Article body


There is a cognitive effect that occurs in the majority of pragmatic and mass-market literary translations, and it plays out at the level of linguistic form. I will call it the “copy effect.”[1] I have, for example, a German-language audiobook version of Stephen King’s popular novel Misery (1987). The translation is titled Sie (2011), and as I listen to its narrative in German through earbuds, I can read along visually in the American source text and get along quite well—not processing word-for-word, of course, but chunk-for-chunk simultaneously. Whenever I want, I can coordinate precise points of semantic and formal invariance in both source and target texts. The American text on the page can be synchronized to the German-language audio like a simultaneous interpreter’s words to those of the conference speaker, or like a film’s subtitles to the words of an on-screen speaker. This copy effect—this bi-textual co-construction and doubling—can only be possible via some kind of copy-orientation in the translation itself.

Such an effect can occur only in the type of translation that strives for similarity with its source text in both linguistic meaning and form[2]—in other words, the type of translation that is literal enough to compel a strong sense of copy adjacency even while the reader remains aware that the translation is a variant in all of its obvious ways.[3] In this study, I hope to demonstrate that current theoretical voices, both in book history and Translation Studies, are reluctant to examine, address, or even acknowledge the real formal similarities that give rise to the copy effect just described, and are therefore arguing at odds with translation’s inherent complexity. My hope is to bring a backgrounded aspect of inter-lingual translation back into evidence, to show that signs of inter-textual identity at the formal level co-exist with those of difference and abide. The idea is not only to add needed theoretical counterweight to positions of inter-textual non-identity, but also to illuminate the secret—academically leveraged but never acknowledged—position of identity from which all variance-oriented arguments must spring.

The larger aim, then, is to help safeguard the paradox that is translation, a paradox within which a postmodern thesis of inter-textual non-identity is currently favoured and its antithesis of inter-textual identity very difficult to find formulated anywhere. Within a paradox, thesis and antithesis are meant to inhabit the same conceptual space with equal value. And so this study is a push to theorize the “sameness” identifying source and target texts at a deeper formal level. And it could also be called a push back toward the productive (if ambivalent) middle space of the paradox, where we should never feel too comfortable calling a translation categorically the same as or different from its source. My own position is squarely in this middle space, but as I have not yet come across any kind of theoretical articulation acknowledging and validating (let alone examining) my copy effect—which is nevertheless very common as cognitive experiences of translation go—I will attempt one here.

First, I would like to discuss the paradox that I have just mentioned. Two opposing book historic positions will help me do this: the first identifies translations with copies in a sweeping way, and the second privileges a position of non-fixity or variance between texts and their translations. Out of the second however, such as embraced by Anne Coldiron (2019) for example, comes the concept of “transformission” as it may apply to translation. When unpacked, this new blend concept helps bring the needle back to the middle space of the paradox, I will argue. Then comes the business of talking about formal identity between texts and their translations, of locating a copy effect conceptually, of distinguishing what type of form precisely is being experienced when one is under its sway. This is a type of form that I am going to call “Pythagorean” because it is strictly quantitative, derived from words as the countable tokens of a text’s material organization.

Then I will move on to translation scholarship itself, which has even in its most text-oriented theories been remarkably silent on this quantitative conception of form, and has made a habit of (1) denying formal similarity as a goal in translating; and (2) eliding the obvious co-presence of both semantic and formal similarity in most translations. To illustrate these habits, I will take my reader through a series of three theoretical intersections with the question of linguistic form in translation, all ordered along an ascending cline toward what I would consider to be the most tenable position. First comes David Bellos’s (2011) statement that formal similarity in translation is irrelevant, then Brian Mossop’s (2017) more qualified statement of the same, and finally Anthony Pym’s (2004) eliding acknowledgement that formal similarity really does matter. In all cases, translation is held at a comfortable distance from the idea of the copy.

As we move through these three critical intersections, I will be giving some depth to the book historic position. Specifically, I will address the materialist stance afforded by the perspective of book history, and then the efforts of Rita Copeland (1991) to connect translation practices to the ideological and institutional contexts of the medieval copyist, and ultimately to the copia verborum, a classical concept encompassing the ideas of (1) a stockpile of stylistic variants, (2) formal and semantic parallelism between these variants, and (3) the transcriptive or rote copy itself. As a concept, the copia verborum is very valuable, I will argue in the final part of this study. Not only does it relate translating and copying conceptually, but we find its iteration within today’s computer-assisted translation (CAT) practices—and specifically within the principles governing translation memories (TMs). Pym (2014) has recently drawn our attention to certain continuities between medieval and contemporary translation practices. I would argue that the principles informing the copia verborum are chief among them, and that contemporary translation practices using TMs are illuminating the connection between copying and translating in a particularly powerful way.

