The Elephant in the Dark: Corpus-Based Descriptions of Translators’ Style

DOI : 10.57086/cpe.205


The corpus-based method of describing translators’ style views style as a matter of “recurring patterns of linguistic behavior” (Baker 2000: 245). Taking the example of a study which applies the “methodology for investigating the style of a literary translator” (Baker 2000: 241), this paper reveals that firstly, in spite of the large amount of data-processing they require, such studies lack descriptive efficiency and yield only partial glimpses of the translator’s linguistic and stylistic habits in a particular translation, and secondly, that this approach diffuses false assumptions about non-corpus resources in translation practice and research. Moreover, as the case study reveals, an approach which applies analytical criteria that are selected without reference to any reading of the source or target texts cannot achieve more than a few lines of the translator’s “thumb-print” (Baker 2000: 245), and can only result in conclusions resembling those which Rumi’s wise men drew concerning the elephant they touched in the dark: each man felt a different part of the animal and came up with a different description of its nature. An application of Berman’s method for describing and explaining a translator’s choices (Berman 1995) will here counter the corpus-based methodology, and will support the idea that a translator’s style may very well lie in the unique choices which the corpus-based approach casts aside. The paper thus shows that a general picture of the translator as “sujet traduisant,” combined with thorough reading of the texts prior to analysis, allows the researcher to move beyond the level of description, and to produce a far more complete picture of the translator’s strategies.

La méthode qui propose de décrire le style des traducteurs à travers l’analyse de corpus présente le style d’un traducteur comme étant surtout une question de « schémas de comportements linguistiques » (Baker 2000 : 245). Prenant l’exemple sur une étude qui applique cette « méthodologie pour examiner le style d’un traducteur littéraire » (Baker 2000 : 241), l’article démontre que, malgré l’importante quantité de données qu’il appartient à ces études de traiter, celles-ci manquent d’efficacité descriptive, ne donnant que des appréciations partielles quant aux habitudes linguistiques et stylistiques d’un traducteur. Aussi cette méthode diffuse-t-elle de fausses présomptions concernant les outils « hors corpus » dont disposeraient les traducteurs et les chercheurs. L’étude examinée ici révèle en outre qu’une approche appliquant des critères analytiques sélectionnés en amont sans référence à une lecture des textes source ou cible ne peut produire qu’une esquisse de l’« empreinte digitale » du traducteur (Baker 2000 : 245). Cette méthode n’est donc susceptible de générer que des conclusions semblables à celles tirées par les sages de Rumi concernant un éléphant qu’ils avaient touché dans l’obscurité : chaque sage n’ayant touché qu’une partie de l’animal, tous en donnèrent des descriptions qui différaient fortement l’une de l’autre. Une application de la méthode proposée par Antoine Berman (Berman 1995) visant à décrire et à élucider les choix d’un traducteur soutient ici l’idée que le style d’un traducteur pourrait très bien s’exprimer dans ces choix uniques que la méthodologie fondée sur les corpus écarte de son champ d’analyse. Enfin, le présent article démontre qu’une image générale du traducteur en tant que « sujet traduisant » (Berman 1995 : 16) complétée par la lecture approfondie des textes constituent des préalables à l’analyse qui permettent au chercheur d’aller au-delà de la description et d’obtenir une image plus complète des stratégies du traducteur.



corpus en traduction, critique de la traduction, Edgar Poe, style en traduction, traduction littéraire, traductologie


corpus-based translation studies, literary translation, Poe studies, style in translation, translation criticism, translation studies



  • PDF


Corpus-based descriptions of translators’ style have existed since the nineties, and this paper will begin with an outline of their initial statements and aims as they were introduced by Mona Baker in “Towards a Methodology for Investigating the Style of a Literary Translator” (Baker 2000). Secondly, some assumptions linked to the corpus-based approach will be critically examined, and thirdly I will present an analysis of Marion Winters’ recent study, which applies Baker’s method in a paradigmatic way. In the last part of the paper, a counterproposal which argues for both a textual and paratextual approach to the description of translators’ style will be put forward, accompanied by some examples of its application. This proposal will support the conviction that the primary basis for any examination of what a translator does in translation must be a repeated reading of the translation and of the source text, and a documented understanding of the context in which the translation came about.

1. The first steps of the corpus-based approach: Mona Baker

Mona Baker is called the “mother of Corpus-based Descriptive Translation Studies” (Laviosa 2002: 18) and her definition of her object of study, namely “the style of a literary translator,” has been copied and pasted onto further applications of the corpus-based approach to investigating literary translators’ style. However, a good part of the “definition” is now ignored by the researchers who apply Baker’s proposals, as their interest in that part of Baker’s object of study seems to have dwindled in equal proportion to the surge of enthusiasm which the corpus-based approach has aroused.

