Essay

Defenders of the Text.

William Clark  ·  August 10, 2026

†The title is borrowed from Anthony Grafton, Defenders of the Text: The Traditions of Scholarship in an Age of Science, 1450–1800 (Cambridge, MA: Harvard University Press, 1994).

Who are the defenders of the text? They are a motley band of heroes, one to be remembered and revered. And so they deserve a real exordium, in the style of Milton’s orations: A lavish introduction. Perhaps though it is a funeral oration, for their time is now past. So we are in a better position, perhaps than ever before, to survey the work of those scholars who defended the text—that is, those who discovered the text; who preserved it; who learned to read it; who transmitted it; who studied it; who understood it; who edited it; who published it; who revised it; who expanded it. In short, the whole clockwork mechanism by which the debris of the past is sifted into history. Far from being invisible, the defenders of the text proclaim themselves, and imprint themselves quite indelibly onto its edifice, through their critical activity, which binds it all together as superstructure. No wonder that so often in the history of the English language, the defenders of the text have been heralds. Those men who attended on the knights of olde, and trumpeted about their fame. Camden; Dugdale; Peter le Neve; Elias Ashmole. Heralds all. Officers of the College of Arms.3 Let us then assume for a bit the role of herald’s herald, and recount their past triumphs a bit.

There is a continuity through the ages, which the antiquaries; historians; and critics have perpetuated through their tireless excavation of the past, which itself is the lower substratum of the very structure they are erecting. And so debates resound through time; they repeat, unrelentingly. The pattern of discourse is both continuous, and discontinuous. Its continuities are hallucinations. Its discontinuities are reaffirmations and reconciliations.4 The question I mean to pose has to do with how we are to continue/salvage/reinvent a modern textual criticism in the age of large language models. It may now appear as if the culture of textual criticism has reached a definite historical end, is that really the case? Opinion essays lament the end of literacy, perhaps rightly so, as now machines do all the reading and writing for us. But is this truly a dis-continuity, or an ever deeper one? If this is true, if text-bound literate culture is coming to an end, then it presents us the perfect object of antiquarian investigation. The antiquary dwells on graves to inspect the tombstones—a certain melancholia prevails among the defenders of the text as a result of this.

To see this one must only imagine Aubrey while at Oxford—which coincided with the King stationing himself there during the Civil War. Melancholia was in the air. Burton had not long since done away with himself. Hobbes was in the midst of the students and regaled them with tales of his conversations with Francis Bacon. Likewise William Harvey was with the King as physician, and injected a current of scientific and experimental excitement. The past came under the microscope of reflection as an object of science in this laboratory, as Harvey demonstrated experiments on chicken embryos to the students. Yet the impending doom of Charles I hung over the whole procession as the young students consorted with the officers engaged in the king’s lost cause, who in turn became enamoured with early Freemasonry and formed secret ambulatory lodges—cells loyal to the soon to be deposed King.5 Aubrey recalls how he and his companions rode out one day, from the city, to make drawings of a cloister outside the walls, fearing that it would be destroyed by the king’s besiegers. Surely enough, his foresight was correct. He had captured the monument at just the right time.6 This is the melancholia that pervaded the generation of antiquaries following the Restoration. The melancholia of a Civil War—just as the melancholia of the Reformation loomed over and pervaded the first-generation of antiquaries7. The shadow cast by the accumulations of the past were so black as to drown John Leland in darkness—who was driven mad by a kind of record fever trying to collect and catalog the material that had been put up on the chopping block of time.8 To further underscore this, we might note that the Academy of Antiquaries which formed around William Camden met on All Souls’ Day, the day of the remembrance of the Christian dead.

History begins with the grave marker; the inscription; the monument, and this doesn’t end with physical artifacts, but carries over to textual ones as well. Documents are often described not just as “records”—but as memorials and monuments. This is further conveyed in the term “muniment,” which refers to documentary proofs of title that were archived in a specially outfitted vault that every great family had installed in their home. So while the antiquary etches the epitaph, he or she also erects the monument—and it is from the monument that we learn literally everything—because everything else about the past is gone. We may bury the defenders of the text, and pen their eulogy; but, in that very act, of melancholy reflection, we also extend them, and we continue their effort anew. We remember here a textual culture now being driven to extinction—but we cannot even contemplate that extinction without thereby inverting the discontinuity. Is, therefore, the current moment really a rupture at all? I was struck by the provocative closing lines of Jonathan Goldberg’s Writing Matter,9, following his final survey of the work of twentieth-century theorists on the oppositions of handwriting and typewriters, where he writes.

But now the fantasy is that somewhere there is (or someday there will be) a program, and rather than having to write oneself, the machine will be able to do it.

Goldberg recognizes that the supposed dichotomy between handwriting, as something personal and individuated; against the typewriter, as something mechanical and anonymous—is critically untenable. Handwriting, as it was taught in the seventeenth century, was also a mechanical, anonymizing system where expert secretaries moved seamlessly between various hands—between secretary and italic scripts; between court hands and the signatures of their masters, which they learned to forge—which they controlled. Perfection was in the perfect copy of the forms distributed by the writing masters—some of whom were zen-like gurus of the hand capable of incredible feats of automatic dexterity. The typewriter and the word processor, therefore, are not discontinuous with the individuating authenticity of handwriting; but rather a reentrenchment and reaffirmation of it. The original large language models are the scribes and secretaries who copied and recopied the text. The scribe made fair copies of authors’ manuscripts for use by the typesetters. The scribe was a scrivener in a law office, copying out the textual legal instruments through which the state perpetuates itself—through which it manifests its authority. The scribe copied works for manuscript transmission, like newsletters. The scribes were the secretaries of powerful men—of the King’s favorites—even the King had his own secretaries, who wielded immense power acting as the monarch’s own hands. By assuming his authority to write—by writing on his behalf.10 To write for the monarch, or the state entailed depersonalization—the secretary gave up his identity, and assumed the identity of power itself—of the state. They thereby participate in this power, and perpetuate it. These men (and less occasionally women) were educated in a kind of rhetoric, handwriting, and textual culture that was often very mechanical. Secretaries kept books of model epistles, which could be refashioned to serve any official circumstance. Secretaries produced and reproduced those texts through a discipline of handwriting. This textual culture—whose passing away we are in the midst of lamenting—it was a generative language machine. The rupture between past and present is perhaps just another hallucination generated by this machine.

This essay will invoke a rather expanded concept of the public textual domain—as the historical corpus of text produced by this same machine; the text that has been generated by it. There is something public about all text—as text is a social performance of common language—and that extends across all paper and digital communications—past & future—as everything ultimately reverts to the public domain; which is the reservoir of the English language, and contains all its forms and resources.11 All contemporary linguistic and textual utterances are rooted in this same soil; every remains of every crop harvested from it ultimately are mulched back into the soil—restoring it. This extends to the code of computer files we create. This is the same soil out of which LLMs are grown—it is their training data as well as ours—it is our common literary and linguistic property. Every paperback edition of every literary classic is ultimately based on a manuscript or early printed edition which is an object in a library or archive somewhere—and these objects are studied and reconstructed by the bibliographer and the textual scholar to create the basis of millions of copies across an untold number of printings and editions. The text is not something abstract, or without body—it is manifest in physical objects—in autograph manuscripts, legal documents, corrected proof-sheets, annotated copies, commonplace books, correspondence—and those objects are often the real property of public institutions. This essay will mostly focus on the English experience, until the early twentieth century, and will therefore focus on institutions like the Public Record Office, the British Museum Library, the College of Arms, the Stationers’ Company.12 But the analogies to the American case can be easily drawn13, and there are further parallels to be drawn between England and Prussia.14 So, even though we confine ourselves to particular historical examples, we should not limit the consequences of our discussion to only a single particular time and place; rather the implications extend to the entirety of our shared English language and the textual corpus in which it’s rooted.

