Wednesday Book Club: Le Ton beau de Marot

Wednesday, 24 September 2008 — 10:01pm | Book Club, Computing, Literary theory, Literature, Science

This week’s selection: Le Ton beau de Marot: In Praise of the Music of Language (1997) by Douglas R. Hofstadter.

In brief: What begins as a comprehensive study of poetic translation evolves into a treatment of human empathy and intercultural understanding, a refutation of John Searle’s Chinese Room argument against artificial intelligence, and a solemn remembrance of the author’s deceased wife. With its exclusive focus on language, Le Ton beau is a substantially less technical and more streamlined tome beau than Gödel, Escher, Bach; the mathematically averse may find it a more accessible point of entry to Hofstadter’s thought, as there is no talk of recursion or formal incompleteness in sight. Those who prefer their poetry devoid of metre and rhyme will take issue with Hofstadter’s conservative aesthetics; those who prize pattern, structure, and wordplay will rejoice.

(The Wednesday Book Club is an ongoing initiative of mine to write a book review every week. I invite you to peruse the index. For more on Le Ton beau de Marot, keep reading below.)


At over 600 pages, Le Ton beau de Marot is as thorough a case as one is likely to find for the argument at its nucleus: that poetic form is an integral conveyor of poetic meaning. Its backbone is Clément Marot’s cigar-shaped poem “A une Damoyselle malade”, 24 lines and 84 syllables of 16th-century Old French to which Hofstadter refers by the first line, “Ma mignonne”. Over the course of the book, Hofstadter presents over 70 different translations and permutations of “Ma mignonne”, from his own multifarious efforts to those of his colleagues, translators, and students. The translations vary in tone, idiom, semantic liberties, and respect for formal constraints: late in the book, we encounter everything from second-order translations (e.g. French-Italian translations juxtaposed with English explications of the Italian text) to the strained attempts of computer programs designed to translate technical documents.

All of this is a basis for a discussion of just about anything Hofstadter can relate to translation. Among the issues in play:

  • To what extent should translators sacrifice fidelity to an original text’s meaning to preserve its formal features—metre, rhyme, puns, lipogrammatic omissions of the letter E—or vice versa?
  • When do we preserve the idioms and cultural connotations of a work’s original language, and when do we opt instead for the transposition (“transculturation”) of those occurrences into the idioms of the destination language?
  • Why do performers of classical music often get top billing over composers, yet translators’ names hide in obscurity under the original authors? How much credit for authorship do translators deserve?
  • How tightly bound to its language must a text be to be truly untranslatable?

This should be different but not unfamiliar territory to anyone familiar with Hofstadter’s earlier works—chiefly, Gödel, Escher, Bach: An Eternal Golden Braid and the compilation of Scientific American columns published as Metamagical Themas: Questing for the Essence of Mind and Pattern. Le Ton beau de Marot is a much wordier tome, as Hofstadter sidesteps the lectures on formal logic and recourse to graphical examples that characterize the other books and specifically writes for a verbally oriented audience. Nevertheless, his predominant interests remain intact: the interplay of form and content, paradoxes of self-reference, and how human cognition grasps the complexities that emerge from interlocking layers of meaning. And, as is characteristic for someone with a deep interest in self-reference, topical puns and anagrams are everywhere to be found.

Le Ton beau de Marot has a very autobiographical bent: there is a story behind every major object of study, and Hofstadter does not shy away from telling us how he first discovered them and imparting his personal aesthetic judgments of their elegance. This appeals to readers like myself who already have a good sense of Hofstadter’s interests from his earlier works and are eager to find out more about the polymath behind the pages, though others reading him for the first time may find him too digressive. But by the end of the bittersweet conclusion, anyone can understand why language means so much to him personally: he maps the marriage of form and content, medium and message, to the joyous marriage that he had himself before it was cut short by his wife’s untimely death.

The strength of the book is the bevy of examples and case studies that fill every chapter. Hofstadter evaluates translations of texts ranging from classics of world literature like Pushkin’s Eugene Onegin and Dante’s Inferno to texts that are defined largely by their formal wordplay—Georges Perec’s E-less La disparition, Hofstadter’s own GEB (beautifully rendered in Mandarin as Jí Yì Bì, “Collection of Exotic Jade”), and sentences like this Dutch specimen:

Dit pangram bevat vijf a’s, twee b’s, twee c’s, drie d’s, zesenveertig e’s, vijf f’s, vier g’s, twee h’s, vijftien i’s, vier j’s, een k, twee l’s, twee m’s, zeventien n’s, een o, twee p’s, een q, zeven r’s, vierentwintig s’s, zestien t’s, een u, elf v’s, acht w’s, een x, een y, en zes z’s.

