Translate: carry across

The metaphor implicit in translation is that of something being carried over from one “place” to another. The most fortunate, the most deserving of saints might be translated into heaven; their earthly remains were often translated from one church or monastery to another, along with the prestige of possessing them. Translate in this older sense bears a clear relation to transfer, of which it is merely the irregular past participle.

In those cases a thing was carried over or across: a person, a relic. In the now most common case, that of translation between languages, what’s carried across is not at all obvious. Decades ago, when theories of meaning—or rather speculation about theories of meaning—were all the rage, one would have said that translation consists in the construction and use of a systematic mapping of the sentences of one language to those of another, a mapping that preserves meaning or truth. The first attempts at computer translation worked from a similar definition.

Source: John Macnamara, Journal of Social Issues; 23.2 (1967) 59.

Translation so conceived “carries across” only abstracta: the meaning or truth-value of the source. For anyone who has taken seriously the task of translation, that is a sort of caricature. The truth in it—what makes it caricature and not outright falsehood—is that something can be captured by the algorithms employed by Google and other automatic translation services. Call it the “gist”.

Capturing the gist will suffice on many occasions. Here’s an example:

L’emploi par Poincaré de la notion de convention au sujet des hypothèses géométriques signale un déplacement par rapport aux problématiques traditionnelles. La découverte des géométries non euclidiennes montre qu’il n’y a pas de cadre spatial unique ; plusieurs systèmes sont possibles. On affirme ainsi l’existence d’un aspect essentiel de la connaissance qui ne dérive pas des faits et ne relève ni de l’inné ni de l’intuition (Anastasios Brenner, “Géométrie et genèse de l’espace selon Poincaré”, Philosophiques 31 (2004) 115–130.

As translated by Google:

Employment by Poincaré in the notion of convention on the geometric assumptions indicates a shift from traditional to the problems. The discovery of non-Euclidean geometries shows that no single spatial framework, several systems are possible. It thus affirms the existence of an essential aspect of knowledge that is not derived from facts and is neither of the innate or intuition.

Quite bad, really. But were you scanning abstracts to find articles relevant to your research, you could glean enough from this to make a decision. That is testimony not so much to Google’s success in translation as to our capacity for extracting sense from degraded signals. (There are much more adequate translation programs for specialized fields that are based on the work of human translators who in contributing to those programs contribute to their own obsolescence.)

The threshold, then, set by the requirement of capturing some part of the gist, needn’t be high. What more does one ask for? What makes certain translations so difficult? Eric Schliesser raised earlier some pertinent questions about translation practice. Why, for example, should the translator endeavor to translate words or phrases by items of roughly the same dimensions? Why not incorporate glosses into the translation itself rather than relegate them to footnotes? What is a gloss, and how does it differ from a straightforward translation?

I don’t pretend to great originality in what follows, or to great rigor. What follows are my thoughts as a translator. The first and most general point is that one is always working within a set of desiderata, distinct, sometimes in harmony, sometimes not, for which there is no reason to believe that they can all be maximally fulfilled at once. The translator knows practically that what he or she values is “multiple, conflicting, and incommensurable”, as Isaiah Berlin said of human goods in general. What is surprising, perhaps, is that sometimes one does hit on a solution that does strike one as the aptest in that circumstance.

Needless to say, questions of this sort haven’t gone unstudied. Walter Benjamin, Jorge Luis Borges, Valéry Larbaud are among the better-known author-translators to have reflected on their craft. There is now a discipline called “translation studies” with its own journals (Target, Meta, Intralinea) and anthologies (Riccardi, Translation studies, Venuti, The translation studies reader). Similar questions arise for the ethnographer and the historian.

It’s best to anchor the discussion to examples. I’ll start with a poem of Hugo which I translated last year.

She’s gone—absent, vanished—too bad!
He may well say: It’s nothing, I’m not sad,
I’ll go to balls, I’ll feast, I’ll stay out late…
He may well spend his days intoxicate,
Play the hero’s, the cynic’s part,
And yet—there stays within his heart
A memory, adrift above the permanent
Shadow, like a torn-off cable’s end.

Elle partie, absente, évanouie, hélas!
Il eut beau dire: bah! c’est bien. J’en étais las.
Je vais aller aux bals, aux fêtes. Je vais vivre.
Il eut beau savourer la coupe où l’on enivre,
Et faire le vaillant, et faire le moqueur,
Il sentit qu’à jamais il lui restait au cœur
Un souvenir flottant sur l’ombre irrévocable,
Comme un arrachement qui laisse un bout de câble.

