Electronic Textual Editing: Authorial Translation: The Case of Samuel Beckett's Stirrings Still / Soubresauts [Dirk Van Hulle, University of Antwerp, Belgium]

Many authors have written in two languages, usually first in their native language and from a certain moment on, often after emigration, in the language of the country to which they moved. Vladimir Nabokov first wrote in Russian, later in English, Julien Green changed from English to French. Beckett's case is more complicated. He wrote in both French and English in turns and translated his own works in both directions. To complicate things further, the switch between French and English sometimes even occurred during the writing process. That is the case in the genesis of Stirrings Still, Beckett's penultimate text.

Stirrings Still was first published in English and subsequently translated into French under the title Soubresauts by the author himself in 1989, the year of his death. A most interesting aspect of this work is its textual genesis. It took Beckett almost six years to complete this short text of only about six pages. That this process took such a long time is to a large extent due to an extreme form of hesitation, which seems to have increased over the years. The first manuscript of Stirrings Still preserved at the Beckett Archives in Reading starts in French, but continues in English. The second page shows the reverse scenario. Here, Beckett started in English and continued in French. The first phase of the writing process is predominantly French, but most of the later manuscripts and typescripts are in English. This long and complex genesis with more than twenty versions makes Stirrings Still a particularly interesting example to discuss the scholarly genetic editing of bilingual writings.

Editorial principles

Since bilingualism characterizes this text both before and after its publication, the aim of the genetic edition is to present the work as both a process and a product, Stirrings and Still. In his typology of genetic documentation, Pierre-Marc de Biasi draws a clear dividing line in a text's production process, taking the bon à tirer moment (all set for printing) as the decisive freezing point between the fluidity of the composition and the frozen shape of a published text (de Biasi, Typology 37). According to French theoreticians of critique génétique a genetic edition provides the reader with all the versions of the work's genesis in extenso. G. Thomas Tanselle also makes a clear distinction between a genetic edition and for instance Hans Walter Gabler's synoptic edition of James Joyce's Ulysses (Tanselle, Historicism 38n72).

Nonetheless, David Greetham suggests that genetic and synoptic editions do not necessarily exclude each other (Greetham, Textual and Literary Theory 20n39). A noteworthy attempt to combine these two types is Charles Krance's editorial model for scholarly variorum editions of Beckett's bilingual works. A French and an English reading text are presented face to face. The synoptic apparatus is remarkable since it marks changes between draft stages, but the sum of these draft stages does not coalesce in a continuous manuscript text. The various stages are represented as different layers, reflecting the order of the writing process. In order to highlight only the variants between versions, undue repetition is avoided in the variant synopses by means of diacritical signs that indicate passages without variants vis à vis the previous version.

One of the reasons why these variorum editions do not include facsimiles and full transcriptions is the codex format (Krance xiii). In 1996, the series' editorial board was not yet convinced of the overall desirability of electronic editions (xiii). In the meantime, electronic editing has evolved, and it is worth a try to see what the possibilities are of creating an electronic, easily searchable environment that meets the requirements and aims of the bilingual editions, i.e., to show the traces of Beckett's work in the composition of his bilingual oeuvre. (xiv)

While Charles Krance readily admits that his methods may not be as all-inclusively ambitious as that of 'hard-line' geneticists (xiv), he rightly argues that the bilingual dimension of this work calls for a clear-text face-to-face presentation of the French and English versions, so as to allow careful comparison. Peter Robinson has pointed out that the number of variants can be so overwhelming that merely presenting a transcription or facsimile of all the documents may not be enough. A reading text is a useful tool and providing the reader with this tool does not imply that this critically edited text is final. Robinson consequently pleads for a presentation of The One Text and the Many Texts. The fact that in the case of Stirrings Still / Soubresauts one has to speak of the two texts and the many texts is all the more reason to provide the reader with this tool. That this double text is indeed meant as a working instrument in the variorum edition is evident from Charles Krance's suggestion: it will be to the user's advantage to photocopy the text(s) under scrutiny as working copy, thus obviating the need to page back and forth between text(s) and synopses (xii). While it may seem remarkable that an edition suggests its own reproduction, the idea of a working copy is essential and serves as a valuable suggestion underlying the interface design in the form of a digital equivalent of the face-to-face representation, to be consulted at all times.