The public domain of writing is the storehouse of precedents upon which English language diction and usage is based. It includes those documents harkening back to Anglo-Saxon times. English, as a language, has been reared up on this material. It is not just a dead record, but also the precondition of literary communication in our language, the fountain from which it flows, or is generated. In the same way creations of Large Language Models are generative and presuppose this textual corpus, so is all literary production generative, and based on the public record. This public record includes lives of saints, medieval chronicles, state papers, court records. All literary work is based in the community of language, which is collective and historical.15

Previously a class of men and women lived who defended this textual domain, for the public good. They were bound, as Knights are by their code of chivalry, to an ideal of textual criticism, and they fought bloody critical conflicts with one another over those ideals. Large language models have now come to defend the text themselves, through their own mathematical, anonymizing, mechanical process. The Public Record Office has been replaced by the data center. And instead of textual criticism being a public rhetoric of controversy, the algorithms which generate and extend the text have been locked away, and become a corporate property and trade secret. The public is shut out of the text in more ways than one. Copyright laws work against the public’s right in its own communal legal and linguistic existence, denying and restricting individual use, while those same repositories are plundered with impunity by AI companies and the entertainment consortium. The discourse of the antiquaries and critics and bibliographers was a machine by which the debris of the past was sifted and deployed to various ideological ends. The rhetorical modes and genres of discourse regulated the operation of the defenders of the text—deindividuating them and bringing their discontinuities into alignment as part of the textual structure of social power. Much of this can also be said about Large Language Models and their function within the text. But is this continuity an hallucination? What separates the kind of textual criticism that we now see passing away? What can be preserved or reaffirmed from it?

This essay will move through three eras of criticism, beginning with the invention of English criticism in the early eighteenth century by Richard Bentley and Alexander Pope. In this section is considered the relationship between the development of authorial copyright and the grounding by the editor of his interventions in the authority of a generalized rationality. Here the text becomes the property of its creator or author; and it obtains its integrity as a piece of property through the authorizing force of that Rationality’s internal generative/creative power. The following section considers the development of ballad criticism and ideas of collective authorship in the later part of the same century. In cases where a textual tradition—such as ballad poetry—was authored by a generative process; rather than an individual, a form of corporate ownership developed. Finally, the next section deals with the professionalization of textual criticism by records scholars employed by the state. In the conclusion, some implications and lessons of this history are then considered. The purpose of this essay is to introduce the reader to a wide array of characters involved in this history, and to do some justice to them as a herald should. Though we may lament the passing away of this culture of textual scholarship in our own brave new era, we may still learn from it, as good antiquaries always must learn from the funeral pyre; or mound; or funerary inscription; or decayed monument. What can be known, first must be buried.

There is a texture to this essay, so I encourage readers to refer to the footnotes, as part of the experience of using antiquarian discursive techniques.16 The argument is not necessarily that LLMs are completely bad; but to focus in on the significance of their dependence on the public domain, broadly conceived. This domain consists of literary remains, legal records, papers of state; both publicly and privately held.17 LLMs can not simply be condemned because they are generative and hallucinatory, on the false basis that the discourse network of human-moderated textual scholarship that underlies public textuality is more “real”. This certainly isn’t the case. Rather we should turn our attention, ever more deeply, to the textual corpus upon which LLMs depend; because that is the basis of our own language, which is our general communal property and responsibility. So we focus here on a series of interconnected issues concerning the public textual domain, the methodologies of textual criticism which curate and control that corpus, constitutional law, copyright law, records administration, forgery. The commons own the text, and it is ultimately subject to a form of democratic control. But, per Goldberg, we cannot dismiss the secretarial use of LLMs, only on the basis that it is deindividuated and mechanical, as systems of handwriting are also of the exact same nature. But we must be aware of the power involved.

— Alexander Pope, Dunciad, Book IV18

Textual criticism is a discourse of reconstruction, and of fabrication, of polishing and normalizing texts within certain constraints of what is reasonable, probable, or arguable within the critical marketplace. In the nineteenth century Karl Lachmann is associated with the use of genetic stemma to reconstruct the archetype of a manuscript, that is, the non-extant exemplar from which all surviving copies were derived.19 This archetype, though, is not the original, and may only itself be a very intermediate copy, already hundreds of years removed. In the early twentieth century, Lachmann’s method was adapted by the New Bibliographers, in England, led by W.W. Greg, to the field of printed books, and the concept of the archetype of a family of manuscripts was extended to the reconstruction of the copy-text which the typesetters of a book had before them.20 In either case, whether works of antique literature known through manuscript transmission, or modern vernacular literature known through printed editions, editors base their restorations not exactly on any document that does exist, but one that can be argued to have existed, but one whose reality is somehow guaranteed by its very absence, by the existence not of text, but of no-text—of a lacuna, or gap, or effaced part of the paper, or a sheet that is missing, or an entire edition whose title is known to us through the Stationers’ Register but that does not physically exist as anything beyond that. The Public record is identical or equivalent to the physical objects—the documents themselves—but not only to what is there, but to what is not. What they did contain. What they could contain. Documents (whether handwritten or printed) that only exist as possibilities, or probabilities, or as illusions, or a system of arguments expounded by the critic. What is “real” here is what can be shown, by argument, and demonstrated rationally, by method. Lachmann and Greg for a long time provided the textual critic with a sense of security, that that argument could be scientific, that it could reconstruct accurately what once was. The method of Greg however rests on the problematic notion of Authorial Intention, a concept which the textual critic unavoidably must have at least some recourse to sort out the many variant readings. Though the validity of authorial intent as a normalizing value in criticism has been challenged, it is in some way inescapable, and has been reasserted by Bowers and Tanselle. In this way even the modern, and more philological and scientific forms of textual criticism, are not entirely removed from the presuppositions made by the innovative critics of the seventeenth and eighteenth centuries.

Whereas, following Lachmann, and Greg, textual criticism reached a high height in the twentieth century, we can trace its history much further back. English criticism was scant before the eighteenth century, that is, until Richard Bentley appeared on the scene.2122 A bizarre character, Bentley was the imperious master of Trinity College, Cambridge, and he was frequently mired in legal controversy with the university and his own students. He infamously claimed that the text of Paradise Lost was corrupted through the forgeries and interpolations of the blind Milton’s scribe, and heavily emended the work according to his own genius in such a preposterous way as to draw the laughter of many contemporaries. Bentley was a powerful man however. He prevailed against his opponents in court, who wanted his Regius Professorship revoked, and bent them to his will, just as he bent the text to his will. He moved in lofty circles and supped with Newton. Besides the Trinity Mastership and the Regius Professorship he also held the prestigious post of King’s librarian. And we can humorously imagine him, when the Cotton library burned in an adjoining building, spiriting away irreplaceable bibliographic treasures from the King’s collection, still wearing his nightgown, climbing out a window down a ladder.23 For escapades like this and his edition of Milton Bentley may have earned the ridicule of Alexander Pope, still though his genius has resisted all effacement, and stood the test of time; for his accomplishments in the editing of classical texts he is still seen by many as the greatest philological mind of the century, not just in England, but in the whole of Europe.

Genius. Here is the key word to take away about Bentley’s approach to the text. It was the method of “conjectural emendation,” the method of making it up. In the editing and printing of any text, and through every iteration and further reproduction it undergoes, errors accrue, and are compounded by the slips of the typesetter or scribe’s hand. The scholarly editor of a text, like an Elizabethan play, may not often have an autograph original of the work in question extant to study. Reconstructing the error, how it came to be, how a word was misread by the typesetter, is an art which is never reducible to a simple scientific procedure. It takes genius. It takes a reconstruction of the author’s original inspiration, a simulation of it, constructed as a hyper-rational object by the critic. The critic becomes the author, or even a higher version of him, and can repair the errors and restore esoteric passages to meaning. So Bentley’s disciple Lewis Theobald emended “a Table of greene fields” to “a’ babbled of green fields” in Henry V. But the role of generative, creative genius in textual reconstruction can perhaps be best demonstrated by Bentley’s own words:24

You also need an incredibly keen judgment; you need sagacity and shrewdness; you need what the ancients ascribed to Aristarchus, a certain faculty of divination and prophecy. These can be acquired by no quantity of labor or length of life, but they come purely as the gift of nature and by happiness of birth. This is why although in other disciplines one’s results answer to one’s labors and prayers, in this arena of Criticism, out of so many who have rolled this stone their entire lives, making themselves old with books and study, there have been so very few who have managed to attain the reputation they hoped for among their successors. And yet, if you don’t already know it, it’s far, far harder today than it was in earlier times to compose emendations. For all the readings that obviously and willingly offered themselves when a Manuscript was collated have already been captured and anticipated: there is virtually nothing left but what you can excavate from the force of the meaning and the form of the language, solely by means of your genius. Therefore in these Horatian notes I will produce more emendations from conjecture than from the help of Manuscripts, and, unless I am totally wrong, the greater part of them more certain. For when there are variant readings, authority itself often makes a fool of us, soothing the mistaken itch of those who would emend. On the contrary, when we propose conjectures against the authority of all Books, timidity and shame pull us by the ear, and solely reason, and the light of the meaning, and necessity itself can get the upper hand.

