Enlightenment Links
Theories of Mind and Media in Eighteenth-Century Britain
Collin Jennings




WHEN SCHOLARS TELL THE HISTORY of the internet, they often begin with the earliest plans for hypertextual media—platforms that would accommodate links among text, image, and other media objects.1 A common starting point is Vannevar Bush’s famous piece of vaporware, the Memex machine, a proposed two-screen microfilm desk that would facilitate “associative indexing,” recalling techniques of cross-references, subject indexes, and even manuscript commonplace books.2 Like Bush’s Memex, the first digital hypertext systems delivered techniques of connecting texts and ideas that had been long imagined but never fully realized. Embedded in these techniques was the principle that the mind works by association, which was first fully theorized and debated by British Enlightenment writers, including John Locke and David Hume.3 This book returns to that earlier moment to revisit how Enlightenment writers did not just theorize the connections between ideas but developed new kinds of history, poetry, and fiction that presented complex systems of linking. These writers carefully considered questions about the significance and strength of links among ideas in the mind, phenomena in the natural world, and events in history. Projecting the device of the hyperlink back in time, “Enlightenment links” refer to paratextual systems that made words work differently by pointing readers to other places inside and outside the texts.

Although the period we call the Enlightenment tends to be associated with linear and teleological forms of progress, writers combined text and linking paratexts to produce nonlinear courses of reading and polysemous forms of reference that resist causal models of experience. I argue that the link represents a fundamental Enlightenment gesture—the move from one context to another, from event to principle, from particular to universal. By drawing upon the methods and insights of computational linguists, I reframe what is at stake in common linking elements of printed books, including footnotes in poetry, indexes in histories, and epigraphs in novels. Since Gérard Genette’s influential Paratexts: Thresholds of Interpretation (1987), which examines the elements of a book that surround and frame the text, scholars have paid closer attention to the interpretive consequences of titles, prefaces, epigraphs, and so on.4 However, the digital context of the link troubles any simple distinction between text and paratext, language and media. For Genette, a work’s paratext represented the “threshold” between the text and the world, but linking devices often make direct connections from the text to objects in the world. In other cases, such as with indexes, the devices exhibit views of the text that contrast with the sequential order of pages represented in the table of contents. Links do not just orient the reader to the text; they also orient the reader differently in ways that both support and challenge linguistic order. What is at stake in Enlightenment links is that they produced the connections that the period’s moral philosophical theories attempted to observe. Enlightenment links performed the work of mediating gaps identified by eighteenth-century writers between a particular context—the specific lexical environment of the text—and so-called universal principles of human nature.

The concept of a paratextual link brings together different devices that similarly produce nongrammatical forms of reference. Scholars of the eighteenth century have considered how writers and illustrators developed visual and textual forms for exhibiting new theories of history and knowledge, including the diagram, the timeline, index, and footnote.5 While these studies trace the history of individual forms, this book considers how different devices perform a common linking function by combining language and format. This focus enables me to examine how printed language developed in the eighteenth century as both a symbolic and a material medium. Links make language and media interact in new ways by compelling readers to navigate texts differently. But in doing so, they also help us pay attention to how language itself moves readers in directions other than forward—how an end rhyme sends the reader back to the preceding line, how tropes like chiasmus and zeugma create new forms of grammatical crossing. In producing literal pointers that move us in new directions, links ask us to notice the devices language similarly calls upon to redirect us. And like links, new computational methods of observing lexical patterns make visible the ways language sends readers in different directions with recurring linguistic forms.

In 1965, Ted Nelson coined the term hyperlink as an element of hypertext to describe “a body of written or pictorial material interconnected in such a complex way that it could not conveniently be presented or represented on paper.”6 Although still ubiquitous, hyperlinks evoke the early 1990s when theorists and creative writers were enthralled with the new modes of writing and reading that hypertext seemed to promise. They celebrated how the technology transformed the relationship between writers and readers by allowing the latter to play an active role in constituting the arrangement of a work. As Stuart Moulthrop remarked in 1991, “hypertext systems exploit the interactive potential of computers to reconstruct text not as a fixed series of symbols, but as a variable-access database in which any discursive unit may possess multiple vectors of association.”7 Jay Bolter similarly posited that “hypertext reflects the mind as a web of verbal and visual elements in a conceptual space.”8 Yet as hypertext has become a fundamental element of digital media, the linguistic and literary significance of endowing language with “multiple vectors of association” has not been fully acknowledged. Links transform how language functions. Words refer to things, but words encoded as links refer to things both in what the words mean but also in what they connect to. Of course, linguists have long distinguished between the sense and reference of a word. Saul Kripke offers the example of the phrase “the president of the United States,” which denotes different meanings, or senses, but also refers to the proper name of the president at that moment in time.9 “Enlightenment links” indicate cases in which writers found this linguistic affordance insufficient and supplemented it with devices to direct readers to other places within and beyond the text. Links supplement and complicate the capacity of language to mean and refer by ascribing a new medial form of reference to words. Attending to this function of links allows us to consider how writers and printers in history challenged the linear and grammatical conventions of language. This is what scholars in the digital humanities and those using computational text analysis, or distant reading, in particular have sought to find in new methods of historical inquiry that are “not conveniently . . . represented on paper.”10 If digital media treats language differently than paper does, how does language on paper look from our contemporary digital standpoint?