The choice of the base text for the English reading text is complicated by the fact that there is a limited first, signed de luxe edition (New York: Blue Moon Books / London: John Calder, 1988, 44 pp., with nine illustrations by Louis le Brocquy), and a non-limited first edition, published in a newspaper (The Guardian, March 3, 1989). Starting from de Biasi's typology and the prominence of the bon à tirer moment, one could argue that the 1988 edition should provide the base text as it represents the text's first public appearance.

The question, however, is how public this appearance is in this case of an expensive edition, limited to only 226 copies. Even though the avant-texte and the texte are characterized by their respectively private and public character, the limited edition was nonetheless a rather private enterprise. It was meant as a way to support Barney Rosset, who had been Beckett's American publisher for decades and was sacked from Grove Press in the mid-1980s. The newspaper version was the first truly public edition. In the same newspaper, Frank Kermode wrote a review of the limited edition of Stirrings Still, pointing out the irony of this prose, rich only in its unmatched parsimony, being published in a limited edition, available only to a happy few: The purchasers can henceforth meditate the destitution of their existence and simultaneously take pleasure in their privilege... It's a bit like buying a Porsche to mitigate angst (Kermode 26).

Moreover, a leaflet in the American facsimile review copy preserved in the Beckett archive in Reading indicates April 13, 1989 as the actual publication date of the limited edition. Its public appearance is therefore proverbially deceptive, for the so-called 1988 edition of this book was actually published on Beckett's eighty-third birthday, more than a month after the Guardian publication.

One of the consequences of this being a newspaper publication is that the text has to make room for an advertisement. In this case, Beckett's short text, worth £1,000 or $1,500 in hardback, is draped over an advertisement of the London Review of Books, which says: Read the softback before you shell out for the hard—followed by an image of a softback copy of the LRB and the slogan The best of all possible words. If the slogan of the LRB advertisement were to serve as an editorial guideline, the question would be whether this newspaper publication contains indeed the best of all possible words. It might have been interesting if the most democratic version would have allowed for the most conservative of editorial methods, i.e., a best-text approach. But unfortunately the newspaper editor has not proven himself the best guardian of the text. Apart from introducing an unnecessary capital letter and leaving out a necessary word, he has divided the text into four, instead of three sections (possibly to make it comply with the newspaper house style). Especially this last intervention makes it unsuitable to serve as the base text of a scholarly edition, since it is the only version that has four sections instead of three.

Still, from a sociological perspective, one might argue that this newspaper publication is a Beckettian equivalent of the 1922 edition of Ulysses. When Joyce's text was reprinted in the 1990s, Fritz Senn advocated this paperback edition of Ulysses: The 1922 Text and the decision to present the reader with those words and phrases, misprints and all, that set the literary world astir (Senn 461). If any version of Stirrings Still set the literary world astir, it would be the newspaper edition. But then again it must be admitted that, compared to the stirrings caused by Joyce's text, the literary world remained relatively still after the newspaper edition of Beckett's short text.

Beckett's act of self-translation has the paradoxical effect of fixing a text by reproducing it in another language. It is only thanks to the French translation (or target text) Soubresauts that the English version on which it is based receives the status of original (or source text). This effect might be considered as a criterion to serve as a basis for the choice of base text of a bilingual edition. But Beckett did not always simply take the final version as his source text. In the case of Bing, for example, there are instances where the English version Ping is based on earlier drafts (Fitch 70). Since Beckett not necessarily stuck to one version to make his translations, the idea of an original or a source text is problematized and cannot serve as a general principle to choose the base text.

An intentionalist approach might be another alternative. Fredson Bowers advises to use a fair-copy manuscript (if any survives) as copy-text in preference to the first edition set from it. The last versions of the three sections preserved in the Beckett archive in Reading are almost identical with the limited edition, but there is one problematic instance that is symptomatic of Beckett's hesitancy. Even in the last typescript of the second section (MS 2935/4/2) the first sentence is marked by an open variant. The sentence starts as follows: As one in his right mind when at last out again he knew not ... The words he knew not are underlined (not cancelled), and Beckett has added the alternative no knowing above. So, even at this late stage in the writing process, Beckett's intentions appear to be multiple. He apparently needed the act of publication (and therefore necessarily a publisher) to make an end to the endless hesitancy. It is even doubtful whether the octogenarian author would still have published anything after Worstward Ho, had it not been for his American publisher's unfortunate situation. This circumstance was a major impulse to try and finish the Fragment for Barney Rosset (as Stirrings Still was still called in 1986). It is remarkable that neither the Guardian publication, nor the text in the Beckett Shorts (vol. 11) published by John Calder mention the dedication for Barney Rosset. It was Beckett's empathy with Barney Rosset's situation that brought about the publication. This dedication is as crucial as the dedication (For Mrs. Henry Mills Alden) in the poem Trees by Joyce Kilmer, discussed by Jerome McGann in Radiant Textuality (42).

