Native Speech Is a Myth

Part of the valuation, and devaluation, of translation relies on the idea of the native speaker as a source of authenticity. The figure of the native speaker projects coherence: it defines a language as a discrete, stable, and uniform system used in a single way by every speaker. Yet the speech attributed to the native speaker is always an abstraction, not necessarily practiced by actual speakers. In truth, many speakers of a language speak it differently—and all authentically (Rosa). Manners of speaking change, dialects abound, ideolects too. Native speech and native speakers are infinitely variable. The myth of native speech is further exposed by the figure of the expert medievalist: no one today is a native speaker of a medieval language and yet these third-hand learners are the only authorities we have (Warren, “Translation” 66). Speech, moreover, is only one mode of languaging, alongside signing, ideograms, and other graphic forms.

Medieval vernacular languages also give the lie to the native-speaker myth. Because the boundaries between languages were somewhat fluid, with active mutual borrowings, words can't always be firmly attributed to a single language (Warren, “Translation” 58). In Chrétien's text, for example, Cligès is the name of a knight with Greek and British parents: the language of his name is not entirely clear. The text as a whole has also been recorded in a variety of linguistic forms. Most manuscripts of Cligès are written in some version of a language group called “langue d'oïl,” named for the way that people wrote, and presumably pronounced, the word yes (Galderisi and Agrigoroaei). Across the regions now called France, these so-called dialects cluster in the north. And yet any individual person copying the text in the Middle Ages might inflect the received text with details from their own language habits. It so happens that the dialect of Chrétien's region greatly influenced the development of modern French because local nobles became kings. French as it exists today emerged relatively recently: regional speakers exerted centralized political authority, which included the modern creation of government-sponsored dictionaries and grammars; those reference works and textbooks became the basis of a compulsory education system beginning in the nineteenth century; politicians exported that system around the world as an imperialist language pedagogy (Warren, “Politics”).

Approaching translation without an idea of a fixed native speech is extreme because it disrupts the priority of sources. It also resists normative definitions of languages, which tie legitimacy to a single speech pattern. From this perspective, a translation is not a lesser expression of a prior source, nor is it entirely constrained by the norms of its own language.

Language Betrays Empire

Medievalists in the European tradition often start discussions of translation by explaining the Latin phrase translatio studii et imperii—“transfer of learning and empire.” The theory that language, knowledge, and power move together from east to west is an influential medieval trope with a long postmedieval legacy. The pithiest version might be Antonio de Nebrija's, published in the fateful year 1492: “language was always the companion of empire” (Armillas-Tiseyra 202). This idea derives from a combination of Hebrew and Roman writings that define knowledge and authority as persistently moving toward better forms (Webb; Smalley). Linguistic translation has itself played a pivotal role in the elaboration of translatio studii et imperii—beginning with the Latin translation of the Hebrew bible by Jerome (d. 420 CE) and continuing on through centuries of Latin teaching with the text of the Roman historian Sallust (d. circa 35 BCE), De conjuratione Catilinae (The Catiline Conspiracy), which includes the famous sentence “Ita imperium semper ad optimum quemque a minus bono transfertur” (“Empire is always transferred from the lesser to the better”; Harkness 78; translation mine). Nineteenth-century Latin textbooks that include Sallust are still widely in print—including a 2015 print-on-demand edition in the War College Series (Harkness). Even today, translating Latin teaches students imperial values alongside grammar.

In the twelfth century, Chrétien de Troyes created an influential literary expression of translatio studii et imperii in the prologue to Cligès. The prologue first introduces the writer, then the bare outlines of the plot, before turning to the story's source in an ancient book and the theory of translatio transmitted by similarly ancient books. In Chrétien's version of the theory, “learning and chivalry” have moved from Greece to Rome to lodge in France. The prologue, however, has already contradicted this claim. The opening lines state that the hero lives in Greece, which implies that superlative chivalry hasn't moved at all. The hero's father did leave Greece, but traveled straight to England, not to Rome or France. The subsequent narrative has each knight traverse the whole of Europe more than once. In the end, neither knight represents the theory of translatio as linear improvement. The very idea comes to seem ridiculous. In this way, Chrétien crafts an illustration of translatio that self-destructs, undermining the whole concept of imperial transfer. He turns empire into chivalry, and then turns chivalry into a joke. Thus do learning and empire part ways.