This question helped spread early excitement surrounding hypertext in the 1990s beyond writers and artists. Scholars in the social sciences and humanities began to explore how hypertext media might be used to reimagine forms of arrangement inside and outside of their disciplines. Jerome McGann’s work throughout the 1990s and 2000s has provided language and interpretive frameworks for using digital media to reexamine our perspectives on historical media. On hypertext, McGann observes, “when a book is translated into electronic form, the book’s (heretofore distributed) semantic and visual features can be made simultaneously present to each other.”11 Hypertext format and digital media transform text from a sequential medium to a nonsequential one, even if the mind still encounters words and images successively. Scholars have also looked to the capabilities of digital media to reimagine older philosophical theories. In 1994, Geneviève Teil and Bruno Latour referred to the “unfashionable philosophical program of associationism” (the theory that successive ideas follow different kinds of connection) of the Enlightenment philosophers Hume and Étienne Bonnot de Condillac in describing the model for a computer program, dubbed the “Hume Machine.”12 This program could record co-occurrences of key terms in sentences across a body of texts. In their essay, Teil and Latour make the case that the “circumstantial facts” of association, defined as instances of co-occurrence, are more powerful than formal rules in representing the features of social interaction between humans and nonhumans, derived from key terms in the texts.13 They posit that reading a corpus of texts as a network of terms offers a way of reflecting social systems. In other words, linguistic patterns of association correspond to social patterns of association.

In the same year, literary scholar Alan Liu developed the first version of the Voice of the Shuttle, a web repository of humanities research and teaching resources, to facilitate connections across disciplines. Like hypertext fiction, the project foregrounds the creative and connective work of users in navigating from one resource to the next, forming chains of items that reflect the “ceaseless reconfiguration of humanities knowledge assisted by the new technologies of dynamic information.”14 The site presents an array of interwoven links and pages, displaying what Liu calls the “ordering of things.” The Voice of the Shuttle illustrates how the work of imposing references among words or textual objects that do not appear in the same grammatical context creates meaning through association. In contrast, Teil and Latour suggest grammatical proximity itself produces association among terms that can be translated into links as co-occurrence networks. Both approaches consider the relationship between grammatical sequence and hypertext structure: the way that written language can be represented as links and the way that links can produce meaning typically ascribed to language. In exploring the relationship between language and links, these projects reveal that the most basic function of the link—its technique of reference—has been overlooked.

Enlightenment Links demonstrates how eighteenth-century writers spoke through the “voice” of print arrangement as much as through language. The name of Liu’s Voice of the Shuttle website refers to a phrase that, in the Poetics, Aristotle takes from a lost Sophocles play on the story of Philomela and Tereus.15 In the myth, Tereus imprisons Philomela after having raped her and cut out her tongue. Without another way to communicate, Philomela weaves a tapestry depicting the crime and furtively delivers it to her sister. She speaks through the voice of the shuttle with which she weaves the image of the scene. The pattern of threads in the tapestry produces a visual narrative without requiring language. Liu chose this phrase for his project to indicate how the website displays patterns of hyperlinks that reveal various logics of association. The voice of the shuttle trope represents the remediation of one form of communication (speech) into another (image) as Liu’s site exhibits the remediation of text into hypertext. For Geoffrey Hartman, the voice of the shuttle trope illustrates what tropes do to language more broadly by eliding the middle term between the instrument (the shuttle) and its effect (to give voice).16 Patricia Klindienst argues that the trope reveals how women are forced to speak through the objects of their confinement; the remediation of the trope represents how Philomela uses the shuttle to speak through violence and imprisonment.17 Across these different readings, the critics note how the voice of the shuttle challenges the linear, sequential form of a stream of speech or rows of words on a page. It illustrates how meaning is made through crossings, recalling that the word text derives from the word for weaving.

In a gothic novel that I consider at length in chapter 4 on epigraphs, Ann Radcliffe’s The Mysteries of Udolpho (1794), the character Blanche encounters a decrepit tapestry that evokes the voice of the shuttle trope. She is at first amused by the tapestry’s dusty and threadbare appearance, which depicts a scene from Homer’s Iliad, but as she approaches it, the material of the worsted calls forward the idea of the weaver’s hands: “She laughed at the ludicrous absurdity she observed, till, recollecting, that the hands, which had wove it, were, like the poet, whose thoughts of fire they had attempted to express, long since mouldered into dust.”18 Blanche considers the tapestry, the hands of the weaver, the fact of his death, and finally the poet’s same condition. The threads of the tapestry send Blanche’s mind through a series of associations from material to hands to thoughts of the grave. This sequence of thought follows a different course from the image of the Trojan War, but it is produced by the interaction between image and material—composed by the poet and the weaver. While the faded state of the tapestry seems to oppose the energy of the depicted scene, the evidence of age corresponds to the passage of time since the tapestry was woven and the even greater distance from when Greek poets sang the poem.

Figure 1 Chapter epigraph header from Ann Radcliffe’s novel The Mysteries of Udolpho, vol. 3 of 4 (1794), 419. Courtesy of the British Library Board, General Reference Collection 1607/4139, vol. 1, p. 272.

Like the appearance of an image from the Iliad in Blanche’s chamber, the chapter epigraphs of Radcliffe’s novel confront the reader with fragments of past poetic voices (figure 1. The choice to append poetic extracts before each chapter signals something to readers. Considering linguistic patterns of novels featuring chapter epigraphs provides a perspective on how the works engage the past. Radcliffe’s Udolpho and the broader archive of novels with chapter epigraphs present a high prevalence of prepositional phrases orienting characters in their physical environments. Characters observe features on the “side of the” or “part of the” wall, tower, tree and so on. These novels also exhibit a significant concentration of modal verb phrases indicating characters often speculating on the meaning of the objects they observe around them. In other words, the scene of Blanche considering the faded tapestry in her room is quite typical. The novels depict worlds in which the natural and built environments are inscribed with marks from the past by hands “long since mouldered into dust.” The material artifact of the printed book contributes to these worlds by confronting the reader with a web of poetic extracts. Even when an epigraph seems to bear little relation to the chapter that follows, the network of passages contributes to a broader perspective on how one encounters history. In other chapters of this book, I explore how different kinds of Enlightenment links similarly supplement the linguistic patterns of the texts or, just as often, challenge those patterns by exhibiting contrasting forms of order. Yet in each case, text and format, language and links, interact to present different possibilities of producing meaning.