As a consequence, the choice of the limited edition for Barney Rosset as the base text is inspired in the first place by this social circumstance: not because it is de luxe, but because it was meant to help a friend. This case shows that authorial intention and social orientation do not necessarily exclude each other. The only misspelling in this edition (withersoever instead of whithersoever on page four) was unintended, as S. E. Gontarski explains in his reader's edition of The Complete Short Prose. In a review of the limited edition in the Irish Times (April 15, 1989) Gerry Dukes pointed out the typographical error, which the actor Barry McGovern in his turn brought to Beckett's attention. Beckett subsequently corrected the error in the actor's copy (Gontarski 285).

Therefore, the English base text of the electronic genetic bilingual edition is the one represented by a single document: the text of Barry McGovern's copy, in which Samuel Beckett made this one correction. This particular case could be regarded as an extension of Hans Walter Gabler's principle of the last version containing additions in the author's hand, opening up the idea of a continuous manuscript text (which is Gabler's description of the left-hand pages in his synoptic edition of James Joyce's Ulysses) to what Jerome McGann calls a continuous production text (Textual Condition 30).

This idea of a continuous production text is reinforced by the French texts, since the translations took place both before and after the dividing line of the bon à tirer moment (in de Biasi's typology). The first French edition includes the dedication Pour Barney Rosset. Although the Beckett archive in Reading keeps a computer print-out of the translated version on which Beckett has written Final, this version differs in some instances from the text published by Editions de Minuit (both in a limited edition of 99 + 10 copies and in a non-limited paperback edition). The kind of changes suggests that, even though Beckett was as fluent in French as he was in English, he counted on his French publisher Jérôme Lindon to correct minor mistakes.

Methods of text representation

The transcription of the documents preserved in Reading is encoded in TEI-compliant XML. The advantage of this non-proprietary format is the resulting transclusive flexibility of the textual material. Depending on the user's focus, the draft material can be rearranged in several ways: 1. a documentary approach, based on the catalogue numbers; 2. in chronological order; 3. per language; 4. with a focus on translation; 5. in retrograde direction, starting from the published texts:

1. Documents: In a menu, the list of documents is probably the most neutral starting point. The archive catalogue number can serve as unique ID of what the Text Encoding Initiative refers to as the body of the different texts. Each transcription is preferably linked to a facsimile of the relevant manuscript page, so that it can always be checked against the scanned image of the manuscript. This possibility will immediately make the user aware that one single document often contains several versions of the same paragraph. The majority of the manuscripts contain units of text, but especially in the early stages of the writing process and on the left-hand pages of the so-called super Conquérant notebook (MS 2934), Beckett wrote loose jottings. The Documents approach therefore also contains a special subsection for notes (as in Charles Krance's variorum edition).

2. Chronology: Although the catalogue numbers reflect the chronology of versions, some documents, notably the super Conquérant notebook, contain more than one version. The versions in this document are not successive, since Beckett made alternate use of this notebook and loose sheets of paper. The chronological order of the writing process can be indicated by means of a logical succession of ID numbers.

3. Language: In the same super Conquérant notebook, Beckett already made some translations. At a certain point he turned the copybook around and started from the back: Under the heading Repeat in different order he opened with a French translation of the most current English version and continued with an English translation of one of his first French versions. Unlike Joyce in Finnegans Wake, Beckett did not mix languages. In the case of Stirrings Still, the paragraphs can be arranged or rearranged per language by means of the lang attribute in the div, p, and seg tags. On the basis of this attribute the versions can be divided into two sections: French and English.

4. Translation: A simple indication of the language may not be sufficient for a thorough genetic investigation, for not all French versions are translations. To highlight the ways in which self-translation can have a generative creative power, those versions that are translations are marked with the attribute ana set to Tr. This way, they can be visualized, facing the version on which they are based.

