Among my dirtiest scholarly secrets is that I still feel Derrida’s work prefigured many of the debates about writing and media we are currently having in the examination of digital culture. I have no wish to take another spin on the language turn, recenter critical theory in the literature classroom, or read new batches of the sometimes leaden deconstructionist criticism. But, especially after reading and then re-reading Juliet Fleming’s recent book Cultural Graphology, I can’t shake the feeling that Derrida was correct about a lot of things about pre-digital writing that continue to be true about digital texts of all kinds. In his final collection of interviews and essays, Paper Machine, Derrida, among other things, addresses the effects that freely available digital technology has had on his and others’ writing practices and the production and reception of texts in digital formats. As much as I feel his ideas here to be valuable, its some of his older texts I found to be the most relevant to current scholarship on digital technology. However, more than simply connecting Derrida’s work to these newer ideas, outlining functions of digital media that adhere to dynamics Derrida describes arising in print also reveals previously invisible features of those pre-digital media. I’m most interested in examining the overlaps between writing for humans in literary texts and writing for machines in programming or coding.
In her 2017 book, Coding Literacy, Annette Vee compares computer code to human language in order to arrive at a definition of literacy that applies to both (that description doesn’t do justice to her nuanced and convincing argument). Vee uses literacy to examines programming as more than simply a “problems-solving tool,” but also as “a species of writing” (loc231). In arriving at her definition of literacy for both code and writing, she enters into a discussion that bears striking similarity with a debate that took place decades earlier between Derrida and proponents of Speech Act Theory. Vee says, J.L. Austin, in How to Do Things with Words, “ultimately fails to find a basis to separate performative statements from descriptive statements. By the end of the lecture series, he proposes that all language implies a kind of action” (loc3379). The distinction between performative and non-performative language, in other words, isn’t quite air tight and therefore its usefulness as a category of speech is diminished. Definitions of code, Vee says, that rely on executability confront the same problem. Ultimately, Vee rejects the notion that code can be defined as a mode of discourse through its executability because the execution and the writing of code are not identical:
While the relationship between code language and execution is slippery, code’s status as simultaneous description and action means it is both text and machine, a product of writing as well as engineering. Programming became a kind of writing when it moved from physical wiring and direct representation of electromechanics to a system of symbolic representation. (loc599).
Between the writing of the code and its execution, in its transition from the symbolic notation to the physical action, a gap emerges. The action, it turns out, even when it is to be carried out by a computer, requires interpretation, and that opens up entirely new range of considerations.
In his book, The Stuff of Bits, also from 2017, Paul Dourish seeks to ground the seeming immateriality of information in the material objects which process and present them. Like Vee, Dourish complicates the idea that code can be understood based on its ability to carry out tasks. He shows how executability of a given program–the written instructions computers carry out–depends on the machine that runs it:
The materialities of digital systems lie not least, then, in this gap between what is denoted and what is expressed, or between the specification and the execution. A computer program may be a precise series of instructions, and yet the experience that results from the program’s execution is radically underspecified by that program. (loc519)
Programs, texts written with machine audiences, require precise procedural language that makes their instructions as explicit as possible, yet the way these instructions are executed varies from one machine to the next. This is because, despite its explicitness, “There are ways of carrying out the tasks specified in the program that are not part of the program’s text; and the computational platform has properties that affect the program’s execution that are also not described in the program itself” (loc509). Between the written text and the execution of its instructions, is a gap in which the computer’s properties influence how the program is executed.
Where I’m going with this is pretty obvious. Derrida makes the same argument about all written symbols and their interpretation. His obsession with limiting cases and extreme examples means he focuses on the aporia that emerge between the writing and the interpretation. Though written and spoken language works very well at communicating information every day, Derrida, like Vee and Dourish, foregrounds the failures (Austin would say infelicities) that emerge between a text’s intention and the interpretation by its receivers. Derrida says, “The conscious presence of speakers or receivers participating in the accomplishment of a performative, their conscious and intentional presence in the totality of the operation, implies, teleologically that no residue escapes the present totalization” (14). Of course, writing would not be writing if both the speaker and receiver needed to be present in order for its intention to be conveyed. And even if that condition were met, as those of us who teach well know, being in the same room with an audience and trying to clearly express yourself fails with discouraging frequency.