Conjectural Emendation took ingenuity, and creativity. The precondition of this, however, was the elevation of the author. As the universe follows a rational, and knowable plan, so do the artist and poet. This is the standpoint of Bentley and Theobald, as textual critics of the Enlightenment. The critic can apprehend and grasp this plan, just as well as the author, and follow it, even when the author himself did not. Thus Bentley elevates the critic to the status of God over the text, and thus the whole operation has a dual effect, to at once elevate the author to a position of Godhood and then elevate the critic of the author even higher above that, creating a veritable Olympus around the summit of the text. Bentley can know better than Milton, in fact, by sympathizing so deeply with his creative genius so as to re-embody it in his critical activity; remaking the text itself, rewriting it all again, but as if it were the author’s own second draft. Creating a version that is both a forgery, and simultaneously, more authentic than what the author actually themselves wrote.

There is some parallel in all this with the operation of Large Language Models, which likewise generate plausible, yet non-existent texts. Texts that could be, or should be. Texts that exist at the intersection of other texts that have not even been written. Everything that LLMs create is hallucinatory. They reconstruct texts just as the antiquarian scholar and editor does, but under their own mathematical principle. Whereas, for the textual critic, like Bentley, or Edmond Malone, or W.W. Greg, different standards have operated to normalize the activity of critics. The difference between LLMs and the history of traditional textual criticism is not that the one involves falsification and the proliferation of fictive-documents, and that the other preserves something that is objectively real and actual; rather, forgery is inherent to them both, and the difference lies in the nature of the normative standards governing each.25

This is the tension within textual criticism, between the objective, and scientific restoration of errors; and evaluative judgement. Between genius and pedantry. There is some relationship between these two poles, and the text is generated out of their interaction. Both Bentley and Pope see themselves as geniuses. For Bentley this entails a philological and linguistic talent; for Pope and Swift, it involves a critical discernment of poetic quality. Both claimed to represent the true author within the editorial debate, by inhabiting and partaking of his creative genius. Both claimed to speak for the author; as Bentley did for Milton, and Pope did for Shakespeare. But both were also wrong.26 For this reason we must separate somewhat the ideas of scholarly and critical practice for Bentley and for Pope. Philological criticism should restrain itself from evaluative judgement, as its purpose is to provide an accurate and impartial text. The literary critic, on the other hand, engages in evaluative judgement of what is right and wrong, both stylistically, and morally in a work. These two tasks however are hopelessly intertwined, as the textual critic unavoidably must make evaluative critical judgements.

There exists a rhetoric of textual criticism. It is a network of discourse, which has its own meta-historical standards. There often are not clear right or wrong answers to problematic texts. Debate circles around the same issues over and over again. Results are inconsistent. The implementation of paradigms advocated for on a theoretical level is complicated by thousands of details in practice. The editors and antiquaries are a quarrelsome bunch. They arrogate over one another, while all being guilty of the very same errors they accuse one another of. We can see this in the battle between Pope and Bentley all too clearly. Above we saw, in Bentley’s own words, how he privileged the critic’s genius, taste, and judgement above manuscript variants as a source of emendations. His approach to textual criticism is creative, bold, innovative, arrogant. Yet here is how Pope characterizes Bentley in Book IV of the Dunciad:

— Alexander Pope, Dunciad, Book IV27

The whole of the Dunciad, which is full of parodic critical footnotes, is a response to editors like Bentley, and his follower Lewis Theobald. Bentley is presented by Pope as a dull pedant, bending over the shoulder of the author to correct him. As Derrida has Plato dictating to Socrates, so Pope has Bentley dictating to Milton.28 Pope was a master of strategy, and could be quite nefarious. No other author did more to, in the courts, drive the recognition of author’s copyright. He set the footman Robert Dodsley up as a publisher, and worked quite closely with his publishers to manufacture and promote controversies. The Dunciad is one such case, where Pope’s first ploy was to publish the poem without a critical apparatus, in the guise of a piracy done in Ireland of a work Pope had not finished or authorized. As word got out about the book, and reviews came out, Pope released his own authorized version of the work, the Dunciad Variorum, which chronicled and responded to the criticisms made against the first edition. In doing all this Pope essentially weaponized his authorial control over matters of bibliographic primacy in establishing the correct text.29

Here The Dunciad is a work also about the moral dimension of criticism, as it surveys a ghastly situation Pope believed had set over London—there was a crisis of bad taste. In this way Pope was much closer to Ben Jonson than Shakespeare, who also favored the kind of meta-textual techniques of satire and authorial intervention into the critic’s response that Pope so often used. Is Pope’s Shakespeare any less ridiculous than Bentley’s Milton? Jonson’s criticisms of his peers are in their way some of the founding documents of English criticism, and these opinions come down to us through William Drummond, who kept a collection of anecdotal literary pen portraits of notable men he entertained at his estate, much in the manner John Aubrey would later do.30 Jonson told William Drummond “That Shakespeare wanted art”. And in his dedication to the folio: “And though thou hadst small Latine, and lesse Greeke…” Elsewhere Jonson wrote:

I remember the players have often mentioned it as an honour to Shakespeare, that in his writing (whatsoever he penned) he never blotted out line. My answer hath been, “Would he had blotted a thousand,” which they thought a malevolent speech. I had not told posterity this but for their ignorance who chose that circumstance to commend their friend by wherein he most faulted; and to justify mine own candour, for I loved the man, and do honour his memory (on this side idolatry) as much as any. He was, indeed, honest, and of an open and free nature; had an excellent fancy, brave notions, and gentle expressions, wherein he flowed with that facility that sometime it was necessary he should be stopped. Sufflaminandus erat, as Augustus said of Haterius. His wit was in his own power; would the rule of it had been so too. Many times he fell into those things could not escape laughter, as when he said in the person of Caesar, one speaking to him, “Caesar, thou dost me wrong,” he replied, “Caesar did never wrong but with just cause,” and such like, which were ridiculous. But he redeemed his vices with his virtues. There was ever more in him to be praised than to be pardoned.

In these comments we see the basis of the Art vs. Nature debate that would become so fundamental to English criticism. Shakespeare lacked learning, Jonson claimed; but really the educations each received generally followed the same template, common to the grammar schools of the time, which followed Erasmus and held up poets like Terence as a model. Jonson however likely received a higher version of that education, at a better school: Westminster School, where William Camden was headmaster. Among his classmates was also Robert Cotton, who was a close intimate of Jonson’s. Jonson certainly associated himself with a higher “arte” which derived from classics-based humanist rhetoric taught through the grammar schools. Beyond that Jonson is said to have even attended Cambridge for a time before returning home to work as a bricklayer for his step-father. Art, we may assume, in Jonson, is responsible for providing the more self-reflexive and hyper-conscious elements of Jonson’s work; his penchant for direct addresses to the audience, or the way in which Bartholomew Fair blurs the lines between performers and audience and performance and life. Like Pope, for Jonson a mastery of art manifests in a highly self-critical form that challenges the boundaries between scholarship and poetry. This is the same impulse that gave rise to the Dunciad, which mocks the commentaries of the learned in favor of a kind of supreme authorial primacy, where the author controls both the copy and its meaning.

This is the paradoxical nature of textual criticism. It is a never-ending debate, where the authentic or original text is only ever a prop. A regulative pre-condition of the activity of criticism itself. Every critic argues that their own editorial activity results in a text that is purer, or closer to a supposed original. Every scholar presents themselves as reconstructing something that did, or should have existed. But in actuality this is never the case. The original, pure text is a negativity. A black hole at the center of discourse around which it orbits. The rhetoric of criticism is a system of trajectories charting those orbits, and not a science by which the original text can be reliably reconstructed. It is a genre of literature unto itself, which encompasses parody and forgery as well as scholarship and criticism.

The text is a ruin. It extends o’er all. It covers the Earth. Our speech is made up of it. Everywhere crop up little spires of a buried castle. The original is always lost to us in some way. Often in many ways. The antiquary stops and contemplates the wreckage of the text, as he contemplates the crumbling facade of an old church or the remains of a decayed monastery. The antiquary has come too late, and only some of the past is left behind, but always only ever in fragments.32

Always, in textual criticism, is there this dimension of reconstruction of what is missing or no longer available to us. This extends from the reconstruction of lost manuscript readings, but further, to the reconstruction of the intention of the author, or speaker. There may be physical gaps in a document. In some cases writing may have been effaced, or scratched out. Some entire documents may be known to us from copies or cross references, but may no longer exist in an original or complete form. Fires burned away portions of the Cotton Library. But the identity of lost books and manuscripts may be known to us through catalogs, or resources like the Stationers’ Register, where titles were recorded for copyright. But many titles therein recorded no longer exist in any known copy. Fragments are heaped upon fragments. The body of the public records, in a sense, consists of what we can know about the records, that is, it extends to the gaps, to what can be shown, through rational argument, to have been there. Sometimes what can be reasonably shown is limited, but this hasn’t stopped critics from trying.

The entire textual public domain is rooted in physical objects, so conceived in this sense—as what is both present, and what is missing or lost in them. This extends to everything from medieval charters and writs, to manuscript copies of Beowulf and Piers Plowman, as well as all later literary works such as those of Shakespeare, which are ultimately all based upon the collation of variants found in early printed editions.33 For more modern literary works, the texts of those are based on manuscript drafts and proof sheets of the author, documents that have been willed to large institutional collections for safekeeping.34 But whether we’re dealing with works of the Renaissance or the Victorian age, the texts of all these works, from Shakespeare to Scott, are rooted in the existence of physical objects. Texts are not self-sustaining; they are instantiated through manuscript transcriptions and printed editions. The text must always return again and again to the early editions and the autograph beyond them. Without this periodic restoration, the text can only degrade, and destabilize, and totter, and collapse.

Since World War II, the orthodoxy of New Bibliography has been challenged, not entirely successfully, by New Historicist approaches, which de-emphasize the significance of authorial intention. It can be argued that the text of every early edition is a work not just of the author; but that the text is significantly shaped by the work of editors and typesetters—not to mention the audience. The final text does not rest on the author’s certifying power. Prior to the eighteenth century, and to Pope, indeed there is something anachronistic about using authorial intention as an editorial principle. What is ironic about this is that New Bibliography already, to some extent, did this, by looking at the book much more as a material object—as a result of the work of typesetters and editors. Variants in the text, and peculiar features of the early editions could be explained by recourse to a reconstruction of the manufacturing process. Greg’s successors have underscored this in responding to New Historicist critiques, noting that Greg’s methods are not exclusionary of other approaches and perspectives.35 Still, though orthodox schools of textual criticism hold that they already partake in flexible concepts of authorship, it may still be argued that perhaps those concepts are not expansive enough. This is what D.F. McKenzie argued in his seminal Printers of the Mind, where he used new empirical data about labor organization within printshops around the year 1800 to challenge many of the baked-in assumptions of the New Bibliography about bookmaking.36 In the case of early printed literature, where we may not have author’s manuscripts, we must recover the author’s intent through a reconstruction of book manufacturing processes whose obscure and esoteric nature separates us from the authorizing power of the creator’s preferred text. They are ciphered by this process. Locked away by it, as if into a casket. The task of the textual scholar is to decode the text, through an intimate knowledge of the encrypting machinery. Only then can the original genius of the poet be uncovered and rescued and restored. What about texts though that have no author? What about cases where the machinery of transmission has also served as the machinery of composition? When the work isn’t the product of a physical machine, but of a much larger social or cultural one?

In 1753, the young cleric Thomas Percy discovered at the home of his friend a curious manuscript volume, one whose pages were being used by the servants as kindling to start fires.37 Percy begged his friend to let him have the volume, as it contained a large collection of traditional ballads, though this kind of literature was not valued particularly highly by critics at this time. Some years later, in 1765, and after much research into the other major ballad collections in the country’s libraries, Percy brought out his magnum opus: The Reliques of Ancient English Poetry. This work went through four editions during Percy’s lifetime, and even though he attempted to distance himself from it in his later years, after he had been promoted to bishop, it remains perhaps the defining, landmark work in ballad scholarship and collection.

There was however one slight problem. Percy had himself altered and sophisticated many of these ballads.38 The extent of Percy’s manipulations of the texts became clear only a century later, when finally the original volume of ballads he had discovered in 1753 was edited and published by F.J. Furnivall, who revealed the extent of Percy’s fabrications.39 The first to raise suspicions though that Percy was hiding something about his editorial methods was Joseph Ritson, the barrister, who did battle extensively not only with Percy, but with Edmond Malone and other members of Samuel Johnson’s circle.40 Percy, in his later life, jealously guarded his volume of ballads, from which he had supposedly taken many of his texts. Sometimes his manuscript treasure was the only source of particular ballads, of which no copies existed in other collections with which to collate. Ritson accused Percy of withholding the volume from inspection by third-party editors in order to conceal the extent of his interventions into the text, and he was not wrong in this, as Furnivall was later to demonstrate beyond all doubt.

Scholarly standards in the later eighteenth century were hardly as strict as they would come to be in the later Victorian era, when Furnivall especially promoted an absolute textual fidelity to the manuscript sources. This however was not standard practice a century earlier. Percy was involved in a project to rehabilitate a rather disreputable form of literature, that of the ballad, and even Percy’s own close friend Dr. Johnson made a host of rather insulting remarks about those enthusiasts of the genre.41 This move to rehabilitate the ballad began earlier, however, with Joseph Addison, who devoted several numbers of The Spectator to the matter, especially to the ancient Ballad of Chevy-chase. The ballad, as a genre, was equated with a kind of crude naturalism, versus the more polished “gothic” style of Enlightenment poetry.42 Even though imitations of ballads became a popular genre of poetry in the middle of the eighteenth century, especially with the work of Thomas Gray, they were considered to be too crude and rough for a polite audience, and therefore were in need of some polishing to present them to the public.43

To illustrate Percy’s position better we can look at an analogous critical attitude towards contemporary landscape gardening, as embodied best of all perhaps by Percy’s friend, the poet William Shenstone.44 Shenstone’s garden, The Leasowes, was a highly artificial affair, replete with plaster ruins erected on the site, and many strategically situated benches, urns, and inscriptions designed to foster contemplation and work in conjunction with the scenic views. For Shenstone “nature” was pre-given, but required polishing and intelligent staging to make the most of. This was in contrast to the French style of garden, which was geometric and Cartesian, and based around repetition and highly sculpted topiaries. Shenstone’s garden smacked also of a fair degree of artificiality, though he did not seek to overwrite nature, to make it perfect or symmetrical. He took what was given, and sought to maximize the effect of the pre-existing landscape with all its peculiarities, giving his garden a more rustic, and sentimental aspect. Percy’s approach to the editing of his ballads was analogous to Shenstone’s philosophy of landscape design. The ballads represented a crude, imperfect nature, which required intelligent presentation to get the most out of. Thus Percy had few scruples about augmenting the ballads in his manuscript whenever he deemed them too crude for polite consumption, and he took extensive liberties with his text, as Furnivall would ultimately demonstrate. In his notes he gave the appearance of transparency regarding his editorial interventions, and noted generally when he had engaged in some “restoration” of a particular text, but upon comparison Furnivall discovered Percy’s interpolations to be much more significant in scope and number than Percy had let on. Despite this Percy’s work has fared better than the work of contemporaries like MacPherson, who is generally characterized as a wholesale forger. Percy is not usually denounced in such harsh terms, and his work is more or less within the bounds of contemporary scholarly practice, yet for Ritson and Furnivall, the liberties he took were inexcusable, as often undeclared, silent, and significant in scope.

Ultimately there is no final, or set text of our olde ballads. Textually they are an uncultivated landscape without any original individual creator who can be named or identified. They exist as a corpus of variants. They evolve like a patch of wilds, subject to erosion and the weathering forces of time. In order to ground the poems in some single, generative authorial genius or persona, Percy turned to the institution of the minstrels, and held that the ballads were originally written by members of this class, and then transmitted and corrupted over time.45 He endeavoured to work backwards, to the authorial intention of the minstrel composer, yet this was only a historical construct. A fiction. Especially in Germany towards the end of the eighteenth century we see a movement towards a conception of collective authorship, which, following Herder46, finds its greatest summation in the work of Wolf.47 Within the tradition of the ballad, each singer, each transmitter has some right to improve; to edit; to polish; to shape; to sculpt; to innovate. The ballad form is impossible to disentangle from this mechanism of transmission and collective composition. Ballads exist in myriad valid versions, each a product of unique historical circumstances; of an era or epoch in the history of a particular ballad.

We have seen how the invention of English criticism was bound up with the origins of author’s copyright and the general idea of authorial intent, which is bound up with the emergence of the rationalized subject in contemporary metaphysics.48 Traditional literature—ballads, fairy tales, nursery rhymes—predates the appearance of such a rationality. Ballads are rather a collective composition—that is, they are the byproduct of a cultural, social, and legal-institutional mechanism. In the same way that the copy-text is reconstructed by New Bibliography, or the manuscript archetype by Lachmann; reconstructing the text of a ballad involves the critic in the reconstruction of a sociological mechanism of transmission/composition. The authorizing subjectivity that the critic must partake in moves from the individual rational agent to the historic-linguistic community. If the individual author of a work—whether Shakespeare or Pope—enjoys a privilege of deciding matters of bibliographic primacy due to their legally recognized private ownership of the book’s copy; what does this mean for works of a collective and historic authorship? Bibliographic control must reside with the collectivity, which retains a kind of democratic right through posterity to determine the horizons of textual possibility through ardent public controversy.

The text is a collective property, never a final thing. There is no “original” that can ever be with certainty recovered. Rather, it is a byproduct of collective critical debate. The text itself is inseparable and impossible to understand outside of reference to this mechanism. It is not simply given. As ballads are the result of a collective process of authorship, so too are the texts we consume of Shakespeare and all our classic poets. It is easier said than done to simply disengage these texts from the editorial debates that have raged over them for centuries. We cannot return to a version of the text uncomplicated by controversy. Emendations prevail over time because they are reaffirmed by successive generations of editors. No editor can ever decisively or finally demonstrate the superiority of their judgement against all alternatives. Rather, there is a process of collective authorship occurring within the community of textual critics, through which, like the reciters of ballads, the text is sculpted and polished and presented in a manifold of equally valid forms that can never be easily collapsed into a single entity. Rather, they exist within a horizon of socially authorized possibilities. The text is ultimately grounded within a kind of collective authority. It is a communal property. It is just this procedure of collective authorship which lends the text its social validity. Only within the matrix of this debate does the text assume some kind of stable form. Without reference to it, it has no ground, but is voided of historical reality.

From the time of Bentley to the time of Edmond Malone, responsibility for the collective property of public textuality was farmed out to individual scholar operators. Critics who could support themselves through the publishing industry (often by drumming up controversy) but who were not institutional scholars. The rise of the civil service in the nineteenth century represents a turn towards large-scale collective management of the text according to a rationalized, institutional textual regime.

Had the Parliamentary Collections been completed … they would when added to the already published Journals of the Lords & Commons have formed a work affording materials without parallel in historical Literature—materials elucidating the uninterrupted progress and complete development of the Constitution from the Middle Ages to the present day. No other Country in Europe can offer such Materials because in none other does there exist so unbroken a series of Records as those we possess—and the work would not only have become the sole authentic foundation for the constitutional History of this Country but would also have thrown light upon the constitutions of Estates, Diets, and other legislative bodies in the various portions of the Medieval Common Wealth of Latin Christendom.

— Palgrave remonstrating with the Record Commission for discontinuing his publication of Parliamentary Writs at a meeting December 4th, 1837.

The scope of the authority given to the Master of the Rolls by the Record Act is now considerably enlarged. Your Majesty’s Order in Council renders him the Keeper, not merely of the Records in a legal or historical sense, but all of Your Majesty’s Public Documents and Archives whatever, and the Repository will ultimately become (if the expression may be allowed) the strong box of Empire.

— Francis Palgrave’s Thirteenth Report, 3 June 1852

There is a productive tension in the nineteenth-century records administration between enclosure/control and openness/access. The archive is an interface, just as the LLM is, layered atop the text, mediating access. In this section we will briefly explore the ideology of the archive as software; and the conflicting principles governing its design and deployment. These principles will be demonstrated most sharply by the dichotomy between the managerial attitudes and practices of two men: Francis Palgrave and Henry Cole. So we will deal with each in turn, beginning with Palgrave:

Entrenched in parchment vestments, brooding in his dungeon, the Deputy Keeper of Public Records reigned over the Constitution of the Common Wealth like a cruel master wizard. In a rapture of melancholy he composed his annual reports of the Public Record Office’s progress through the muniments of the nation. In Chancery Lane a suitable castle was constructed for the Deputy Keeper to haunt, a Benthamite-Gothik sarcophagus so massive it had to be constructed in three separate phases for a period of twenty years. The Dark Lord the Deputy Keeper served, the Keeper himself, was the Master of the Rolls: Lord Langdale, one of the powerful judges of Chancery, the main equity court of England. The day-to-day administration of the business of the evil empire of documents however fell to the Deputy Keeper: Francis Palgrave, a mean and miserly old man, who was detested by his underlings. Legend has often maintained that Thomas Duffus Hardy, Palgrave’s successor, once decked the lord high deputy wizard of records at the Tower of London during a heated dispute. Palgrave ruled with an iron fist.49

Following the Public Records Act of 1838, records were transferred by the ton and trainload to the central repository built in Chancery Lane. Here the Master of the Rolls had, by ancient rite, a chapel and certain facilities at his disposal. These structures gradually were converted to record storage and overwhelmed by the never-ending flow of paper. These old structures were demolished to construct the new Record Office. Iron cages were fitted up to store the records, like torture racks. Lighting and heating were limited to guard against fires. Here was the constitution itself under lock and key. That key had been closely guarded by the Deputy Keeper, who also lived on the premises.

Besides Rolls House the Public Records had been traditionally held in a number of customary repositories,50 most prominent of those being the Tower of London. Here, in the Tower (which served simultaneously as a prison, it should be remembered51), overflow records of the Exchequer and Chancery had been stored since at least the fifteenth century. Another jurisdiction was created however by Elizabeth or James I, with the establishment of a distinct State Paper Office, which was to house the archives of the monarch’s correspondence.52

This whole situation was upended in 1838. Prior to this year an irregular parliamentary sub-committee was used to administer the public records. But it proved ineffectual. Repositories were semi-autonomous, and their respective keepers independent. Following the act of 1838 however, all repositories were placed under centralized administration, as branches of a rationalized records bureaucracy. All the other branches were systematically closed. Records were transferred to the central strong box of Empire and entombed in Chancery Lane. The grand wizard of records Francis Palgrave ruled supreme, and had beaten out Hardy and his other rivals for control over the public textual record of England.53 For Palgrave, this was the English Constitution itself. Now here was a treasure worth keeping. Here were the real jewels of Empire. Not in the East Indies, but in the dirt and grime of England. In the moldering, filthy, rat-eaten rolls that constituted the records. Some masses of documents were soaked and seeped through by raw sewage, and became poisonous to the assistant-keepers, who were all the same still tasked to organize and salvage them. One of these went permanently blind from a fecal infection of his eyes. These were the prisoners locked in Palgrave’s dungeon and tied up on his torture racks: the rolls. Enormous bundles of parchment sheets that had been wrapped around one another to form a giant mass. Some were so large they required two workmen to be moved. Here were indentures, charters, pleas, writs, and many other kinds of documents besides. Here was the Constitution of England in chains.