1. Positions in book history

Book historians are now visiting the paradox of inter-textual identity and difference that has preoccupied Translation Studies for the past 50 years. Like translation scholars, they are split in their positions: some prioritize identity and treat translations as copy-like, and others signal the inter-textual variances brought about by inter-cultural displacement. In the first group, discourse passively conflating translating with copying is abundant[4], but I will bring forward here an example that is more actively formulated. When asked to define “translation” from the perspective of the book’s material supports and frames, Roberta Capelli asks and then answers a question of her own:

[…] est-il possible d’opérer une distinction entre la phénoménologie de la copie et celle de la traduction? La réponse nous semble devoir être négative, car toute oeuvre source et toute version intermédiaire – en tant qu’exemplaires définitifs et reproductibles d’un original hypothétique – sont des « copies », et parce que leurs traductions en français sont à leur tour des « copies de copies ».

2011, p. 225

This is a conservative position in material book history, and it stands in sharp contrast with Coldiron’s front-line position with its focus on variation as it signifies target cultural intervention:

We have begun to see variation and variant in translation, especially at what would have been thought the ‘too-free’ side, as important signs of cultural intervention, as symptoms, or as keys to understanding the encounter with alterity that any given translation witnesses and represents

2019, p. 210

There is an obvious challenge here to the reductivity of a statement like Capelli’s, but not to its lack of foundation. By this I mean that a translation can certainly be like a copy, and it can have the effect of a copy, but it should never be categorically reduced to one with a sweeping erasure of inter-textual difference.[5] The inverse applies as well: for all the differences that might distinguish it from its source, a translation should never be reduced to pure non-identity. The only thing that appears to be categorical here is the need for a paradox, and Coldiron formulates one with the help of Randal McLeod (2009; Clod, 1991). Out of “transformation” and “transmission,” McLeod (2009; Clod, 1991) has created a portmanteau with epistemological potential: “transformission.” Coldiron sees room enough within it for “translation” as well, and specifically a good framing concept for the study of early modern translation practices (see also Belle and Hosington, 2019).

A blend like “transformission” urges us away from binary thinking. Demonstrable identity through space and time (transmission) and demonstrable non-identity (transformation) occupy the same conceptual space, and Coldiron’s chosen descriptor for translation attests to this: “Translations, after all, are variant versions of a work.” (2019, p. 205, my italics). “Variance” by any definition relies as much on the premise of deeper axes of identity unifying texts and their versions as it does on any foregrounded inter-textual difference. Variance (transformation) would be the divergence—accidental or purposeful, superficial or more thoroughgoing—that becomes possible only if we are experiencing a text concomitantly as something iterated (transmitted): a cultural object with quotation marks around it, identical to itself wherever it may travel and by whomever it may be read, and whose versions may present tangible signs of this identity alongside those of difference.[6]

Together, Capelli’s and Coldiron’s positions offer in one snapshot the conservative view and the emerging front line of book historic views on identity and variance in translations. More importantly, Coldiron proposes a path for research: “Transformission asks us in particular to consider material textuality as a co-factor in translation, concomitant with verbal or linguistic factors” (2019, pp. 205-206, my italics). I see here a direct invitation to talk about translation in a truly material way. However, before I go further, a vital semiotic distinction needs to be made regarding Coldiron’s “verbal and linguistic factors,” which appear in this formulation to fall outside of the materialist’s remit. And, of course, they would, if the “verbal and linguistic” were being invoked in their usual symbolic function as vehicles of the semantic, conveyers of denotations and connotations. However, the “verbal and linguistic” within a text are by no means strictly symbolic. They are also fully material: they have a hard existence in the world, fill up the pages of books and make them physically larger or smaller. They are formatted spatio-geometrically on the page and within all matrices of technological production, and they are commodified in charge-by-the-word scenarios. So I will assume that the strictly material order of “words”—“words” not signifying linguistically but rather indexically and mathematically through the logic of their contiguity on the physical plane—is indeed of concern to a materialist.

I will move forward with this assumption and try to respond to Coldiron’s call, because I see it as an invitation to propose that the “words” of a source text are the first, the most regulatory, and the most pre-conditioning of all material givens within the translator’s immediate working environment—again not “words” as signifiers in the linguistic sense, but “words” as one of a book’s countable components (exactly like its two covers, its binding, etc.). The linear count of words (but also of spaces, characters, syllables, sentences, and any other kind of textual chunk set off materially for rhetorical or formatting reasons, etc.), as well as the spatio-geometrical measure of textual chunks (like columns, etc.) within codiciological frames like margins and pages—I am calling these things “form” in the Pythagorean sense of a limit imposed upon the material world by the rule of numbers.[7] In other words, I do not want to read language and texts; I want to measure them quantitatively, whatever they are—if I can count them somehow, then they count as relevant.

Despite overwhelming evidence that form in this sense regulates translators’ work and often shapes target texts toward a parallelism permitting bi-textual formatting, much of translation scholarship (1) argues as if this type of parallelism simply were not relevant, or (2) narrows its conception of form to exclude the Pythagorean, as if the latter did not apply as any kind of constraint. This is a shame, because acknowledgment of a strong sense of formal sameness in this quantitative sense would open a gate to the long-standing genealogy of critical literature on analogy, which is being re-vitalized in a recent cognitive turn in the Humanities (see Guldin, 2015; Itkonen, 2005; Hofstadter and Sander, 2013).