The definition which Mona Baker gave as a working statement for a description of “a literary translator’s style” is partly derived from Leech and Short’s descriptions of style in English fictional prose (Leech and Short 1981), and is here divided in two parts in order to underline the bifurcating nature of the statement:

I understand style as a kind of thumb-print that is expressed in a range of linguistic – as well as non-linguistic – features. As such, it covers the notion of “voice” as defined by Hermans above, but also much more. In terms of translation, rather than original writing, the notion of style might include the (literary) translator’s choice of the type of material to translate, where applicable, and his or her consistent use of specific strategies, including the use of prefaces or afterwords, footnotes, glossing in the body of the text, etc. (Baker 2000: 245)

Though Baker begins her conception of a translator’s style with activities which are very visible and open, we will see in a moment that she overtly excludes “instances of open intervention” from her field of interest. Still, most of the aspects just mentioned, such as the “choice of the type of material to translate” and the “the use of prefaces or afterwords, footnotes,” can hardly be called hidden interventions. Baker’s definition of her translator’s thumb-print continues with:

More crucially, a study of a translator’s style must focus on the manner of expression that is typical of a translator, rather than simply instances of open intervention. It must attempt to capture the translator’s characteristic use of language, his or her individual profile of linguistic habits, compared to other translators. Which means that style, as applied in this study, is a matter of patterning: it involves describing preferred or recurring patterns of linguistic behavior, rather than individual or one-off instances of intervention. (Baker 2000: 245, my italics)

Baker thus insists that we should ignore “individual or one-off instances of intervention” to the benefit of studying “preferred or recurring patterns of linguistic behavior,” which contrasts with what she has previously announced. The confusion is cleared up, however, by Baker’s further emphasis that she is “… interested in patterns of choice (whether these choices are conscious or subconscious), rather than individual choices in isolation” (Baker 2000: 246). In other words, the literary translator as a decision-making subject who carries out open interventions has been brought onto the stage for an appearance that will be too short to be remembered, as our examination of a recent application of Baker’s proposals will confirm. This raises the question whether Baker really aimed to formulate a complete definition of “a literary translator’s style” that would serve as a basis for a complete description of that “style,” or whether she was preparing the ground on which to launch a methodology that whose main reason of existence is derived from the practical possibility of applying pattern counts to literary translation. In any case, the diverging nature of the definition of object explains why its first part no longer appeals to those who base their research on it.

2. Assumptions related to the corpus-based approach to translators’ style

Baker’s working statement has now allowed us to single out the object, namely recurring patterns of linguistic behavior, and the aims, namely selecting and describing recurring patterns of linguistic behavior, of corpus-based descriptions of translators’ literary style, and I will now have a closer look at two assumptions which the corpus-based approach diffuses. The following sections do not constitute a rejection of the use of corpora in descriptive translation studies (or elsewhere, for that matter) – corpora and their tools are very useful as a complementary tool in the description of existing translations, and to some extent they can be used in a meaningful way in translator training. The aim is to warn against the careless propagation of assumptions which sustain the much-lamented miscommunication between literary translators and translation theorists, whose dismissive attitudes of the practitioners' expertise and competence can only widen the gap.

The first of these assumptions is taken from Jeremy Munday’s 1998 article “A computer-assisted approach to the analysis of translation shifts,” where Munday shows his optimism concerning the utility of corpora in translation practice and in translation studies, stating that

No doubt there will be growing use of corpora in the work of practicing translators themselves as a means of attaining consistency and as a complement and counterbalance to the uncertainties of intuition. (Munday 1998: 15, my italics)

In reality, the spectacular rise in translation practice of translation memories and other CAT tools, complementary tools that are not corpora, prove Jeremy Munday to have been wrong in his optimism. Moreover, as any professional literary translator will confirm, this branch of translation practice has so far gained little from the above-mentioned tools. More to the point, a similar expression of the assumption that corpora are needed to counterbalance “the uncertainties of intuition” has been carried forward by Lynne Bowker:

In the past, translators […] have had to work with conventional resources (e.g. dictionaries, printed parallel texts, subject field experts, unverified intuition) which are not always very conducive to provide the conceptual and linguistic knowledge necessary to objectively evaluate a translation. (Bowker 2001: 345)

Translated into Tarzan tongue (in-cheek), the two assumptions in Munday’s and Bowker’s statements would be: “INTUITION: SUBJECTIVE and BAD; CORPORA: OBJECTIVE and GOOD.” In spite of these warnings, however, literary translators still don’t use corpora any more than they did fifteen years ago, and continue to rely on monolingual dictionaries and a series of tools and methods to which their intuition as professional or semi-professional translators directs them, such as cross-readings, parallel texts and paratextual investigations. Moreover, Bowker’s suggestion that corpora guarantee greater objectivity would be rejected by corpus linguists of any school on the basis that the objectivity of a corpus is always debatable, since corpora are composed of texts that have been selected by individuals using criteria formulated by individuals. Lastly, the descriptive efficiency of corpus studies concerning literary translators’ style has so far not been proven, as the following discussion will demonstrate.