Links between Eighteenth-Century Philosophy and Twentieth-Century Information Theory

One premise of this book is that the invention of the hyperlink did more than introduce a new technology; it also gave a name to a range of preexisting devices that combine language and format to produce nonlinear arrangement across textual media. Devices including cross-references, footnotes, and indexes resemble one another in how they create associations outside of the stream of writing. While Nelson conceived of the hyperlink as facilitating connections beyond the capabilities of paper, Bush envisioned the Memex as a kind of supplement to paper and other analog technologies. The microfilm screens of the Memex would magnify tiny photographs of printed sources, which would be connected in the memory of the machine according to the user’s “associative trails.”19 Bush claimed that this system of connecting textual material would make it possible to arrange information according to how it was processed and accessed in the mind—that is, by association rather than by, say, chronological or alphabetical sequences.

Bush drew upon the language and premises of associationism in representing how the mind forms connections among ideas. Indeed, the name for the machine is a portmanteau of “memory extender,” and his plan for the Memex even resembles eighteenth-century metaphors of mind. Locke famously compared the mind to a camera obscura. He described it as “a closet wholly shut from light, with only some little openings left, to let in external visible resemblances, or ideas of things without.”20 Sean Silver elaborates Locke’s image in a description that could also work for depicting Bush’s device: “It is as though the understanding stood within a miniaturized chamber, viewing images cast upon a two-dimensional surface, contemplating them simply as they come and go.”21 Locke even could be credited with creating a new kind of “memory extender” in his popular New Method of Common-placing (1703), which transformed the commonplace book from a simple tool for organizing rhetorical topics into a technology that, as Silver suggests, “worked knowledge up into patterns and systems for their instant, motivated use” via the addition of a subject index (33). Bush similarly intended the electrical mechanisms of the Memex to match the speed of the mind in moving among associated texts.

As Bush and other theorists of hypertext imagined techniques of connecting text, early computer scientists examined the statistical dimensions of language itself. Bush’s doctoral student, Claude Shannon, developed the first theory for representing language digitally, as combinations of ones and zeroes. Shannon created the field of information theory by reframing any communication event as the transmission of a message from a finite set of possible components (say, the twenty-six letters of the alphabet).22 As Shannon’s explicator, Warren Weaver, put it for a lay audience, “information is the measure of one’s freedom of choice when one selects a message.”23 Information theory explained the “role which probability plays in the generation of a message” (9). The probability of any given symbol occurring in a message is determined by the symbols that precede it. Shannon developed his theory of information by analogizing it to entropy as a measure of disorder in thermodynamics. The amount of information a message conveyed corresponded to its unpredictability, or disorder. If a message is written in English and the first symbols are th_, only a few letters are likely to follow, and so the next letter conveys little information. In contrast, if the message is written in an undeciphered code, there may be many potential symbols, and so the next symbol in the message conveys a lot of information. As James Gleick observes, “a message . . . can behave like a dynamical system whose future course is conditioned by its past history.”24 In other words, language does not just represent the world; language acts in the world like complex physical systems. While hypertext made new connections between language and media, Shannon revealed how language already resembled processes in the natural world.

For many humanities scholars, computational approaches to language and texts may seem like a recent phenomenon, an approach taken in response to new technologies like hypertext. However, linguists and literary scholars took up Shannon’s theory almost immediately. The field of computational linguistics emerged in the 1950s as scholars including Zellig Harris, J. R. Firth, and Margaret Masterman used Shannon’s insights as a starting point for measuring the statistical properties of language. In 1954, Harris posited the “distributional” hypothesis that states that “difference of meaning correlates with difference of distribution.”25 Firth phrased this idea more pithily in his dictum, “You shall know a word by the company it keeps.”26 Subsequent computational linguists have built upon this premise by developing techniques for modeling language according to how words tend to co-occur, producing analog versions of Teil and Latour’s “Hume Machine.” In the context of literary studies, Yohei Igarashi has demonstrated how I. A. Richards developed his pedagogical practice of close reading in relationship to research on statistical analysis in literacy studies.27 Similarly, Michael Gavin has examined the resonance between William Empson’s theory of the “compacted doctrines” of words and the development of methods of modeling word frequencies in computational linguistics.28 Empson thought that words “carried their senses as latent semantic potential,” and measuring the statistical distribution of words across contexts has offered a precise way of observing how those different senses manifest (641). The techniques that linguists developed from the 1950s forward are now applied in natural language processing and machine translation. In Enlightenment Links, computational methods of observing lexical patterns make it possible to examine how linguistic forms interact with the organizing structures created by different linking devices. Attending to the probabilistic structure of language allows us to consider what’s at stake in suspending the succession of letters and words by linking out—say, with a footnote or cross-reference—to a new lexical context.

This is one side of the connection between the eighteenth century and today that I develop in the following chapters. Computational text analysis produces a new perspective on linguistic patterns of the past. However, we can also start from the eighteenth century to recontextualize current digital methods. Eighteenth-century moral philosophers developed theories of association that culminated with new statistical techniques for viewing the world. At the end of the seventeenth century, with the publication of Locke’s Essay Concerning Human Understanding (1689), British moral philosophers foregrounded questions regarding how knowledge of the world and human nature could be produced from empirical observation. Early in the 1700s, philosophers including Francis Hutcheson and George Turnbull argued that an invisible order in effect guaranteed the correspondence between observed particulars and general truths.29 Yet by the 1730s, Hume demonstrated that a strictly empirical analysis could never prove a “necessary connection” between causes and effects, or circumstantial facts and principles of knowledge. Subsequent philosophers have called this idea the “problem of induction.”30 Mary Poovey explains that the “problem of induction turns on the gap between one’s ability to observe discrete qualities by discrete perceptions and the impossibility of explaining why we believe that qualities belong to the same object.”31 Poovey notes that, at the end of Hume’s Treatise of Human Nature (1739–1740), this problem “stalemates the moral philosophical agenda of producing general and demonstrable knowledge about the human mind from observed particulars” (202). Even as experience was elevated as the foundation of knowledge in Enlightenment Britain, it was also found to be insufficient for proving laws regarding how the world worked.