To my delight, Hofstadter recounts his own attempt to devise a Polish-English translation of the first story in Stanislaw Lem’s Cyberiad, the one about the machine that can create anything starting with the letter n. He compares it to the standard English translation by Michael Kandel (who contributes his own rendering of “Ma mignonne” to the book) to illustrate different solutions to the problem of preserving, among other things, a joke about natrium (sodium), a satirical passage about nauka (science), and the crux of the story, when the machine is commanded to create Nothing.

I was surprised that Hofstadter did not mention one of the later stories in The Cyberiad (and my personal favourite), “Trurl’s Electronic Bard”. The story tells of a poetry-writing machine that spits out alliterative poems about haircuts and romantic pastorals in the language of tensor algebra, and is in some respects a microcosm for the entirety of Le Ton beau de Marot in a fraction of the space. It covers everything: man/machine hostilities, the avant-gardist rejection of crossword-puzzle poetry composed within tight formal constraints, the construction of poetic language upon the edifice of an entire cultural history—it’s all there. As Hofstadter has obviously read The Cyberiad, I am shocked he never mentions “Trurl’s Electronic Bard” once.


My regular readers may recall that about a year ago, I wrote a post about The Cyberiad that lamented the decline of poetic forms that have yet to be exhausted; Hofstadter dedicates an entire chapter to the topic, sparing no harsh words for modern translations of Dante’s Inferno that cast aside its delicate terza rima architecture as if it were a dispensable accessory to the poem’s meaning. “Should [Dante translator Robert Pinsky] someday wind up in one of the nine circles of Hell,” he writes, “I know just the punishment that matches his sins committed on earth: He should be condemned to construct, unto eternity, one tercet after another in perfect, non-slanted English rhymes.”

Central to Hofstadter’s aesthetics is a belief that art is pattern; his personal hero is Frédéric Chopin. In one of Le Ton beau‘s most controversial passages, entitled “The Trendy Pooh-poohing of Pattern”, he expresses his bafflement at how the anti-structural movements of the twentieth century have taken over the tradition of high art:

The collective message radiated by a large clique in the club of today’s poets is that the devices of rhyme and rhythm and such things were nice back in the quaint old days, but in our infinitely more serious and sophisticated day and age, it would be horribly kitschy to resort to such easy, fluffy, childish sound-games.

Over our century the esoteric and obscurantist tendency grew and grew, till it became the norm. As this happened, of course, rhyming and rhythmicality grew tainted with suspicion: to indulge in such things was the signature of doggerel, of light verse, of Ogden Nash-style frivolity. It thus became de rigueur to avoid catchy sonorous patterns, so as to avoid being labeled “light”, “popular”, “accessible”. What a kiss of death! Whereas poets and composers in previous centuries didn’t make being inaccessible a high goal, in our century failing to achieve wide appeal has become a mark of elitism and success.

It is important to note that this is not a populist argument: Hofstadter reserves an equal amount of disdain for modern popular music (rock in particular), which he sees as a far and vulgar cry from the intricate lyrical stylings of W.S. Gilbert, Cole Porter, and Ira Gershwin. What Hofstadter values in his art is complexity and sublimity of formal architecture, in no small part because he conceives of human intelligence in general as the ability to detect patterns and create analogies.

Readers familiar with Gödel, Escher, Bach will remember the passages in which Hofstadter criticizes the works of John Cage (“Composition of Aleatorically Generated Elements”) as lacking in intrinsic meaning and structural qualities, as opposed to pieces by J.S. Bach (“Beautiful Aperoidic Crystal of Harmony”). Cage’s compositions may possess a lot of extrinsic, phenotypic meaning in the academic conversation over the history of Western music, but as genotypic structures they do not stand alone apart from external explanation; for Hofstadter, art should stand on its own merits.

The irony, however, is that most of the translations of “Ma mignonne” that Hofstadter puts on display in Le Ton beau in Marot are incredibly context-dependent. It is easy to appreciate their cleverness if you regard them as solutions to a puzzle, but without reference to the Marot poem and some of its more esoteric formal constraints (among them, a requirement that the poet name himself in the text), most of them are but lightweight trivialities. If you were to set the translations apart from the original poem, they would hardly be worthy of consideration. Hofstadter concedes as much: he acknowledges that his playful style comes off as juvenile next to some of the profound submissions by other translators, which he also includes in the book. As he admits in the chapter on untranslatability, there is a point at which the imposition of constraints comes at the expense of meaningful standalone coherence.