3. A case study: Marion Winters’ application of Baker’s approach

3.1. Scott Fitzgerald’s Die Schönen und Verdammten?

I will now examine a study by Marion Winters, whose exemplary application of Mona Baker’s methodology to two translations of Scott Fitzgerald’s The Beautiful and the Damned is presented in an article entitled “F. Scott Fitzgerald’s Die Schönen und Verdammten: A Corpus-based Study of Speech-Act Report Verbs as a Feature of Translators’ Style” (Winters 2007). It seems relevant to dwell on Winters’ use of the possessive case in her title, which illustrates how some translation theorists perpetuate the long-standing tradition of forgetting that the translator, the “sujet traduisant,” as Berman put it (Berman 1995: 16), constitutes an essential factor in what they are describing. It is unlikely that any reader needs the assurance that Fitzgerald did not write Die Schönen und Verdammten, but the remark is made here because Marion Winters does not justify the possessive case in her title, and thus implicitly perpetuates the tradition of conferring a secondary status on translation and on translators. This leads to the observation that Lawrence Venuti’s work on the scandalous invisibility of translators (Venuti 1995 and 1998) still has a long way to go before it will have permeated the consciousness of some researchers in translation studies.

3.2. Winters’ aims and scope of investigation

In the introduction to her article, Marion Winters proposes to analyse two contemporary translations of Fitzgerald’s novel which both came out in 1998, one by Hans-Christian Oeser and one by Renate Orth-Guttmann. The aim of the paper is “to investigate the styles of the two German translators of Scott Fitzgerald’s novel” (Winters 2007: 412) with an understanding of style as “preferred or recurring patterns of linguistic behavior, rather than one-off instances of intervention” (Baker quoted in Winters 2007: 412). Winters then presents her specific criterion of analysis, namely the translation of speech-act report verbs, a choice which appears to be wholly independent of any reading of the translations or of the source texts. Nor is the selection of this particular item (why not, for instance, qualifying adjectives?) explained through any specific discourse function it might have shown to have in the narratives in which it occurs. More surprising and worrying still is that nowhere in her article does Winters refer to any partial or complete reading of these texts.

3.3. Winters’ method

Once Winters has explained what “speech act report verbs” are, she narrows her selection down with criteria which seem worthy of examination against the above-mentioned claim, made among corpus-based-approach proponents, of guaranteeing greater objectivity and complementing the uncertainties of the translator’s / researcher’s intuition. Winters indicates that among the “most frequently used speech act report verbs in the German translation” she has carried out a “qualitative analysis of these cases, investigating whether the options chosen were also the most obvious” (Winters 2007: 415, my italics), without explaining what she means by “obvious,” nor giving any scale or measure of obviousness. The only tool which Winters mentions here is a bilingual dictionary (a tool for which many professional translators harboura constitutional distrust), and we are thus led to assume that Winters has also relied on her own intuition, though we do not know whether she is a native speaker of German. Winters subsequently asks a question which will supposedly allow her to draw conclusions concerning the translators’ respective styles: “Finally, the instances where the translators did not opt for the obvious translation were examined with regard to the semantic meaning of the alternative verb used. “ (Winters 2007: 415, my italics). The jump from pragmatic to semantic meaning is not explained, though the difference seems relevant after the discussion which Winters has presented on the pragmatics of speech-act report verbs, and the phrase “semantic meaning” thus stands as a mysterious tautology in an analysis which selects parts of speech on the basis of their pragmatic meaning.

3.4. Winters’ findings

Winters then proceeds to take the reader through the pattern counts which she has listed in a series of tables, and after six such tables concludes with answers to questions which she has not asked, without, however, answering the question which she did ask at the outset of the study. Indeed, the question of what characterizes the style of the translators Oeser and Orth-Guttmann when translating The Beautiful and the Damned into German, i.e. which are the observed “preferred or recurring patterns of linguistic behavior, rather than individual or one-off instances of intervention” (Winters 2007: 412), is answered through three “possible interpretations” or “tentative conclusions”:

(i) Oeser uses lexical repetition more than Orth-Guttmann, who tends to avoid repetition; (ii) Oeser stays closer to the source text than Orth-Guttmann; and (iii) Orth-Guttmann tends to explicitate. (Winters 2007: 423)

These answers show nothing beyond the fact that Marion Winters can correctly count occurrences of what she has designed as “obvious” choices. In spite of one interpretative sentence which describes the consequences of the observed repetitions and explicitations, namely “Orth-Guttmann also appears to use speech-act report verbs to emphasize character’s emotional states to the extent that she influences the reader’s attitude to the characters, ” (Winters 2007: 424) Winters ends the paper with the admission that “The question remains, however, as to whether she [Orth-Guttmann] does this [avoiding lexical repetition] systematically and to such an extent that the readers of her translation would perceive the characters differently from those of Oeser’s translation ” (Winters 2007: 424). The reader who was expecting the use of corpus tools to at least result in a systematic and complete description of linguistic patterns throughout the text is, by now, rather disappointed.

3.5. Winters’ conclusions

However, the conclusive interpretations deduced from the observed “significant differences in the use of speech act report verbs between the two translators” (Winters 2007: 424) are definitely the most worrying aspect of this study:

Orth-Guttmann was aware of the accepted wisdom that repeated use of speech-act report verbs, particularly sagte (said), is considered bad style, or as Orth-Guttmann (personal communication, July 2005) put it, that German readers do not like the repetition of sagte. Thus Orth-Guttmann’s strategy conforms to the perceived target language conventions by avoiding too much repetition. Oeser uses repetition as a stylistic device which runs counter to current conventions. (Winters 2007: 424)

These concluding sentences raise a number of issues. Firstly, the corpus analyst expands her conclusions far beyond the question she has asked, and does so on the basis of a personal opinion which one of the translators has communicated to her. This opinion, which she introduces as an “accepted wisdom,” subsequently becomes a “perceived target language convention” and then obtains the status of “current convention” (no longer “perceived” by an individual translator).