The history of British empiricism has been well covered by scholars of the eighteenth century. Rather than rehearse it further, I am interested in how the failings of empiricism ultimately influenced subsequent probabilistic ways of thinking about the world and language. In the essayistic Enquiries (1748, 1751), Hume repackaged the ideas of the Treatise for a broader audience, and in the process, he laid out a pragmatic solution to the problem of induction: probability. He explains:

When we transfer the past to the future, in order to determine the effect, which will result from any cause, we transfer all the different events, in the same proportion as they have appeared in the past, and conceive one to have existed a hundred times, for instance, another ten times, and another once. As a great number of views do here concur in one event, they fortify and confirm it to the imagination, beget that sentiment which we call belief.32

Hume suggests that the mind intuitively computes the likelihoods of particular outcomes. By attributing belief to the consequence of repetition, he locates it in the domain of calculation.

This shift was picked up by contemporary writers and theorists of mathematical probability. Jesse Molesworth has demonstrated how developments in probability theory can be observed in the midcentury fiction of Henry Fielding and Laurence Sterne.33 The namesake of Bayesian probability, Thomas Bayes, appears to have engaged Hume’s argument regarding association and probability as early as 1749.34 He later presented a mathematical formula for determining the probability of chances in his posthumous piece, “Essay towards Solving a Problem in the Doctrine of Chances” (1761). Bayes’s response to the problem of induction addresses Hume’s framing of the problem in the Enquiry Concerning Human Understanding in 1748. Hume’s philosophy and Bayes’s essay subsequently informed the great French mathematician Pierre-Simon Laplace’s “rule of succession” formulated in the 1790s. As Ian Hacking notes, in the following century, theories of chance changed from “what Hume called the ‘superstition of the vulgar’ to become the foundation of the physical world and the cement of the social universe.”35

By making degrees of certainty quantifiable, Bayes, Laplace, and their successors laid the groundwork for the computational methods of analyzing the statistical properties of language. Like Hume’s account of probability, Bayesian statistics relies upon information about prior events in contrast to the frequentist school of statistics.36 Computational linguists adapted Bayesian probability for determining the likelihood of successive letters and words, rather than natural events. The resonance between eighteenth-century moral philosophy and twenty-first-century computational linguistics moves in both directions. On the one hand, computational methods of observing lexical patterns can be brought to bear on texts in which Enlightenment writers consider questions of order and connection, and, on the other hand, these techniques can be traced back to the theories of probability laid out by those very writers.

Developing a Digital Critical Method from Eighteenth-Century Theories of Mind and Media

In the past thirty years, scholars of eighteenth-century literature and culture have frequently examined the resonance between the proliferation of digital media at the end of the twentieth century and the similar explosion in print at the end of the eighteenth century. Andrew Piper’s Dreaming in Books (2009) is organized around media operations like “networking,” “copying,” and “processing” that a person commonly encounters using the internet. He remarks, “The digital provides us a critical lens to see the bibliographic with fresh eyes” even as he emphasizes a desire to show how the history of books “can help us contextualize our understanding of the digital.”37 Similarly, in their edited essay collection, This Is Enlightenment (2010), Clifford Siskin and William Warner redefine this intellectual and cultural period of European history as “an event in the history of mediation.”38 Other scholars, including Johanna Drucker, Jerome McGann, and Elaine Treharne and Claude Willan, have produced more expansive histories and definitions of concepts such as writing (or what Drucker calls “graphesis”), textuality, and text technologies, respectively, that bring disparate media environments into relation with one another.39

My approach in this book differs from these examples because the link represents a guiding concept as well as a method for reconsidering how association works in the interaction between language and media. Eighteenth-century philosophers and writers were interested in both how we perceive order in the world and how we present order in books. This book offers a way to examine works using methods from a new technological standpoint but derived from a similar theoretical origin in theories of probability. In Local Transcendence: Essays on Postmodern Historicism and the Database (2008), Liu imagines how scholars might use the affordances of digital technology to reencounter technologies of the past. He envisions an approach that is “like holding a microphone up to the far past: a way of committing ourselves to hearing the past but, through the conscious election of a media paradigm from a different era.”40 The payoff of such a method would be “to make us see history as a compound relation of proximity and distance between past and present” (25). This combination of proximity and distance is precisely what text analysis makes possible. Piper similarly describes the new forms of historical relationality afforded by quantitative methods, “To place language within this logic of number is to move from a system of substitution (x means y) to one of succession (x is so much more or less than y).”41 In graphical and tabular forms, words or texts (depending on the unit of analysis) are represented as units closer or farther away from each other according to the variation between their co-occurring lexical contexts. In a similar way, Gavin has examined how representing words in this manner enables us to reapproach Julia Kristeva’s theory of “intertextuality.” He explains that for Kristeva, “words are defined by their ‘status,’ their position within language conceived as a multidimensional space . . . Words become links that mediate among texts.”42 Kristeva offers a framework for relating language to space. She observes, “The word as minimal textual unit thus turns out to occupy the status of mediator, linking structural models of cultural (historical) environment, as well as that of regulator, controlling mutations from diachrony and synchrony, i.e., to literary structure.”43 Computational methods make it possible to visualize words in multidimensional space, and in doing so, they change our perspective on how words associate.

Text analysis provides a collection of techniques for measuring how likely patterns of language are in a given text or genre relative to a larger corpus. My analyses tend to focus on patterns of common words and phrases (articles, pronouns, and prepositional phrases, among others) because they are difficult for humans to isolate, and because the results can guide close readings of specific passages by indicating the prevalence of linguistic features across collections of a given genre. I describe my computational methods in detail in the appendix, but in broad strokes, I take two primary approaches in the chapters that correspond to the genres being examined. Chapters 1 and 3 address new kinds of philosophical writing—Ephraim Chambers’s Cyclopaedia (1728) and the philosophical histories of the Scottish Enlightenment, respectively. For these chapters, I compare the associations among terms created by linking systems to the patterns of word co-occurrences in the broader eighteenth-century print archive, represented in word vector spaces and topic models. I observe to what extent the series of cross-references (in chapter 1), or pages included under an index entry (in chapter 3) correspond to how words tended to appear together in the period. For chapters 2 and 4, which concern more traditional literary genres—poetry and the novel—I employ clustering and classification tests to observe how the use of linking devices in genres correlates with other stylistic differences from larger corpora of contemporary works. I ask, what other characteristics do poems featuring footnotes or novels with chapter epigraphs share? The first set of methods foregrounds what the links do to language in producing new forms of knowledge and history. The second set focuses on how the links interact with the rhetorical and grammatical aspects of the genres in achieving broader aesthetic effects. In implementing these methods, I draw upon recent work by Piper, Gavin, and Ted Underwood. Framing my analyses in relation to paratextual features of the works also reflects Katherine Bode’s argument regarding how book history can inform computational approaches to literature.44

