GENETIC JOYCE STUDIES - Issue 14 (Spring 2014)


Finn’s Hotel and the Joycean Canon

James O’Sullivan


Ithys Press controversially published Finn’s Hotel in June 2013, describing the collection as “almost certainly the last unpublished title by James Joyce”.[1] Edited by Danis Rose, the publication divides opinion amongst Joyce scholars, and considering the profile of the author, this debate has, thus far, been played out in the public domain. Ithys and Rose contend that the fragments warrant consideration as a standalone collection, the style in which they are written suggesting that such was Joyce’s intention: “The prose pieces of Finn’s Hotel … are written in a unique diversity of styles, much more so than Ulysses. Taken together, they form the true and hitherto unknown precursor to the multi-modulated voices of the Wake—but these first utterings from Finn’s Hotel are far easier to understand.”[2] This view is not unanimously accepted, with some scholars countering that the writings are merely early drafts for what would later become Finnegans Wake, and thus should not be published as an independent addition to the Joycean canon. Terence Killeen notes, in The Irish Times, that “the pieces scream of Finnegans Wake itself”.[3] He states: “It is true that one or two of them did not end up in the final text, but it is quite normal for a writer to draft and then abandon various passages in the initial stages of a major work.”[4] The debate revolves around Joyce’s authorial intentions, which Rose believes are evident when one considers chronology: “Joyce wrote these pieces in 1923 some six months after his disengagement from writing his encyclopaedic Ulysses and well before he had as yet conceived of the plot, structure, or sheer immensity of his epic Finnegans Wake.”[5] Rose claims that Joyce’s correspondence demonstrates that the stories can be considered a “special set”, and that they remain “absolutely distinct” from Finnegans Wake, which at the time, he argues, Joyce had not even conceptualised.[6] “In a sense, Finn's Hotel is to Finnegans Wake as one of the Dubliners stories are to Ulysses. They are not the same, but thematically intertextualised,” he tells The Guardian.[7] Killeen views this as insufficient justification for a standalone edition: “Although Joyce may have toyed with the name Finn’s Hotel as a possible title for the final work, there is no evidence that he ever envisaged these pieces as a separate publication, and his intentions should surely be respected by any serious scholar.”[8] Derek Attridge supports Killeen: “Rose’s theory, which has not, to my knowledge, been accepted by any other Joyce scholar, is that these pieces were intended as a collection of stories entitled Finn's Hotel. Unless Rose has now come up with some substantial new evidence that this ‘work’ is not an editor's fantasy, publication can only distort and damage Joyce's reputation.”[9] John Nash also criticises the decision to publish the collection, remarking that Rose’s circumstantial evidence, found in correspondence between Joyce and Harriet Shaw Weaver, remains largely without credit: “They are certainly interesting transitional pieces but not a 'short story cycle' like Dubliners.”[10]

            While there is some limited discussion of the relevant manuscripts in scholarly publications, the majority of such pre-dates the Ithys edition. Undoubtedly, further research will surface in the near future, but for now, much of the debate is presented in the aforementioned public forums. Rose’s editorial decisions have long been the concern of critics, his intentions in relation to the Finn’s Hotel manuscripts initially questioned some three decades ago. In 1998, Sidney Feshbach branded Rose’s edition of Ulysses as “ridiculous”; his remarks on the supposed emergence of Finn’s Hotel from the Finnegans Wake archive being of a similar sentiment: “His idea is no worse than the claim that Giacomo Joyce is a novel. A rumor was that Rose wanted to publish the archivist’s ‘discovery’ as though it were by James Joyce with Rose seated at Joyce’s elbow.”[11] Much of the debate is in a similar vein, with the argument founded upon a seeming lack of any significant circumstantial or literary evidence. This very lack is seen by many as sufficient justification for the common consensus that Finn’s Hotel should not have been published as a standalone collection. Hans Walter Gabler aptly summarises the issue in his review of Rose’s The Textual Diaries of James Joyce. He notes, in advance of the critical edition substantiating Rose’s claims, that scholars are “challenged to accept on faith” the “autonomy” of Finn’s Hotel.[12] Some two decades later, it would seem that the majority of scholars remain unconvinced, while the resolve of Rose appears only to have strengthened. Engaging with the scant evidence that does exist would serve little purpose here. Firstly, I do not yet pretend to possess a working knowledge of Joyce sufficient to enter into a discourse with either Rose or his critics. Secondly, I do not think there is sufficient evidence for such a debate to even be carried out in an appropriately scholarly fashion. From what I have seen presented, there is simply insufficient material to argue, conclusively, what the intentions of Joyce might have been. Regardless of the credentials of the scholar, Joyce is not an author to be second-guessed. In the absence of significant circumstantial evidence, we must seek alternative methods of determining the place of Finn’s Hotel within the Joycean canon.

            It is only on rare occasions that style fails to provide a marker from which we can discern the position of a text within a corpus. Thus, it is to style that I look in my effort to contribute something novel to the Finn’s Hotel debate. For this task I adopt various computational methods, all of which have been demonstrated by literary scholars to be accurate indicators of authorial signatures. The purpose of this approach is not to prove beyond doubt the position of Finn’s Hotel within Joyce’s oeuvre, but rather, to add some quantitative data to the discussion, so that the debate may proceed with evidence that is statistically significant. In the interest of transparency, I should point out that the electronic edition of Finn’s Hotel used for my analysis was kindly provided by Ithys Press, with the permission of Rose. The cost of the edition puts it beyond the reach of many scholars, and thus I am grateful that they saw it fit to furnish me with a digital version. Regardless of how one views his editorial decisions, Rose cannot be accused of standing in the way of critical debate. The remaining texts came from my private collection; the digital files necessary for this approach were prepared using optical character recognition, which converts scanned material into machine-encoded text. Ulysses is the Bodley Head edition, which uses Hans Walter Gabler’s 1993 revision, while the remaining texts are all from the Oxford World’s Classics series. Of course, as statistical sampling comes into play, and relative frequencies are used, the specific editions should not have a bearing on results. Computational stylistics is a means of analysing style using statistical methods. Armed with an authorial signature based on the most frequent words in a text, critics can establish the stylometric distance between texts. This approach is demonstrated as being highly effective by a variety of scholars, including, but not limited to, John Burrows,[13] David Hoover[14] and Matt Jockers,[15] as well as Jan Rybicki and Maciej Eder.[16] A variety of computational methods are used in this analysis, which will be outlined as they are presented. Analysis is conducted using the “stylo” library for the statistical package, R.[17] A free and open environment, R is a command-based software package designed for statistical computation and data visualisation. The purpose of my analysis is to form an authorial signature of Joyce’s oeuvre so that I might determine, at a macro-level, where it is that Finn’s Hotel resides within such a context. In doing so, I hope to demonstrate whether the collection is, from the perspective of style, quite distinct, or alternatively, closely aligned to Finnegans Wake. If style can be considered a determinant of what makes a text, then I believe that the results of such an analysis should be accepted as an indicator of whether Joyce intended Finn’s Hotel to be a standalone publication, or whether the relevant manuscripts are indeed the earliest incarnations of what would eventually come to be Finnegans Wake.

            Initially, I conduct a stylometric analysis of Dubliners, A Portrait of the Artist as a Young Man, Ulysses, Finnegans Wake, and Finn’s Hotel, using the relative frequencies of the 100 most frequent words in each text to form an authorial signature. The most frequent words in a text, typically function words, are proven to be accurate measures for classification by style. I opt for Classic Delta distance, a mathematical algorithm, outlined in several of the aforementioned articles, that is largely accepted as the approach most suited to conducting literary classifications based on most frequent words. While there is some disagreement on the issue, it is generally accepted that a text should be at least 5,000 words in length before its authorial signature can be considered statistically significant. In this particular instance, this requirement is, to an acceptable degree, satisfied.


Fig. 1. Joyce texts clustered by 100 MFW


When treated as complete texts, Finn’s Hotel and Finnegans Wake form a distinct cluster which is separated from Ulysses, A Portrait and Dubliners (Fig. 1). The clusters are represented here on a dendrogram, which visualises stylometric similarity based on the distance that one must traverse across the lines of the graph to reach one text from another. This dendrogram shows two distinct clusters, with Finn’s Hotel alongside, and thus similar in style, to Finnegans Wake.

            A criticism of the aforementioned results might be founded upon the fragmented nature of Joyce’s style. Thus, to demonstrate that Joyce’s authorial signature remains largely consistent across each of his texts, I conduct further analyses using separated episodes and chapters. However, the issue of length comes into play, Finn’s Hotel being only ~7,187 words. As can be seen as follows, the word counts of individual stories are arguably insufficient for the formation of statistically valid samples:


“The Irish Chinchinjoss”                     205

“Kindness to Fishes”                           132

“A Tale of a Tub”                                174

“Issy and the Dragon”                         520

“The Big Kiss”                                     1,087

“The Staves of Memory”                     537

“Skyward to Stardom”                         1,082

“The House of a Hundred Bottles”      593

“Here Comes Everybody”                   1,043

“Here’s Lettering You”                       1,814


Even when calculating results using relative frequencies, the length of samples is nonetheless significant. Research by Eder[18] shows that there is no conclusive answer to the question of sample size, but suggests that modern novels, irrespective of language, require a sample of 5,000 words. Eder’s seminal paper argues: “Until we develop more precise style-markers than word frequencies, we should be aware of some limits in our current approaches, the most troublesome of these being the limits of sample length. As I have tried to show, using 1,000-word samples will hardly provide a reliable result…”[19] The majority of texts in Dubliners are less than 5,000 words in length, but they are all greater than 1,000 words:


“The Sisters”                                        3,083

“An Encounter”                                   3,243

“Araby”                                               2,328

“Eveline”                                             1,819

“After the Race”                                  2,230

“Two Gallants”                                    3,902

“The Boarding House”                        2,632

“A Little Cloud”                                  4,907

“Counterparts”                                     4,086

“Clay”                                                 2,632

“A Painful Case”                                 3,618

“Ivy Day in the Committee Room”      5,204

“A Mother”                                          4,508

“Grace”                                               7,488

“The Dead”                                         15,603


While Eder outlines that 5,000-word texts ensure a more robust study, his suggestion that the lowest reliable threshold is 1,000 words appears to have been confirmed by the consistent clustering of Dubliners (see Fig. 2). So as to conduct a robust analysis as per Eder’s findings, Finn’s Hotel has been grouped as a singular text throughout this study. In addition to Eder’s research into length, there is another, far simpler reason as to why Finn’s Hotel should be kept as a single text for this experiment. We are concerned with comparing the fragments against Joyce’s entire canon for the purposes of identifying their similarity to his other works. Thus, it makes sense to group the vignettes as they were published - as a collection. Otherwise, they would only cluster as a group themselves, which would be meaningless, as all this would tell us is that the fragments within Finn’s Hotel are stylometrically similar, which is to be expected. We do not want to compare Finn’s Hotel to Finn’s Hotel, but rather, Finn’s Hotel with Joyce’s remaining oeuvre.


Fig. 2. Cluster analysis of Joyce corpus


When segmented, Joyce’s texts typically cluster (see Fig. 2), demonstrating, stylometrically, that his style remains consistent across particular novels. Close reading is necessary to delineate those exceptions that do exist, an undertaking that is beyond the scope of this particular study. To validate these results, I conduct a “bootstrap consensus tree” with a consensus strength of 0.5, availing of Burrows’ Delta once again to perform a cluster analysis over maximum frequency words ranging from 100 to 1,000, in intervals of 100 (see Fig. 3). This approach is based on the same principles as the previous analyses, but by “repeating” the process of classification in an iterative fashion, ensures that results are more accurate. Clusters are then visualised in a tree-like manner.


Description: Description: Description: Description: Description: Macintosh HD:Users:josullivan:Desktop:Screen Shot 2013-12-07 at 18.25.56.png

Fig. 3. Cluster analysis of Joyce corpus as bootstrap consensus tree


There is little difference between the initial cluster analysis as represented by the dendrogram (see Fig. 2), and the subsequent consensus tree (see Fig. 3), correlating these findings.

            As the majority of Joyce’s texts cluster together, it is significant that Finn’s Hotel aligns with particular sections of Finnegans Wake and Ulysses. When clustered as complete texts, Finn’s Hotel is closer to Finnegans Wake (see Fig. 1), suggesting that, when taken in their entirety, Finn’s Hotel is stylometrically more similar to Finnegans Wake than any of Joyce’s other works. When texts are fragmented, Finn’s Hotel continues to cluster with sections of Finnegans Wake, but it also aligns with the “Oxen of the Sun”, “Eumaeus” and “Ithaca” episodes in Ulysses. Were it not for the distance between these three episodes and the remaining fragments of the novel, focusing on this portion of the results may potentially serve to confuse matters. My interpretation of this particular clustering would be that these results do not demonstrate that Finn’s Hotel is stylometrically similar to both Finnegans Wake and Ulysses, but that the three episodes from Ulysses, like Finn’s Hotel, are statistically similar to the style adopted by Joyce in Finnegans Wake. Again, a complete literary interpretation of these findings would only serve to detain us unnecessarily. What we can conclude from the macro-analyses conducted in this study is that Finns Hotel, in relation to style – while not conclusively – is certainly closely aligned to Finnegans Wake. What is more certain, is that “the multi-modulated voices of the Wake” are just that, “voices”. When one considers style, Finnegans Wake remains consistent throughout, a trait which appears typical across Joyce’s canon. These findings show, statistically, that Joyce’s style is not “multi-modulated”, and thus, Finn’s Hotel is particularly similar to Finnegans Wake. These results do not conclusively resolve the issue of how we situate Finn’s Hotel within the Joycean canon, but they do contribute to the debate, supporting the view that these works should not be treated as a standalone collection. If we accept style as a tool suited to literary classification, then the fragments within Finn’s Hotel are most likely drafts for what became Finnegans Wake. As I have attempted to outline, there are limitations to this approach which would be supplemented by close, as opposed to purely distant, reading. Statistically, however, the criticisms of Ithys Press are seemingly upheld.




With thanks to Dr Órla Murphy and Prof Graham Allen, School of English, University College Cork, as well as Maciej Eder, Pedagogical University of Krakow, and Jan Rybicki, Jagiellonian University, Krakow. This research is generously funded by the Higher Education Authority, under the Programme for Research in Third Level Institutions, Cycle 5, and supported by the School of English and College of Arts, Celtic Studies and Social Sciences, UCC.


[1] “FINN’S HOTEL by James Joyce.” Ithys Press. Web. 15 Oct. 2013.

[2] Ibid.

[3] Pamela Duncan. “Publication of James Joyce Collection Divides Scholars.” The Irish Times. 15 June 2013. Web. 15 Oct. 2013.

[4] Ibid.  

[5] “FINN’S HOTEL by James Joyce.” Ithys Press. Web. 15 Oct. 2013.

[6] Pamela Duncan. “Publication of James Joyce Collection Divides Scholars.” The Irish Times. 15 June 2013. Web. 15 Oct. 2013.

[7] Alison Flood. “James Joyce’s ‘Last Undiscovered’ Collection to Be Published.” The Guardian. 14 June 2013. Web. 15 Oct. 2013.

[8] Pamela Duncan. “Publication of James Joyce Collection Divides Scholars.” The Irish Times. 15 June 2013. Web. 15 Oct. 2013.

[9] Alison Flood. “James Joyce’s ‘Last Undiscovered’ Collection to Be Published.” The Guardian. 14 June 2013. Web. 15 Oct. 2013.

[10] Ibid.

[11] Sidney Feshbach. “Why so Much Fuss?” James Joyce Quarterly. 35.4/36.1 (1998): 824-830. Print. pp. 826.

[12] Hans Walter Gabler. Rev. of The Textual Diaries of James Joyce, by Danis Rose. James Joyce Quarterly 33.4 (1996): 621-625. pp. 624.

[13] John Burrows. “‘An Ocean Where Each Kind. . .’: Statistical Analysis and Some Major Determinants of Literary Style”. Computers and the Humanities 23.4-5 (1989): 309–321. Print.

---. “‘Delta’: a Measure of Stylistic Difference and a Guide to Likely Authorship”. Literary and Linguistic Computing 17.3 (2002): 267–287. Print.

[14] David L. Hoover. “Frequent Collocations and Authorial Style”. Literary and Linguistic Computing 18.3 (2003): 261–286. Print.

---. “Frequent Word Sequences and Statistical Stylistics”. Literary and Linguistic Computing 17.2 (2002): 157–180. Print.

---. “Testing Burrows’s Delta”. Literary and Linguistic Computing 19.4 453–475. Print.

[15] Matthew L. Jockers. Macroanalysis: Digital Methods & Literary History. Urbana: University of Illinois Press, 2013. Print.

[16] Jan Rybicki and Maciej Eder. “Deeper Delta Across Genres and Languages: Do We Really Need the Most Frequent Words?” Literary and Linguistic Computing 26.3 (2011): 315–321. Print.

[17] Maciej Eder, Mike Kestemont and Jan Rybicki. “Stylometry with R: a suite of tools.” Digital Humanities 2013: Conference Abstracts. University of Nebraska-Lincoln, Lincoln (2013): 487-89.

[18] Maciej Eder. “Does Size Matter? Authorship Attribution, Small Samples, Big Problem.” Literary and Linguistic Computing (2013): fqt066. Web. 5 Jan. 2014.

[19] Ibid.