Besides pointing at this revelatory verbiage, the question must also be asked whether this conclusion provides answers concerning the question asked at the beginning of the study. In other words, how do these findings describe, as the title and working hypothesis announced, the translators’ “styles” as expressed in patterns or habits of linguistic behavior? The answers given by Winters do not concern style, but concern Winters’ personal perception of translation norms in Germany: Orth-Guttmann, pre-concludes Winters, “may be motivated by a desire to avoid what is seen by some commentators as bad style” (Winters 2007: 423), while Orth-Guttmann’s colleague is seen to have “deliberately” retained the repetitions that were present in the source text. We are thus presented with answers which do not match the question, at least not in the framework of a definition of “style” as given by Baker and Winters, but which do answer the question of the position of the translators with regards to what the researcher perceives as norms governing literary translation in Germany. Indeed, Winters findings tell us very little in the way of the translators’ respective linguistic habits, and instead seem to lead to the situation described in Rumi’s story, where five wise men who have never seen an elephant are asked to touch it in the dark and report on its nature. The one who has touched the ear describes it as a large rug-like thing, whereas the one who has touched the trunk talks about a tubular pipe-like object, etc. All have felt a part and each comes up with a different description of the whole. One of the meanings of this teaching story is that partial perceptions are apt to give distorted pictures of the whole.

3.6. A few suggestions for improvement

I will conclude this section on Winters’ study with a few suggestions for improvement, of which even a non-specialist of corpus-based studies may conceive. The first major advance would lie in an understanding that the corpus is and remains a tool, not a methodology or an end in itself. This affirmation is put forward in several instances in Maeve Olohan’s work (Olohan 2004), but clearly needs to be more widely adopted. Secondly, whichever definition one applies to the translator’s “style,” the point at which to formulate the specific question(s) to be asked of a translation in order to “investigate” it can in no way precede the reading of the texts. What good can any study of a text hope to do which does not allow for any information supplied by a reading of that text? The selection of the tools of comparison must be informed by thorough and repeated reading, both of the translations and of the source texts, preferably as independent texts. Only after such readings can the researcher formulate working hypotheses that will efficiently guide the comparative analysis. As we will see in the following sections, such hypotheses are the foundation for a complete examination and description of literary translations, and can help reveal aspects and shifts of a variety which no corpus-based study would be able to discern, though corpus tools can of course be used at a later stage to confirm and broaden the observation of these occurrences.

4. Counterproposal: A paratextual approach to reading and analysing translations

In order to preempt accusations of unproductive criticism, I will now present a different kind of comparative approach to reading and analyzing translations, an approach which aims both at describing and interpreting a translator’s manner of translating, and thus allows the researcher to go beyond aspects of the ill-defined object of “style.” The translation scholar who has so far offered the most complete paradigm for such descriptions is Antoine Berman, whose method was proposed in Pour une critique de traductions: John Donne (Berman 1995). Having elsewhere brought to the surface the philosophical bases which underlie this approach (Wallaert 2004), I again introduce Berman’s work as a paratextual approach, for reasons which will shortly become obvious.

5. Research method: pre-analytical reading and enquiries

In Pour une critique de traductions Berman puts great emphasis on the importance of reading the target and source texts intensively and independently1. However, he also proposes a number of pre-analytical enquiries which allow the researcher to consider the translator as an individual agent of choice, because, as he puts it, “… une traduction est toujours individuelle … parce qu’elle procède d’une individualité, même soumise à des “normes” (Berman 1995: 60), and this is where paratexts come into play. The question “qui est le traducteur? ” is of the utmost importance, says Berman (Berman 1995: 73), and will result in a “théorie du sujet traduisant” (Berman 1995: 75) which finds a large part of its information in paratexts. This theory of the translating subject can be subdivided in four points of inquiry.

Firstly, Berman asks us to look at the translator’s linguistic position (his “position langagière,” Berman 1995: 75), i.e. what is his position as a user of both the source and target languages. This question remains relevant today, for the assumption that all translators are and should be native speakers of the target language is under siege, as Nike Pokorn’s work so powerfully illustrates (Pokorn 2005). Berman links this “position langagière,” this “being-in-language,” to a “being-in-literature,” and claims that the researcher should also inquire into the translator’s personal literary preferences.

The second group of inquiries involves anything that can reveal the translator’s general stance in translation (his “position traductive,” Berman 1995: 75), as it can be deduced, for instance, from what is said in prefaces and footnotes or in comments which the translator may have produced regarding his work in translation in general. Such information can lead to informed interpretative statements, and not conjectures, concerning the translator’s manner of translating. Thirdly, the translator’s stance is usually adapted to each particular translation project, and the question needs to be asked what the translator’s particular project may have been for the source text at hand. Again, these data are often found in the paratexts that accompany translations, such as footnotes and prefaces, and sometimes in reviews of the translation.