My digital critical approach enables me to contribute to a persistent concern in eighteenth-century studies: how do we understand the relationship between Enlightenment philosophy and the literature of the period? To put it in more specific and current terms, what does eighteenth-century theory of mind have to do with eighteenth-century theory of media? Ian Watt’s The Rise of the Novel (1957) remains the most famous study to tackle this issue in its exploration of the connection between philosophical realism and the formal realism of the early novel, but more recently scholars have attempted to reenliven these questions.45 Across multiple works, Jonathan Kramnick has examined how eighteenth-century philosophy, fiction, and poetry represent actions that “extend the mind into the world.”46 He positions his work as reorienting older accounts of the relationship between British philosophy and literature, which previously had seen them developing “a new language of inwardness or subjectivity.”47 Instead, he emphasizes the role of external factors and environments conditioning the operation of the mind in the world. Indeed, Kramnick has contributed to a growing body of work examining what Silver calls the “cognitive ecologies” of eighteenth-century philosophy and media.48 Silver and Christina Lupton have drawn upon the theories of Bruno Latour to examine how the material things of the eighteenth-century world provided the language for the empiricist theories of mind articulated by Locke, Hume, David Hartley, Thomas Reid, and others. Silver looks to the organizational models of libraries and museums as resources from which these philosophers developed their models of thinking, as observed in, for instance, Locke’s metaphor of the mind as a room or cabinet. Lupton foregrounds the consciousness of mediation exhibited in the language of eighteenth-century texts themselves.49 She focuses on works that make print and mediation their subject matter. Together, these studies and others have done much to break down traditional oppositions between mind and material, philosophy and literature, and language and media in histories of Enlightenment Britain.

I am suggesting that we can advance this line of inquiry by examining how language interacts with print formats in nonthematic ways—how frequencies of terms and structures echo or challenge how the text is arranged. Eighteenth-century writers asked the same questions of the world that twentieth-century computational linguists have asked of language: What is at stake in association? How should we interpret the relationship between successive events, ideas, or words?

Progressive and Recursive Histories

Observing points of contact between eighteenth-century Britain and our own media moment, however, does not mean that any specific aspect of Enlightenment links—from most of the paratextual devices, including indexes and footnotes, to the theory of mental association—originated in the eighteenth century. Ann Blair has examined how Renaissance humanists developed complex systems of information management that included marginal notes, indexes, and other linking devices.50 However, the eighteenth century represents the moment in which practice and theory converged. Eighteenth-century writers recognized the specific problem of bridging the divide between observed particulars and abstract principles that came down to questions of connections—ones among different kinds of objects, ideas in the mind, occurrences in the world, and even words on the page. Enlightenment links represented a way for writers to step outside the history of the text: to suspend the succession of words with an element of the work’s medium and to point the reader to a different place inside or outside of the work. To draw upon the comparison to the physical world, we might say that linking moves the act of signifying outside of the immediate, local environment of the text to a new place that is not beholden to the same semantic and syntactic demands.

One unexpected place I observe this use of links is in poetic footnotes. In chapter 2, I examine the use of footnotes in progress poems, which became a popular genre in the late seventeenth to early eighteenth century for tracking the movement of learning from ancient cultures to modern Europe. As writers and philosophers compared classical and modern accomplishments in learning, the progress poem became a model for tracing the translatio studii, the transfer of arts, from one country to another. Alexander Pope used the progress convention explicitly to foreground the problem of reference in the 1743 edition of his anti-progress poem The Dunciad in Four Books. In Book II, he creates an imaginary poet, depicted as an immaterial figure: “A Wit it was, and call’d the phantom, More.”51 The footnote for “More” (see figure 2) observes that contemporary bookseller Edmund Curll “affirmed” More to refer to James Moore-Smyth, a poet whom Pope had accused of plagiarism. Yet the next note, referring to the same term and attributed to the fictional editor, Martinus Scriblerus, disputes the attribution (“this is not the name of a real person, but fictitious”) and traces the etymology of More to the Greek word for plagiary. The term More refers both historically and abstractly to the notion of plagiarism. Pope allows the term to denote both proper and general signifieds.

Scholars of Pope including Aubrey Williams and Blakey Vermeule have offered different interpretations regarding why the successive versions of the Dunciad tended toward increasing ambiguity and complexity of reference, culminating with the addition of the fourth book in 1742.52 However, these readings operate from the still common premise that words work through signifier-signified relationships. A different way to think of what the Dunciad produces is not as one-to-one associations among terms and the historical figures or abstract ideas to which they refer. Instead, we might think of Pope as triangulating terms like More between figure and idea to produce webs of links among associated terms. Pope’s map of Dullness, as it attempts to advance Westward from London to Westminster, extends across multiple axes of meaning. That is to say, the Dunciad works diagrammatically as much as it does sequentially by combining constellations of terms with a linking system of notes, and this turns out to be a key way in which the poem resembles more traditional progress pieces. By visualizing the linguistic differences between progress poems and other contemporary texts as well as training a machine learning classifier to distinguish between the two groups, I find that progress poems feature few distinctive verbs that serve to depict causal relationships. The irony of progress poetry is that it is characterized by deictic phrases and relative clauses that work to organize the material of history diagrammatically, rather than trace the effects of specific actions and events.

