Bitstreams

The Future of Digital Literary Heritage

Matthew G. Kirschenbaum

• Publisher: University of Pennsylvania Press
• Year: 2021
• ISBN:
  9780812298048
• Topic: Letteratura
• Subtopic: Critica letteraria nordamericana

What are the future prospects for literary knowledge now that literary texts—and the material remains of authorship, publishing, and reading—are reduced to bitstreams, strings of digital ones and zeros? What are the opportunities and obligations for book history, textual criticism, and bibliography when literary texts are distributed across digital platforms, devices, formats, and networks? Indeed, what is textual scholarship when the "text" of our everyday speech is a verb as often as it is a noun?These are the questions that motivate Matthew G. Kirschenbaum in Bitstreams, a distillation of twenty years of thinking about the intersection of digital media, textual studies, and literary archives. With an intimate narrative style that belies the cold technics of computing, Kirschenbaum takes the reader into the library where all access to Toni Morrison's "papers" is mediated by digital technology; to the bitmapped fonts of Kamau Brathwaite's Macintosh; to the process of recovering and restoring fourteen lost "HyperPoems" by the noted poet William Dickey; and finally, into the offices of Melcher Media, a small boutique design studio reimagining the future of the codex.A persistent theme is that bits—the ubiquitous ones and zeros of computing—are never self-identical, but always inflected by the material realities of particular systems, platforms, and protocols. These materialities are not liabilities: they are the very bulwark on which we stake the enterprise for preserving the future of literary heritage.

1 | Archives Without Dust

It’s startling to see it on the Dell flat-screen set up lengthwise to duplicate the orientation of a sheet of paper: “BELOVED A NOVEL BY TONI MORRISON.” These words are typewritten in the center of a badly burned page, barely legible through the discoloration of smoke and water damage. The edges are blackened and irregular; most of the top left quadrant is scorched away entirely.

I am looking at a high-resolution digital image of a hard copy printout of a working draft of the Pulitzer Prize–winning book. It is one of the manuscripts that survived the house fire that engulfed Toni Morrison’s Grand View-on-Hudson residence at the foot of the Tappan Zee Bridge on Christmas Day in 1993. The crumbling pages have been conserved and scanned by the Department of Rare Books and Special Collections at Princeton University Library. Here all access to Morrison’s papers—damaged or otherwise—is via digital surrogates viewed at a dedicated workstation in the reading room, network and USB ports disabled. (A note in the finding aid informs patrons that this mode of access is by request of the author.) Pace Carolyn Steedman, there is no dust in this archive and thus no risk of archive fever, at least not in any pathological sense.¹

The haunting digital image in front of me registers as a testament not only to the resilience of paper and ink in the face of fire, smoke, and water but also to the efficacy of the collective array of memory institutions that ensure that treasures such as this are—in the novel’s own parlance—passed on. Despite the screen that separates a researcher from the painstakingly preserved originals, it is easy to picture scholars doing much the same work they have always done: scrutinizing drafts for variants, poring over the author’s correspondence, squinting to decipher handwriting on the pages of yellow legal pads. And the truth that such passings-on don’t just happen by happenstance is one of the most basic lessons Beloved has to teach us. (Appended to a different draft in the digital collection is a note from someone at Morrison’s publisher, Knopf: “Dear Toni, Look what I came across in the deepest darkest recesses of the bookshelf behind my door,” it begins. “I thought I should send off this heirloom that should be passed down through the generations.”) Their digital pedigree notwithstanding, however, a scholar who wishes to view these heirlooms must still travel to Princeton, New Jersey—perhaps as I did, first by riding Amtrak and then local commuter rail to nearby Princeton Junction and then clambering aboard the one-car “dinky” train to the campus. One then makes their way to the looming gothic façade of the Firestone Memorial Library. Only after my visitor badge had been procured, a security checkpoint passed, two flights of stairs descended, my bags stowed in a locker, a check-in at another desk, and my hands scrubbed at a plain porcelain basin could I enter the reading room with its green-shaded bankers’ lamps, where most collection items (unlike Morrison’s) are still wheeled out in Hollinger boxes on trolleys.

2 | Fire-blackened typescript pages of Toni Morrison’s Beloved with a glitched portion of one of the Microsoft Word files created during the novel’s revision process displayed together on screen in the reading room at Princeton’s Firestone Library. Photograph by the author. By permission of the Princeton University Library, Special Collections.

As compelling as the fire-damaged page images were, they were not what had brought me to the archives. Among the 200 linear feet and some 325 boxes of material that comprised Morrison’s papers when they were first acquired in 2014 were 150 floppy disks, most of them having passed through the hands of one of her office assistants. The majority of the disks were of the 3½-inch variety, but there were 33 in the older, more cantankerous 5¼-inch format. Of these, 4 had been identified as containing files pertaining to Beloved. I knew that those files had been recovered using a process that involved connecting a vintage 5¼-inch drive to a modern computer using a purpose-built controller card as the hardware bridge.² Despite their being nearly thirty years old, the stored data had proven legible. I was also lucky in that Morrison’s assistant had been using a DOS PC running Microsoft Word, so the files were saved in a familiar format, and that the diskettes themselves had been kept at a university office, thereby sparing them from the ravages of the fire. Nonetheless, when opened on the reading room workstation at which I sat, some of the electronic documents were shot through with junk code, the artifacts of formatting instructions long out of date and thus meaningless to the WordPad application the computer was outfitted with. These artifacts had the effect of scrambling certain lines of prose, usually where the page breaks in the original document had been.³ No dust in this archive then, but instead what is sometimes called bit rot—the inevitable decay of digital information in whatever its stored form. When these glitched electronic documents were juxtaposed with the scanned images of the fire-damaged manuscripts, the screen in front of me bore witness to two very different kinds of textual trauma.

The most basic thing I wished to accomplish during my visit was to understand the relationship between the files on the four diskettes and the other Beloved manuscripts in the collection. Were the contents of the Word files duplicated in the page scans of hard copy materials like the one that had already arrested my attention? Or did the recovered files represent unique states of the text? And at what stage of the composition process? Questions like this are becoming increasingly commonplace as so-called born-digital materials take their place alongside manuscripts, typescripts, and printouts in archival collections. Practically speaking, there is no easy way to obtain the answer—digital files must be examined on a case-by-case basis alongside their analog counterparts in order to arrive at a complete view of the extent to which any given text’s composition history is available to us through extant versions. That Morrison’s paper-based materials are themselves delivered as digital page images is incidental in this regard as there is no easy way of automating searches or other comparative operations. And yet the homogeneity of the means of access in this case—both the digital photo of the damaged manuscript and the glitched text of the legacy .DOC file are each just a click away in the online finding aid—underscores the necessity of treating printed and digital materials as each mutually constitutive of the work’s textual condition.

I could see from the date-stamps that all the recovered files—there are five of them across the four diskettes, each containing a different portion of the novel—had been created between July and September of 1986, the period just before the text went to Knopf for presswork.⁴ The files closely resembled the so-called setting copy of the text delivered into Morrison’s hands from her publisher, which she then penciled with final corrections and changes (the setting copy was not damaged in the fire and is itself accessible in its entirety at Princeton). Nonetheless, at least one of the electronic files—a document named BELOVED3.DOC— appeared to capture a minor variant state of the text not otherwise represented in the collection materials.

Readers will know that the novel ends with its title word, “Beloved,” which is also the name of the two-year-old girl whom her mother, Sethe, mercy kills with a handsaw rather than allowing her to be taken by slave-catchers. In her earliest drafts, however, Morrison had instead been ending the book with “A hot thing,” a phrase that appears in a monologue tendered by Beloved herself in an earlier chapter—it refers to Sethe’s furnace-like love. Sometime late in the composition process, Morrison had also hand-penciled “Beloved” just above, thus making the ending read (now on two separate lines): “Beloved. / A hot thing.” Morrison’s editor at Knopf, Robert Gottlieb, then struck out “A hot thing,” leaving “Beloved” as the final line and the final word.⁵

This was not the only consequential change effected by Gottlieb. In an essay she would publish a decade later, Morrison reflected back on the novel’s editing process, drawing attention to what was to become its penultimate line, “Certainly no clamor for a kiss.” She had originally written “Certainly no clamor for the join.” That word—likewise resonant throughout the novel, and accompanied by a definite rather than indefinite article—was chosen by Morrison because it “connected everything together from epigraph and the difficult plot to the struggles of the characters through the process of re-membering the body and its parts, re-membering the family, the neighborhood, and our national history.”⁶ Gottlieb, however, felt that such an unusual word was a distraction on the last page and asked for something less “esoteric.” Morrison reluctantly settled on “a kiss” instead, but she had remained uneasy about the decision: join may indeed have been the “wrong” word, as she would later put it, but it was also the only right word.⁷

And yet, despite revisiting this editorial episode at some length a decade further on, Morrison makes no mention of the removal of “a hot thing,” which had remained in the closing array of lines even longer. Indeed, Morrison had continued to work with the last two lines right up until the novel went to press. She appears to have first tried inverting them, and the BELOVED3.DOC file captures the text in this unique sequence: “A hot thing. / Beloved.” (Here “the join” has also not yet been changed to “a kiss.”) It is not until the setting copy produced soon after to create the first proofs of the book that “a hot thing” at last disappears, still printed as the penultimate line (despite Gottlieb’s initial cancelation) but struck out there once again with a final, decisive pencil line. What this means is that the digital file BELOVED3.DOC most likely captures this portion of the text somewhere in between Gottlieb’s editorial work (and his request for a word other than “join”) and the setting copy, by which time “the join” had become “a kiss” but “a hot thing” still persisted—as best I could determine, a unique state of the text not otherwise represented in the manuscripts available at Princeton.