A fourth and last group of enquiries concern the translator’s horizon, both on a translational and literary level. We are here asked to examine both the target literary system into which the translation is introduced (which in the case of authors/translators includes their own “being in literature”), and the norms in translation that govern translation practice at the time when the translator is working. A serious study of the norms that govern translation of a particular type of literature in the target culture will again avoid speculations of the type which Marion Winters makes regarding translation norms in Germany.

What is essential to note at this point is that being-in-language, being-in-literature, stance, project, and literary and translational horizon, do not have any pertinence without a demonstration of their traces in the target text(s), in other words, the main aim of these enquiries is to provide information which will help the researcher explain and interpret the translation shifts she/he observes in the target text.

The researcher can also rely on what Genette calls “public epitext,” (Genette 1987), i.e. reviews and commentaries which other readers may have written on the source and target texts. The use of this public epitext is indispensable and very useful, both to better understand the source text, and to observe how others have read the translation. Indeed, readers of translations often review these as independent texts, and to this day there is an abundance of profound critical statements which supposedly concern the work of an author but are really based on translations of that author’s work. Such reviews and studies provide invaluable complementary views of how a translation is read as an independent text in the target system.

Once the data have been gathered, in as much detail as possible (admittedly a time-consuming activity, though not necessarily more so than turning a complete novel and its two translations into a corpus), we can return to the impressions which the pre-analytical reading of the target texts and the source texts as independent pieces of literature have yielded. These impressions, whose subjectivity and speculative nature is completely allowed for, are grouped together as a series of hypotheses, which will guide the researcher in her/his comparative analysis, and which are there to either be cast aside as mere impressions, or to be confirmed by what the researcher finds in the course of a detailed comparative analysis. When enough textual data bear out the hypothesis, the researcher can go back to what the pre-analytical inquiries have revealed about the “translating subject,” and interpret the findings in an informed way.

This method thus compensates for shortcomings which affect other approaches to reading and analyzing translations, since it allows one to move beyond description into explanation and interpretation, and besides covering all aspects of the translatio of a text from one language into another, its main strength lies in the fact that translations are not evaluated by external standards, but by standards which their own coming into being has created.

6. Applications of the paratextual approach

6.1. Baudelaire as “translating subject” of Poe

What follows are some applications of this method to the translation which Charles Baudelaire produced for Edgar Allan Poe’s “The Fall of the House of Usher.” Concerning Baudelaire, the scope of the “theory of the translating subject” fills a volume in itself, and can obviously not be repeated here (see Wallaert 2004). To put it succinctly, Baudelaire’s name first became known to the general public through his translations of Poe’s work, and as a translator-poet he had a very specific and well-documented being-in-language in the source and target language, and obviously a well-known being-in-literature in the target language and literature. Baudelaire also expounded his specific project for the translation of Poe’s work in a number of paratextual writings that accompanied his translations, and in other texts which he produced during his lifetime. Moreover, the case of historic translations such as Baudelaire’s allows a detailed description of his literary and translational horizon, since both these factors have been widely discussed.

6.2. “La chute de la maison Usher”: Reading target and source texts and formulating hypotheses

The pre-analytical readings of “La chute de la maison Usher” and “The Fall of the House of Usher” led to the distinct impression that there is a shift in a number of key aspects of the story. In the French version, the reader is confronted with a narrator who seems more involved in what goes on, and seems on the whole more emotionally vulnerable and more hysterical than the English narrator. The French narrator also seems less imaginative and more rational than his English counterpart, and is felt to be more sceptical, even slightly pedantic, when it comes to the esoteric fancies of his friend Usher. Between the two versions of this story, which is woven through with a play on symmetry creating a series of suggestive undertones, ambiguities and uncertainties, Baudelaire’s version also appears the less ambiguous of the two.

Several guiding questions were thus formulated after the pre-analytical readings, among which I have here selected:

(i) Which psychological and emotional reactions cause the French narrator to come across as more emotionally vulnerable and more hysterical than the English narrator?

(ii) Which elements tend to present a French narrator who is less imaginative, more rationalizing and skeptical concerning Usher’s esoteric interests?

(iii) Which elements may have caused a change in the symmetry or in features that work to sustain ambiguity and uncertainty?

Among a great number of occurrences which the complete comparative analysis yielded, a few will be presented here, selected both for their eloquence in confirming the hypotheses, but also because they illustrate that the mark of the translator does not necessarily lie in patterns of linguistic behavior, but may very well lie in those one-off textual interventions which so many researchers insist on ignoring. The comparative analysis was carried out simply by laying both texts in their printed version next to each other and comparing every sentence, clause, phrase, word and punctuation mark. The tools used were my own intuitions as a native speaker of English and as near-native speaker of French, supported by monolingual dictionaries which include 19th-century meaning and usage (Littré, the Trésor and the OED).