2. Outright and then qualified denial from Translation Studies

Instead of moving in this direction, unfortunately, Translation Studies has either turned away from texts altogether to focus on the question of hegemony in larger cultural and social contexts, or has appealed to contextual relativism for the purpose of opposing the very idea that translators do anything remotely similar to copying. The most convincing critical perspective here comes from Mossop, who argues that translation should be viewed more like indirect or reported discourse than verbatim reproduction. And Mossop’s (1983) work inspired Barbara Folkart’s Conflit des énonciations (1991).

I do not wish to be mistaken: both Mossop’s and Folkart’s arguments are convincing to me; in fact, I find them unassailable. The problem is that they conflict with another equally unassailable truth: indirect citation and reported discourse are long-standing communication practices with their own conventions, and translation makes no use of them.[8] A translation is not set up discursively like any conveyed message whose continuity with an anterior message might be challenged. On the contrary, it moves directly into and occupies the first-person position of source texts, and channels the enunciation of this other person often right down to its pattern. In light of this, to ask readers of translations to dwell upon the metaphysics of difference may well be to ask something at face-value unfair, namely (1) that they interpret against the continuities that translations promise by their discursive configuration to ensure; and (2) that they disable their sense of the generic, ignore their most basic cognition of formal and conceptual sameness.

One of the fronts on which such requests emerge consistently is the question of formal similarity in translation. To prise apart and keep separate the ideas of translating and copying, scholars seem ready to argue that similarity in linguistic form is not something that translators deem relevant, let alone privilege. The first two types of denial along our gradient are the outright and qualified. Outright denial is the simplest, and can be detected in the following by Bellos: “A competent translator with a lot of time on her hands could easily reproduce the word order and character count of a source by paragraph, sentence, or line, but these kinds of sameness are not considered relevant to the translator’s task” (2011, p. 319). What I perceive here, first of all, is “sameness” in things pertaining to form (word order and character count) rejected out of hand as irrelevant. Later on, a need for “likeness” in these things is acknowledged (are they relevant, then?), while the lion’s share of the translator’s energy is devoted to recovering source text meaning. A conscious effort toward “sameness” then seems reserved for semantics and an “irrelevant likeness” for form, assuming we can understand how any “likeness” that is the result of a conscious effort could be deemed “irrelevant.”

Mossop (2017), for his part, offers the best example of what I am calling a qualified denial. This is the type that opens with an unacceptable statement that then becomes understandable through an elaboration qualifying it. In his recent contribution to the Translation Studies forum, Mossop proposes the concept of “invariance orientation.” Most professionals, he argues, translate with a mindset toward producing sameness in their target texts—but semantic sameness only (2017, p. 331). Oddly enough, even within an article that pushes for recognition of the types of similarity that translators strive for, formal similarity is left explicitly out of the equation. Mossop’s invariance orientation, we find out, is for semantics only:

Invariance-orientation has nothing to do with linguistic form: it allows but in no way favours lexical, syntactic and rhetorical choices which are formally “close” to the source; it does not typically manifest as a word-substitution exercise; it is compatible with either a foreignizing, source-oriented project or a domesticating, target-oriented project.


The unacceptable statement precedes the colon; the elaboration follows it and mitigates what comes before by narrowing the concept of linguistic form. In Mossop’s view, a “formally close” translation would seem to range from acceptable source-oriented strategies to the type of unacceptable word-substitution exercise that can de-nature the target idiom and produce a calque.[9] The problem is that Mossop is arguing as if linguistic form as a concept were restricted on the inside to the correctness of lexis and grammar (linguistic expression) and on the outside to an optimization of style guaranteeing discourse quality. I will call this translation-scholarly conception of form “stylistic form”[10] because it is fundamentally qualitative (concerned with performability and discursive decorum) and stops short at an effect of matching or clashing that occurs when the patterns of a proposed translation are tested against the expectations of a target readership (with respect to things like diction, sentence composition, figures of speech, etc.). Mossop makes a pars pro toto argument here, ignores the vastly greater part of form to argue as if the concept begins and ends in style.

But what of Pythagorean form and its numeric measure? Invariance orientation in translation really would have quite a lot to do with this kind of form. It would have to, if a version like my German-language Misery exists and is the industry standard. This much must be acknowledged if we are not to create untenable contradictions with the evidence. As an example, I will refer back to the “peregrine falcon” translations that Mossop cites in his 1983 article, “The Translator as Rapporteur: A Concept for Training and Self-Improvement.”[11] Two reasons compel me to use them: (1) I believe that they would stand (translation 2, specifically) as an example of the type of professional translation that invariance orientation would produce; and (2) they offer two English versions of a French source text to compare—the first source-oriented to a fault, and the second much improved:

Figure 1

Mossop, 1983, pp. 247-248

-> See the list of figures

Translation 1 is too close to the source lexically and syntactically, a denaturing of the target idiom. Translation 2, by contrast, is styled congenially for the Anglophone target readership. There is no question that translation 2, the rapporteur type, is the more serviceable. Given the re-formulation in translation 2, and given Mossop’s method of providing two target texts to compare—the second (stylistically distanced and superior) contrasting with the first (stylistically close and inferior)—it might seem reasonable to conclude that the translator responsible for the much-improved second version has in fact excluded the linguistic form of the French source from consideration altogether, and effectively free-formed the translation. A conclusion of this type seems to be asked of us.

It would not be reasonable, however, simply because translation 2 and the French source text are still, from the standpoint of quantitative proportion, much too close to each other. Translation 2 identifies strongly with both translation 1 and the source by any Pythagorean measure of form. Its extension through time (oral reading) and/or space (across the page) demonstrates a closeness passing well into parallelism, and nearly verging on symmetry. Mossop himself has underscored this closeness by numbering the lines of the three texts and using italics to facilitate bi-textual coordinate comparison.[12] If we take only the first paragraph, we see that all three texts share the same line count of 20; the word count, again in the first paragraph, is 129 (source):144 (translation 1):131 (translation 2); the character count (with spaces) is 885:879:780; the paragraph divisions and count are nearly the same, barring the one fused paragraph. Formal closeness by this measure has not only been “allowed for,” but very much favored, privileged, and performed discursively at least as much as semantic closeness.

3. Acknowledgment from Translation Studies, but with elision

The last step along our cline is an explicit acknowledgement that formal invariance not only matters in translation, but is also a norm defining it. I am calling it acknowledgment with elision, however, because it involves a conspicuous omission, and demonstrates a reticence to explore its own implications. Pym’s “Propositions on cross-cultural communication and translation” (2004) contains no fewer than four propositions that address the question of quantitative formal representation directly. The first acknowledges quantity as one of two maxims of representation that define translation’s nature: “The maxim of translational quantity holds that a translation represents an anterior text quantitatively. If a translation is longer, a corresponding anterior text is presumed to be longer as well” (ibid., p. 8, prop 8.4).

This is the same position that Umberto Eco defends in two books (2001, 2003) that critique Roman Jakobson’s “On Linguistic Aspects of Translation” (1959). Eco takes Jakobson to task for following Charles S. Peirce (1931-1948) in equating the term “translation” with “interpretation” in its broadest sense and then disregarding the strictures (qualitative and quantitative) constituting translation’s material economy and enforced within its actual practice:

Although it is admitted that in translation proper the substance of the expression changes—since we are shifting from the sounds of one language to the sounds of another—even in the most practical cases there is a sort of implicit stricture by which a certain ratio between substances must be respected […] If to interpret always means to respect the spirit (allow me this metaphor) of a text, to translate means also to respect its body.

Eco, 2003, p. 136

Eco does a formal analysis on industry standard translations ranging from literary (poetic and prosaic) to pragmatic texts (Casio Keyboard instructions in several European languages), and the same kind of tight ratio observable in the peregrine falcon text invariably pertains. There is a very specific type of respect that translators pay to the material body of a source text—a respect for its mathematical abstraction, its “rationale,” and within the latter its “ratio.” And so two conflicting energies seem to conspire: the first a practical one concerned with sloughing off the body of the source text at certain qualitative levels of form (phonetics, lexis, syntax, style); the second a more rational one concerned with preserving and mapping over this same body at other more quantitative and summative levels.

4. Book history: the prototypical translation displays formal invariance

It is in the two propositions where Pym talks about prototype concepts that he most clearly articulates the material research perspective of book historians like Capelli (2011). The criterion of quantitative proportionality between source and target texts is, of course, a material one, and it would be responsible for distinguishing the text type “translation” from any and all other types of formulated re-statement. Consider proposition 8.10:

A prototype concept would be, for example, a view where quantitative differences between a translation and its anterior text were ideally held to be non-significant […], and less-than-ideal translations tapered off in both directions, along continua where texts became significantly longer or shorter than the anterior text, perhaps becoming expansive commentaries at one end, and summaries at the other. In certain grey areas, receiving subjects would hesitate to give the attribute “translation,” or would disagree on the issue […].

Pym, 2004, p. 9, prop. 8.10

This statement finds a compelling echo in one of Jacques Derrida’s (2005) more straightforward essays on translation. Derrida echoes Pym, and then situates the above-cited proposition directly into a material research perspective shared with the field of book history:

Pour qu’on se serve légitimement du mot « traduction » (translation, Übersetzung, traducción, translaciôn, etc.), dans la rigoureuse acception que lui aura conférée, depuis quelques siècles, une histoire longue et complexe dans un ensemble culturel donné (plus précisément, plus étroitement dans une Europe abrahamique et post-luthérienne), il faut que, hors de toute paraphrase, explication, explicitation, analyse, etc., la traduction soit quantitativement équivalente à l’original.

ibid., p. 22

Each of the text types expressed above—“translation, paraphrase, explication, analysis” (but this could also include things like summary, definition, encyclopedia entry, commentary, gloss, etc.)—is a form of mediated re-statement handed down historically with its own pre-determined economy, Derrida argues. This economy is bi-partite, consisting of metaphysical guidelines concerned with “propriété” or semantic recall, and then physical or material guidelines governed by quantitative proportion. Translation never gained distinction from other recognized modes of re-statement by virtue of its metaphysical attributes, simply because all forms of restatement trade in semantic recall. Rather, it is only by virtue of its formal distinction via inter-linguistic crossing and quantitative alignment—as well as its discursive distinction in displacing the first person—that the prototypical translation becomes knowable.

This is the material researcher’s and book historian’s dictum: the specimen’s physical form in culture (like a body in nature for the zoologist or botanist) matters, and it achieves definitional clarity only by contrast with other forms submitted to similar measure. This materialist epistemology—the operative mindset that “matter matters,” as Karin Littau (2016) puts it—informs Roger Chartier’s The Order of Books (1994 [1992]): text form determines codiciological form, which then determines the book’s construction. Once a standard form is established, Chartier argues, a chain of production develops to anticipate it, and a material culture then develops around this chain. Operating in deference to the chain, this culture gains sway in co-determining what type of textual body does or does not meet the standard. Chartier has only recently begun talking about translation specifically, but case studies that I have seen give me no reason to think that he is affording any surplus latitude to the concept.[13]

And yet unlike other types of restatement such as “summary” or “definition,” “translation” seems subject to re-definitions emerging in blatant disregard of the idea that it ever had a body in culture measured to a certain standard to begin with. Jakobson’s idea (1959), following Peirce (1931-1948), of translation as a kind of free-forming emanation driven by semiosis collides with a material culture trading in its own restrictive mathematical code exemplified here by Eco in the following case scenario:

I ask you to imagine (you are an editor), and have given a translator a printed manuscript in Italian, format A4, font Times Roman 12 point, 200 pages. If the translator brings you back, as an English equivalent of the source text, 400 pages in the same format, you are entitled to smell some form of misdemeanour. I believe one would be entitled to fire the translator before opening his or her product.

2003, p. 4

Note that not a single properly semantic justification for this sacking is put forward. Only numbers seem to be guiding our editor’s decision-making, specifically the countable measure of some critical proportionality between a source and target text’s respective material extensions through the line, page, or book.

5. Book history: translation, the copia verborum, and the medieval culture of copying

Book history has exercised enough pressure on the concept “translation” to help Pym (2004) acknowledge the principle of quantitative similarity explicitly and assertively. However, as I suggested earlier, he stands at a threshold that he does not cross. He elides something important, buries it in implicature: he offers no proposition on semantic similarity. This strikes me as odd in an article addressing the elementals of translation. Meanwhile, the far less popular idea of quantitative or formal similarity appears in no fewer than four elaborations. While I would like to thank Pym for speaking directly to the truth of a formal invariance orientation, I am beginning to wonder whether Translation Studies scholars, when addressing the thorny question of translation’s nature, feel compelled toward a kind of rhetorical gambit whereby the acknowledgement of one type of similarity entails a denial or elision of the other. All of our theorists have done this so far. My sense is that we have trouble putting these two facts of translation forward both together and explicitly. The moment we do, a translation begins to seem a lot like a copy, and the idea raises affect.

Book history puts this uncomfortable idea before Translation Studies all the time, but few Translation Studies scholars pick it up and pursue it. There is, for example, a historical work whose central claim invites Translation Studies to explore the connection between translating and copying in the context of medieval culture. This is Copeland’s Rhetoric, Hermeneutics, and Translation in the Middle Ages (1991). Copeland equates translation practices historically to a very specific rhetorical re-statement exercise bound by both semantic and quantitative similarity: the copia verborum, or copia of words (also copia of expression). Copia[14] is, of course, the Latin origin of the modern “copy” (also of “copious”), and it evolved out of the stylistic exercise originally coined by Quintilian as amplificatio. In the late classical period, the copia verborum—a term emerging at the beginning of Book X of Quintilian’s The Orator’s Education and translating into English as a “stockpile” and “capital” of words (2001 [circa 95 CE], pp. 254-255)—was a scholastic exercise that consisted of creating lists of stylistic variants for the purpose of building an inventory for use in discourse.

What is most important to note here is that the copia verborum (copia of words) has a corollary: the copia rerum (copia of ideas). The most obvious difference between the two is a perfect split between the quantitatively word-bound and the more free-forming nature of their respective currencies: the copia verborum traded in the physical extension of the word, phrase, or proposition and was restrained to a 1:1 quantitative ratio of expression and replacement expression. The copia rerum, on the other hand, was idea-centered, describing modes of variation in subject matter (the res). A potentially vast disproportion of verbal material was allowable between a statement of res and its variant re-statement. So their difference is not only in the type of “matter” they address, but also in the quantitative relationship between the matter of the statement and that of the variant restatement. In the copia verborum we see a situation where the source “word”—extending perhaps to a slightly larger fractional unit like the “phrase” or “clause”—sets very clear quantitative parameters within which the proposed variant was expected to fall.