5. Teleology: A major problem with genetic editing of prose texts is to find a way to compare one particular passage in one version with the corresponding passage in another. Charles Krance has found an elegant solution for this problem by dividing the text into small units or segments, preceded by line-numbers corresponding to the reading text. A practical example of an electronic edition that applies a similar division into smaller units is the electronic-critical edition of Stijn Streuvels' De Teleurgang van den Waterhoek by Edward Vanhoutte and Marcel De Smedt. Every paragraph in the reading text can be linked and compared to other versions of the same paragraph. Edward Vanhoutte has called this linkable unit a linkeme, i.e., the smallest unit of linking in a given paradigm (Linkemic Approach).

By making the links bidirectional, this system can be expanded and applied, not only to the reading text, but to any version the reader happens to be reading. In Beckett criticism, the later texts are often referred to by means of paragraph numbers (Cohn 380). Stirrings Still consists of three sections, the first of which is subdivided into seven paragraphs. The paragraphs that made it into the published text may start with either 1, 2, or 3 (the numbers of the three sections of the text), followed by a letter indicating the paragraph.

The difficulty with reference to Stirrings Still is that some paragraphs are too long to serve as a workable unit. Charles Krance therefore employs a flexible criterion to divide the text into segments or units: The length of these discrete units of text varies, depending on the degree of complexity and/or extent of their variants and revisions (xiv). If a paragraph is divided into n segments, each of these can be indicated by means of the seg tag. In order to indicate where a particular segment in a manuscript eventually ended up in the published version of Stirrings Still, the id attribute in the seg tag may be used. This way, all the segments indicated with e.g. the id SS1.03.05 (i.e., all the units corresponding to the fifth segment of the third paragraph of Part One in the published text) can be arranged in vertical juxtaposition. Apart from these segments, the texts can also be compared at p and div level, as these are also encoded with an id (e.g. segment SS1.03.05 is part of paragraph SS1.03 and of section or div SS1).

The main danger of a teleological perspective is the neglect of passages (especially in the early manuscripts) that did not make it into the published text. Nonetheless, there may be several versions of these passages, so that they should be comparable too. For instance, the first paragraph of the first few versions starts with the phrase Tout toujours à la même distance, translated as All always at the same remove. After a few versions, Beckett abandoned this path in the writing process, so that it remained a dead end. By taking the last stage in such a cul-de-sac as a reference text for the division into segments, marking them with a special initial number (distinct from section numbers 1, 2, and 3), these dead ends can be mapped as well, so as to show how Beckett's famous motto I can't go on, I'll go on also applies to the textual process.

Form of textual apparatus

Traditionally the notion of variants applies to either variation between copies of ancient or medieval documents by scribes, or between different editions of the same work. When dealing with modern texts, a distinction has to be made between these transmission variants and genetic (or composition) variants. Moreover, the edition of bilingual works requires an extra category of translation variants.

1. For transmission variants, i.e., variants between published versions of the text, the special section of the Guidelines for Electronic Text Encoding and Interchange TEI P4 devoted to the apparatus can be applied. Chapter 19 offers three methods. Since the Guidelines advise to avoid the location-referenced method [w]here it is intended that the apparatus be complete enough to allow the reconstruction of the witnesses and since the double end-point attachment method is lengthy and difficult, the so-called parallel segmentation method is the most suitable in this case.

2. Genetic variants: The difference between genetic and textual criticism, according to Daniel Ferrer, is connected to their respective focus on invention and repetition. Recognizing that this is too much of a black-and-white division, he immediately qualifies this statement and emphasizes the dialectic of repetition and invention (Ferrer 54) characterizing any writing process. In this context, Brian Fitch's characterization of Beckett's bilingual work as a double form of répétition is particularly relevant: One might say that while the first version is no more than a rehearsal for what is yet to come, the second is but a repetition of what has gone before, the two concepts coming together in the one French word répétition (Fitch 157).

The advantage of a synoptic apparatus is its focus on what has changed between versions, or in Ferrer's formulation: the invention. But in order to represent the dialectics of invention and repetition, it may be useful to present the whole textual context (i.e., also the repetitions). A traditional apparatus is usually designed for transmission variants in versions that are characterized by more identity than variation. In modern manuscripts, especially in the earliest draft stages, this proportion is often inverse. According to the French school of genetic criticism a genetic edition in the true sense of the word presents the complete texts of all the documents in chronological order and in their entirety. But evidently an important concern of the genetic edition of prose texts is not to overburden the user with a mass of texts. By offering the user the possibility to adapt the size of the textual unit s/he wishes to compare (large, medium, small — i.e. the unit of the section div, the paragraph p, or the sentence seg) the result is already a refined form of versioning. But it is possible to go further and make it into a critical genetic edition, which implies that the editor indicates the genetic variants explicitly.