6.3. Analysis and comparison

(i) Psychological and emotional reactions of the French narrator

Several instances occur in the French translation where the narrator is more excited in his reactions to the strange events that occur in the story. The following extract presents an example of this type of shift, and as it occurs at the very beginning of the story, it also sets the tone for the narrator’s feelings and perceptions throughout the story:

Qu’était donc ce je ne sais quoi qui m’énervait ainsi en contemplant la Maison Usher? C’était un mystère tout à fait insoluble et je ne pouvais lutter contre les pensées ténébreuses qui s’amoncelaient sur moi pendant que j’y réfléchissais. (Le Dantec 1951: 337, lines 26-31)

What was it – I paused to think – what was it that so unnerved me in the contemplation of the House of Usher? It was a mystery all insoluble; nor could I grapple with the shadowy fancies that crowded upon me as I pondered. (Mabbott 1978: 397‑398, lines 21‑24)

While the French narrator asks himself what “énervait” him so, the English narrator is not agitated or nervous, but unmanned and/or bewildered. For “énerver” the Trésor gives a series of figurative explanations: “Faire perdre à quelqu’un ses forces physiques ou morales” is the first of these, but the more common explanation is certainly “Exciter, irriter les nerfs de quelqu’un; rendre nerveux” (Trésor 7: 1082). The French narrator thus begins the story in a state of nervousness and psychological agitation, while OED does not mention, for “unnerve,” anything resembling the French dimension of agitation or irritation (OED 11: 279). Moreover, the French narrator is trying to “lutter contre” a number of “pensées ténébreuses.” “Lutter contre,” in its figurative sense, means “Combattre, résister” (Littré 3: 3608), and the Trésor gives the very appropriate example of “lutter contre une impression, une sensation” (Trésor 11: 75). The French narrator is thus trying to get rid of certain “pensées” that bother him – which explains his heightened nervousness and agitation, but does not coincide with the English version, where the narrator is trying to “grapple with” things, in the sense given in OED, namely, “to try to deal with (a question, etc.); to try to solve (a problem, etc.)” (OED 5: 362). Moreover, what the English narrator is trying to come to terms with are not “pensées,” but “fancies.”

The world of difference that lies between “fancies” and “pensées” involves a detailed explanation of the ways in which Poe, following Coleridge, saw the difference between “fancy” and “imagination,” and the significance of this recurrent non-translation of terms as crucial to Poe’s thought as these, will have to be taken up on a different occasion. Last but not least, in the above extract Baudelaire omits the interjection – “I paused to think “ – which gives the English passage a much calmer pace, and translates “pondered” by “réfléchissais,” and activity which is again much less contemplative than musing or pondering. The scope of this article does not allow me to add any more of the numerous other examples which were found of this first type of shift.

(ii) A skeptical, less imaginative and more rationalizing French narrator

The extract which shows a more rationalizing and less imaginative French narrator was selected among the numerous other examples for the way in which it also reveals the general shift towards the gothic which the story undergoes (see also Wallaert, 2009)

Peut-être m’impressionna-t-elle plus fortement, quand il me la montra, parce que, dans le sens intérieur et mystérieux de l’œuvre, je découvris pour la première fois qu’Usher avait pleine conscience de son état, – qu’il sentait que sa sublime raison chancelait sur son trône. (Le Dantec 1951: 346, lines 394-398)

I was, perhaps, the more forcibly impressed with it, as he gave it, because, in the under or mystic current of its meaning, I fancied that I perceived, and for the first time, a full consciousness on the part of Usher, of the tottering of his lofty reason upon her throne. (Mabbott 1978: 406, lines 309-314)

Besides the questionable translations of “fancied” and “mystic,” the French narrator does not hedge his observation about Usher’s state at all, but simply talks about “découvrir” something, whereas the English narrator states “I fancied that I perceived.” This produces a largely less certain and confident English narrator than the one we find in the French version, who cognitively discovers, instead of doubting his perception. Interestingly, for Todorov, a reader of the French version, “J’en vins presqu’à croire” is “la formule qui résume l’esprit du fantastique” (Todorov 1970: 35), and by annulling this key element, Baudelaire turns his narrator into a rational thinking being, and thus produces a clear shift towards the gothic. It is therefore not surprising that Todorov is also (to my knowledge) the only specialist of fantastic literature who excludes Poe’s work from the fantastic mode of writing.

(iii) Less ambiguity in the French text

The third hypothesis concerned elements that could have caused changes in the ambiguity which pervades the story. A first set of such changes can be found in the French translation in places where the English version presents a possibility of amalgamating Roderick Usher and his twin sister Madeline, who dies and is buried in the vaults of the mansion in the middle of the story. This possibility is annulled by Baudelaire on several occasions, of which I will here give two examples. Firstly, when the narrator finally arrives at Usher’s chamber, and is “ushered in,” he hardly recognizes his childhood friend and says:

Ce n’était qu’avec peine que je pouvais consentir à admettre l’identité de l’homme placé en face de moi avec le compagnon de mes premières années. (Le Dantec 1951: 341, lines 195‑197)

It was with difficulty that I could bring myself to admit the identity of the wan being before me with the companion of my early boyhood. (Mabbott 1978: 401, lines 150‑151)

Baudelaire here cancels the vagueness concerning the identity of the “wan being” which the narrator is facing, and in the following example he again makes Usher’s gender explicit:

Mais actuellement, dans la simple exagération du caractère de cette figure, et de l’expression qu’elle présentait actuellement, je doutais de l’homme à qui je parlais. (Le Dantec 1951: 342, lines 210‑213)

And now, in the mere exaggeration of the prevailing character of these features, and of the expression they were wont to convey, lay so much of change that I doubted to whom I spoke. (Mabbott 1978: 402, lines 161‑164)

The extract not only confirms that omissions were part of Baudelaire’s translation strategies (nothing in the French text reproduces the emphatic “lay so much of change that”), but more to the point is Baudelaire’s translation of “whom” by “l’homme,” which again annuls an ambiguity which is created twice in this manner, and twice again elsewhere in a more implicit way, in the original text.