By virtue of being in the Public Record Office, overflow records of offices of the state, like the Treasury, when transferred, became public. And so a schedule was established, under which records older than eighty years would always be entering into the public domain. The Public Record Office also had itself granted the authority to destroy and dispose of records it deemed redundant, or historically useless.54 The first generation of assistant keepers of the PRO became also the first generation of professional historians under Palgrave’s discipline. The historian became a specialist during the Victorian era, a specialist in the records.55 To be a historian was to have special training in the records. The assistant keeper needed to develop a facility with historical styles of handwriting, not just the olde secretary hand—that Shakespeare used—but the older Gothic forms—not to mention the court hands, which were specialized handwriting styles unique to each government bureau/court. But not just in handwriting did the assistant keepers need to be learned, but also in the many languages which have been used to administer England: Latin, Anglo-Saxon, French, and formes of Inglysshe, wyth dyuers and sondrye variauntes of spellynge multiplyede. It was Palgrave who reared up the corps of men whom the state would use to comprehend its own confusing and nettled history.

For Max Weber a bureau is defined by its papers.56 The archive of the bureau’s own activity encloses the bureau itself, and delimits it. So a government, made up of bureaus, is really then a leviathan, made up of archives. If this is all the case, then the Public Record Office has a special place within this system: it is a bureau whose papers are the sum total of the historical records of all the other bureaus. The PRO not only took over the administration of historical records, from before the Glorious Revolution, but also became the deposit place for government offices like the modern treasury department. Treasury, now there is an important word. For the archivists and keepers of the olde records, in days long past, were the keepers of the keys to the treasury rooms where the records were stored. Treasuries of the exchequer. So always has it been the case, since the fifteenth century at least, that there has been a connection drawn between the records and jewels.

Palgrave was accused by The Athenaeum in its review of his Thirteenth Report (1852) of his “empire building” for monomaniacal control over the records. Cantwell describes his overall style as highhanded. He opposed giving free access to the office’s finding aids to searchers, and any plans to lower or remit fees charged to the public. In his later years he grew isolated and detached, and haunted the palace of records like a nightmare spectre. The public records grew into a twilight realm, where the textual heritage of England was fully enclosed. The fist of antiquarianism tightened around its object as its body withered and gave rise to professionalization, progressive reform, and centralized political control. Upon his death none of his employees seemed to mourn much for Palgrave, as the office of Deputy Keeper was passed along to the more liberal and popular Thomas Duffus Hardy, one of Palgrave’s students who relaxed some of his teacher’s strict policies. Still among Palgrave’s pupils there was yet a greater reformer though: Henry Cole.

Henry Cole.57 Old King Cole himself! The greatest alumnus of the PRO, who would go on to a spectacular career far beyond musty old manuscripts. Yet an appreciation for the power and methods of records history in some way extends through all his many projects. Cole was wrongfully terminated by the Records Committee, and suddenly, when faced with the necessity, discovered in himself a great aptitude for public relations during his (successful) effort to have himself reinstated. Traipsing over hill and down dale with his young friend, John Stuart Mill, Cole and his companion drank in the spirit of Romantic Poetry and mulled over matters of life and death when Mill was in a suicidal rut. With Mill’s help, and with the help of the Benthamite reformers (a faction of MPs known as the “Philosophical Radicals”) in Parliament, Cole was able to orchestrate the passage of the 1838 act, and was given considerable leverage in drafting up plans for the structure of the new archival bureaucracy.

Soon after, while still employed in the New Public Record Office, Cole turned his attention to the Penny Post. Up until this point, postage for a letter was paid by him who received it upon delivery. Postage was also by the sheet, and sheets were folded up and sealed; envelopes would have necessitated postage for a whole other sheet. Cole along with Rowland Hill changed all this, as the penny post introduced envelopes and prepaid stamps.58 Another, less successful effort was made by Cole to standardize railroad gauges.59 Though not always victorious, in everything he still proved himself first and foremost a great public organizer, and his greatest triumph was to curate and organize the 1851 Great Exhibition for Prince Albert.

This was only the culmination of years of labor by Cole to modernize British design. At the Board of Trade Cole gained control over the administration of trade and craft education throughout the country, and he saw with razor-sharp precision that in order to make British goods more competitive on the increasingly internationalizing market, the taste of British manufactures would have to be raised. So he began implementing exhibits of antiquities and artifacts to spur the imaginations of British craftsmen. The Great Exhibition was a natural extension of these earlier exhibits staged by Cole, only on a much larger scale. Still at the heart of the Exhibition was a sense of internationalism, progressive reform, utilitarian public education, and a desire to promote and improve trade; industry; and craft. Once the Exhibition had closed, Cole salvaged what remained into what he referred to as the South Kensington System of museums and educational institutions.60

There is an integrating force of imagination in Cole’s activities that we need to draw out. The reform of the public records brought administrators under salary. Fees were charged by the Office to members of the public (or their agents) to search the records, and these devolved directly upon the officers as their primary income. These positions were bought and sold, and were treated sometimes as an investment or a guaranteed annuity to be secured. The professionalization of the records bureaucracy made the records more accessible to the public, by means of regularizing these fees, and sometimes even waiving them. The records themselves were of use to members of the public as evidence of precedent in court cases. Cole imagined all of his projects integrating. The Public Record Office would make the public textual domain more accessible to historians and legal petitioners. The penny post and railways would render the archives more accessible still on a national level, by allowing people to write in from around the country to seek material.

Cole grew very fond of rail travel and promoted it to the public in the form of sight-seeing pamphlets authored by his fictional persona: Felix Summerly. The Summerly character was a great success, and spawned a whole sub-genre of educational literature written not only by Cole but by other imitators as well. Certainly his love of English antiquities played some role in Cole’s infatuation with railway jaunts, as he outlined in his travel brochures how England’s emerging rail network could be used to visit the old Roman and Medieval architectural points of interest. In this way we can hear an echo of the old antiquaries: of William Lambarde, Camden, Dugdale, and Aubrey; who toured around the counties of England in their imaginary perambulations and county histories. In days of olde it had been for the traveling herald to pause along the road to contemplate the old churches and their fading inscriptions—for Cole, this cultural heritage was universalized and democratized by rail; by post; by new public-spirited interfaces of bureaucracy.

The archive is a kind of interface that separates us from the past, by many layers. The finding aid, index, and catalog are one such layer. But we may also think of the bureaucratic strata regulating public access as another; e.g. the fees searchers paid, the structures of public accommodations, the attitudes institutions had towards the public they served.61 In this they are a deeper level of software undergirding the digitized texts we interact with today. Today, digital editions can be said to comprise a whole “family of new editions and impressions comprising at least six parts: an archival TIFF, a JPG, a PDF, an OCR-derived text file, an XML file, and a web interface.”62 The archival layer is another, extended part of this software. Policies, grants, the content of available archives all influence why and how texts are digitized—how they become part of the training data. The LLM is only another layer on top of all of these. A meta-user, exercising the unlimited rationality of the searching subject—reducing all lower interfaces to one: text.

There is always an ideological dimension to the text. This has been true from the very beginning, as we’ve seen how Robert Cotton brought his antiquarian resources to bear on problems of law and policy. In his position papers he searched the public records for precedents for how the king could raise monies, while the Society of Antiquaries delivered papers on the history and powers of the Earl Marshal (head of the College of Arms). There was always a connection between antiquarianism and politics, and therefore even the scholarship of the early seventeenth century is to some degree ideological. In the eighteenth century, however, we see a much more sophisticated ideological view of the text articulated by Edmond Malone, who understood the stability of English society as depending on the continuity of English records. Malone counted Burke among his closest friends, and the two viewed their respective projects as closely interrelated.63 This was not far-off from how Dr. F.J. Furnivall viewed the matter, as a Christian Socialist, nearly a century later. In the wake of the Revolutions of 1848 and the Chartist movement in England, Furnivall hoped that educating the workers in olde English literature and making early English texts widely, and affordably available would help stabilize the country against violent revolutionary upheaval.64

Institutionally, meanwhile, we must view the collective work of the PRO and the entire British archival administration during the nineteenth century in even more purely ideological terms. The entire civil service was reformed around Benthamite ideals during the century, and this rationalization extended into and through the development of archival administration and textual scholarship as practiced within the records institutions.65 This same process can be observed within the British Museum Library, during the highly contested battle over the printing and publication of its catalog during the same period. The PRO undertook the publication of the historical manuscripts that make up the English constitution itself, and therefore we may speak of the records administration as a significant organ in the state’s self-reproduction—as a large-scale, industrially-organized mechanism for publicly propagating the textual foundation of the English state. As the text is now enclosed by a consortium of AI companies like Anthropic and OpenAI, we might rightly wonder what role the ideologies of Rationalism and Effective Altruism play in contemporary textual control.