Figure 2 Notes on “James-More Smith” in Alexander Pope’s mock-epic poem, The Dunciad in Four Books (London: 1743). Source: Courtesy of the Watkinson Library, Trinity College, Hartford, Connecticut, p. 80.

This is what I observe across all the genres and their corresponding links in the book. Although each contains elements of history (and purports to produce new perspectives on the present’s relationship to the past), the interaction between language and media creates nonlinear representations of historical experience—ones that, for instance, are schematic, palimpsestic, or typological. Such arrangements recall the associative connections between ideas in the mind that move across chronological or causal routes but also, according to Hume, contiguous and comparative ones.

The Case Studies

This book is organized around different case studies of Enlightenment links at work in particular emerging genres. That being said, these cases reveal a broader narrative about the period. In moving from the 1720s to the end of the eighteenth century, I trace how writers went from exploring new ways of arranging language in print to using links to reach beyond the book and blur the boundary between the text and the world.

I begin in chapter 1 with the print device most associated with the hyperlink: the cross-reference. Chambers introduced the first comprehensive system of cross-references in the original edition of the Cyclopaedia, which reimagined the universal dictionary in the form of the modern encyclopedia.53 But the connections were designed to serve a very different task than the one envisioned by new media technologists. Chambers imagined the cross-reference as affording a course of reading that corresponded to the structure of human knowledge, which he depicted in a tree diagram and catalog of headwords in the preface to the work. The cross-reference was not meant to open a democratic, reader-centered mode of engaging the text. Rather, Chambers’s account of his cross-references draws upon Locke’s critique of language. The cross-reference represented an answer to the ambiguity of language by isolating a specific sense of a word in relation to a network of other terms. However, my analysis reveals a pattern of discrepancies between the positions of terms in Chambers’s tree of knowledge and in the connections of cross-references. While the tree abstracts arts and sciences from contemporary social and political issues, the cross-references and language of the entries betray the myriad ways in which bodies of knowledge had become tangled up in the politics of the day. In working through the countervailing tendencies of paratext and text, I demonstrate how the Cyclopaedia helps to frame important questions for Enlightenment links that persist for digital media: How do links index the tension between an author’s planned course and the contingent circumstances of one’s social context? And, how do links both purport to escape the entailments of language and yet often become caught up in those entailments?

Chapter 2 continues my exploration of reference by examining the distinctive features of progress poetry, building upon my discussion of Pope’s Dunciad in Four Books above. By analyzing the language of progress poetry in relation to its frequent use of footnotes, I find that the form is characterized by an indexical poetics that relies on organizing objects across history, geography, and ontological categories. While progress poetry may seem to resemble Biblical genealogy, the primary work of the genre is arranging figures, places, events, and ideas in a kind of palimpsestic tableau. Ironically, a form that purports to depict progress emphasizes arrangement rather than action, organization rather than causation.

The second and third chapters consider genres that produced new views of history, first in poetry and then in the hybrid form of philosophical history. The clustering and classification analyses provide a way of comparing how language works in genres featuring Enlightenment links in relation to comparable texts without them—for instance, how Scottish philosophical histories compare to contemporary narrative histories. The differences I observe primarily concern words and two-word phrases with a particular emphasis on function words, in part because I am interested in how the medial connecting function of links corresponds to the linking function of prepositions, conjunctions, and articles. At other times, as in chapter 3, I use linguistic dictionaries to group words into broader social and psychological categories to gain a better perspective on the types of activities that philosophical histories foreground in comparison to narrative histories.54 By building models from genre and paratextual differences, this approach serves to, as Underwood suggests, “describe a relationship between social and textual evidence.”55

Midcentury Scottish authors developed philosophical history by combining chronological and systematic, or linear and nonlinear, arrangements in both the texts and paratexts of their works. One popular approach for bringing together linear and nonlinear material was to include supplemental finding devices and texts, most notably the index, which offered the reader alternative courses for navigating the book. Unlike cross-references or footnotes, indexes offer a comprehensive representation of the text in one place. Indexes to philosophical histories provide readers a tool for observing the relationships between different principles of social advancement and tracing their occurrences across the body of the work. Relative to classical narrative histories, the texts and indexes of the philosophical histories feature a higher prevalence of language describing abstract processes as well as a relative dearth of language referring to the passage of time. This is what stadial theories of history accomplished: the removal of contingency but also the schematizing of history that, in effect, flattens its diachronic dimension. When actual historical events do appear, they frequently produce tension with the abstract principles of the works. Adam Smith’s comprehensive subject index for The Wealth of Nations (1776) provides a clear example of the friction in these works between textual instances of chronological narrative and the systematic organization of the index. Scottish Enlightenment indexes work by abstracting principles of development and locating them within a static historical schema.

The first three chapters consider how writers developed links to explore the possibilities of print arrangement for exhibiting new forms of knowledge and history in emerging genres of the period. The final chapter turns to the novel to consider how late eighteenth-century novelists incorporated historical material to emphasize the sensory experience of confronting the past. While links in the Cyclopaedia and philosophical histories support the move from concrete particulars to abstract bodies of knowledge and principles, gothic and historical novels reveal how links may foreground the details of context. What a link accomplishes depends upon both the language that constitutes it and its material situation—its physical location and context. The chapter epigraph is a paratextual feature defined by its position set above and apart from the text that follows. At the end of the eighteenth century, led by the success of Radcliffe’s works, gothic novels exploded in popularity, and they popularized the convention of appending poetic extracts as epigraphs for each chapter. Like other kinds of chapter headings (such as “bill-of-fare” summaries often found in picaresque novels as well as the dates and addresses of epistolary novels), epigraphs indicate a genre difference. They are textual artifacts from the literary past that function as landmarks over the course of narratives that shuttle readers across time and space. As the tapestry scene from Udolpho illustrates, the narration of coming to know the environment represents the process of language making material legible, and from the countervailing direction, the position and orientation of the epigraph reveals how a material context produces meaning. Epigraphs and other fragments of the past in gothic and historical novels produce webs of British literary history that call out to readers.