Associated as it is with both the strength of Sethe’s love and her decision to end her daughter’s life, “a hot thing” is a key phrase in any reading of Beloved. Its presence in the closing lines of the text throughout nearly the whole of the composition process—persisting even longer than “the join”—underscores the extent to which Morrison used it to bear the emotional weight of the novel, offset only by the title word itself. Moreover, this particular sequence of variants raises the possibility of other, less conspicuous changes elsewhere in the BELOVED3. DOC file. Why does there not appear to be any corresponding printout of it that would have been digitized for researcher access by the Princeton team? Perhaps one simply did not survive, or perhaps the file was never printed in that exact state. Regardless, it is clear that Beloved’s final lines remained in flux right up until the moment the book went to press, and that any complete account of its textual history would have to draw upon the evidence of the surviving digital files alongside the evidence of the surviving notes, jottings, printouts, galleys, and proofs.

This was as far as I could get in the course of a day: the determination that the contents of at least one of the electronic files appeared to be very close to but not exactly identical to any of the hard copy manuscript materials available to researchers as page scans.⁸ Given that others of Morrison’s novels in the Princeton collection are represented by born-digital materials even more extensively—for example, Paradise (1997)—the value of examining these files alongside other manuscript materials should be plain. Most immediately for our purposes, the ultimate incommensurability of the scanned images versus the actual digital document files is an object lesson in the non-self-identicality of the bitstream. Both forms of content, after all, were right there on the screen in front of me, presented as digital objects. But whereas the page scans of the manuscript materials are simply high-resolution photographs with little capacity for user manipulation beyond zooming or rotating or applying some stock filters, there is the potential to explore the word processing files in more diverse ways, exposing the bitstream to various kinds of automated analysis and interpretation. This difference is an outgrowth of the fact that the underlying binary code can be operationalized in very different ways depending on whether the data in question is string-based (like text) versus a pictorial bitmap. The former is searchable, reconfigurable, and potentially even executable; the latter lacks the capacity for manipulating discrete elements at a meaningful semantic level. There is thus no way to make comparisons between these two different data types short of manual collation, which is to say by reading and eyeballing the scans and the document files side by side on the screen.⁹ File formats and their underlying data representations—JPEG and DOC, TIFF and TXT—create many of the conditions that enable, as well as constrain, the bibliographical encounter with the bitstream.

Manuscript archives have been receiving born-digital materials for more than two decades now, an inevitable outgrowth of the widespread shift to word processors and personal computers that swept the global North in the 1980s. When J. G. Ballard’s papers arrived at the British Library in 2011, the Guardian opined that it was perhaps “the last solely non-digital literary archive of this stature.”¹⁰ This is not strictly true of course, but it captures the reality that an author’s papers nowadays are very likely to also include digital media, everything from diskettes and CDs to hard drives and even entire computers.¹¹ Archivists, for their part, have become increasingly adept at the intake of such materials, at least the more common types: there are procedures and best practices for lifting the bits from different types of media and devices, for indexing their contents, for safeguarding the files on up-to-date servers, and for making copies available to researchers—although provisions for that last can be imperfect, as we have already seen. Yet whatever technical challenges they presented to the staff at the Firestone Library, Toni Morrison’s diskettes (dating from the 1980s and 1990s) were still a more straightforward case than the scenario I depicted in my introduction. At Princeton, all the digital content was on physical media the archivists were able to process without regard for any conditions or terms other than those originating with Morrison herself. There were no re mote servers, cloud services, or social media trails to untangle. All the files were also from easily identifiable software products in common commercial formats. Nonetheless, even in this instance, the bitstream twists and torques beneath one’s fingertips, as manifest in the glitches we observed when the files were opened in WordPad.

The challenges archivists now face when confronted with the digital legacies of writers or other public figures are, of course, just a reflection of a much ...

Table of contents

• Cover
• Title Page
• Copyright
• Dedication
• Contents
• Preface | Actual Facts
• Introduction | The Bitstream
• 1 | Archives Without Dust
• 2 | The Poetics of Macintosh
• 3 | The Story of S.
• Coda | The Postulate of Normality in Exceptional Times
• Notes
• Acknowledgments
• Index