Lastly, Baudelaire often diminishes the suggestive undertones of certain images, and opts for terms that are apt to create a more explicit gothic type of horror. When Roderick talks about his sister’s approaching death, for instance, Poe uses the words “dissolution” (Mabbott 1978: 403‑404, line 226‑227) and “decease” (Mabbott 1978: 404, line 229), which Baudelaire twice translates by the more prosaic “mort” (Le Dantec 1951: 344, line 289 and 292). Sometimes the more outspoken morbidity is constituted by a pure addition, as in the translation of “an excited and highly distempered ideality” (Mabbott 1979: 405, lines 268-269) by “une idéalité ardente, excessive, morbide” (Le Dantec 1951: 345, lines 339‑340), and a “tempestuous yet sternly beautiful night” (Mabbott 1979: 412, line 512) becomes “une nuit d’orage affreusement belle” (Le Dantec 1951: 352, lines 638-639), while “some bitter struggle” (Mabbott 1979: 416, line 659) is translated as “quelque horrible lutte” (Le Dantec 1951: 356, line 817). This is only a small sample of the numerous occasions where Baudelaire diminishes the suggestive ambiguity of the narrator’s descriptions, and where he sprinkles the text with more explicit expressions of morbidity. Such additions and shifts, resulting in an increased atmosphere of gothic horror, also occur in other translations which Baudelaire made of Poe’s tales in a consistent way (See Wallaert 2001, which features a discussion of Baudelaire’s “Le scarabée d’or”).

Conclusion: Describing an ear for an ear

The three hypotheses which my repeated reading of the translation and the source text of “Usher” helped to formulate are borne out by a series of translation shifts which can be observed on numerous occasions. These shifts create a variety of effects of which I have only presented a small sample here, and can be explained by referring to Baudelaire’s translation project and his own being-in-literature, which included a strong liking for the prose writings of the more gothic type represented by writers such as Pétrus Borel. The shifts concern diverging lexical choices, and omissions or additions which appear in different discourse functions. The disparate and varying nature of these functions entails that these shifts would not have been discerned through a corpus analysis which searches for a limited set of discourse features, and obviously even less so if these features had been selected arbitrarily without reference to the texts themselves. This observation also leads to the conclusion that a translator’s mark or thumbprint is not only expressed through his linguistic behaviour, and moreover, that important aspects of this linguistic behaviour do not lie in recurring patterns, but might better be discerned through one-off instances of intervention which the corpus-based approach rejects.

What this paper mainly aimed to show, however, is that only thorough reading of the target and source texts can convey impressions which may be borne out by further analysis, whichever shape that analysis chooses to take. The case of the a-priori determined criteria which we saw Marion Winters use illustrates that, unless such reading takes place, all we can hope to describe both on a stylistic level and beyond, is the elephant’s ear, and this is not a basis from which we can extrapolate a general picture of the animal. As we have seen, Rumi’s story illustrates that partial perceptions such as the one which Winters so painstakingly produces, often result in general pictures that are spectacularly distorted indeed. The corpus-based approach, though claiming to have seen the whole elephant which it has wrongly named “style,” has only touched a small part of the beast, and can in no way make informed statements about things whose nature it has not managed to completely define. Moreover, an approach which asks the question of patterns of linguistic behaviour, must in all honesty limit its answers to describing patterns of linguistic behavior – it must describe an ear for an ear, so to speak – and when its proponents wish to go beyond these limits, they need to carry along tools that are adequate to tackle the complex issues which they bring to the surface.

In a paratextual approach, which allows for reading not only of the texts, but also of any available paratextual and epitextual materials, some of the hypotheses will also be inspired by our knowledge of the translator’s position and literary preferences, and can therefore help to explain and interpret the shifts we observe, thus allowing us to move beyond the stage of description. Finally, an approach based on reading can make use of corpus tools, in the sense that when a translational shift is found to centre around a concept which occurs in the shape of a finite set of lexical items in a definable set of discourse functions, the researcher can proceed to use corpus tools to facilitate and expand the analysis of the translator’s treatment of these terms and their corresponding concepts. Corpora and their devices are thus not discarded as tools, but their use as sole method of analysis seems quite inefficient, and their claim to being a “method for analysing a translator’s style” is rejected as being too ambitious by far.

1 Berman believes that we should give precedence to the translations, and thus begin by reading and rereading these, because it enhances our


BAKER Mona, 2000, “Towards a Methodology for Investigating the Style of a Literary Translator”, in Target 12:2, p. 241‑266.

BERMAN Antoine, 1995, Pour une critique des traductions: John Donne, Paris, Gallimard.

BOWKER Lynne, 2001, “Towards a Methodology for a Corpus-based Approach to Translation Evaluation”, in Meta XLVI, 2, p. 345‑364.