The turf war over the centrality of structure is not a new debate in poetics by any means. If we wind the clock back to 1668, we can observe John Milton’s prefatory comments on the verse form of Paradise Lost:

The Measure is English Heroic Verse without Rhyme, as that of Homer in Greek, and of Virgil in Latin; Rhyme being no necessary Adjunct or true Ornament of Poem or good Verse, in longer Works especially, but the Invention of a barbarous Age, to set off wretched matter and lame Meter; grac’t indeed since by the use of some famous modern Poets, carried away by Custom, but much to their own vexation, hindrance, and constraint to express many things otherwise, and for the most part worse than else they would have exprest them.

Surely, even Hofstadter would agree that the prosody in Paradise Lost is beyond reproach (in good measure because of its strict metricality), though he would probably prefer that it rhymed. As a connoisseur of structurally elegant art that speaks for itself (as we would expect him to be, coming as he does from a background in classical music, where there is no content without form), Hofstadter does not appear to put great stock in literary history. Linguistic history, yes—he has a keen ear for the associative effects produced by the dialects of Shakespearean English and inner-city rap English alike—but as for the evolution of poetic constraints and verse forms themselves, not so much.

I would argue that, for better or for worse, the erosion and eventual collapse of metre and rhyme came about as a result of conventional verse forms accumulating so much cultural baggage that a descent into self-parody was all but unavoidable. We can see this happening to the English heroic couplet as early as 1714, in Alexander Pope’s The Rape of the Lock. Mind you, this did not preclude Romantics such as Coleridge and Byron from continuing the tradition of rhymed verse with a straight face a century later.

The fault that Hofstadter finds in the avant-gardist abandonment of past forms as passé is the notion that constrained compositional spaces have been fully explored. (I’ve addressed this quite frequently myself: see my post on The Cyberiad and my review of jazz critic Stanley Crouch’s Considering Genius.)

My personal position on anti-structural art is conflicted: I, too, am none too pleased that beauty and elegance have yielded to social transgression as the most cherished value of critical discourse. On the other hand, I firmly believe that modern and postmodern works of art, music, and verbal composition are at least worthy of objective analysis that takes them apart, explains how they work, and sheds light on why they command the tremendous academic capital they presently do.

Conversely, no matter what your opinion is on modern music and poetry, the public rejection of it is also too systematic to ignore or attribute to the ignorance of the masses; it should be open to investigation in the field of cognitive science. Contrary to the apologists for obscurantism in the humanities and social sciences, I do not believe that Hofstadter is being merely dismissive of something he doesn’t “get”.


We should also remember that Hofstadter’s rejection of anti-structural art only becomes prescriptive—a dictation of how others should behave—in the context of translation. His dictum is clear: write your own poetry if you must dispense with form, but if you are translating a carefully structured work, treat the medium of expression as part and parcel with the message (as opposed to a dispensable container) or risk losing the distinctive flavour of the source text entirely.

In this debate in translation theory, the individual Hofstadter identifies as his chief opponent is Vladimir Nabokov, whose idea of a translation of Eugene Onegin (originally composed in Onegin stanzas) was a prose explication of the Russian text accompanied by 1,000 pages of notes. A handy reference for scholars of Russian, perhaps—but not, in Hofstadter’s view, a translation: at least, not a translation presentable to an English-language readership that wants to savour the lilting qualities of the Pushkin sonnet, the very property that distinguished the work for its original Russian audience.

One of the unexpected discoveries buried amidst Hofstadter’s excoriation of Nabokov (and, specifically, of Nabokov’s derision towards other Onegin translators who deigned to attend to form) was the light it shed on why Nabokov’s novel Ada, or Ardor bothered me as much as it did. (You may recall from a fortnight ago that Ada was a book I desperately wanted to love, but couldn’t.) The culprit, directly or indirectly, was Nabokov’s attitude towards translatability, which no doubt informed Ada‘s excessive untranslated trilingualism as well as its overwhelming allusive dependence on antecedent literature—the antithesis of the standalone ideals that Hofstadter promotes.

(The latter impediment was the same problem I had with Umberto Eco’s Foucault’s Pendulum when I read it years ago. In one of Le Ton beau de Marot‘s most amusing moments, Hofstadter points out how the name Umberto Eco could be anglicized as “Humbert Echo”—in other words, Humbert Humbert, the protagonist of Nabokov’s Lolita! It would be easy to regard this as one of Nabokov’s characteristic allusions were it not for the fact that Lolita was published well before Eco produced anything of note.)


Readers with an keen interest in artificial intelligence, as I do, will find an additional delight in Le Ton beau de Marot: the manner in which Hofstadter parlays issues of translation into a refutation of John Searle’s Chinese Room argument. For those of you new to the ongoing battle between proponents of artificial intelligence and the philosophers who insist computers can never really think, I’ll provide the briefest summary I can.