— Victor Hugo, in Océan (Paris: Robert Laffont, 1989) 359
(fragment ms 13 422, f76, dated 1860–1861)

The desiderata here were to maintain the rhyme (but not the meter, only an approximation of it), the sense, the ethos. The hardest part was the last two lines with their image of the suspended torn cable-end. The rhyme (or near-rhyme) did not come easily, but I wanted to preserve the sting, the abruptness with which Hugo closes off the octet. The result preserves the shape and dimensions of the original. Had I incorporated as a gloss-in-translation a dissertation on 19th-century sexual mores, I would have destroyed the poem, however much that gloss would have helped the reader to understand why the “he” in the poem would want to appear to have forgotten his lover even while being unable to do so. On the other hand, that he wants to do so should be apparent, or I have failed.

It is customary to distinguish, in talking about literary translation, content from form—and equally customary to hold that they cannot be separated. As regards form, at minimum something like the shape, as I’ve said, is to carried over. A haiku ought not to be translated even into a sonnet, let alone into forty lines of blank verse. Is this mere prejudice? I don’t think so. The point of haiku is brevity. Lose that, and you’ve lost the poem. What you would have might be good poetry or good prose in its own right, but it would have failed, in one important respect, as a translation.

I distinguish translation proper from commentary. The translation proper presents itself as (the gist or substance of) what the author of the original said; the translator functions as an instrument, a proxy for the author. If I translate a text full of curses I do not myself curse, if I translate a text full of falsehoods I do not myself lie.

Or rather I should say: not in the first instance, not primò as an Aristotelian would say. If I do curse or lie, it is by means of the text I have translated, in the same way that if I quote your utterance of a four-letter word I am not myself uttering it, but I can, slyly, quote your utterance in order to “say” what I could not say in propria persona.

In straightforward instances, the translated text is an utterance, a product of its author, not its translator. The diminished responsibility of the translator for what is said in the translated text carries with it diminished credit, so that in the limit the translator may be elided, and text presented as if original (such is the case with treaties and the like).

Source: Virgil, Publius Virgilius Maro varietate lectionis et perpetua adnotatione ed. C. G. Heyne, G. P. Eberard Wagner (Leipzig, 4th ed.,1830). Virgil’s poem occupies the first two lines of the page. The rest is commentary, first on the text, and then on its content.

The commentary even if it purports to put forward “what the author really meant”, is the product of the commentator, who is responsible for the assertions made therein. Because it does not go proxy for the author of the original, it need not take account of the desiderata of translation. It need not preserve the shape of the original, nor its ethos. One speaks of free translation but not of free commentary. The commentary makes available to the reader of a text (not necessarily in translation) what the commentator thinks the reader needs to know or be acquainted with in order to understand the text. That is a task distinct from providing the reader with a version of the text in the reader’s own language. Above we see Heyne’s Latin commentary on Virgil. His reader did not need to descend into the vernacular, but did need, as we do, a great deal of information to understand the poem well.

When I translate philosophy, I have many of the same desiderata in mind. The shape of the text matters. Wittgenstein’s Tractatus would hardly make sense were it translated into continuous prose. Lucretius’s De rerum natura presents, to my mind, a more difficult decision. It would take a translator of the calibre of Pope to pull off the feat of putting the poem into blank verse (our equivalent of Lucretius’s hexameter); a lesser light may content himself, as the Loeb translator did, with prose.

The lines that follow are quoted from De rerum 1:73–75 (with slight errors) in Douglas Bridges’ memorial to Errett Bishop (From sets and types to topology and analysis, Oxford, 2005, ix):

Ergo vivida vis animi pervicit, & extra
Processit longè flammantia mœnia mundi,
Atque omne immensum peragravit mente animoque […]

Bridges then quotes an uncredited translation from the Oxford Book of Quotations:

The vital strength of his spirit won through, and he made his way far outside the flaming walls of the world, and ranged over the measureless whole, in both mind and spirit.

By way of contrast, here is a verse translation in couplets (from W. Francis H. King, Classical and foreign quotations, London, J. Whitaker & Sons, 3rd ed., 1904, p273):

Thus, the keen vigour of his mind prevailed
And the bright bastions of the world outsailed.
His reason and his soul’s intelligence
Swept the whole area of the void immense.

One sees from the prose translation that the verse of King had to take some liberties to achieve its rhymes. The walls of the world become bastions, going forth (processit) becomes outsailing. On the other hand, King manages to keep not only the verse form but some of the feel of Lucretius’ lines—the alliteration of “vivida vis” is lost (“keen vigour”) but “bright bastions” makes up for it somewhat while echoing also “mœnia mundi”. In English one would be hard pressed to maintain the intricacy of Lucretius’ repetitions while conveying the sense; the effect is comparable to that of Shakespeare at his most intense.

But, you might say, surely the philosophy can be carried across without all this fuss. What have philosophical concepts to do with alliteration or rhyme? Not much, perhaps: but because we don’t know, the underlying principle, that of preserving structure, ought to be observed by the translator of philosophy too.

For example, many philosophers rely on control of vocabulary to cut down on ambiguity, and to keep the reader in mind of stipulations, distinctions, and so forth. In some prose works, it may be reasonable to vary the translation of a word; after all, a word may vary in sense from one context to another, and those various senses need not, and perhaps cannot, be conveyed by the same word in the target language. Nevertheless in the work of a philosopher who is careful with their lexicon, the translation of a given term should not vary without good reason. In Suárez’s Metaphysical disputations, virtually every word is a technical term—a “reserved word” as the jargon of programming would call it. Gratuitous variation, or even motivated variation, would obscure the structure of his thought.

At one extreme in the preservation of structure would be William of Moerbeke’s translations of Aristotle and some of his ancient commentators.

τὸ ὂν λέγεται πολλαχῶς, καθάπερ διειλόμεθα πρότερον ἐν τοῖς περὶ τοῦ ποσαχῶς: σημαίνει γὰρ τὸ μὲν τί ἐστι καὶ τόδε τι, τὸ δὲ ποιὸν ἢ ποσὸν ἢ τῶν ἄλλων ἕκαστον τῶν οὕτω κατηγορουμένων. τοσαυταχῶς δὲ λεγομένου τοῦ ὄντος φανερὸν ὅτι τούτων πρῶτον ὂν τὸ τί ἐστιν, ὅπερ σημαίνει τὴν οὐσίαν (Metaphysica 7c1, 1028a, ed. Christ, Teubner)

Ens dicitur multipliciter, sicut prius divisimus in hiis quae de quotiens. Significat enim hoc quidem quid est et hoc aliquid, illud vero quod quale aut quantum aut aliorum unumquodque sic predicatorum. Totiens autem dicto palam quia horum primum ens quod quid est, quod significat substantiam (C. E. D. Buckner, The Logic Museum, ad loc. Moerbeke, Meta 7c1).

Even without knowing Latin or Greek it is fairly easy to see what Moerbeke’s principle is: for each Greek word a Latin; the Greek word order is preserved as much as it can be subject to basic features of Latin grammar.

For example: Aristotle can put the definite article τὸ in front of ὂν “being”, but Latin has no definite article, so Moerbeke does not try to match Aristotle’s phrase exactly (nor would one say “the being is said in many ways” in English; but in French one would translate τὸ ὂν quite naturally as “l’être”. —Moerbeke does occasionally vary the translation of a word: the Greek δέ, which in English could be translated clumsily as “on the other hand”, he translates once as vero, once as autem.

The result is a Latin that Renaissance critics, not without reason, thought barbarous. Here’s a flat-footed translation into English of a bit of Moerbeke’s Latin: “Being is said in many ways; so first we divide [it] into these according to how many. For it signifies this, namely, what is and this something, or again what sort or how much or each in this way of the other predicables”. Moerbeke’s barbarisms were replaced by translations into good Ciceronian Latin (see Paul Botley, Latin translation in the Renaissance, Cambridge, 2004).

Moerbeke’s procedure looks like the simplest, most naïve possible: word for word, the same each time, preserve word order. You might think it “primitive”, but I think perhaps it isn’t. What Moerbeke gives his Latin reader is a translation that preserves structure as much as possible: not only the “distributional” structure à la Zellig Harris (“Discourse analysis”, Language 28 (1952) 1–30), but also the sequential, hence the syntactic, structure. The result stands to the original in roughly the relation that a photograph in which every color has been systematically replaced with another stands to the original: with a judicious choice of replacements, the reader can, so to speak, still see the shape of the original in the translation.

The result will annoy the readers used to the suavity of Cicero, but it can be understood. Thomas used Moerbeke’s translations while writing his great commentaries on Aristotle. We now tend to prefer translations that accommodate themselves more gracefully to the expectations of readers (though opinions continue to differ: witness the controversy between Nabokov and Edmund Wilson on Nabokov’s translation of Pushkin). The virtue of doing so is that it yields an approachable text; but it also intrudes into the relation of text to reader an element, deriving from the translator’s sense of his or her readers’ expectations, which is independent from the source. Moerbeke is in that respect the most self-effacing of translators: he has no style of his own except that which is reflected in the choice of correspondences between Greek and Latin lexical items.

The desiderata remain. A controlled vocabulary in the original calls for a controlled vocabulary in the translation. A careful writer marks distinctions by the various terms he or she uses; and if the translator without compelling reason translates the same term differently, that will introduce into the translated texts expectations of distinctions where there were none in the original. The same of course goes for translating different words into the same, only then distinctions are not being fictitiously introduced but rather fictitiously erased.

The translator ought, moreover, to preserve the connotations, the resonances of an author’s vocabulary. Texts are written from within some sort of tradition or perhaps a mélange (and translations are made from within a tradition too, of course). Even a dedicated re-fashioner of words like Kant does so with their prior uses in mind. In Descartes, for example, there are many harkings-back to Scholastic terminology, most evidently in the Principles, which were intended to be a textbook, but also in the Meditations. Especially if he uses a technical term like denominatio, it is not helping but rather impeding the reader’s understanding to an up-to-date, reader-friendly substitute. The point of using the word is in part to recall to the reader its prior usage (and perhaps then to revise it, but not simply to ignore it). The same goes for Newton’s De gravitatione and for Leibniz’s self-conscious revivals of Scholastic usage.

See for example, the interesting discussion of the gender-marking of translation in the eighteenth century, as described in Mirella Agorni, “A Marginal(ized) Perspective on Translation History: Women and Translation in the Eighteenth Century”, Meta : journal des traducteurs / Meta: Translators’ Journal 50.3 (août 2005) 817–830 [http://id.erudit.org/iderudit/011598ar]. In general, translation from classical langauges was a male prerogative; women, by and large, translated from (and to) modern languages; Mme Dacier was thus able to make her mark as a translator of Homer. On the context of translation, see Kwame Anthony Appiah, “Thick Translation,” Callaloo 16:4 (1993) 808–19.

One last observation. Only speakers in a relatively monolingual environment are afforded the luxury of thinking that translation occurs rarely, that it is an optional activity. In the more usual situation, translation is frequent and requisite to people’s daily existence or working lives.

Kant’s texts are peppered with Latin expressions. They function in several ways: some are carried over into German, printed in the Fraktur face, inflected as German, and are thus fully domesticated. Kant at one point in the Critique of pure reason remarks that the word ‘absolut’, which is borrowed, of course, from the Latin, has no equivalent (B381), so that its having been borrowed was (I presume) out of necessity.

Other Latin expressions retain their Latinity but are integrated into the German (e.g.,, “a parte priori”, at B389). Still others are inserted into the texts as glosses of Kant’s German terms—as if the text had been, or were going to be, translated into Latin (which was the language in which dissertations and other scholarly works continued to be written, and a still-living presence for educated readers). The table of the four flavors of Nothing, one of my favorite pages in the Critique (B348), includes with each of the four German names a Latin phrase: “Leerer Begriff ohne Gegenstand”, for example, is glossed as “ens rationis”.

The Kritik is “in” German, but the German it is in is itself environed by Latin and by French, the French in which Frederick the Great wrote the lines quoted by Kant by way of compliment in the Critique of judgment (§49, Akademie-Ausgabe 5:316). One might say that it is written in the German dialect of the virtual language of the Republic of Letters.

The task of translating it into the contemporary English dialect is thereby made both simpler and more difficult. Simpler because the translator can take for granted, even now, the grounding of that language in the classics and in a common philosophical vocabulary. More difficult because the text, in presupposing that Republic in its late eighteenth-century form, requires that the translator—I should say the ideal translator—should carry over enough of its culture as to make the words and sentences of her translation comprehensible to readers whom she can no longer assume to be its citizens.