What is absent in writing and speaking is the full expression of context or intention, both for the speaker and for the receiver. It is impossible to express that context in the text and equally impossible to anticipate all the contexts in which it will appear again and the audiences that will consume it. Derrida says,
[T]his essential absence of intending the actuality of utterance … prohibits any saturation of the context. In order for a context to be exhaustively determinable, in the sense required by Austin, conscious intention would at the very least have to be totally present and immediately transparent to itself and to others, since it is a determining center of context. (18)
The writer and coder or programmer cannot include the context in which their text was written or account for the new contexts in which it will be read. Dourish says that no matter how complete the specification, the machine will interpret the program in unpredictable ways: “The mechanics of devirtualization–of the production of actual effects based on digital specifications, be that the running of a program or the rendering of an image file–inherently exceed the reach of the specification” (loc521). Vee, Dourish, and Derrida all insist on attending to the space between writing and reception which opens space for interpretive ambiguity. What is novel about pointing this out, to my mind, is that what Derrida describes in communication between humans, also takes place when a series of precise and painstakingly procedural instructions are interpreted by a machine.
What does Derrida add to Vee’s discussion of coding literacy? What does he contribute to Dourish’s discussion of digital materiality? Not that much, really. They both make convincing arguments without, to my knowledge, once referring to the French thinker. The value in seeing Derrida expressing ideas similar to these studies of digital media is that he was, for the most part, talking about pre-digital media. This underscores that all types of writing have certain features in common. But Derrida had been saying that for years. If all writing is in some sense the same, how does code differ from human language? If we can’t base the difference between human and machine writing on executability, then is there a way to differentiate the two types?
Vee offers some answers to this question, but also shows how difficult it is to definitively separate human from machine language. She says, “Programming and writing are not the same thing, but they have a lot in common and can even merge into each other” (loc3348). This merging is what’s interesting to me. Vee argues that human writing and code, by both being discretized forms of communication, are productively constrained by formal requirements: “Constraints in form such as those that discretization provides can foster creativity, even poetry, in both programming and writing. In programming, these poetic forms often highlight the dual audience for code: computers and humans” (loc3389). Vee’s point here, I think, is very exciting. The reading of code also requires understanding the constraints, the formal requirements of a given computer language or platform for instance, but also the computer as audience. The language of code is addressed, in some sense, to the machine, speaking in its idiom so that it will understand. Reading code as a human requires understanding how a machine understands. It is not correct, however, to say that code is solely addressed to the computer. In some sense, it is always also addressed to the human reader, or at least the human reader willing to learn the computer language. Is the reverse true? Are the constraints, the implicit rules according to which literary aesthetics operate, similarly addressed at once to a human and non-human audience?
My contention is that the avant-garde experimentation with language taking place in the modernist period can be better understood through comparisons with code. It is easy to suggest that the formal requirements of poetry, Petrarchean Sonnets or Terza Rima, operate according to certain rules and therefore resemble code. I’m arguing something slightly different. Interpreting certain literary texts requires reading with this dual audience in mind. Like code, the poetry or prose is not only addressed to the human reader, but also to another, more difficult to define audience. If code is addressed to the machine, literary language in these texts is in constant dialog with other language also appearing there. Interpreting these texts requires understanding what certain elements mean, but also how they interact or “speak” with each other. Unpacking the portmanteau words in Finnegans Wake, tracing local and remote patterns in The Cantos, or understanding how the final rhapsody of The Making of Americans relates to the 900 pages that precede it, requires overhearing and understanding the conversation taking place between elements of information. Though it is a far cry from the explicit machine language that appears in computer programs, I argue this these dynamics resemble the “dual audience” at work in code that Vee describes.
This is the value in connecting the discourses around print and digital writing of various kinds. We can situate the texts that make computers work within the ancient practices of writing, but also reveal features of print texts that have only become visible in digital contexts. The above comparison between Vee, Dourish, and Derrida seeks to crosshatch the discourses of code, literacy, digital materiality, and literary language. If code and human language have these features in common, if they are subject to the same limitations and affordances, there is value in comparison from one to another.