Links locate language in the world. What is lost when such connections are rendered invisible? I conclude by considering how the social and historical conditions of text seem to be inescapable in the formats by which language and media, both print and digital, interact. For more than any previous paradigm, the language of digital media is a product of the surrounding linguistic and, by extension, social context. What appears on your timeline or in your search results is equally determined by the terms of your query, the previous links you have clicked, whom you friend or follow, and whom you resemble in terms of age, race, sex, class, and so on. All writing depends to some extent on the context of the writer, and writers frequently imitate, allude, or even plagiarize without acknowledgment. Digital media, however, increasingly presents text for users that has been generated by replicating the patterns of previous writing without noting the sources. While the link has been a hallmark of digital media since the introduction of the internet, its simple legibility has been gradually superseded by applications that resist interpretation—the algorithms that operate behind the interfaces of search engines and social media sites. Attending to the evolving relationship between text and its techniques of reference means interrogating new technologies that do not acknowledge their links.


1. For instance, see The New Media Reader, ed. Noah Waldrip-Fruin and Nick Montfort (Cambridge, MA: MIT Press, 2003). The anthology opens with two introductory essays that begin their accounts of new media with early hypertext plans. Janet H. Murray, “Inventing the Medium,” 3–13; and Lev Manovich, “New Media from Borges to HTML,” 13–28. The selections in the anthology also begin with Jorge Luis Borges’s short story “The Garden of Forking Paths,” which is cited as an early model for hypertext fiction, and Vannevar Bush’s 1945 article, “As We May Think,” in which he describes the plan for the Memex machine.

2. Vannevar Bush, “As We May Think,” Atlantic Monthly 76, no.1 (1945), accessed September 17, 2021,

3. I address this history at greater length here, Collin Jennings, “Of Calendars and Graphs: Transformations of the Succession Concept,” The Eighteenth Century 60, no. 1 (2019): 87–106.

4. Gérard Genette, Paratexts: Thresholds of Interpretation, trans. Jane E. Lewin (Cambridge: University of Cambridge). For work building upon Genette’s theory, see Anthony Grafton, The Footnote: A Curious History (Cambridge, MA: Harvard University Press, 1997); Janine Barchas, Graphic Design, Print Culture, and the Eighteenth-Century Novel (Cambridge: Cambridge University Press, 2003).

5. See John Bender and Michael Marrinan, The Culture of Diagram (Palo Alto, CA: Stanford University Press, 2010); Daniel Rosenberg and Anthony Grafton, Cartographies of Time: A History of the Timeline (New York: Princeton Architectural Press, 2010); and Grafton, The Footnote; Brad Pasanek and Chad Wellmon, “The Enlightenment Index,” The Eighteenth Century 56, no. 3 (2015), 359–82.

6. T. H. Nelson, “Complex Information Processing: A File Structure for the Complex, the Changing and the Indeterminate,” ACM Proceedings of the 1965 20th National Conference (August 1965), 96.

7. Stuart Moulthrop, “You Say You Want a Revolution? Hypertext and the Laws of Media,” in Essays in Postmodern Culture, ed. Eyal Amiran and John Unsworth (Oxford: Oxford University Press, 1994), 71.

8. Jay David Bolter, Writing Space: Computers, Hypertext, and the Remediation of Print (New York: Routledge, 2001), 25. Gunnar Liestøl similarly declared in 1994, “Hypertext reconfigures the way we conceive of texts. The facilities of manipulation, individual navigation, and freedom from given, authoritative structures provide us with new practices of writing and reading.” Liestøl, “Wittgenstein, Genette, and the Reader’s Narrative in Hypertext,” in Hyper / Text / Theory, ed. George P. Landow (Baltimore, MD: Johns Hopkins University, 1994), 88.

9. Saul Kripke, “Frege’s Theory of Sense and Reference: Some Exegetical Notes,” Theoria 74 (2008): 182.

10. I primarily use the term computational text analysis as opposed to distant reading or text mining because the techniques I use are not solely for looking at large collections of texts. I also produce what Michael Gavin has called “computationally assisted close reading.” Gavin, “Vector Semantics, William Empson and the Study of Ambiguity,” Critical Inquiry 44 (2018): 642. For the origin of the term distant reading, see Franco Moretti, Distant Reading (New York: Verso, 2013).

11. Jerome McGann, Radiant Textuality: Literature after the World Wide Web (New York: Palgrave, 2001), 56–57. McGann also has demonstrated the scholarly benefits of digitizing historical multimedia archives in the Rossetti Archive (, accessed May 15, 2023. More recently, McGann has addressed the limitations of markup frameworks for capturing the different “dimensions” of texts that scholars want to examine. See, McGann, “Editing and Curating Online,” Textual Cultures 15, no. 1 (2022): 53–62.

12. Geneviève Teil and Bruno Latour, “The Hume Machine: Can Association Networks Do More Than Formal Rules?” Stanford Humanities Review 4, no. 2 (1995): 47.

13. Teil and Latour, 47.

14. Alan Liu, “The Voice of the Shuttle,” accessed September 15, 2021,

15. See Liu, “What the Allusion Means,” accessed September 15, 2021,

16. See Geoffrey Hartman, “The Voice of the Shuttle: Language from the Point of View of Literature,” Review of Metaphysics 23, no. 2 (Dec. 1969): 240–58.

17. See Patricia Klindienst, “The Voice of the Shuttle Is Ours,” Stanford Literature Review 1 (1984): 25–53.

18. Ann Radcliffe, The Mysteries of Udolpho (Oxford: Oxford University Press, 2008), 474.

19. Bush, “As We May Think.”

20. John Locke, An Essay Concerning Human Understanding, ed. P. H Nidditch (Oxford: Clarendon Press, 1975), 163; Bk. II.xii.17.

21. Sean Silver, The Mind Is a Collection: Case Studies in Eighteenth-Century Thought (Philadelphia: University of Pennsylvania Press, 2015), 31.

22. Bruce Clarke, “Information,” in Critical Terms for Media Studies, ed. W. J. T. Mitchell and Mark B. Hansen (Chicago: University of Chicago, 2010), 158.