This translation lesson reveals how claims of fixed linearity rely on their opposite—the dispersed, erratic, recursive, fragile movement of ideas and people in many directions at once. It points to the ironic structure of translation itself: preservation requires change. Under these conditions, language tears down the empires it seems to build.

Beauty Is Trash

Trashing aesthetics is the beginning of translation's fame. By which I mean, notions of beauty are inherently hierarchical because they elevate certain forms above others. And such hierarchies of judgment underlie the most common first question that people ask of a translation: “Is it good?” For the most part literary scholarship is well past this question as the main basis for meaningful criticism. Translation studies too has moved on to more nuanced approaches. And yet, the question hovers in both popular and academic discourse. The very idea that judging a translation means comparing it to a source also defines translation as a lesser aesthetic mode. In this comparison, faithful is usually good but could also be derivative; freedom shows independent creativity but could also be a betrayal.

Medieval studies can show the way past these dichotomies. Some of the most celebrated medieval authors were first and foremost translators. What's more, writers regularly claimed to translate or merely copy even when they were inventing: truth and authority derived from continuing established tradition rather than from originality. Translation made a work more valuable, not less. The Cligès narrator invokes this trope when he claims that he found the story in a library: “The book is very old in which this story is told, and this adds to its authority” (Comfort). In this framework, readers should value all versions equally. This approach extends to modern translations of medieval texts (Warren, “Translating”). After all, in Cligès it turns out that the very old book may have been lying, if it even existed at all.

English translations of the statement about empire's westward movements from Cligès illustrate well how any and all translations can bring novel cultural insights. Four translations each render a distinct version of the hope that the “height of learning” will remain in France:

In each case here, a singular pronoun refers to learning, not to both learning and chivalry (translatio studii et imperii). Each text implies that chivalry may have remained in Rome. With this syntax, knowledge systems might critique political systems rather than only supporting them. The one feminine pronoun, meanwhile, brings a personification into the story that calls into question the masculinist values of knighthood. Finally, each translator's verb choice conveys distinct relationships with knowledge: hoarding (retain, prevent future changes), love (cherish), stability (maintain), and further improvement (advance). Each of these expressions translates a different truth about the translatio topos. I note, too, that two translators locate the narrator outside France (“there”) and two within it (“here,” “our part”). These conflicting decisions render the narrator's position ambiguous, undermining readers’ ability to pinpoint which values are being endorsed and which critiqued.

The variable presence of fame across these four translations further demonstrates the fragile bond between empire and knowledge. Newell uses fame twice—once for the ambition of Cligès's father and once for what the Greeks and Romans have lost. Kline only uses fame for the first situation—and Comfort and Gardner not at all. Thus does Newell construct a new version of Chrétien's famous irony: the fame that defines cultural achievement is either a hope for the future or a regret about the past, never a present reality. By comparison, the loss of the word fame in two versions performs a different kind of irony: a key concept of translatio has been erased by translation. These real acts of translation thus undermine the very ideology of transferring knowledge and empire.

The prologue of Cligès ends with metaphors for the reputation that the Greeks and Romans have both lost in favor of the French—all related to fire:

Each version draws from a different aspect of fire—flame, embers, ash. And each version provides a subtly distinct interpretation of how memory and fame disappear. Comparison across them affords ample insights into literary technique and the mechanisms of metaphor. Newell and Gardner make the absent situation the most memorable (“vivid flame,” “bright glow”); Comfort combines the past and present into an impossible image (“glowing ash”); Kline invokes a specific method for putting out a fire (water that quenches). With so many ways to end an empire, France can't be far behind Rome and Greece.

Treating all translations as potentially valuable overturns the priority of sources, original languages, and “very old books.” Comparison of multiple versions in a single language can advance literary insights even without access to any source languages. At this extreme, even a bad translation is good for something. This approach resists the specialness of literature as a category in order to make the practices of literary interpretation more relevant to all kinds of reading.

Machines Are Literary

It's probably second nature for literary scholars to belittle machine translation. If even highly talented human translators have struggled for recognition, how could machines garner praise? Moreover, the recent expansion of large language models (LLMs) such as ChatGPT has brought new levels of naturalistic imitation to automated text production, generating new ethical and aesthetic problems (Raley and Rhee; Kirschenbaum; Wolfram). The result may be a devaluing of text as a uniquely human product—and of the literary as a textual feature. Machine translation, however, still produces glitches that call for literary interpretation. When machines, like human poets, break language norms, even generic texts produced by algorithms can invite readers’ creativity.

In the history of computing, machine translation began with the lofty goal of “fully automated high quality translation.” Eventually, engineers settled on “fully automated useful translation” (Warren, Holy Digital Grail 76). This focus on use-value makes machine translation seem fundamentally antiliterary and antihumanistic. And yet, machine translation has become integral to many literary and humanistic experiences in the twenty-first century. In fact, in the shift from quality to utility, poetry has returned.

Since “Old French” isn't an option on automated translation tools, I used a modern German translation of Cligès to explore machine renderings of the statement about the westward movement of learning and chivalry. I started by copying the German text into two widely used open-access tools—and later added some experiments with ChatGPT:

The variations across these translations are revelatory. Rendering the German Bildung as both “education” and “learning” points to subtly different aspects of knowledge transmission: how to share established knowledge (education) and how to acquire new knowledge (learning). “Knighthood” and “chivalry” are also distinct, the first a social status and the second a set of values. “Flowering” and “blossoming” are synonyms that evoke two different linguistic and political histories—one Latinate and the other Germanic. The “regenerate” function in ChatGPT brought more variations: from “blossoming” to “blossom” to “bloom;” from “emerged” to “arose” to “originated.” Each variant brought new ways to think about knowledge transmission. The regenerations also brought a demand for comparative judgment: “Was this response better or worse?” This question revives the hierarchies aligned with empire, reducing linguistic nuance to a simple binary choice.

After five regenerations on ChatGPT, though, Rome was still a migrant. Rather than chivalry or learning moving to France, a misplaced modifier put Rome itself in motion (“Rome, which has now migrated to France”). This error recurs in every version across all three tools. The algorithms have produced a metonymy: Rome, associated with chivalry and learning, takes their place. This error is likely apparent even to monolingual readers of English, who can infer that a city doesn't move to a new country and so are prompted to make their own interpretive moves. In this mode of extreme translation, machines produce literature precisely when they produce inaccuracies.

Canons Are Fictions

The power of literary canons to define cultural values, social identities, and political ideologies has been thoroughly exposed across many fields of literary study. Their power, though, is hardly fixed: the very contingency of creating and disseminating canons reveals their potential fragility. And if canons can change, so can the fixed values that they seem to sustain. In other words, language and literature may be companions of empire, but they also encode the unraveling of power structures.

Translation amplifies this reality of canons by simultaneously reinforcing them and undermining them. On the one hand, a translation expands a text's audience and confirms that the source deserves that broader audience. On the other hand, a translation introduces differences that destabilize the source's authority. Translation thus sews contradiction into the fabric of canonicity. Theories of translation inhabit this tension where “language change is simultaneously meaningful and meaningless” (Warren, Holy Digital Grail 47). As translation preserves and distributes texts, it moves values through time and across cultures. Translations must appear to change nothing important, and yet no text can survive translation unchanged.

Cligès foregrounds these dynamics by beginning with the author's bibliography—the canon of texts that frame translatio as the preservation of knowledge from language to language, place to place:

With this first sentence, the narrator sets out to fix the canon of Chrétien's work. The effort has been partly successful, in that Erec and Enide has been lauded as the first modern French novel (Uitti and Freeman 36). And Chrétien remains a vibrant starting point for even the most incisive critiques of literary tradition, such as Thinking through Chrétien de Troyes, by Zrinka Stahuljak, Virginie Greene, Sarah Kay, Sharon Kinoshita, and Peggy McCracken (Stahuljak et al.). At the same time, Chrétien's canon and his own authorial reputation have fluctuated across the centuries (Hult). What's more, Chrétien's authorship of the other works named in these opening lines is now considered questionable—and possibly even fictional. Have they been lost or were they created right here to bolster the author's credibility?

The role of translation in this bibliography further complicates the canon. Second in fame to Erec and Enide (which does exist) is Chrétien's translation of Ovid's Art of Love (which does not exist). In naming this text, the translator elevates himself by highlighting an already famous author—but also eliminates the need for that author. The translation confirms Ovid's canonicity but also displaces Ovid with the translator's own text. Other texts mentioned here might also be translations from Ovid, as several seem related to The Metamorphosis (Roustant). Regardless, Ovid is the only author named here: the French translator who is the subject of the sentence remains anonymous—another joke about his canonicity. Chrétien has erased himself in order to enhance his reputation, a risky ploy that can lead to the loss of reputation all together. Indeed, some translators suppress his name where it does occur later in the prologue (Kline). From the first line, the audience is divided into insiders who know the author's name and outsiders who don't. Among these feints and reversals, we can never be sure who wrote what.

Translation takes canons to their extreme contradiction: they are ubiquitous and unavoidable—and also impossible and fragile. Chrétien's possibly fictional statement about his own translations is just an extreme case of the fabrication that attends all canons.

Media Make Meaning

The idea that form and format impinge on the meaning of texts has broad currency. Marshall McLuhan's formulation is perhaps the best known: “the medium is the message” (7). A close second, at least in some circles, is Donald F. McKenzie's “forms effect meaning” (4). Translation or transfer from one form or format to another is a particularly meaningful feature of medieval literature. Each surviving manuscript is a unique creation, copied from some prior version (even if only one copy exists today). French translations of Sallust's Latin history of the Catiline conspiracy, for example, brought the idea of imperial improvement to new audiences in the fifteenth century—a message also conveyed by the text's visual translation into deluxe illustrated manuscripts (Hedeman). Media translations of many kinds continue today with editions, photographs, and networked interfaces.

Cligès, like many medieval texts, now exists in many different forms: medieval manuscripts, digitized medieval manuscripts, printed editions, digitized printed editions, microfilms, digitized microfilms, HTML pages, PDFs, and Internet addresses with shifting interfaces. All together, these forms constitute a new kind of multidimensional and multitemporal book that combines manuscript, print, and digital media (Mak; Kirschenbaum and Werner; Cordell; Drimmer; Warren, Holy Digital Grail 28–31). Each format translates from another, providing a specific way of knowing texts and books. Editing translates handwriting, photography translates dimensions. Each format opens new perspectives; each can respond to different kinds of questions. All this copying keeps rewriting literary history, redefining canons, and restructuring access. In a unique manuscript, preserved in a library with restricted access, a copy of Cligès is a rare and remote artifact of a distant time; in an open-access English translation on a public website, a copy of Cligès is an entertaining story anyone might peruse (based on Warren, Holy Digital Grail 237). In each case, the medium effects meaning.

From the perspective of media, even bad copies have something to teach us. Tellingly, the “bad translations” that I've used in this essay are readily available in what some might consider the worst book format currently littering Google.com, Amazon.com, Walmart.com, and other aggregators: print-on-demand (POD) books made from PDFs of out-of-copyright editions. Thousands of titles appear with the same simple cover design, offering seemingly new editions of works both popular and obscure. Like Chrétien, POD marketers sell a fictional canon—the “aura of literary tradition” defined by the prosaic fact of copyright law (Warren, Holy Digital Grail 198). In their book blurbs, POD marketers tout the stability of cultural value, turning outdated texts into accessible treasures (230–31). Occasionally, a supposedly cheap paperback appears at an astronomical price when an algorithm mistakes the single copy of a POD listing for a rare book. Such unreliable pricing reveals again the truly fictive value of canons (235). In translating from digital to print, POD reverses the so-called progress of knowledge, disrupting again the logic of translatio studii et imperii.

At this extreme end of translation, form and content create each other's meaning. Even this essay, which is available in PDF, HTML, and printed paper, partakes of these dynamics. Formats are not empty vessels for a stable text but partners in producing every kind of meaning.

I have one more extreme translation lesson for you. As I completed this essay, I learned that PMLA does not include translations in its tally of word limits (“Submitting Manuscripts”). The purpose, of course, is to promote multilingual citation so that authors can broaden their readership without having to shorten their arguments. And yet the concept of not counting translated words throws words themselves into existential limbo. How many words does an essay full of translations even have? This playful take on word counts circles back to the world of extreme translation. In this world, medieval literature can be as mundane (worldly, down to earth) as anything else people read. Go ahead. Give any old translation a try.