Our best examples of copia do not come from the classical or medieval periods, however, when it was largely a scholastic exercise not considered worthy of greater attention. They come from the Early Modern period, when it was elevated to an object of contention among humanists, as Terence Cave argues in The Cornucopian Text (1979). On the one hand, there was an undeniable didactic functionality, and a kind of polyphonic beauty (Von Koppenfels, 2013), in the parallelism on display within the verbal stockpile. On the other hand, the pile itself could produce an effect of excess, a verbosity (verbum) swamping and displacing the idea (res). The copia, in other words, was “duplicitous,” to quote Cave’s expression.[15] Erasmus himself, in his school manual On Copia of Words and Ideas, was quick to articulate a disclaimer of sorts in a brief fourth chapter entitled “To whom unrestrained copia has been attributed as a fault” (2012 [1512], p. 14). In the copia verborum specifically, the most famous example is his 150 variations on the expression “your letter has delighted me very much” of which I will limit myself here to six:

How exceedingly your letter has delighted my spirit; what you wrote has given me incredible pleasure; the reading of your letter imbued my mind with singular joy; what you wrote was the keenest delight to me; you would scarcely believe how greatly I enjoy what you wrote; on receiving your letter, I was carried away with joy […].

ibid., p. 39

The quantitatively bound nature of this kind of variation is undeniable. An extension of textual material has been chosen (in this case a 7-word clause), and all of the variants hover around this same extension (between 8 and 11 words). There is also the sense that this type of restatement exercise is within a hair’s breadth of prototypical translation. A stylistic principle of variance within the bounds of quantitative and semantic similarity is evident. All that is needed is for variants to be in some other language.

Copeland sets the stage historically for the conceptual overlap of copying and translating. Translation, she argues in chapters 1 and 2 (1991, pp. 9-63), experienced a demotion in status as the classical world transitioned into the medieval one:

In the first chapter I consider how Roman disciplinary debates and practice created a space in which a rhetorical theory of translation could emerge; and the second chapter considers how that space could be redefined in the early Middle Ages by the force of new disciplinary directives. The rhetorical value of translation is lost in the very discourses that carry over Ciceronian theories of translation.

ibid., p. 6

In the early medieval period, then, there came a shift in the perception and experience of rhetoric. The latter—and specifically its subdomain elocutio, which encompasses discursive style, performance, and translation—lost its connection to the public forum and became cloistered in the monasteries in a pared-down version of its former self. Copeland refers specifically to the early 5th century treatment of translation by Consultus Fortunatianus in the Artes rhetoricae libri III (cir. 435). Here is where a decidedly medieval privilege afforded to inventional doctrine (inventio), along with an obvious demotion of discourse in performance (elocutio) becomes evident. A scant fraction of this third book, Copeland points out, is dedicated to elocutio, which is defined as “quantitate verborum et structurae qualtitate” (“verbal amplitude and stylistic decorum”) (1991, p. 40).[16] Fortunatianus then explains elocutio almost entirely through the lens of the copia verborum, and finally identifies inter-lingual translation (conversio) as an exercise in its achievement (ibid.).[17] Simply put, translation is deemed to be an exercise in copia verborum, and the latter becomes a restrictive conceptual funnel through which elocutio or style is channeled.

Due to its divorce from the dynamics of public performance, Copeland argues, classical rhetoric—and most of all the rhetoric of elocutio—went “textbook” and “mathematical” (privileging the quantitate verborum) in its medieval incarnation:

While rhetoric strengthens its affiliation with dialectic, its practical engagement with discourse, and hence with the ethics of human performances, has seemingly weakened, and it verges on becoming a textbook art. The divorce between invention and the rhetorical performance signals in effect a divorce between meaning and language, as rhetoric ceases to grapple with discourse as totality, from discovery to mediation to determination of meaning. Cut off from application to such issues, rhetoric’s orientation seems to shift almost to that of a fixed science such as mathematics or even grammar, from which it had so purposefully distinguished itself in the past.

1991, p. 40

The loss of performed discourse as a principle, then, had the effect of a kind of myopic, text-bound mathematization of style, which would begin fixing upon text-quantitative criteria over discourse-qualitative ones (verbal amplitude or quantitate verborum begins to trump stylistic decorum or structurae qualitate). Style, Copeland argues, became increasingly a “quantitative aim” (ibid., p. 41). So here we have the classical copia verborum (and translation as its signal exercise) impoverished, reduced toward (of course not “to”) a matching of material quantities with a diminished regard for the discursive-performative power of the result, and in many cases even for its intelligibility.

This impoverishment would see the copia in its classical sense fall and become the medieval “copy.” It would also potentialize the copy-effect in interlingual translation, which would become cryptically literal: less burdened by the requirements of qualitate structurae, and therefore by the need to know or understand language at the higher discursive and performative levels, translators became more inclined to build their versions from the bottom up, in an accumulation of inter-lingual material transactions at the lexical level (Mossop’s “lexical substitution” from earlier). This, of course, facilitated the production of target texts of striking Pythagorean symmetry with their sources, but which were all but unintelligible.