Whereas Francis Palgrave ruled over the records with an iron fist, the centralized control of the public textual domain brought the archive under professionalized bureaucratic governance that secured its use and availability to the public. It became subject to collective management, on behalf of the commons, whereby the government administered public textuality via a corps of salaried textual scholars. New and improved editions were no longer representatives of individual labor, or the property of a single absolute critic; they were now official government statements establishing the text for all comers. A monopoly on textual violence—the obelus (†) and the penknife—was established by the state. The same state whose constitution is predicated upon and based in those documents. The text thereby was stabilized, in the same way justice was under the administrative constitution that resulted from liberal reform during the Victorian era. Weber characterizes the administration of justice prior to rationalization as having an individuated quality. Justice is now cold and impersonal—it is a machine. The courts are a machine into which one feeds cases, that produces consistent verdicts out the other end. Whereas the early eighteenth century critic had authority through his individual judgement, the nineteenth century textual scholar in the PRO was a manifestation of collective judgement—an instrument of it.

Now all these layers of interface are being subsumed and collapsed by LLMs. Just as the PRO ate up the memory of the government through the records it shed; LLMs now eat up the textual memory of society, assimilating it and reconstituting it as a new interface of textual control dominating the public’s access to our linguistic heritage. But there is always an invisible apparatus of ideological control governing the design and implementation of such layers of institutional interface. There is always a tension between enclosure and control; and openness and access as represented in the extremes of Francis Palgrave and Henry Cole. Power seeks to establish itself through the structure of access it establishes to its own memory. Those structures must balance the public good against their own consolidation of textual control.

In the age of the Large Language Model, who defends the text? We have set out to introduce, briefly, some points about the history and development of textual criticism and the administration of the public records. Having done so, we are now in a position to draw some lessons from that history, and from the work of past generations of antiquaries and archivists, who were once entrusted with safeguarding the text.

There is a relationship between the critic and the corpus; as there is a relationship between the model and its training data. Machine reading only grasps a percentage of its corpus, at least when it is provided in the form of PDFs and must be assimilated via OCR. Other data may come from more easily readable text, in the form of websites, social media posts, etc. Yet this category is especially derivative, as it is all predicated upon the larger linguistic-textual corpus we have discussed, which is the foundation of our language, law, and literature. For newspapers, books, and manuscripts, and all older kinds of public records the LLM grapples with the public textual record through photographic scans, which are themselves reproductions. This entraps us in a serious web of complications that should be considered more seriously.

The first of which is that the LLM is therefore entirely dependent upon photographic facsimiles, many of which are very imperfect in terms of their reproduction.67 An LLM therefore cannot “see” a book or a manuscript for itself—with its own eyes. It cannot feel the texture of the page, or hold up a leaf to the light to check the paper’s watermark. Color, light, shadow, the LLM is unable to perceive the subtleties of these through scans. These qualities, like the feeling of a book’s paper to the tips of the fingers, are not in the training data. There is a human pleasure in such things that the LLM will never experience, and it can be argued that they are valuable in and of themselves, as a fundamentally human mode of interaction with our physical world. That they possess a value beyond the comprehension or grasp of a model.68 Yet the consequences extend much farther, in bibliographic terms at least. All these tactile and visual details that are available to the textual critic who examines a physical document or book are inaccessible to LLMs. On a case-by-case basis, this may amount to a small margin of error with respect to the individual document—yet this error is accumulative and compounding across the entirety of the corpus, which, we may remind the reader, amounts to the sum total of all historical public records. As we saw at the Public Record Office, in the nineteenth century this amounted to a truly enormous volume of texts, which were measured by the ton. Per Clanchy, it is this volume of material that is the sea of literacy into which the people of England were plunged, it is the life-water of our language. All later digital communication is fundamentally predicated upon this corpus, as a derivative textual body. Yet the LLM, across the entire scope of this foundation, is separated from it by millions upon millions of lost details—details which eluded the scanning machines which assembled the training data.

The dangers which threaten us can also be demonstrated in the problems with destructive scanning. That is, not only in the scans themselves, but in the bibliographic destruction involved in producing those scans. In order to feed the LLM, books are torn apart to be scanned into the training data. This practice is not new, and has been used prior to the training of large language models; but the problems it creates have been exacerbated by the need to augment the training corpus to keep pace with the needs for technological innovation. As G. Thomas Tanselle argued in A Rationale of Textual Criticism:69

Whenever any copy of any book is mutilated or destroyed, the total stock of bibliographical evidence is lessened by that much, and our ability to reconstruct the history of the edition from which it came is correspondingly diminished. No matter how large or how recent the edition, the loss of a single copy reduces the degree of certainty with which one can generalize about the edition as a whole (and therefore about the text it contains). If we care about history, in other words, we should treat every copy of every edition with the respect for physical evidence generally accorded only to books in so-called rare book collections…

It follows that all librarians and other collectors should think twice before disposing of, or rebinding, a tattered copy of any edition of any work, recognizing the grave significance of these destructive acts… How absurd, then, are those library programs (ironically called preservation programs) in which deteriorating books are microfilmed or photocopied and the books themselves destroyed. A reproduction, whether produced photographically, xerographically, or in some other way, can no more be a substitute for the thing reproduced than another printed copy from the same press run can be.

In order to train the LLM we are engaged in a self-defeating, self-undermining process. The bibliographical evidence is destroyed. We lose knowledge of variants within editions of books. We lose marginal notes and marks when pages are trimmed. Physical characteristics of those marks are no longer perceptible through a scan. The public record is diminished. The libraries of the monasteries are burned anew. But what does the public get in return for the text on which our very language is based being degraded? The evidence is destroyed, and the knowledge gleaned is fed back to us through a black box over which the public has no oversight or control.

This whole concerning train of reasoning is extended even further however, to the corruptions brought into the text through the digitization of these scans. To extract the actual text from the images—the text is processed by Optical Character Recognition methods—these methods exhibit varying degrees of performance on different types of texts based on the quality of the original scans. Basic OCR processing may not only misrecognize many characters of the text, but it also struggles to reproduce and capture aspects of typography; page layout; and white space, which are features that may not be translated into the output. At this point it should be obvious why this is problematic for the purposes of textual purity. This compounds and continues the problem seen above, as OCR even further removes the model from the text by a potentially significant margin of error. What is going into the model to train it may ultimately be significantly degraded from the original, with not only the general loss of tactile and visual detail in scanning, but also the further degradation of the text through the misrecognition of characters.70

It may be argued, however, that this technology is always improving, and that recently VLM techniques have emerged to provide an enhanced pipeline. Consequently, one might hope that old scans and training data may always undergo a kind of iterative reprocessing, securing their accuracy over time.71 This would be to follow Theobald’s conception of textual criticism, where the corpus of editions of a particular work forms a progressive body of knowledge, with error giving way over time to truth—to the pure or accurate text.72 Both the OCR pipeline and textual criticism then are responsible for error checking, for securing the text which underpins language (in both cases). But against Theobald we can use also, against the model, the response of Alexander Pope to this, who tore apart the critics in The Dunciad. In appendix VI of the Variorum, Pope provides a long parallel catalog, of criticisms that were made against both himself, and Dryden. He did so to demonstrate how criticism is, contra Theobald, not a progressive body of accumulating knowledge which safeguards the text, but, as Michael Gavin puts it, a “recurrent pattern of misreading that deforms literary discourse in predictable ways.”