Except in the case of the first extant version, the previous version can always serve as temporary invariant against which the genetic variants can be measured, even if the writing was eventually aborted and never published. This could be called retrospective collation. If the work did reach the stage of publication, it is only natural that readers will also be interested in another kind of genetic variants: the difference between a version and the published text. This teleological perspective requires a prospective collation. In the XML encoding, therefore, the genetic variants are marked in two directions by means of rdg tags with a type attribute: retro for differences vis-à-vis the previous version and pro for differences vis-à-vis the edited reading text. The code can be visualized in different ways; for instance on the smallest level (seg) all versions of one single sentence can be displayed in vertical juxtaposition with the retrospective variants in italics and the prospective variants in bold.

This chronological arrangement is suitable for a collation of versions in their order of composition. If the user wishes to compare two not necessarily consecutive versions of a particular sentence, s/he can select these two versions from the chronological sequence and focus on just these two in parallel frames, which facilitates comparison.

Because this part of the edition tries to offer a teleological perspective, the absence of a word or word string that appears in the published text is indicated by a rend attribute, mentioning absence: rdg type="pro" rend="absence". The absence can for instance be visualized by means of a vertical bar. This genetic approach has the advantage that the number of diacritical signs can be reduced to a minimum, and that variants can be indicated in their context, leaving the structure of the whole sentence intact. For bilingual works, the same system can be expanded (using for instance a type attribute trans to indicate variants in the authorial translation).

3. Translation variants: Variants in the translation further complicate the distinction between genetic and transmission variants. In theory, there is no difference between making a self-translation of a text that is already out in the open and translating a draft, since in both cases the author translates his own texts. But inevitably the act of publication has a petrifying effect. Rainier Grutman makes a distinction between simultaneous auto-translation and delayed auto-translation (Grutman 20). In the case of Murphy, for instance, the English version was published before the Second World War, but the French translation appeared only a decade later. Because of this delay, the English text had already led a public life, which may have limited the possibilities of further invention.

But publication or even performance (in the case of his plays) did not prevent Beckett from introducing considerable variants in the translation. In extreme cases, this may raise the question whether a self-translation should be regarded as a separate work or as a version of the same work. Klaus Gerlach argues in favor of the latter option, suggesting to enlarge Siegfried Scheibe's definition of version to include the notion of equivalence: Textual versions ... are related through textual identity [and equivalence] and distinct through variation (Scheibe 207; emphasized passage added DVH). In translation studies, however, the notion of equivalence is central, but also controversial, as Dorothy Kenny notes (77). The concept's problematic nature immediately appears in a bilingual edition. Non-equivalent instances are called mismatches by Magessa O'Reilly. Words or word strings which have no match in the facing text are underlined; a vertical bar indicates the position that would most likely have been occupied by the matching segment, were it present (O'Reilly xiii). O'Reilly readily acknowledges that these are obviously not the only instances of non-equivalence. He mentions shifts of verb tense, of singular and plural, of person or register, etc., and adds that it would be impossible to point out such an endless array of mismatchings (xiii).

Analogous to the pro- and retrospective genetic variants another type attribute value can be used to indicate translation variants. This has some theoretical implications, for it means that the publication of an authorial translation is not regarded as the end of a genesis, but rather as the continuation of the writing process after publication. In the Translation approach (see number 4 supra) the English source text Stirrings Still and the French target text Soubresauts are each other's versions, comparable to all the other versions at div, p, and seg level.

What this genetic bilingual edition wants to emphasize is that the creative power of translation continues to be operative after the first publication, to the effect that the bon à tirer moment is less decisive than in non-bilingual works. Beckett was well aware of what Jerome McGann calls the algorithmic character of traditional text (McGann, Radiant Textuality 151): text generates text and in Beckett's case, translation played a crucial rule in the exploitation of this self-generative power of texts. Authorial translations give evidence of an enhanced textual awareness. As a consequence, their textual examination and scholarly editing are a crucial part of their critical interpretation.