CORNWELL Neil, 1990, The Literary Fantastic, London, Harvester Wheatsheaf

GENETTE Gérard, 1987, Seuils, Paris, Seuil. English translation: Paratexts: Thresholds of Interpretation, by Jane E. LEWIN, Cambridge, CUP.

JACKSON Rosemary, 1981, Fantasy: The literature of subversion, London, Methuen.

LAVIOSA Sarah, 2002, Corpus-based Translation Studies.Theory, Findings, Applications, London, Rodopi.

LE DANTEC Yves (éd.), 1951, Edgar Allan Poe: Œuvres en prose. Traduction française de Charles Baudelaire, Paris, Gallimard – Pléiade.

LEECH Geoffrey, SHORT Michael, 1981, Style in Fiction. A Linguistic Introduction to English Prose Fiction, London, Longman.

MABBOT Thomas Oliver (ed.), 1978, Collected Works of Edgar Allan Poe: Tales and Sketches (18311842),vol. 2 & 3, London, Belknapp Press of Harvard University Press.

MUNDAY Jeremy, 1998, “A Computer-Assisted Approach to the Analysis of Translation Shifts”, in Meta XLIII, 4. p. 1‑16.

OLOHAN Maeven 2004, Introducing Corpora in Translation Studies, London, Routledge.

POE Edgar Allan, 1840, “Thomas Moore”, in THOMPSON, Geoge R. (ed.), 1984, Edgar Allan Poe: Essays and Reviews, New York, Library of America, p. 333‑341.

POKORN Nike, 2005, Challenging the Traditional Axioms: Translation into a non-mother tongue, Amsterdam, John Benjamins.

TODOROV Tzvetan, 1970, Introduction à la littérature fantastique, Paris, Éditions du Seuil.

VENUTI Lawrence, 1995, The Translator’s Invisibility. A History of Translation, London, Routledge.

VENUTI Lawrence, 1998, The Scandals of Translation: Towards an Ethics of Difference, London, Routledge.

WALLAERT Ineke, 2009, “Du fantastique au gothique: ‘La chute de la maison Usher’ de Charles Baudelaire”, in Traduire,n° 219, Revue française de la traduction (SFT). p. 47‑69.

WALLAERT Ineke, 2004, Baudelaire’s Rewriting of Poe: A Para-Textual Critique of the Translations, Unpublished Ph.D Thesis, University of Edinburgh.

WALLAERT Ineke, 2001, “The translation of Sociolects: A Paradigm of Ideological Issues in Translation?“, in COTTERILL Janet, IFE Anne (eds), British Studies in Applied Linguistics,n° 16,British Association for Applied Linguistics, London, Continuum, p. 171‑184.

WINTERS Marion, 2007, “F. Scott Fitzgerald’s Die Schönen und Verdammten: A Corpus-based Study of Speech–act Report Verbs as a Feature of Translators’ Style”, in Meta,LII, 3, p. 412‑425.


LITTRÉ Paul Émile (éd.), 1970, Dictionnaire de la langue française, vol. 14. Monte Carlo, Cap.

IMBS Paul (éd.), 1980‑1994, Trésor de la langue française – Dictionnaire de la langue du XIXe et XXe siècle (1789-1960), Nancy, Centre National de Recherche Scientifique, Institut de la Langue Française (vol. 1‑10) & Paris, Gallimard (vol. 11‑16).

MURRAY James A. H., BRADLEY Henry W., CRAIGIE William A. & ONIONS Charles T.,1961, Oxford English Dictionary, vol. 1‑12, Oxford, Oxford University Press.


1 Berman believes that we should give precedence to the translations, and thus begin by reading and rereading these, because it enhances our perception of the linguistic and textual qualities of these texts as independent pieces of writing. My own experience confirms this claim.


Electronic reference

Ineke Wallaert, « The Elephant in the Dark: Corpus-Based Descriptions of Translators’ Style », Cahiers du plurilinguisme européen [Online], 2 | 2010, Online since 01 janvier 2010, connection on 21 juillet 2024. URL :


Ineke Wallaert

Ineke Wallaert a obtenu son diplôme de maîtrise en traduction à la Provinciale Hogeschool voor Vertalers en Tolken (École supérieure de traducteurs et interprètes) de Gand en Belgique. Elle a poursuivi ses études à l’université d’Édimbourg : master de linguistique appliquée et doctorat en traductologie : Baudelaire’s Rewriting of Poe : A Para-Textual Critique of the Translations, 2005. Elle a travaillé et enseigné en Belgique, Californie, Chine, Écosse et à la Martinique. Depuis 2006, elle est maître de conférences à l’université de Strasbourg où elle dispense des cours de grammaire anglaise, de traduction ainsi que de traductologie au sein de l’UFR des langues et sciences humaines appliquées. Aujourd’hui, ses intérêts de recherche se portent sur les aspects philosophiques, culturels et idéologiques de la traduction, la traduction de la littérature pour la jeunesse et les méthodes de recherche sur corpus en traductologie.

Author resources in other databases


By this author


Licence Creative Commons – Attribution – Partage dans les mêmes conditions 4.0 International (CC BY-SA 4.0)