In the eighteenth century, between Pope; Swift; Addison; Bentley; and Theobald, textual criticism operated within a delicate matrix between concepts such as taste; wit; judgement; and genius. For Theobald it was a compounding body of knowledge; for Pope it was a stagnant debris of recurrent error that the true critic and man of taste could easily cut through. Yet for Bentley conjectural emendation of the textual scholar depended upon his individual genius, just as it did for Pope. Pope ridiculed Bentley and his school of criticism as a kind of pedantry. Through the eighteenth century the publisher Tonson held exclusive (and for a long time, it was argued, a perpetual) rights to publish the works of Shakespeare, and this led to the succession of editions from Rowe; to Pope; to Samuel Johnson’s.73 There is no stable, or singular text. There is an evolving one that mutates within the critical matrix described. Textual criticism is a public rhetoric of controversy operating within this matrix. It was through this mechanism that the text of Shakespeare was supposedly secured in its accuracy from corruptions. Yet there is no way to objectively rank the editions which were put together in this time into a clear hierarchy. Each has its own peculiar value. For the LLM, the text is secured through a pipeline which is closed. Where the critical matrix is replaced by a series of secret statistical weights rather than publicly contested ideas like wit or judgement. The model is indifferent to the variety of approaches and the different readings of different editors; amalgamating them statistically, rather than evaluating the strengths and weaknesses of all their editors’ individual arguments and deciding between them according to a rational, articulable, and defensible standard employed in a critical apparatus.

The LLM is dependent upon the same corpus of texts as the textual critic: the public domain. These texts, though assimilated by the LLM as digital files making up its training data, are ultimately rooted in a substratum of physical documents stored in archives. Every edition of a literary work in the training data is the result of editorial effort. A cheap paperback reprints the text of a critical or scholarly edition which has been prepared by the editor through careful collation of manuscripts and early editions. The fragmentary corpus of legal public records has been reconstructed and printed in easily consultable volumes.74 All these books have been digitized and used to train LLMs. And as the editor and antiquarian reconstructs the public textual corpus using genius, parody, forgery, wit, ingenuity, reason, creativity, invention, and imagination; the LLM reconstructs probable texts through the mechanism of finding the next likely token. Thereby creating a large phantom corpus that is parasitic on the public domain; that swells it, and overstuffs it with fabricated pseudo-documents.

The store of our public records is a jewel house, in which we all have a vested interest and right. We previously discussed the enclosure of the records that took place in the nineteenth century under the auspices of the PRO, and we can observe a parallel enclosure happening today. Copyright law today disunites much public knowledge, with publishers prostituting their content libraries to different AI companies while charging heavy subscription fees to the public and its institutions. This has in turn driven the creation of shadow libraries, like Anna’s Archive, Libgen, and Sci-Hub, which scrape many enclosed repositories on an alternate model—of providing free access to individuals, but making the material available on an enterprise level as well. Where the PRO had been the driving organizational force of so much large-scale activity, current public archival institutions—especially in America—have not managed to keep pace. The public has ultimately been disinherited of its own property—the linguistic corpus which is borne out of the collective social use of that language. That property is then sold back to us by AI companies in the form of models; models trained on the public textual corpus which is not officially open to the individual except through illicit shadow libraries. The administration of the content, which was previously conducted by textual scholars in a public discourse of controversy, is now aligned with the corporate bureaucracy of the AI companies in a closed off business arrangement between publishers and techno-industry. The public and its rights over the public domain have been left out of the equation.75

It was Victorian administrators who first seemed to dream of some formal international system of cooperation between libraries and museums. Henry Cole, for instance, wondered in his particular way: “Might not a system be established of an international lending of those portions of a collection that would not be likely to suffer in travelling? This might easily be done by Railways now?” Librarians are always dreaming about establishing a more and more complete and totalized collection. Behind this dream is the dream behind every library—of collection, curation, organization, systemicity. Aristotle has, fittingly therefore, often been said to be the inventor of the library. Robert Cotton was another such dreamer, who dreamed of a national-level, publicly accessible library for the promotion of English scholarship. Even earlier, the wizard John Dee proposed schemes along similar lines.76 Those dreams were only ultimately realized by the formation of the British Museum Library, which brought together Cotton’s library along with those of Sloane and Harley to create such a national resource—one whose existence was to promote scholarship and learning within the realm. Further building out the institutional infrastructure necessary to achieve a complete national library involved much more work than mere founding however. It required many modifications to copyright law and the curation of an internal institutional discipline of professional librarianship. This was all overseen by Antonio Panizzi.77 Panizzi argued in favor of public access thusly:

I want a poor student to have the same means of indulging his learned curiosity, of following his rational pursuits, of consulting the same authorities, of fathoming the most intricate inquiry, as the richest man in the kingdom, as far as books go, and I contend that the Government is bound to give him the most liberal and unlimited assistance in this respect.

This is much the same credo as espoused by the administrators of shadow libraries, with one Library Genesis admin writing:

Here’s what I see as a consequence of free educational book distribution: within decades, generations of people everywhere in the world will grow up with access to the best scientific texts of all time. The quality and accessibility of education to the poor will grow dramatically too. Frankly, I see this as the only way to naturally improve mankind: we need to make all the information available to them at any time.

There is a unity to all knowledge. It is built up, like Theobald’s understanding of criticism as an accumulating variorum; as an aggregation rooted in textual transmission and the communal use of language. The duty of a state should be to protect that textual domain, and provide it back to the community which generated it. It is a tradition like that of the bard’s song. It is the property of the commons. It is therefore fitting that one important early Russian shadow library was called The Kolkhoz, a type of Soviet collective farm.78 The text is something collectively cultivated and harvested. Greg said the text is a “living organism which in its descent through the ages, while it departs more and more from the form impressed upon it by its original author, exerts, through its imperfections as much as through its perfections, its own influence on its surroundings.”79 The formation of our entire archival and library infrastructure has been the collective work of generations. It is the dream of the antiquaries. It is a dream they bequeathed the administrators of today’s shadow libraries, like a ballad half-remembered.80 With the advent of LLMs though, the shadow libraries have turned data brokers—as they now sell their corpora to AI companies to feed models. Now public support—through the state’s activity—is not driving librarianship and digital bibliography; private exploitation is. Exploitation against the rights of the community to their textual heritage. Not the rights of publishers—as those rights are also ultimately similarly based on an exploitation of public textuality.

The public record is the work of ages. It is the result of the curation and efforts of textual critics, bibliographers, librarians, archivists, antiquaries. It exists as the basis of our law and our language. Therefore the public has a vested interest that should be considered primary, above the interests of copyright monopolists and LLM companies that exploit it for training data. There is a long, well-established legal precedent for the public control of the text, and this should above all else be reasserted. The public is being left only what is discarded. AI companies train using the public domain of text with impunity, whereas the public and the individual are more than ever shut out from access to the material which constitutes that domain. It is in turn being sold back to the public in the form of a new system of literacy mediated entirely by proprietary and corporate-controlled writing machines. Per Goldberg our participation in that system serves to entrench its power over us, by extending it through the depersonalizing systems of perfect copy. In the case of the LLM, that is its perfectly averaged and statistically coordinated style.

Today, to confront these technological changes, we must turn again to the antiquaries and the defenders of the text. We must follow their example. We must reassert our rights to the public domain, through a recognition of the complex, and deep-seated processes which have governed and shaped it. When we look at the situation today, between shadow libraries; copyright law; AI company exploitation of the textual commons; publishers’ monopolies—things seem quite bad. My goal with this essay has not been to declare all LLM use immoral. The evil is rather in us becoming lax, in surrendering the text, and in giving up our communal property. I have therefore endeavoured to provide a deeper look at how the text is managed; collected; published; used; owned; generated; curated; and pruned—with my desire being that this will instill in the reader a greater sense of appreciation for those antiquaries and textual scholars who created everything upon which the text today depends. The text is not neutral. It is a product of collective effort. Of law, of criticism, of rhetoric, and discursive procedures inherent to literature as a medium. We must all take stock of that collective effort, and take an interest in it, and return to it. We must revive it, and insist upon our collective ownership of the text itself. We must insist that the public record be used to the public good, and that the public has a right to whatever is produced out of it or derived from it.

The British Museum wasn’t built in a day. These systems of textual control take lifetimes to design and install. We should take heart in that. We must peer through the discontinuities of history, and lay hold of a positive textual tradition that can see us through the current moment. If one culture of textual criticism has passed away, another must be born in the inscription of its epitaph.