Littau has argued the need to look within the material working environments of historical media cultures as significant co-determiners of translation practices:

If material carriers (human body, tablet, roll, codex, book, computer) and their hardware (voice, clay, papyrus, parchment, paper, screen) make a difference to practices of writing and reading, as historians of the book have demonstrated, then, surely, the same carriers of hardwares also make a difference to practices of translation.

2011, p. 262

I said before that the countable words of a source text are the first and most important material given in the translator’s working environment. Another tool of medieval scribe culture was used to frame words physically and split them off from the flow of the text. Its use supports the idea that the copyist or translator built texts in transactional units from the bottom up. Stephen Greenblatt describes it: “A sheet with a cut-out window generally covered the page of the manuscript being copied, so that the monk had to focus on one line at a time” (2011, pp. 41-42). This cut-out window suggests a hard material limit forcing an intersection with a fractional unit of text. It suggests an approach where the eye, in the moment of execution, is not free to range over the text, gather information, and process it in a discursively viable way, but rather in the way of lexical substitution, upward from the measured segment.

6. The copia verborum of the digital age: The TM

This brief historical look at medieval copy orientation in translation is, of course, meant to cycle back to my bilingual reading of Misery. My final aim is to point out a compelling continuity between medieval translation practices and contemporary ones. And I would like to rejoin Pym here, who has recently argued that “many of the ideas and models most in tune with medieval translation are nevertheless reappearing in certain fields, in new guises, via deviously fashionable detours, and mostly without knowledge of their past” (2014, p. 1). The argument comes with the appropriate caveat that this kind of comparison is always by nature superficial, but is nevertheless a good thought experiment. Specifically, he proposes that the dynamics of textual innovation introduced by medieval commentators to mitigate the cryptic literalism of the period’s translations have a kind of superficial comparability with today’s processes of localization. Now I am not interested in localization here, but I am compelled by the major driver of innovation cited by Pym, namely the effort to mitigate excessive literalism: “The many literalist translations of authoritative sources, often only dimly understood, result in a Latin that must have sometimes been so opaque that the secondary adaptation discourses became necessary” (2014, p. 3). I will paraphrase the premise here, as I have read it: in the Middle Ages, a systemic cryptic literalism in the translation process created a need for innovative editing, for recovering unreadable translations after the fact—namely secondary adaptations such as scholia, commentary, glosses, and the like.

Now I would like to hold Pym here and suggest another compelling continuity between medieval and modern-day practices, which he does not mention because of his focus on localization, but which suggests itself when literalism is the driver of innovation: translators now in the digital age have also systematized a form of excessive literalism, and have then systematized innovative modes of “translation recovery” after the fact. In the modern-day context, I am referring to CAT (computer-assisted translation), which produces a new type of hyper-literalist translation typically requiring postediting. Now, of course, I am in full agreement with Pym that medieval and modern digital practices can only ever be compared superficially, but a comparison upon a single axis and toward a single point is all I need here: translators in the professional sphere both then and now worked and are working deliberately backwards from a systematized hyper-literalism, a fact that in itself suggests a gravitational force exercised by the copy on translation.

I will limit myself here to drawing an analogy between the copia verborum, a translation memory (henceforth a TM), and a “translation unit” (henceforth a TU) as it is constituted within the TM and retrieved during translation. And my point in doing so is to show that a TM uses rote transcriptive copying to unlock a copia verborum, out of which premeditatedly literal translation solutions are retrieved. In other words, in a TM rote transcriptive copying and translating work together. I want to make this point alone and hold it here—at the fact of a technical integration of true copying into the translation process. Until now copying and translating have been identified only conceptually and intuitively, but here is an integration that is concrete and material—and evidence to validate a copy effect resonating in the finished product of any translation produced in this way. Lynne Bowker’s and Michael Barlow’s (2008) work explaining the inner workings of TM to academics in other disciplines will be helpful here.

A TM is analogous to a copia verborum in that it is a stockpile of variant expressions of the same, which is available for translators to draw upon when they need an alternative way of saying the same thing (which of course is their profession). If we re-visit Erasmus, we see that there are some 200 variant expressions listed for the titular expression “your letter has delighted me very much” (2012 [1512], pp. 39-42). Each variant forms a pair with the titular expression and is understood as a stylistic variant of it: “Your letter has delighted me very much = How exceedingly your letter has delighted my spirit.” As I pointed out earlier, the variants tend not to range quantitatively beyond 4 words more or fewer than the titular expression. Now any single pairing of the titular expression with a variant could easily create what a TM would recognize as “translation unit” (TU), except of course in a TU, the titular expression would be a source language fragment, and the variant a target language one: “Your letter has delighted me very much = C’est avec grand plaisir que j’ai lu votre lettre.” So TUs are equations of roughly commensurate source and target language fragments set typically at the length of a sentence. And potentially millions upon millions of them are in fact stored in the TM as a copia verborum, a stockpile of formally commensurate variants waiting to be drawn upon.