23. Warren Weaver, “Some Recent Contributions to the Mathematical Theory of Communication,” in The Mathematical Theory of Communication, ed. Warren Weaver and Claude Shannon (Urbana-Champaign: University of Illinois Press, 1949), 9.

24. James Gleick, The Information: A History, a Theory, a Flood (New York: Vintage, 2012), 226.

25. Zellig S. Harris, “Distributional Structure,” in The Structure of Language: Readings in the Philosophy of Language, ed. Jerry A. Fodor and Jerrold J. Katz (Englewood Cliffs, NJ, 1964), 43.

26. J. R. Firth, “A Synopsis of Linguistic Theory, 1930–1955,” in Studies in Linguistic Analysis (Oxford: Oxford University Press, 1962), 11.

27. Yohei Igarashi, “Statistical Analysis at the Birth of Close Reading,” New Literary History 46, no. 3 (Summer 2015): 485–504.

28. Gavin, “Vector Semantics, William Empson and the Study of Ambiguity,” 641.

29. Mary Poovey, A History of the Modern Fact: Problems of Knowledge in the Sciences of Wealth and Society (Chicago: University of Chicago Press, 1998), 149.

30. David Hume, Treatise of Human Nature, ed. David Fate Norton and Mary J. Norton (Oxford: Oxford University Press, 2009), 105, bk. 1.3.14.

31. Poovey, A History of the Modern Fact, 202.

32. David Hume, An Enquiry Concerning Human Understanding and Other Writings, ed. by Stephen Buckle (Cambridge: Cambridge University Press, 2007), 56.

33. See Jesse Molesworth, Chance and the Eighteenth-Century Novel: Realism, Probability, and Magic (Cambridge: Cambridge University Press, 2010).

34. Sandy L. Zabell, “The Rule of Succession,” Erkenntnis 31, no. 2/3 (1989): 290–92.

35. Ian Hacking, The Emergence of Probability: A Philosophical Study of Early Ideas about Probability, Induction and Statistical Inference (Cambridge: Cambridge University Press, 2005), xix.

36. For a useful comparison between Bayesian and frequentist statistics, see Branimir K. Hackenberger, “Bayes or Not Bayes, Is This the Question?” Croatian Medical Journal 60, no. 1 (2019): 50–52.

37. Andrew Piper, Dreaming in Books: The Making of the Bibliographic Imagination in the Romantic Age (Chicago: University of Chicago Press, 2009), 8.

38. Clifford Siskin and William Warner, “This Is Enlightenment: An Invitation in the Form of an Argument,” in This Is Enlightenment, ed. Clifford Siskin and William Warner (Chicago: University of Chicago Press, 2010), 7.

39. See Jerome McGann, Radiant Textuality; Johanna Drucker, Graphesis: Visual Forms of Knowledge Production (Cambridge, MA: Harvard University Press, 2014); Elaine Treharne and Claude Willan, Text Technologies: A History (Palo Alto, CA: Stanford University Press, 2019).

40. Alan Liu, Local Transcendence: Essays on Postmodern Historicism and the Database (Chicago: University of Chicago Press, 2008), 25.

41. Andrew Piper, “Reading’s Refrain: From Bibliography to Topology,” ELH 80, no. 2 (2013): 373–99.

42. Michael Gavin, Literary Mathematics: Quantitative Theory for Textual Studies (Stanford, CA: Stanford University Press, 2023), 65.

43. Julia Kristeva, “Word, Dialogue, and Novel,” in The Kristeva Reader, ed. Toril Moi (New York: Columbia University Press, 1986), 37.

44. See Andrew Piper, Enumerations: Data and Literary Study (Chicago: University of Chicago Press, 2018); Ted Underwood, Distant Horizons: Digital Evidence and Literary Change (Chicago: University of Chicago Press, 2019); and Gavin, Literary Mathematics. For an analysis on the significance of book history for text analysis work, see Katherine Bode, A World of Fiction: Digital Collections and the Future of Literary History (Ann Arbor: University of Michigan Press, 2018).

45. Ian Watt, The Rise of the Novel: Studies in Defoe, Richardson, and Fielding (Berkeley: University of California Press, 2001), 9–34.

46. Jonathan Kramnick, Actions and Objects from Hobbes to Richardson (Palo Alto, CA: Stanford University Press, 2010), 3. See also Kramnick, “On Beauty and Being at Home,” in Paper Minds: Literature and the Ecology of Consciousness (Chicago: University of Chicago Press, 2018), 74–99.

47. Kramnick, Actions and Objects, 2.

48. Silver, The Mind Is a Collection, 13–16.

49. Christina Lupton, Knowing Books: The Consciousness of Mediation in Eighteenth-Century Britain (Philadelphia: University of Pennsylvania Press, 2012).

50. Ann M. Blair, Too Much to Know: Managing Scholarly Information before the Modern Age (New Haven, CT: Yale University Press, 2010), 3.

51. Alexander Pope, “The Dunciad, Variorum,” in The Dunciad, ed. James Sutherland, The Twickenham Edition of the Poems of Alexander Pope, ed. James Sutherland, vol. 5 (London: Methuen, 1943), 107, II.44–46; Aubrey Williams, Pope’s Dunciad: A Study of Its Meaning (Baton Rouge: Louisiana State University Press, 1955), 69.

52. See Williams, Pope’s Dunciad, 131–58; Blakey Vermeule, “Abstraction, Reference, and the Dualism of Pope’s ‘Dunciad,’Modern Philology 96, no. 1 (1998): 16–41.

53. Jeff Loveland, The European Encyclopedia: From 1650 to the Twenty-First Century (Cambridge: Cambridge University Press), 191.

54. See J. W. Pennebaker, R. L. Boyd, K. Jordan, and K. Blackburn, K., The Development and Psychometric Properties of LIWC2015 (Austin: University of Texas at Austin, 2015), DOI: 10.15781/T29G6Z.

55. Underwood, Distant Horizons, 24.