The Chinese Room argument runs thus: a computer program that demonstrates complete linguistic facility indistinguishable from humans is still, under the hood, nothing more than a syntactic symbol-pusher. It no more “understands” a language like Chinese than a non-Chinese speaker would if placed in a room to mechanically produce a translation based on dictionaries and set procedures. Searle argues that syntactic manipulation can never give rise to semantic comprehension; ergo, computers are inherently unable to think “about” anything in the way humans do.

The pertinence of translation to the Chinese Room argument is obvious. Hofstadter, a proponent of artificial intelligence and abstract models of consciousness (albeit one who recognizes AI’s current practical limitations), initially defers to the Systems Reply: that it is fallacious to identify with the symbol-pushing human in the Chinese Room when he or she is merely one infinitesimal cog of a complex system that, taken as a whole, is intelligent.

What Hofstadter’s study of translation adds to the discussion is its dismantling of Searle’s clean division between syntax and semantics. As Hofstadter shows, any machine of sufficient linguistic aptitude to pass for a native human speaker (as per the Turing Test) must account for semantics. If it plays with syntactic patterns alone at the naïve, grammatical level, it would fail to pass for an entity of humanlike intelligence in the first place. To demonstrate this, Hofstadter turns the tables on the Chinese Room, and compares a machine translation of “Ma mignonne” (needless to say, with complete disregard for the formal qualities of the poem) with a translation performed by a Chinese speaker with little to no knowledge of French or English, armed with a stack of dictionaries and rule books.

This isn’t to say that Hofstadter has unrestrained enthusiasm for machine translation: his refutation of Searle is a defence of the validity of the Turing Test, but in no way a commendation of MT research’s present direction. Hofstadter criticizes efforts in MT for their exclusive dedication to industrial applications, which has led to odd stylistic guidelines like Xerox’s directive that its technical writers compose in a specific manner designed to avoid tripping up the translation engine. For Hofstadter, this privilege of denotative meaning at the expense of higher-level structures exhibits a lack of respect for the subtleties of natural language, and precludes MT from reaching a stage where the Turing Test is remotely within reach. As the hype-driven enterprise of automated translation reaches out to wider problem domains without pause for refinement in any of them, we gain as little insight into human cognition as we did when Deep Blue defeated Garry Kasparov with brute-force search.


Seeing as how Le Ton beau de Marot was published in 1997, long before the advent of Google Translate’s large-scale implementation of a statistical methodology, I thought I’d plug “Ma mignonne” into Google Translate to get a sense of our present state of affairs.

Here is Clément Marot’s original:

Ma mignonne,
Je vous donne
Le bon jour ;
Le séjour
C’est prison.
Puis ouvrez
Votre porte
Et qu’on sorte
Car Clément
Le vous mande.
Va, friande
De ta bouche,
Qui se couche
En danger
Pour manger
Confitures ;
Si tu dures
Trop malade,
Couleur fade
Tu prendras,
Et perdras
Dieu te doint
Santé bonne,
Ma mignonne.

And here is “Ma mignonne” according to Google Translate:

My cute,
I give you
The good days;
The stay
It is prison.
Then open
Your door
And so we
For Clement
The request you.
Will fans
Your mouth,
Who goes to bed
To eat
If you hard
Too sick
Color fade
Thou shalt take,
And lose
Being overweight.
God you doint
Good health,
My cute.

The similarities to the machine-translated samples in Le Ton beau de Marot are striking, and many of the stumbling blocks are the same—the use of “cute” as a noun, for instance; or the archaic formations vitement and doint, which speakers of modern French should still be able to parse. What surprises me is that the corpus-based statistical translation provided by Google Translate far more closely resembles the poem generated by the syntactic, rule-based Globalink Translation System (#66 in Le Ton beau, “My Cute”) than the one produced by the corpus-based, statistical Candide (#67, “My Flapper”).

At this stage in the game, it’s no wonder that the opponents of AI would seize upon these rudimentary, syntactic efforts and extrapolate, from them, the impossibility of machines attaining humanlike levels of linguistic comprehension and production. Of course, Hofstadter also produces examples of how good generated prose can be if one develops and refines it within a specific problem domain. Consider Anthony Davey’s “Proteus”, a program developed to play tic-tac-toe and supply a running strategic commentary that produced lucid sentences like this one: “If you had blocked my edge, you would have forked me, but you took the middle of the one opposite the corner I had just taken and adjacent to mine and so I won by completing my edge.”

The human capacity for language, too, may ultimately be a reflection of our versatility in switching between limited contexts and domains. We read and write in one way when it comes to business letters, and another way when it comes to rhymed verse; one way in French, and another way in English. Not so different, after all, from the architecture of digital computers—the universal Turing machines, the programs that can simulate all other programs.


submit to reddit

3 rejoinders to “Wednesday Book Club: Le Ton beau de Marot

Say something interesting: