Is It Joyce We Are Reading? Non-Fiction, Authorship, and Digital Humanities

Abstract

This chapter reflects on possible updates to Barry’s Occasional, Critical, and Political Writing in relation to the discovery of new manuscript materials, in particular the Paris and Pola notebook; the restrictions imposed on publication by the James Joyce Estate; probable errors of attribution; and the reconsideration of unpublished English translations of Joyce’s Italian journalism in the James Joyce Archive. Finally, it makes use of new methods of computer-based natural language analysis to offer a definitive statement on the corpus of Joyce’s non-fictional writing, especially in relation to the contested essay “Politics and Cattle Disease.”

The ca non of Ja mes Joyce’s non-fictional writings has been at once a subject of dispute, an awkward moment in the history of literary publishing, and an opportunity for new editorial responses to recently discovered Joyce materials. The Critical Writings, edited by Richard Ellmann and Ellsworth Mason (1959), and the Occasional, Critical, and Political Writing , edited by Kevin Barry (2000), provide readers with a set of texts that requires updating. The discovery of new manuscript materials, in particular an “Ea rly Commonplace Book”; the restrictions imposed on publication by the James Jo yce Estate; probable errors of attribution; and the reconsideration of unpublished Eng lish translation s of Joyc e’s Italia n journalism in the James Joyce Archive—all of these issues justify a new testing of what might be included in our definition of Joyce’s corpus of non-fi ctional writing. This chapter will identify the relevant range of problematic texts, outline the difficulties for readers who seek access to a complete and reliable collection of Joyce’s non-fiction, and answer some questions of attribution using new methods of computer-based natural language analysis.

Research in computer-based natu ral language processing has revealed that particular authors have characteristic patterns in their use of language. What is surprising about these patterns is that they are expressed most strongly in the sequences of characters that authors use, rather than the words that they use. That is to say, any particular author can be considered to have a signature in terms of the frequency of letter combinations that they habitually employ, known as character n-grams, where n is a number. By computationally analyzing a large corpus of known works by a particular author and identifying their characteristic patterns of 3-grams, 4-grams, and 5-grams, we can create a n-gram signature of the author. We can then compare the n-gram signature of works whose authorship is disputed with the known signature of the proposed author to identify the probability of the authorship of the disputed work. In this chapter we apply character n-gram analysis to several non-fictional items that may have been written by Joyce, and compare the character n-gram of these items to known works of a similar kind by Joyce, on the one hand, and to a selection of comparable contemporary writing on the other.

This analysis provides valuable objective evidence as to the authorship of several disputed texts. For example, an essay heretofore included in the Joyce canon, “Politi cs and Cattle Disease,” has been in dispute since 2007, when Terence Matthews published persuasive circumstantial evidence that supports questions about Joyce’s authorship and the provenance of the text.¹ Furthermore, Joyce proposed to an Italian publisher in 1914 the publication in book form of a revised arrangement of the essays he contributed to Il Piccolo della Sera from 1907 to 1912. A typescript of English translation s of these revised essays exists in the James Joyce Archive, and suggests an intended English p ublication by Joyce of the same materials. To bring us up to date: Hans Walter Ga bler first attributed these transl ations to Stanislau s Joyce,² while Giorgio Mel chiori cast doubt on that attribution.³ Gabl er has since returned to the texts and proposed an alternative to the binary Joyce/not-Joyce model: a collaborative project in which James Joyce played no small part.⁴ In the present project, n-gram character analysis will provide new evidence towards establishing probable authorship of these transla tions, and offer a quantitative, evaluative basis for estimating the extent to which these translation s of Joyce’s Italian can be said to be themselves Joycean, non-Joycean, or part-Joycean.

This chapter approaches the ca non of James Joyce’s non-fiction in two distinct ways and is therefore divided into two separate sections: first, an argument in favour of a more coherent and complete publication of a full corpus of relevant texts; and second, an analysis of certain doubtful texts that may or may not have a place in that corpus of non-fiction. The purpose of the chapter is to bring together editorial scholarship (which derives from the work of Kevin Barry as editor of Joyce’s non-fictional wri tings) with new approaches in digital humanities (which derive from recent innovations in computer science by Kevin Feeney, Gavin Mendel-Gleason, and Bojan Božić) in order to redefine the ways in which Joyce’s non-fiction is categorised, arranged, and studied.

Towards a Joycean Corpus

It is extraordinary how poorly served are the innumerable students who read James Joyce in English. Those who opt to read him in French, for example, are far better off. They can beg, borrow, or purchase a two-volume set under the imprint of the Bibliothèque de la Pléiade.⁵ With these in hand, the reader has access to almost all of Joyce’s writi ngs. Moreover, published respectively in 1982 and 1995, both volumes remain in print. Of most relevance to readers of Joyce’s non-fiction is Volume I, with its generous list of contents and more than 2000 pages of text, including a full range of Joyce’s poe try; the complete set of the epiphanies; Dubliners; the 1904 essay “Portrait o f the Artist”; St ephen Hero; A Portrait of the Artist as a Young Man; “Giaco mo Joyce”; Exi les; Joyce’s jo urnalism, essays, and lectures, along with a selection of his notes on English dramatists and his translat ions of foreign-language authors; a large number of le tters; plus analyses of his own and No ra’s dreams. The English-language reader who wishes to become familiar with such a range of Joyce’s writi ngs must search out a long list of books from many publishers: Penguin for the poems and Ex iles; Oxford for the journa lism, essays, and lectures; Faber (with Poems and Shorter Writings) for the epiphanies, “Giacomo Joyce,” and the 1904 essay “Portrait of the Artist”; and Garland’s sixty-three volumes of The James Joyce Archive (1977–1980) for the notes on English dramatists, the translation s of foreign-language authors, and the analyses of dreams. Of that long list of materials, some are out of print and not a few are beyond the means or the reach of the ordinary student. In all events, the English-language reader of Joyce needs a coat with large pockets, a gift for trawling the internet, a portable scanner, some airline tickets, and a robust credit card. While American and Canadian readers will soon have access, through Oxford University Press, to a further 1800 of Joyce’s le tters, postcards, and telegrams, their counterparts in Europe will continue to rely on Faber, once again, for access to roughly half of Joyce’s extant correspondence.⁶

I have no explanation for the fact that the James Joyce Est ate issued one kind of licence for the French market and so many other, more restrictive licences for a much larger market: those who read Joyce in his own words. The consequences are especially serious for any edited collection of Joyce ephemera. By their nature, such writings survive in disparate locations: notebooks, newsp apers, fair copies, incomplete typescripts. They include published and unpublished writings. Unpublished items comprise materials never intended for publication, for example “The Centenary of Ch arles Dickens,” as well as materials that Joyce may have intended to publish, for example the typescripts of English-language trans lations of his Triestine news paper articles. The task of an editor of such diverse kinds of ephemera is to provide access to the most complete range of texts that are known to have survived. My experience as editor of the Occasio nal, Critical, and Political Writings made it clear that no such complete edition would be allowed, and that students would need to find their way to very many other books if they wished to read the full variety of Joyce’s non-fictional writin gs.

In the first instance my contract had been with Penguin, following an invitation from the general editor of the Penguin Joyce series, Seamus Deane. During the preparation of galley sheets, the Joyce Estate withdrew permission for that collection of occasional writings. The only explanation that I could discover was the Estate’s irritation with Penguin for having published in a single volume both Joyce’s po ems and the play Exil es, which the Estate had wished to see published in separate volumes of the series. Matters remained in suspense until, following a report in the Irish Times of a lecture I had given at the James Joyce Summer School in July 1995, I received a phone call from Stephen Joyce offering the possibility that Oxford University Press might be permitted to publish the collection.⁷ What was to be excluded and included changed with the change of publisher, and was determined as follows: items for which Faber held licences (for example, the epiphanies, “Giaco mo Joyce,” and the 1904 essay “Portrait o f the Artist”) would not be permitted; the Italian originals of articles from Il Piccolo della Sera must accompany new translati ons; the transl ations of these newsp aper articles that already existed as typescripts in the James Joyce Archive were not allowed, nor indeed was there permission to include any materials that were solely available in the Archive (for example, Joyce’s notes on English dramatists, and his analyses of drea ms). In other words, the range of permissions given to the Pléia de edition would not be made available to Oxford. Furthermore, this Oxford edition of the Occasional, Critical, and Political Writings must not publish any notice or complaint about these restrictions.

Whatever their differences of scale and completeness, all these editions of Joyce in English and in French have one trait in common: they predate the discovery in 2002 of what is now called the “Early Common place Book”: 1903–1912. Previous to its arrival in the manuscript collection of the National Library of Ireland (NLI MS 36,639/02/A), this inexpensive school copybook, from the Papeterie-Imprimerie F. BÉNARD, had been known only through some famous transcriptions that appeared in Herbert Gorman ’s 1939 biography of Joyce, of certain items that Joyce himself had reworked in the discussions of beauty and perception in A Portrai t of the Artist as a Young Man. All editions of Joyce’s wri tings on art (The Workshop of Daedalus, The Critical Writings, the Occasional, Critical, and Political Writings ) had followed Gor man’s transcriptions.⁸ The manuscript was usually referred to not as one but as two items, “The Paris Notebook” and “The Pola Notebook,” whereas it is a single copybook recording Joyce’s accounts; notes from his reading of, among others, Aristotle, Aqu inas, and Ben Jonson; booklists transcribed from the United Irishman ; notes on Arab musical notation; notes for Step hen Hero; and Thomistic definitions, signed and dated by Joyce himself, on art, beauty, and perception. It is these definitions that were more or less inadequately transcribed by Gorm an. The copybook, rich in its structure and diversity, was revisited by Joyce as late as 1912.⁹ For anyone who reads the early comm onplace book alongside the transcriptions made from it by Gorm an, what is especially striking is the persistence with which Joyce pursues in the pages of this copybook a personally defined aesthetic begun on 13 February 1903 and still in process some twenty months later. The persistence of Joyce’s attention is less visible in Gorman ’s transcriptions, not least because those ignored the extent of Joyce’s revisions: for example, his cancelling a paragraph that fails to satisfy him, on 15 November 1904, only to replace it with another, “The Act of Apprehension,” entirely rephrased, autographed, and dated the following day.¹⁰ Subsequent to the 2002 discovery of the early commo nplace book, we have access to the 21-year-old Joyce’s day-to-day accounts, including his notice of receipts from the Daily Express and the Irish Times, and also to his copying out of booklists of Irish culture and history from the pages of the United Irishman of 1903. I now share with other readers of the early co mmonplace book a stronger awareness of Joyce’s edgy, early, and unbroken conversation with varieties of Irish political discourse.¹¹ Joyce’s systematic omission from his transcriptions of the United Irishman booklists of titles in Gaelic is part of the edginess of that conversation.

At a single stroke, the decision by the NLI t o publish online its collection of Joyce manuscripts (the early commonpl ace book can be found at http://catalogue.nli.ie/Record/vtls000356987/HierarchyTree#page/1/mode/1up) has transformed the landscape for readers of Joyce’s non-fictional writings.¹² It is now possible for most readers to have access to an impeccable ur-text of materials that are essentially unfinished, and also to witness Joyce’s “ephemeral” acts of self-revision and textual construction. In this landscape, some of the textual content of my edition of the Occasional, Critical, and Political Writings becomes wonderfully out of date, and will be all the more so when other libraries follow the example of the N LI and publish their Joyce collections online. However, the lasting value of the edition will lie less in the selected non-fictional writings it makes available to students of Joyce than in its mapping of Joyce’s omnifarious reading, the heterogeneous material out of which he made fresh texts: his reading in Rich ard Wagner, in Tim Harrington, in Susannah Centlivre, in Ernest Renan, in Sinn Féin/ United Irishman , in the Comte de Montalembert, in James Hardiman, in Bernard Bosanquet, and in Richard J. Kelly, author of the pamphlet on Galway as a Transatlantic Port that Joyce deciphers in the twilight as his ferry returns from Aranmore.¹³ The sources are inexhaustible and to map their outlines is to display the extraordinary opportunism of what, for lack of a better word, we can call Joyce’s “method”: writ ing even non-fiction from the ready-made words of others.

Given the restraints imposed by the Joyce Es tate on what I could include in the Occasional, Critical, and Political Writings , it is all the more frustrating to me now that my collection both includes materials of doubtful attribution, and excludes others of hitherto uncertain provenance. Chief among these are Go rman’s faulty transcriptions of the early com monplace book to which most readers can now have full access by consulting the manuscript online at the NL I. It is timely to reconsider other inclusions and exclusions: in particular the probable error identified by Terence Matthews of my allowing the 1912 essay on “Politics an d Cattle Disease” to remain part of the Joycean canon,¹⁴ and the lingering uncertainty about the provenance and authorship of the typescripts, gathered together in the James Joyce Archive, of English-language trans lations of Joyce’s Italian journalism.¹⁵

The drama of who might be the author, or authors, of the English-language translation s of Joyce’s Italian j ournalism has developed over many decades. Hans Walter Gab ler first proposed in 1978 that Stanisl aus Joyce might have taken a significant role in the translation s.¹⁶ In 1985, Giorgio Melchi ori played down the possibility that Stanis laus was the translator, while allowing that he “may have been involved in this job.”¹⁷ Melchio ri subsequently discovered that most of the typescripts of these English translation s “bear penciled figures on their first pages,” and convincingly argued that this sequence of numbers corresponded to a restructuring of the translat ed essays for publication. Joyce may well have intended, after the failed effort to publish the Italian essays as a single volume with the Genoese publisher Angelo Fortunato Fo rmiggini, “to address this time an English audience.”¹⁸ Evidence that Joyce intended to publish for an English-speaking readership a collection of essays under the title “Irelan d at the Bar” lends further importance to the question of who is the author of these English translation s. In 2004, Hans Walter Gabler revisited the issue and vividly demonstrated the dilemma of authorship: he began with the phrasing in Italian and English of Joyce’s article on Oscar Wi lde (“Oscar Wilde: il poeta di ‘Salomé,’” Il Piccolo della Sera, 24 March 1909).

The Italian text is as cadentially rhythmical in the article’s last sentence: “Il futuro potrà forse scolpire là un altro verso, meno altiero, più pietoso: “Partiti sunt sibi vestimenta mea…”, etc. The rhythm is virtually identical in Italian and in English: “The future perhaps will carve there another line, less proud and more piteous.” Or, let us take the opening of the last paragraph: “Nell’ultimo suo libro ‘De Profundis’ si inchina davanti ad un Cristo gnostico, risorto dalle pagine apocrife della ‘Casa del Melagrani’ ed allora sua vera anima, tremula, timida e rattristata, traluce attraverso il manto di Eliogabalo.” “…allora sua vera anima, tremula, timida e rattristata, traluce attraverso il manto di Eliogabalo”: we know that this is Joyce’s own text, he wrote the Italian contributions to Il Piccolo della Sera […] But who, we must stop to wonder, wrote the English equivalent: “and then his real soul, tremulous, timorous and sorrowful, shines through the mantle of Heliogabalus”; and who speaks of the future line on Wilde ’s gravestone, “meno altiero, più pietoso” as “less proud, more piteous” (for all that it is the cognate of “pietoso”, “piteous” is very much an adjective in the style of Joyce, the artist as a young man)? Who did truly write and rhythmicize in English the peroration to “Irela nd at the Bar” […], giving it not only its irresistible rhythmic drive, but also varying the double “i messi” … “i messi” of the Italian to “the messengers” … “the envoy”? Who, similarly, was responsible for driving home the article on “The Shadow of Parnell” with “In his intimate fiery appeal to his nation he implored his countrymen not to throw him to the English wolves who were howling around him. It redounds to the honour of his countrymen that they did not fail him at that desperate appeal. They did not throw him to the English wolves; they tore him to pieces themselves.” And who was it that worded Dublin ’s readying itself for the Annual Horse Show: “La città … si veste da sposa novella” as “the … town arrays itself as for a bridal”? Or who decided, for the essay “The C ity of the Tribes”, and in talking about the house of the Lynches in the central street of Galway  – “il triste e scuro castello che ancora nereggia nelle via principale” – to abandon the Italian description of it as “blackening the street” and instead to say “a bleak, dark castle which still stands a black mass in the main street”?¹⁹

Hans Walter Ga bler’s alert ear and intertextual reading of turns of phrase, idiom, and rhythm that carry a Joycean trace must be placed against an awareness that there is much that is maladroit and inexpert across the full range of these translation s. Indeed, it is not Gab ler’s argument that James Joyce is the author of these texts. However, he wishes to do more than persuade us that the translation s are a collaborative project among members of Joyce’s Triestine circle. The phrasing and rhythms he discovers persuade him that James Joyce had a significant role in the writing of these texts, which “are very likely related genuinely to the Joycean oeuvre.”²⁰ That is, they might be placed in a middle position between Joyce and non-Joyce, and stand instead in the position of a partial or collaborative Joyce.

Separating the Sheep from the Goats

Our task, then, is to confront questions about the authorship of two distinct bodies of text in the corpus of Joyce’s non-fictional wr itings:

• 1. The typescript translation s of James Joyce’s Ita lian journal ism, gathered together in the James Joyce Archive and probably intended by Joyce for publication in book form for an English readership.

• 2. The essay “Politic s and Cattle Disease,” published in the Freeman’s Journal, 10 September 1912, and traditionally attributed to James Joyce on circumstantial grounds: in particular, a le tter of 6 September 1912 from Charles Joyce to Stanisla us Joyce that states that James Joyce wrote a sub-editorial on foot and mouth disease for the Freeman; and also the appearance beneath the article of a note on Joyce’s essay on the Galway Harbour Scheme, “The Mirag e of the Fisherman of Aran.” The attribution of the essay to James Joyce has been called into question by Terence Matthews.²¹

Authorship attribution supported by statistical or computational methods has a long history dating back to the nineteenth century.²² In 1964, Frederick Mosteller and David L. Wallace carried out seminal work on the authorship of the “Federalist Papers” that would prove the value of these approaches and provide the base for modern computational approaches to author attribution for short texts or texts by many hands.²³ The fifty years since then have seen a marked increase in the potential for these kinds of analyses. Firstly, the expanding availability of substantial digital literary corpora has improved the quality of statistical material being input into these analyses—simply put, there are more texts available in a form that is ready for computational analysis.²⁴ And secondly, the development of new, high-quality machine learning techniques has expanded our ability to consume and analyse this new data quickly and precisely.²⁵

In order to assess the authorship of the disputed Joycean texts, on objective internal rather than interpretative or circumstantial evidence, a team of computer scientists in Trinity College Dublin (Kevin Feeney, Gavin Mendel-Gleason, and Bojan Božić) used methods of analysis based on research in computer-based natural la nguage processing.²⁶ This research shows that particular authors have characteristic patterns, expressed most strongly in the sequences of characters that authors use, rather than the words that they use. The signature of a particular author can be considered in terms of the frequency of letter combinations that they habitually employ, known as character n-grams, where n is a number. In this analysis we produce an n-gram signature for the works that have a disputed authorship, and compare this with the n-gram signature of those works that are not in any dispute. This comparison allows us to generate a simple statistical probability (from 0 to 1) that the disputed work shares authorship with the works of undisputed authorship.

We used supervised learning to predict the output variable from a training set of existing text segments. Training sets consist of a corpus of texts of a known author. For this natural language processing problem, we developed a novel method in machine learning. The chosen supervised learning method is called “support vector machines” and can be used for classification, regression, and detection of outliers (that is, items differing from all other elements of a set). In this case, the problem can be considered a classification problem: we want to analyse text segments and predict the author according to one of the pre-defined labels of “James Joyce,” “Stanisla us Joyce,” or “other.”

The “support vector machines” approach has the advantage that it is memory efficient, as it uses a subset of training points in the decision function. It has been successfully applied to the question of authorship attribution in numerous domains.²⁷ For the purposes of this project, the scikit-learn²⁸ Python implementation was used and the results were gathered by comparing SVC (support ve ctor classification), Linear SVC, and NuSVC predictions. First the classifier was set up with training data consisting of text segments, which represent the features, and corresponding authors, then the function produces the score, cross-validation mean, and variance, which can finally be used for segment classification. The clas sification itself is defined by the prediction that is the result of using the support vector classifier on text segments from one or multiple documents with an unknown author.

Our training set consists of a corpus with segments from known James Joyce works of:

A::

Non-Fiction (“Trust Not Appearances,” “Subjugation,” “The Stu dy of Languages,” Royal Hibernian Academy: “Ecce Homo,” “Drama an d Life,” “Ibsen’s New Drama,” “The Day of the Rabblement,” “James Clarence Mangan (1902),” “An Irish Poet,” “George Meredith,” “Today and Tomorrow in Ireland,” “The Soul of Ireland,” “New Fiction,” “A French Religious Novel,” “The Bruno Philosophy,” “Shakespeare Explained,” “A Curious History,” “The Centenary of Charl es Dickens,” “Programme Notes for the English Players,” and “From a Banned Writer to a Banned Singer”).

B::

Fiction (Dubliners and A Portrait of the Artist as a Young Man).

The known segments from Stanislau s Joyce’s works that were used are The Dublin Diary of Stanislaus Joyce (1962) and My Brother’s Keeper (1958).

The literature segments that are known to be not-Joyce were the texts of Donahoe’s Magazine (c. 1878–1908); The Jungle Book (1894), The King in Yellow (1895), and An English Garner (1877–1909). These texts were selected as they fulfilled several criteria: they were readily available in electronic form; they include multiple contemporaneous authors who wrote in the same style as the Joyces; they include a similar spread of article types, themes, colonial voices, and various other attributes. The goal is to make this corpus as broadly similar to the Joyce corpora as possible.

We have used the previously described learning algorithm and training sets to answer the following questions:

• 1. Are the translation s of James Joyce Italian articles by James Joyce, Stan islaus Joyce, or neither?

• 2. Is the essay “Politics a nd Cattle Disease” written by James Joyce?

In answer to Question 2, the data clearly indicates that James Joyce is not the author of “Politics and C attle Disease”: the predicted probability that the author was not James Joyce turns out to be 99.0677546157%. Such an overwhelming outcome from the analysis confirms Terence Matthews’s assertion, based on the grounds of style and context, that “not only is there no reason to suppose that Joyce wrote ‘Politics and Cattle Disease’ on 6 September, 1912 as claimed by Mas on and Ellm ann, but we can now say definitively that he did not.”²⁹ Readers of Ulysses c an now more thoroughly relish Stephen’s suspicion that “Mulligan will dub me a new name: the bullockbefriending bard” (U, 2.431).

In answer to Question 1:

• a) The data clearly indicates that there is almost no probability that Stanis laus Joyce is the author of the translation s of James Joyce’s Ital ian journ alism. This is based on the predicted probability of 97–99% across a range of transl ations that the author was not Stani slaus.

• b) We can draw significant conclusions from the data and propose that there exists a material connection between the translation s of James Joyce’s articles and his known wri tings. The analyses of the data are gathered in tabular form in Table 5.1. The headings A: Non-Fiction and B: Fiction refer to the two training sets of known James Joyce writings that we have used. The heading A+B: Fiction & Non-Fiction refers to a combination of both training sets. Table 5.1 can be read as follows: results under either Column A: Non-Fiction or Column B: Fiction provide the similarity of the tested translation to, respectively, that genre of wri ting by James Joyce. The combination in Column A + B: Fiction and Non-Fiction provides more certainty, as it provides additional and broader data, correcting and improving the analyses. A predicted probability of 0.5 or greater would identify authorship.

Table 5.1 Probability of James Joyce authorship with varying training sets

The data of Table 5.1 is set out in ascending order of probability, and suggests that two of the articles (“The Home Ru le Comet” and “Home Ru le Comes of Age”) are not translat ed by James Joyce. Seven of the articles (“Oscar Wilde,” “Ireland at the B ar,” “Fenianism ,” “The Battle b etween Bernard Shaw and the Censor,” “The Mirage of the Fisherman of Aran,” “The City of the Tribes,” and “The Shadow of Parnell”) fall within or close to categorisation as the work of James Joyce. As such, the analysis (while setting at naught previous notions that Stan islaus had a hand in the translation s) provides strong support for Hans Walter G abler’s more recent hypothesis that “the unpublished translation fragments, while collaborative, are very likely related genuinely to the Joycean oeuvre.”³⁰

The results described in this chapter are significant not only for our engagement with James Joyce’s non-fictional w ritings, but also because they demonstrate how modern computational techniques can be used to help resolve significant questions of literary authorship attribution. These techniques are important, because they use objective statistical measures to identify literary styles and produce testable, verifiable probabilistic predictions. Although such predictions are seldom definitive, they are accurate as long as the corpora are genuinely representative of the authors who are being compared. When the computer tells us that a given text is more than 99% likely to be written by “not-Joyce” and less than 1% likely to have been written by Joyce—as it has in relation to “Politics and Catt le Disease”—this is strong evidence based on a purely objective analysis of the corpora, and can be considered accurate barring significant representative failings in the corpora.

These results can immediately help in constructing a more accurate and objective account of the Joycean canon. It may be that within the domain of Joyce’s non-fictional writi ngs, where he employs a limited and largely consistent set of argumentative, wittily allusive, and acidic styles, the task at hand is the attribution of authorship. Beyond this set of occasional writings, some scholars may be interested to test the resources of such computational and statistical methods, as are employed here, for a more ambitious analysis of whether James Joyce, i n Ulysses and Finnegans Wake, achieved the possibility of being in one single text more than one author. For the more modest purposes of this collection of essays, however, it is sufficient to have provided new evidence of how the English-language reader of James Joyce will be better served if an accurate, diverse, and complete edition of his non-fictional writi ngs and translation s is published in a single volume or on a dedicated website.