Renear and McGann, two well-known scholars who have written and spoken extensively on the nature of text. Each will present their views in an opening statement of approximately twenty-five minutes. They will then be given a further five minutes to reply before the topic is open to the floor for further discussion and questions.
I will advocate this theory by defending five theses; I will argue that texts are:
I will rehearse the arguments, which I believe to be decisive, and consider alleged counterexamples, which I believe fail. I do not claim that this view encounters no hard cases, or that it solves all problems. But in the course of the discussion we will see that this account of text is rich in explanatory and predictive power, implied by our modal intuitions about cultural artifacts, and useful both in regulating further inquiry, and in guiding the development of tools and resources. Finally, and very importantly, I argue that this view has no serious competitors. It is the best account of text that we have, and, fortunately, it is a good one.
This theory of text has been developed in collaboration, over the last ten years, with colleagues at Brown and elsewhere -- however most of those involved would demur from the ambitions, and commitments, of my formulations here. The development of this view can be found in the following papers.
Although criticisms of this view from within the humanities computing community have been expressed in conference papers by Claus Huitfeldt, Dino Buzzetti, Fabio Ciotti, and others (see: http://www.hd.uib.no/AcoHum/abs/orel.htm), it is widely thought that very serious problems for this theory can also be found in the influential work of the theorist of textual criticism, Jerome McGann.
In the field of Humanities Computing the idea of text has been dominated by conceptions practically realized in the TEI implementation of SGML markup. Several key theoretical papers published by Steve DeRose, Allen Renear, "et al." explain the ground of that implementation.
This ground, explicitly "abstract" (Renear1997), represents a view of text as essentially a vehicle for transmitting information and concepts (final cause). Text is "hierarchical" (formal cause) and "linguistic" (material cause), and it is a product of human intention (efficient cause).
I invoke these Aristotelian categories because Renear correctly insists upon the "platonic" character of the TEI/SGML approach to textuality. That self-description, traceable to several Platonic works, the Republic in particular, helps to clarify the differential involved in "poetic" or noninformational forms of textuality. There is no question but that most of our textual archive is hierarchically organized. On the other hand, there is also no question but that poetical texts comprise a key, perhaps even a defining, part of the corpus of our humanities archive. When Plato called for the expulsion of the poets from the city, he was arguing for a certain theory of textuality.
Unlike expository text, poetry is not organized in a determinate hierarchy. TEI and SGML markup, therefore, while reasonably adequate vehicles for expository and informational texts, fails to render those features of poetic text that are most salient for its makers and users. Poetical texts are recursive structures built out of complex networks of repetition and variation. No poem can exist without systems of "overlapping structures", and the more developed the poetical text, the more complex are those systems of recursion. So it is that in a poetic field no unit can be assumed to be self-identical. The logic of the poem is only frameable in some kind of paradoxical articulation such as: "a equals a if and only if a does not equal a".
This essential character of poetical text helps to explain why content in poeisis tends to involve more broadly "semiotic" rather than narrowly "linguistic" materials. The sonic and visible features of text are, so far as the poets who make these texts are concerned (or the readers who engage them), nearly as apt for expressive poetical purposes as the semantic, syntactic, and rhetorical features. Each of these features represents a field of textual action, and while any one field may be individually (abstractly) framed in a hierarchized scheme, the recursive interplay of the fields produces works whose order is not hierarchical. Of course a governing hierarchy can be imposed upon such works. TEI and SGML create, as Renear shows, a certain type of "linguistic" hierarchy, one that privileges text as a container for storing information. But even that linguistic hierarchy is highly specialized (it does not consider, for example, the rhetorical structures that overlap and infect the syntax and semantics).
The case of poetry in fact defines a kind of textual ethos, as it were, that may be seen to pervade genres not normally thought of as poetical. Certain kinds of philosophers lend themselves to a hierarchical approach - St. Thomas, Kant, Hegel. Others don't. Not without reason has the Bergen Wittgenstein project abandoned TEI/SGML as a system for marking up the corpus of Wittgenstein's texts; and the scholars setting out now to "edit" the Peirce archive are well aware that TEI/SGML does not lend itself to an adequate treatment of Peirce's work, and least of all to his existential graphs. "Text" in Kant "is" one thing, but in Peirce it "is" something else again.