TMs are not created in a vacuum, however. Human beings create them and direct them toward specific types of discourse. Keeping with Erasmus, we could think of a TM built specifically for epistolary discourse. Furthermore, human beings create TMs from previously accepted translations from the particular discourse in question. Bowker and Barlow (2008) supply the helpful analogy of re-cycling: potentially millions of acceptable translations are broken down into TUs and fed into the TM, in the same way that a car might be broken down for scraps to be stored as potentially useable parts for the construction of some future new car. The result is a copia with optimized potential for discursive recovery in postediting, provided users exercise judgment in using the right TM for its intended discourse. The idea is not to repurpose but to recycle: parts from cars should be used for building cars.

It is in the very particular material mode of translation retrieval from the TM that the process becomes interesting from the point of view of the copy effect: a translation solution is mediated into the target text via a transcriptive copy. When a source text is offered for translation, the TM first breaks it down into sentence-length fragments that are commensurate with its TUs. The source text is now in pieces that are held separate but still in their correct sequential order. What happens next is purely transactional:

The tool starts at the beginning of the new source text and automatically compares each segment to the contents of the TM database. If it finds a segment that it “remembers,” (i.e. a segment that matches one that has been previously translated and stored in the TM database), it retrieves the corresponding TU from the database and shows it to the translator, who can refer to this previous translation and adopt or modify it for use in the new translation.

Bowker and Barlow, 2008, p. 6

The quotation marks around “remembers” are appropriate here. The machine has no knowledge of semantics or equivalence, or of language for that matter. It reduces language to patterned strings among which it can detect identical copies between 0 to 100 percent. And so a source text fragment in L1 can never be matched in an entirely direct and unmediated with an L2 variant. To connect with an L2 variant, the source text fragment must first find its (ideally identical) transcriptive copy on the L1 side of the TU equation. The entered source-text fragment and the L1 component of the TU make a copy-based lock-and-key connection, and then the L2 variant is carried from the copia into the translator’s working environment on the back of this copy. An exact copy will yield an optimal L2 fragment.[18] Failing an exact match, imperfect or “fuzzy” matches will be recommended on a downward cline in value from 100 percent. Match recommendations come in L2 fragments that align with L1 fragments so that the translator can compare them and then accept or reject them. Each accepted match builds the target text piecemeal and paradigmatically.

And the result is hyper-literalism, but of a new type generated by the technology itself—not word-for-word but “sentence-for-sentence,” or even “segment-for-segment.” Bowker characterizes the result once again using re-purposed car parts as an analogy: parts from potentially hundreds of thousands of broken-down cars come together to build up a new vehicle that may have the doors from a Benz, the roof from a Corvette, and the steering wheel from a Toyota, etc. The opacities of this literalism arise from the collage effect of its assembled parts. However, the translations are recoverable with diligent postediting. More tellingly, the translation process is deemed to move forward more efficiently and expediently by passing through this kind of literalism rather than by avoiding it altogether.

So the association of copying with translating goes from largely conceptual in the classical and medieval context to fully realized and technologically functional in the TM context. With TMs, the copy is the place where the whole process begins, the first move of the machine. If I wanted to push the argument further, I could risk suggesting that modern-day translators using TMs are in fact starting from the ground zero of the copy, then proceeding to a hyper-literalist translation, and then finally arriving at a postedited text located at just far enough of a remove from ground 0 to be considered discursively viable. What is clear is that if they are not actually copying, they are hovering very close to it, certainly close enough to instantiate its effect.

If I were asked about what kind of experience this study has been, I would answer that it has been like walking into virgin theoretical territory to talk about one of the most obvious effects of translation imaginable. “Virgin territory” is perhaps not quite right, because I have found in certain classical concepts a path through the question. However, I have had to fashion from these concepts my own epistemological moorings: I have had to (1) approach words and texts from a strictly material rather than linguistic perspective; (2) distinguish sharply between two orders of textual “form”—a qualitative stylistic one familiar to translation theorists and an unspoken but omnipresent Pythagorean one ruled by numbers; (3) define translation materially, which means arguing that the material strictures placed upon it by historical practices matter and have carried through to the present day.

The copy effect in translation—and the question of inter-textual identity in translation more generally—demands a more thoroughgoing theoretical elaboration, one preferably emerging out of attentiveness to our basic cognition of translation’s textual materiality in activities like bilingual reading. These activities pull us back toward inter-textual sameness and allow us to question its nature. And if we are too much in the habit of calling translations different from their source texts, they help us back into the middle space of the paradox where this study began. I am reminded here of Jorge Luis Borges’s Menard: Autor del Quijote (1974), which is essentially a tale about the loss of translation itself through failure to join in its paradox. Early 20th century French symbolist poet Pierre Menard attempts the nec plus ultra of Don Quixote translations. After spending many years obsessing over the historical, cultural and linguistic differences separating him from Cervantes’s Renaissance Spain, he manages a result that would be short work for any schoolchild: a straight copy, letter-for-letter, of the novel’s first two sentences. The absurdity of the result is, of course, in the vacillation between two extremes: from a struggle with differences perceived as insurmountable comes a result of laughable sameness. And “translation,” which cannot find a home at either of these extremes, has slipped unnoticed through grasping fingers.